INTER-CENTRE COOPERATIVE RESEARCH PROGRAMME

Project n°J: Final Report

PROGRAMME DE RECHERCHE COOPERATIVE INTER-CENTRES

Projet n° I: Rapport final

INFANT AND CHILDMORTALITY

IN THE THIRD WORLD•

MORTALITÉ INFANTILEET JUVÉNILE •

DANS LE TIERS MONDE

CICRED WHO/OMSPARIS

1983

INFANT AND CHILDHOOD MORTALITYIN THE THIRD WORLD

PageFOREWORD by Harald HANSLUWKA,of the World Health Organization. 3INTRODUCTION by Jean BOURGEOIS-PICHAT, Chairman of theCommittee for International Cooperation in National Research inDemography 5LIST OF THE PAPERS contributed to the project 7FINAL REPORT by Hugo BEHM, General Coordinator 9

SELECTED CONTRIBUTIONS

La mortalité infantile et de la petite enfance dans les pays africains lusophones : niveaux ettendances, by Carlos A. da Costa CARVALHO, Centro de Estudos Demográficos,Lisbon (Portugal) et Nations Unies (Rwanda) 37

Influencia del sector salud en los niveles de la mortalidad infantil cnilena, by BertaCASTILLO M., Fresia SOLIS F., Graciela MARDONES A., Escuela de Salud Publica,Facultad de Medicina, Universidad de Chile 49

Community Variations in infant and Child Mortality in Peru: A Social Epidemiológica!Study, by Barry EDMONSTON and Nancy ANDES, International PopulationProgram, Cornell University, Ithaca, N.Y. (U.S.A) 71

Some Factors Associated with Infant Mortality in Mexico, by Irma Olaya GARCIA Y.GARMA, El Colegio de Mexico (Mexico) 91

Mortalidad infantil en Cuba ; su comportamiento en el decenio 1970-Ï979, by RaulRiVERON-CoRTEGUERA, José A. GUTIÉRREZ-MUÑE, Francisco VALDES-LAZO,Instituto de Desarrollo de la Salud, La Habana (Cuba) 129

Child Mortality in Different Contexts in Brazil: Variation in the Effects of Socio-economicVariables, by Diana Oya SAWYER and Elidimar Sergio SOARES, CEDEPLAR,Federal University of Minas Gérais (Brazil) 145

La mortalité aux jeunes âges : un essai d'approche explicative inter-disciplinaire, byDépartement de démographie et Unité d'épidémiologie (Université Catholique deLouvain) and Unité de nutrition et Unité de santé publique (Institut de médecinetropicale d'Anvers) (Belgique) 161

SUMMARIES

Differential Infant and Child Mortality in Costa Rica, ¡968-1973 (Michael R. HAINES andRoger C. AVERY) 177

Differentials in Infant and Child Mortality and Their Change over Time : Guatemala,1959-1973 (Michael R. HAINES, Roger C. AVERY and Michael A. STRONG) 178

Health Problems in Perinatal Period and Infancy in a Rural District of Thailand (PensriKHANJANASTHITI and Vilai BENCHAKARN) 180

Effects of Fertility on Fetal, Infant and Child Mortality in Bangladesh (Ingrid SWENSON). . . 181

FOREWORD

Infant mortality has traditionally been viewed as an indicator of thesocial and economic well-being of a society. It reflects not only themagnitude of those health problems which are directly responsible for thedeath of infants, such as diarrhoeal and respiratory infections andmalnutrition, but the net effect of a multitude of other factors, includingprenatal and postnatal care of mother and infant, and the environmentalconditions to which the infant is exposed.

Recently, the Member States of WHO have pledged to work together"to attain the goal of a level of health for all the people of the world by theyear 2000 that will permit them to lead a socially and economicallyproductive life". A strategic consensus has been reached on how toimplement this policy. In order to measure and monitor progress towardsthis goal, a set of 12 global indicators has been agreed upon; the inclusionof infant mortality in this list underlines the importance attached toreducing the gap in infant loss between the developed and the developingcountries. It is in the developing countries where, even nowadays, one outof ten newborn will die before the first birthday. Indeed, in 52 countriesinfant mortality even exceeds this figure and there are still some20 countries where every fifth or sixth newborn dies in the first year oflife. Based on the current state of medical knowledge and technology, thenumber of "preventable" infant deaths in developing countries can beconservatively estimated to be of the order of 5 million annually. Thesefigures speak for themselves!

The new social health-oriented strategy of WHO, with its emphasison "Primary Health Care" and the "Risk Approach", i.e. the identificationof population groups with comparatively high risks of maternal and childloss, attaches great importance to statistical support for the properplanning, implementation and evaluation of maternal and child healthprogrammes. It is for this reason that WHO welcomed the initiative ofCICRED to sponsor a research project on "Infant and Child Mortality inthe Third World" and was glad to co-sponsor it. The results of thesestudies convincingly demonstrate that despite the serious constraintsresulting from inadequate and/or defective data sources, valuable insightsconcerning the magnitude and determinants of infant and child loss can beobtained by proper statistical data evaluation and interpretation. At thesame time, it is essential to stress the fact that the results have important

implications for the formulation of appropriate social policies. However,the studies also show that evidence concerning the effects on mortality ofmodern technologies for the prevention and treatment of diseases underadverse socio-economic conditions (which prevail in many developingcountries) is still, in many respects, woefully inconclusive. It is here thatwe are faced with a challenge for follow-up activities in order to obtainmore appropriate and sensitive information and thus overcome thisobstacle to efficient and effective health intervention. The CICREDresearch project is an important milestone and pointer in the rightdirection.

H. HANSLUWKAChief Statistician

Global Epidemiological Surveillanceand Health Situation Assessment

World Health Organization

INTRODUCTION

The CICRED Programme of Inter-Centre Cooperative Research waslaunched by the CICRED Fifth General Assembly, held in Mexico City inAugust 1977. The Programme is aimed at strengthening collaborationamong population research institutions having similar scientific interests.It is based on the principles of mutual assistance and self-help. Theparticipating institutions in each of the Programme's projects join on avoluntary basis and take charge of their share of the work and costs. Theprice of coordination at the inter-institutional level is kept at a minimum.

The project on Infant and Childhood Mortality in the Third Worldwas begun in 1979 - the Year of the Child - when an initiating meetingwas convened at the Carolina Population Center at Chapel Hill, NorthCarolina (U.S.A.). The concluding meeting of the project took place inManila (Philippines) in December 1981. Given the specific interest of theproject, the World Health Organization has kindly agreed to co-sponsor itwith CICRED. The Project's Coordinator is Dr. Hugo Behm and, at theconcluding meeting, Dr. Lado Ruzicka acted as rapporteur.

The present volume provides the project's report, prepared by Dr.Behm, seven reports from national research institutions and abstracts ofreports from four other institutions. The number of papers contributed tothe project is actually 27, but they are at various stages of completion.Those papers which are at a very early stage, or those already publishedelsewhere have not been taken into account. As a matter of fact, theproducts yielded by the project are abundant, but highly diversified. Thisis not surprising, since the Third World countries themselves are deeplydiversified, both from the point of view of mortality and from that of dataavailable. Furthermore, as may be expected, the effort that theparticipating centres have devoted to the project is not. and in fact cannotbe, evenly distributed: the human and financial resources vary accordingto each national institution, both in size and content: the discipline,scientific interest, specialization of the research workers assigned to theproject are multi-faceted: etc. Differences in national situations and inresearch resources contribute to differentiating the state of achievement.Nevertheless, the results of the undertaking allow us to consider theproject globally as a success. Such an achievement comes wholly from thewillingness of population research institutions to cooperate with eachother.

It is the privilege of the CICRED Chairman to express the gratitudeof all the population community to the participating centres, andespecially the Carolina Population Center, the host of the project'sinitiating meeting. The moral and financial assistance of the UnitedNations Fund for Population Activities and the World Health Organiza-tion has enabled CICRED to fulfil the coordination activities of the projectand to publish the present volume.

Jean BOURGEOIS-PICHATMarch 1983

LIST OF THE PAPERSCONTRIBUTED TO THE PROJECT

BEGHIN Y.. BARTIAUX F., BORLEEI.. BOULANGER P.M.. MASUY-STROOBANTG., NZITAD.,SALA-DIAKANDA M.. TABUTIN D.. VANDERVEKEN M., LERBERGHE W., VUYLSTEKEJ. : La mortalité aux jeunes âges : un essai d'approche explicative interdisciplinaire.Département de démographie. Université Catholique de Louvain + Institut demédecine tropicale. Antwerp. Belgium.

BEGUM S. : Bangladesh: Levels and Differentials. Centre for Population Studies, LondonSchool of Hygiene and Tropical Medicine, UK.

BLACKER J. : Experiences in the Use of Special Mortality Questions in Multi-PurposeSurveys: The Single-Round Approach. Centre for Population Studies. LondonSchool of Hygiene and Tropical Medicine, UK.

BLACKER J.. BRASS W. : infant and Child Mortality in Kenya. Centre for PopulationStudies. London School of Hygiene and Tropical Medicine. UK.

BUTANA E. : Child Mortality in British Solomon islands. Centre for Population Studies.London School of Hygiene and Tropical Medicine. UK.

CAMPBELL E. Mortality in Liberia: Method, Levels and Policy Implications. DemographicUnit. University of Liberia.

CASTILLO B., SOLIS F.. MARDONES A. : Chile : factores de salud asociados à la mortalidadinfantil por causas evitables. Escuela de Salud Publica, Universidad de Chile.

CASTILLO B.. SOLIS F., MARDONES G. : Chile •. atención médica y mortalidad infantil en los27 Servicios de Salud del pais : ano 1979. Escuela de Salud Publica. Universidad deChile.

CHAO D. : The Effects of Socioeconomic Development and Fertility Change on InfantMortality: An Econometric Study of Taiwan, 1964-1975. Research TriangleInstitute. North Carolina. USA.

da COSTA CARVALHO C.A. : La mortalité infantile et de la petite enfance dans les paysafricains lusophones : niveaux et tendances. Centro de Estudos Demográficos,Lisbon, Portugal.

EDMONSTON B., ANDES N. : Community Variations on Infant and Child Mortality in Peru:A Social Epidemiológica! Study. International Population Program. CornellUniversity, USA.

GANDOTRA M.M., DAS N. : Infant Mortality and Its Causes in Gujarat State. PopulationResearch Centre, Faculty of Science, Baroda, India.

GARCIA I : Determinants of infant and Childhood Mortality in Mexico. Centro de EstudiosDemográficos y de Desarrollo Urbano, El Colegio de Mexico, Mexico.

HAINES M., AVERY R., STRONG M. : Differentials in Infant and Child Mortality and TheirChange Over Time: Guatemala. 1959-1973. International Population Program.Cornell University, USA.

HAINES M., AVERY R. : Differential infant and Child Mortality in Costa Rica: 1968-1973.Wayne University and International Population Program, Cornell University.USA.

8

HILL A. : Indirect Estimates of Infant and Childhood Mortality in Kuwait. London Schoolof Hygiene and Tropical Medicine, UK.

HILL A., CALLUM C. : Childhood Mortality in Syria. Centre for Population Studies,London School of Hygiene and Tropical Medicine, UK.

HILL A., MOSER K. : Childhood Mortality in Jordan. Centre for Population Studies,London School of Hygiene and Tropical Medicine, UK.

KHANJANASTMTI P., BENCHAKARN V. : Health Problems in Perinatal Period and Infancy atBang Pa In. Faculty of Medicine, Ramathibodi Hospital, Bangkok, Thailand.

KIM O.K. : Childhood Mortality in Korea. Centre for Population Studies, London School ofHygiene and Tropical Medicine, UK.

MARDONES G., CASTILLO B.. SOLIS F. : Chile .• integridad del registro de nacidos vivoscorrespondientes à los 27 Servicios de Salud del pais, anos 1969-1978. Escuela deSalud Publica, Universidad de Chile.

RAHMAN Sh. : Seo-natal Mortality Patterns in Rural Bangladesh. National Institute ofPopulation Research and Training, Bangladesh.

RIVERON R., GUTIERREZ J.. VALUES F. : Mortalidad infantil en Cuba : su comportamientoen el decenio 1970-1979. La Habana. Cuba.

SAMAD S. : Infant and Child Mortality in Pakistan. Centre for Population Studies, LondonSchool of Hygiene and Tropical Medicine. UK.

SOLIS F.. CASTILLO B.. MARDONES G. : Chile .• mortalidad infantil por grupos de causasevitables en los 27 Servicios de Salud del pais, ano 1979. Escuela de Salud Publica.Universidad de Chile.

SWENSON I : Effects of Fertility on Fetal, Infant and Child Mortality in Bangladesh.University of North Carolina, Chapel Hill, USA.

SAWYER D.. SUAREZ E.S. : Child Mortality in Different Context in Brazil: Variations in theEffects of the Socio-economic Variables. Centro de Desenvolvimento e Planeja-mento Regional. Minas Gérais. Brazil.

COMMITTEE FORINTERNATIONAL COOPERATION

IN NATIONAL RESEARCHIN DEMOGRAPHY

INTER-CENTRE COOPERATIVERESEARCH PROGRAMME

Final Report on the Research Project onInfant and Childhood Mortality in the Third World

Hugo BEHMGeneral Coordinator

INTRODUCTION

The Committee for International Cooperation in National Researchin Demography (CICRED) is currently engaged in strengthening thecooperative efforts among population centers having similar researchinterests. In response to an earlier communication from CICRED, anumber of demographic centers agreed to participate in a cooperativeresearch on "Infant and Child Mortality in the Third World". A meetingof the representatives of 14 centers was held from September 3rd to 6th1979 at the Carolina Population Center of the University of NorthCarolina, Chapel Hill, U.S.A.

The objectives of the research programme were defined as follows:a) identify problems and suggest improvements in data collection,

methods of estimation, analysis and international comparison ofresearch findings in the field of infant and child mortality in the ThirdWorld;

10

b) study and compare on an international basis the levels, trends andmain differentials on infant and child mortality in the developingcountries;

c) investigate the mutual relationship between fertility behaviour andchild mortality;

d) attempt to explain the determinants of the infant and child mortalitytransition in developing countries and derive the policy implications.

At the meeting, it was agreed to form five working groups of theparticipant centres for the different subjects.

A work program was defined and centres which would act ascoordinators on each subject were also designated.

The Concluding Meeting of the project was held at Manila,Philippines, on December 17th., 1981. The participant centres contributedwith eighteen papers and W. Brass provided selected information on nineadditional researches carried out at the Centre for Population Studies,London School of Hygiene and Tropical Medicine.

This report summarizes and comments the most significant aspectsof the contributions received as well as the discussion of the ConcludingMeeting, separated in each of the initial objectives'1'. It also includes abrief evaluation of the project.

DATA SOURCES ON MORTALITY

The sources of data used by the authors show that, in developingcountries, mortality registration statistics can rarely be used for mortalitystudies. Only in Chile, Cuba, Taiwan and the small African territories ofCape Verde and Sao Tome and Principe, registration has been consideredreasonably reliable to be used; in Kuwait it is mentioned that since the late60's, this source appears to give trustworthy results.

The alternative source of data most frequently used is the populationcensus, applying indirect methods for the estimation of mortality. Thispreference is explained by the growing availability of population census,improvement of new methods of estimation, and the recent inclusion ofthe necessary questions. This source has been used in 12 studies,particularly in population censuses carried out during the 60's and 70's.Usually estimation have been derived on the levels and trends ofmortality, and on various important differentials. In the studies ofGuatemala, Costa Rica and Brazil, the information provided by thepopulation census and other sources has been used to investigate, with

(1 ) The list of papers contributed is detailed on p. 7-8. The papers are quoted in thetext under the author's name or (he country.

11

some type of multivariate analysis, the relative weight of differentvariables in determining the level of mortality.

In recent decades, the extensive use of fertility surveys, especially inthe frame of the World Fertility Survey, has opened significantpossibilities of analysis on mortality in several countries, in spite of thefact that this was not its main objective(2). The group in charge of thecomparative analysis of WFS could not accomplish it for various reasons.Nevertheless, the project could gather information - at times quite brief- of seven studies of this nature (Mexico, Peru, Jordan, Syria, Kenya,Pakistan and Bangladesh). Besides obtaining indirect estimations, fertilitysurveys permit to obtain direct measures on mortality based on pregnancyhistories. It is also possible to carry out analysis at the microlevel of thefamily, the woman or the child. This possibility has been used, forexample, in Garcia's paper with data of the National Survey on Fertility ofMexico. Finally, fertility surveys enable a more direct study of therelationship between fertility and child mortality.

In spite of all its advantages, the use of fertility surveys for mortalitystudies is limited, among other reasons, because of the fact that mortalityanalysis is not its principal objective. The best alternative are surveyswhich have this specific purpose, where more relevant information formortality analysis is collected. The best results are obtained by prospectivestudies, considering the frequency of errors of retrospective surveys. Thepopulation followed-up can not be small, because death, after all, is not avery frequent event. All this makes mortality surveys more complex andcostly and consequently, of a more restrained feasibility(3). The WorldHealth Organization has conducted child mortality surveys in recent yearsin Afghanistan, Algeria, Sierra Leone, Trinidad-Tobago, Mauritius andSudan'4'. Information on breast-feeding, weaning, morbidity andutilization of health care services is collected in addition to socio-economicdata. As reported by Hansluwka in the concluding meeting, the

(2) Among other papers that show the possibilities of comparative analysis on childmortality with WFS surveys, see: Arriaga E. (1980). Direct Estimates of Infant MortalityDifferentials from Birth Histories. World Fertility Survey Conference. London. CaldwellJ. and McDonald P. (1981). Influence of Maternal Education on Infant and ChildMortality: Levels and Causes. IUSSP General Conference. Manila. Philippines. ChackielJ. (1981). Niveles y tendencias de la mortalidad infantil en base a la Encuesta Mundial deFecundidad. Notas de Población (CELADE). No. 27. and Chackiel J. (1982). Factores queafectan a la mortalidad en la niñez. Notas de Población. No. 28. Taucher E. (1982). Effectsof Declining Fertility on Infant Mortality Levels: a Study Based on Data from Five LatinAmerican Countries. Report to the Ford Foundation and the Rockefeller Foundation.CELADE.

(3) The methods of data collection for mortality studies, with special reference todeveloping countries, were discussed in one session of IUSSP Seminar on Methodology ofData Collection and Analysis in Mortality Studies. Dakar. Senegal. 1981.

(4) Reports are being published by WHO and the respective governments. Summaryarticles on findings from Afghanistan and Algeria were published in World HealthStatistics Quarterly. Vol. 34. No. I. 1981.

12

experience with the country-studies is intended to be evaluated by the endof 1982, before launching new such surveys in the future.

For various reasons it was not possible to have the results of thevarious surveys on mortality carried on in some French-speaking Africancountries. Among the studies of this type, there was available the paper ofKhanjanasthiti and Banchakarn, which consists in a follow-up of 1119new-born in a district close to the capital of Thailand. Swenson analyzesfetal and infant death in a group of pregnancies in 132 villages of a districtnear the capital of Bangladesh, to determine mainly what effect haspregnancy spacing on this mortality. In an investigation of more limitedreach, Rehman studies neonatal mortality in a district of Bangladesh,controlling within 28 days a total of 838 born alive children whose birthhad not technical attendance. These experiences show the advantage ofthe more complete information thus obtained, but likewise show theproblem of the representativeness of local samples and the different bias towhich they are exposed.

The analyzed experience shows that at present and in a near future,mortality studies in developing countries, especially in several countries ofAfrica and Asia, should use information sources other than deathregisters, due to their deficiency. If they are acceptably reliable, theiranalytical possibilities should be fully exploited; this does not alwaysoccur. In some countries, these registers are deficient at a national levelbut acceptable in some regions, frequently in the most important urbanareas. Even with these restrictions, mortality studies of these populationscan be very useful(5).

Without dismissing the possibility of encouraging the developmentof special mortality studies, the collected experience also indicates thatsources of data of more immediate use in many developing countries,mainly in Africa and Asia, are fertility or demographic surveys andpopulation census. As to the former, the possibility of including amortality section should be promoted. Regarding population census it hasgeneralized the inclusion of questions permitting indirect estimations ofchild mortality. It is a matter of taking full advantage of these analyticalpossibilities. The papers of Avery and Sawyer illustrate these perspectivesnot only referred to estimations of levels, trends and differentials, but alsoto the analysis of different determinants of mortality. We will refer furtheron to the importance of elaborating a theoretical framework for analysis.

It is convenient to mention other sources of information for thestudy of mortality. Where health services exist, meaningful information iscollected with better data on the causes of death. The extension of primary

(5) In Brazil, for instance, vital statistics are in general only referred to the Statecapitals. Nevertheless, the Special Group of Demographic Analysis of the SEADEFoundation, has carried out studies of high interest using the death registers of the State ofSao Paulo. See: Segundo Encontró Nacional. Associaçao Brasileira de EstudosPoblacionais. Aguas de San Pedro. 13-19 de Octubro de 1980.

13

health care in many developing countries, mainly in its rural areas, - ifthe program is well accomplished - is generating data on population andthe occurrence of pregnancies, births and deaths. These sources ofinformation are not often used in mortality studies. Such use has theadvantage of linking health professionals who are direct operators ofinterventions aiming to reduce mortality, to demographers and otherprofessional groups interested in the same subject.

The papers presented to the project also show that the problem ofsources of data for mortality studies, less than a controversy onadvantages and disadvantages of different options, should consist in theintelligent use of all the available sources of data on the population understudy. The advantages of an interdisciplinary collaboration is obvious.

THE METHODS OF MORTALITY ESTIMATION

Mortality index can be directly calculated when civil registrar data isavailable, in surveys with pregnancy history and in follow-up studies of agiven population. The need to derive mortality estimation frompopulation surveys or census has promoted the development of a greatnumber of indirect methods. Blacker has made a brief revision of thesemethods. In regard to mortality during the first years of life, the mostusual methods are based on the proportion of dead children out of totalchildren ever born of women classified by age. Sullivan and Trussell havemodified the original method elaborated by Brass in 1968. Furtherrefinements have also been developed by Preston and Palloni, using theage distribution of surviving children, which has the advantage ofeliminating assumptions on fertility patterns.

Indirect estimations are exposed to various sources of error, one ofthem being the nonfulfilment of the assumptions implied in the method.Nevertheless, they seem in general to be quite robust on this respect. Onthe other hand, methods have been developed so as to correct some ofthem(6). It is not important either what is the variant of method used,because they usually lead to similar estimations. The main problemconsists in errors in the declaration of children ever born and of deadchildren. When estimations have been evaluated by comparison withreliable sources, sometimes important differences have been found; forexample, in geographic contrasts of mortality(7). But in general,experience shows that estimations of infant and child mortality are

(6) A description of the several indirect methods of child estimation mortality in: HillK.H.. Zlotnick H. and Trussell J.J. (1981). Demographic Estimation: a Manual of IndirectTechniques. National Academy of Sciences. USA (forthcoming).

(7) Tabutin D. (1979). Mortalité des enfants dans les pays en développement:observations et analyses. Chaire Quetelet 1979. Louvain-la-Neuve.

14

internally and externally consistent(8). The use of indirect methods in thestudy of this mortality has shown that they can contribute with importantfindings on early mortality levels and differentials in countries where nobetter information is available (9>. Naturally, estimations should always bejudged with precaution.

The papers contributed by the centres do not have the purpose of asystematic evaluation of the estimating methods, obviously becauseseveral of them were made in regions where these methods were appliedbecause of lack of better information. When indirect estimates have beencompared with a reliable figure, there has been an acceptable agreement.There are two aspects that should be pointed out.

The method originally developed by Feeney and elaboratedafterwards by Trussell(10) permits to derive estimations of mortality whichrefer to dates prior to the survey or census, that is, the estimation ofmortality trends. This is a significant progress because, in this way, it ispossible to analyze the transition of infant and child mortality in thesecountries at a national level as well as in significant groups of population.Several papers study the consistency of national estimations thus obtainedby comparing estimates derived from two surveys or ceasuses. Resultsobtained in studies of Jordan, Kuwait, Syria and Kenya are in generalquite satisfactory. It is also evident, in this experience, as noted by Blacker,that estimations derived from data of women 15-19 and 20-24 years oldlead to an overestimation of mortality which alters real trends and shouldbe rejected. The reason is probably that children born by mothers of thoseages consist largely of first order births which are generally subject to ahigher mortality than those of higher order. In the first age group, thegreater death risk of early pregnancies is added. It has also been observedthat the use of the North model life tables leads to underestimations ofmortality.

An interesting application is made by Begum in Bangladesh wherepast trends of infant mortality are estimated for subgroups defined by theplace of residence, education, the religious group and other significantcharacteristics. It is not possible to judge on the reliability of the contrastsobserved in levels and trends, except to say that they are coherent withwhat was expected to be obtained. If this type of trend estimations provesto be reliable in further experiences, indirect methods would be providing

(8) Hill K. (1981). An Evaluation of Indirect Methods for Estimating Mortality.IUSSP Seminar on Methodology of Data Collection and Analysis for Mortality Studies.Dakar. Senegal.

(9) Behm H. (1978). Mortalidad en los primeros años de la vida en la AméricaLatina. Notas de Población (CELADE). No. 16.

(10) Feeney G. (1980). Estimating Infant Mortality Trends from Child SurvivorshipData. Population Studies. Vol. 34. No. 1. and Hill K. et. al. Demographic Estimation ...op. cit.

15

a very rich information for the analysis of the factors that affect the courseof mortality in these countries.

The study of Haines et. al. in Guatemala is a good example of theexploitation of census data in order to estimate differentials and theirtrends, as well as to provide information for a multivariate analysis. Theyused the surviving children method, using the age distribution of thesurviving children of the women, obtained by an "own children" styleallocation of children to a mother within the household. The census datawas evaluated and it was possible to correct some errors in the declarationof children ever had. It was also necessary to estimate the number ofsurviving children, information that did not exist in one of the two usedpopulation censuses. It was found that the estimates using the survivingchildren method were higher than those using the Sullivan age model. Aclear explanation of these differences could not be found but it isimportant that "the mortality differentials were usually in the samedirection, although the magnitude changed."

MORTALITY LEVELS AND TRENDS

Table 1 summarizes estimations of national levels of infant mortalityand its trends according to the papers presented by the Centers. Certainly,they do not exactly represent the prevailing situation in all developingcountries; besides, estimations have a variable degree of reliability. In spiteof this, they do emphasize some interesting characteristics.

It is evident that infant mortality is extremely variable in the ThirdWorld, much more than in developed countries, a situation determined bythe many different social, economic and cultural conditions which prevailin those countries. In the 70's, the rate of mortality is relatively low in thethree countries of Latin America: Cuba, Costa Rica and Chile 0". It isestimated that Kuwait, Syria, Korea, Mexico, Sao Tome and Principe havereached an average level of mortality (40-70 per 1000). The ratesapproximate 100 per thousand in Kenya, Peru and Brazil, while inMozambique, Liberia, Bangladesh and Pakistan there is a considerabledelay in the reduction of infant mortality, reaching rates higher than 120per thousand.

With all the restrictions derived from the used methods ofestimation, this table also presents the trends of mortality in the first yearof life in recent decades. They reveal an important fact: in spite of the verydissimilar historic conditions presented in these countries and expressed invery different levels of mortality, there is a general tendency to a decline

(11) Infant mortality rates per 1000 in 1980 are: 19.6 in Cuba. 19.2 en Costa Ricaand 33.0 in Chile.

16

TABLE 1. - Estimates of infant mortality rates, selected developing countries, 1950-1979

Countries

AFRICACap VertKenya (b)LiberiaMozambiqueSao Tome and Principe

ASIABangladeshKorea (a)Kuwait (d)PakistanSyria

LATIN AMERICABrazil (a)ChileCosta RicaCubaGuatemala (a)MexicoPeru

OCEANIABritish SalomonIslands

1950-54

143155

174

112

1238779

146

Infant mortality rates (per 1000)

1955-59

141

140

112149140

14411475

178

129

1960-64

118120

144

82148

10876

39 (e)

100

1965-69

106

86

146

59139

1379066

39 (e)155

75

1970-74

9494

151-176 (c)14270

1585545

12270

120705133

102

50

1975-79

91

54

107472923

60

(a) Mortality index is 2% per 1000.(b) Average of Feeney's estimates.(c) For females, 1964-1974.(d) Mean of kuwaities and non-kuwaitis.(e) Official rates, considered under-estimations.

in rates. Nevertheless, the intensity and chronology of this decline arequite variable. With the exception of Cuba, Costa Rica and Chile, existingrates in the 50's were very high and many of them exceeded up to 140 perthousand. The decline was early and steady with a reduction that canreach up to half of the initial rate, in Kenya and the small territories ofCape Verde and the British Solomon Islands. Of greater significance iswhat we observe in Pakistan, Mozambique and Brazil, which arecountries with a much larger population. The decline of mortality in thesecountries is more belated and of less intensity, in such a way that rates forthe 70's still exceed 80 - 100 per thousand. In Bangladesh, there is noteven a clear trend towards a decline and it could even have a recessiontowards 1970. On the contrary, in Latin America, Costa Rica, Cuba and

17

Chüe - this latter in a smaller measure - were, in the 5O's, in a moreadvanced phase of mortality transition with a level similar to the onereached in recent years by the already mentioned African and Asiaticcountries. These countries in which mortality transition occurs earlier, arealso characterized by a continuing marked decline, which maintains itselfuntil the most recent date.

The analysis of the historic transition of mortality occuring in thepast, in today's advanced countries, shows that infant mortality was thelast to join the process of decline. In several countries of Europe,prevailing rates at the end of the XIX century bordered 200 per thousandand sometimes a clear decline did not take place until the early XXcentury. Nevertheless, there are two fundamental differences between thishistorical experience and what is happening today in developingcountries. The first is favorable to the latter ones and refers to the currentexistence of impressive technological advances in the prevention ofdisease and death. Of course, these means did not exist at the time whenthe developed countries initiated the decline of early mortality. On thecontrary, they then benefited for being centres of colonial empires whosedominions contributed to their own progress. On the opposite, today'sThird World is characterized by its dependency on central economies.Some of the studies of mortality determinants, presented by the Centres,help us to examine how this contradiction is being solved.

As mentioned in the initial document prepared by the Coordinatorof the project(12), the diversity of current situations in regard to mortality,as shown by the mentioned estimates, express the very different historicconditions of the studied populations. Thus, its comparative study shouldbe an important source for its better interpretation. This project achievedthis purpose only in a limited way but in any case, it has been the sourceof important suggestions for further investigation, as will be seen furtheron.

MORTALITY DIFFERENTIALS

The papers presented explore differentials of infant and childmortality in terms of available variables in censuses and surveys. Thefinding of these differentials does not explain "per se" the mechanismsdetermining mortality. Nevertheless, these differentials describe signifi-cant contrasts of mortality and are the basis to elaborate hypothesis on itsgenesis. We will consider the most significant differentials, mainlyassociated to geographic variables and to maternal education. Its

(12) Behm H. (1979). Infant and Child Mortality in the Third World: BackgroundInformation and Proposals for Cooperative Studies Among Demographic Centers.CICRED.

18

significance will be discussed further in relationship to multivariateanalysis.

Geographic contrasts within countries should be interpreted withcaution because basic data has a variable reliability in the differentregions. Furthermore, indirect estimations are subjected to bias linked tointernal migration. Table 2 summarizes geographic differentials. Certainlythese differentials are properly socio-geographic ones, because - with theexception of special cases, as death by malaria-they mainly expressdifferences in social and economic mortality determinants of each region.

The table seems to show that when mortality is extremely high,geographic differentials are of less importance and that high mortality is asmuch a characteristic of urban as well as rural region. This is the situationof Bangladesh in 1960, Pakistan in 1966-76, Guatemala in 1959 andMexico when birth cohorts are included since the year 1940. Generally,the excess of mortality in rural populations is not higher than 10%.

When mortality declines, the geographic differential increasesbecause the decline is higher in urban zones. In this way, rural rates canduplicate urban mortality. It is also noticeable that regions which includelarger cities, especially the national capital, have relatively low mortality.This evolution is quite clear, for example, in Guatemala where estimationsof ^ are compared for 1959 and 1968. The excess of rural mortality overthe urban one has increased from 6 % to 48 %, due to the fact that duringthis period, rural mortality has not a significant decline while in the urbansector it reaches 32%. In the population of the capital city of Guatemala,the decline reaches 48 %. In two countries of relatively low mortality,Chile and Cuba, geographic contrasts are considerably more intense in thefirst mentioned country. In Brazil, the greatest difference is not found inthe urban/rural contrasts; on the other hand, the backward Northeastduplicates the mortality of the Southern Region, which has the greatestindustrial centre of Latin America.

The pattern of change of urban/rural mortality observed in thesecountries is not the same as the one registered in the past in to-dayadvanced countries. Towards the end of the XIX century, for example,the rate of urban infant mortality exceeded the rural one in approximately50 % in Norway and Sweden; a more intense decline in the cities madethis differential disappear in the first third of the XX century.Nevertheless, in Bulgary we can find the pattern of the earlier and mostsustained urban decline since the 1930's(l 3).

The factors that likely determine geographic differentials arediscussed further on. It is convenient to emphasize that rural mortality ishigher and that it makes less progress in countries of the Third World. In

(13) United Nations (1978). Determinants and Consequences of Mortality Trends.ST/SOA/SER.A/50.

19

TABLE 2. - Geographical differentials of infant and child mortality,selected developing countries, 1940-1979

Country

AFRICAAngola

Mozambique

ASIABangladesh

Jordan

PakistanLATINAMERICA

Brazil (b)

Chile

Cuba

Guatemala

Peru

Mexico

Year

1940

1970

196019681972

1966-76

1966

1979

1979

195919681968-77

Cohort1940-76Cohort1972-76

Mortality

index (a)

l i o

2«Jo

IMR

IMR

l i o

Death per

1000CEBfem. 20-29

IMR

IMR

2<JoInf + childMortality

ratio tonation, mort.

i i o

Higher mortality

populations

Luanda(capital)

Centre

RuralRuralTown andlarge vil-lagesMedium +

279

155

142146

93

small villages 95Rural

Northeast

UrbanNortheastruralSouthernprovinces

EasternprovincesRuralRural

145

117

109

47

20-26182172

Andean region4. IOCK

Rural +

Rural

Rural

t*7 /\J

31%

97

97

Lower mortality

populations

MalangaBiéBenguelaSouthNorth

UrbanUrbanAman(capital)Zarka-Irbia

Urban

South urban

South rural

Provinceswith largestcitiesCentral

229)225246)126)137)

134)116)70

73

116

58

68

21

provinces 15-20UrbanUrbanLima (cap.)-

Urban

Urban

Urban

171116

55%

• 9 %

79

70

Ratiohigher/lower

1.20

1.17

1.15

1.31

1.25

NE/S =1.79

2.23

1.31

1.061.48

-

1.22

1.39

(a) Per 1000 births.(b) Effects of other intervenient variables controlled.

IMR: Infant mortality rate. CEB: Children ever born.

20

these countries precisely, rural population is the majority; thus, the courseof child mortality in these regions is determined by the progress that canbe achieved in these rural populations.

Infant and child mortality differentials by maternal education arepresented in Table 3. As extensively described, the level of educationreached by the mother is inversely associated with the child death anddetermines important differentials. In general, children of illiterate womenhave a mortality that is at least 80 % greater than if the mother reachessecondary education; in some cases, it is even greater. Regression isapproximately linear, although reduction of child mortality is greaterwhen one passes from the condition of illiteracy to some level ofelementary education. In contrast with geographic differentials, the onesassociated to maternal education are observed in all the countries studied,whatever the level of mortality may be. Contrasts are always important.

Garcia has analyzed in Mexico first year of life mortality separatelyfor the neonatal and the post-neonatal period. Her information shows thatthe contrast associated to maternal education is higher in the ages of 1-11months. The rates for the groups without any education and withinstruction higher than secondary are, in neonatal mortality, 54 and 34per 1000 and in post-neonatal, 52 and 10 per 1000, respectively. This is inaccordance with a more direct dependency that causes of post-neonataldeaths have in respect to material conditions of life in the family.

Available information enables to analyze another point of interest:the tendency of early mortality by groups of maternal education. InBangladesh the increase of mortality towards 1971, that the authors relateto famine, deteriorate the mortality of the groups that already sufferedhigh rates in relation to the nonexistent or deficient maternal education,while among children of more educated women, mortality apparentlyshows a discrete tendency to decline.

On the other hand, the comparative study of Haines in Guatemala,between 1959 and 1968, points out that the progress made in themortality of children of illiterate women is negligible (7 % of reduction),while the decline is more intense if the educational level of woman ishigher, reaching 60 % among those having middle or higher education.Due to these differential trends, the excess of mortality in the group ofilliterate women in regard to the ones of higher education has raised from1,87 to 4,32 in the mentioned period. In other words, the ones who havebenefited with the progress of preventing child death have been thegroups which were initially favoured with lower mortality.

Studies made in Pakistan and in Guatemala enable to jointly analyzemortality according to urban/rural residence and maternal education. Inboth cases it can be found that rural surmortality diminishes in part if wecompare within an equal level of education, indicating that part of theurban/rural contrast of mortality is explained by the lowest levels ofeducation of the rural population. Both studies also point out that high

21

o f ia5 T3 S)5 «

12ca o

§3o.S

c

I

m — - ^ «

•a

•a > v

pr

3

Is11

I— O v o O >Oo — 1 m 00

O O O §

0 •* "1

z00 cVO O

- z

01c 0

z -o O

- - z— z

•oi5

^ ou< c<2 03

C•gO

Iaa.

a

M3O

00

!

ac

ío

3 O

o E

1*111 Ï

IÎIa. £ gQ ¿¿ a-

22

mortality of children of illiterate women is quite similar in urban andrural sectors. Thus, benefits of large cities apparently do not reach sectorsof less education, where children are exposed to a mortality degree as highas the one prevailing in rural areas. All these facts indicate that, more thanthe geographic place of residence, child mortality is associated to the formin which parents are inserted in the social structure of production of agiven society. This point will be analyzed further on.

In the paper of Pakistan, differentials of child mortality have alsobeen explored as to the level of instruction of the father, finding similardifferences to the ones observed in accordance with maternal education.As is generally known, child mortality is more associated with maternaleducation than to father's education.

There are a certain nomber of other differentials of infant and childmortality which have been analyzed in various papers, and whichgenerally confirm what is already known in this respect. It is found thatthere is a discrete excess of male mortality. It is interesting to note that thisdifferential tends to disappear in Bangladesh among the children ofilliterate women, exposed to an extremely high mortality. This suggeststhat differentials of biological prevalence may have less relevance whensocioeconomic determinants generate a high level of mortality. Rahman,in Bangladesh finds a somewhat higher rate of neonatal mortality ingroups of muslim religion in respect to hindus groups; nevertheless,Begum, in the same country, describes the opposite situation based onrates of infant mortality indirectly estimated.

In Guatemala, Haines compares indigenous and non-indigenouspopulations, the former group showing higher mortality in the first twoyears of age; the differential is more marked in the urban medium. Thecomparison between 1959 and 1968 shows again that the nativepopulation maintains its high mortality while there is a decline of 23 % inthe non-native population. Progress in the native groups has been limitedto the urban sector where it is always lower than the one obtained by thenon-native groups.

In the paper of Mexico, a higher mortality is found when there is nomaternal breast feeding, a condition which nevertheless has occurred inonly 11 % of the interviewed women. In Bangladesh, child mortality is

1 higher if the mother is a widow or divorced as compared with currentlymarried ones and when residence has a worse hygienic conditions. Theassociation of mortality with pregnancy order, age of mother andpregnancy spacing has also been analyzed and those results will beconsidered when discussing the relations of fertility with infant mortality.

The above description of infant and child mortality differentialsaccording to a certain number of variables could hardly constitute acomprehensive and logical explanation of the mortality conditions foundin each population. Among other reasons, there is a collinearity amongseveral of these variables; the observed contrasts are not necessarily

23

generated by the observed variables. On the other hand, variablesinvestigated are a non-organized set of indicators of some of the factorsthat may intervene in the genesis of mortality. They have not beenselected according to a theoretical framework of analysis, but are ratherthe result of data available in census or surveys. This aspect is discussedfurther on.

Nevertheless, even with these restrictions, some of the describeddifferentials and their changes in the course of time are significant forunderstanding the problem of Third World infant and child mortality.They show that the main part of the mortality problem in these regions isconcentrated in groups that are presently defined according to their mostunfavorable conditions as to education and residence, as a consequence ofhaving less access to benefits of social progress. They also point out that inthe conditions in which the transition to lower levels of mortality takesplace, the process has fundamental contradiction. The decline occursmore intensively in the already more favored groups, while it is smallerand sometimes insignificant in human groups which are a majority. If theexplanation given in the case of Bangladesh is correct, they also point outthat when the situation is deteriorated, the groups of higher mortality willbe again the ones to suffer most.

CAUSES OF DEATH

The papers which refer to the causes of death are a minority andshow that the structure of these causes varies according to the levels ofmortality. In the rural district of Bangladesh studied by Rahman, where"women mostly depend on what food is left over... 90% aremalnourished and pregnant women are kept hungry," neonatal mortalityis as high as 70 per thousand. Here mortality by infectious diseases isclearly predominant: 39% by tetanus neonatorium. 17% by respiratorydiseases.

On the contrary, the analysis of the trends of groups of causes inCuba 1970-1978, when the rate of infant mortality declined from 38.8 to22,3 per 1000, shows that 61 % of this decline is due to the reduction ofinfectious causes, mainly diarrhoeal diseases, sepsis and respiratoryinfections. The group of perinatal causes contributes with 30% to thedecline of relation with improvements in the care of the child birth andthe newborn. In 1979 infectious causes represented only 26% of infantmortality in Cuba and deaths due to malnutrition were practically nil.This means that, even within limitations that developing countries arefacing, effective policies contributing to the improvement of materialliving conditions and access of the whole population to the benefits of

24

health care, can produce important reductions of many of the mortalitycauses that are prevalent in other populations(14).

Solis et. al. analyze in Chile, 1979, when the infant mortality rate is38 per 1000, deaths by causes considered avoidable, which total up to84.5 %. The analysis of 27 geographic areas show that in the zones ofhigher mortality, this excess is principally linked to deaths caused byrespiratory infections, diarrhoeal diseases and malnutrition.

INFANT MORTALITY AND FERTILITY

The influence of fertility on mortality during the first year of life wasanalyzed in the papers of Mexico, Costa Rica, Guatemala and Bangladesh.It was confirmed that pregnancy of women less than 20 years old and ofwomen older than 35 years of age, was associated to greater risks of childmortality. This association persisted after controlling other variables.Haines et. al. find in Guatemala as well as in Costa Rica, two countrieswith very different levels of fertility, that general fertility rates and averagechidren born alive per women aged 20-34 have in the multivariateanalysis- a significant and direct association with child mortality. His"major conclusion is that there seems to be a significant causal role offertility on mortality, as well as the reverse."

Swenson presents a research especially designed "to determine howmuch of an effect pregnancy spacing will have on reducing fetal andinfant mortality." The study is carried out in a rural population ofBangladesh, of high mortality and fertility (total fertility rate 6,0-6,6 per1000 and infant mortality rate 115-150 per 1000) where less than 4% ofthe married women between ages 15-49 were using a modern method ofcontraception. Results suggest that if all second and higher orderpregnancies were preceeded by intervals of 12-24 months, post-neonatalmortality would decline by 19 % ; if intervals were longer than 24 months,mortality would be expected to decline by about 44 %. It was found thatpregnancy spacing would have greater impact in decreasing early fetaland post-neonatal mortality than that obtained by averting births in thehigh risk subclasses for other maternal factors (pregnancy order, age ofmother, previous pregnancy and child losses). Garcia, in the analysis ofthe fertility survey in Mexico, also confirms that when the spacing is less

(14) A similar evolution has been described in Costa Rica (Ministerio de Salud(1978). Costa Rica : Extensión de los servicios de salud. Comunicación a la ConferenciaInternacional sobre Atención Primaria en Salud. Alma Ata, Unión Soviética) and with lessintensity in Sri Lanka (Meegama, S.A. (1980). Socio-economic Determinants of Infant andChild Mortality in Sri Lanka: an Analysis of the Post-war Experience. World FertilitySurvey Scientific Reports. No. 8).

25

than 12 months, neonatal as well as post-neonatal mortality increases in2-3 times. This association maintains itself in the multivariate analysis.

As indicated by Swenson, high fertility behaviour and infantmortality are both associated to mutual unfavorable socio-economicconditions. In this sense, policies intended to reduce both variables shouldaim at the change of this context deep rooted in both phenomena. In spiteof it, the described findings suggest that measures specifically directedtowards the reduction of fertility and the increase of pregnancy spacing,can have a collateral effect which will contribute to the decline ofexcessive infant mortality.

None of the papers analyzes the opposite effect, that is, the actionthat high infant mortality can have in maintaining an also high fertility,which was the main subject of a previous CICRED meeting(15).

THEORETICAL FRAMEWORK OF ANALYSISFOR THE STUDY OF MORTALITY

Diseases that directly lead to child death are phenomena of biologicalnature at the individual level. But the health-disease process in acommunity is often and primarily determinated by historic characteristicsof the social formation in which the child is born and living. By manymechanisms these characteristics define the material conditions of life atthe home and the care given to the child. On the other hand, socialpolicies - particularly health policies - which affect the frequency ofdiseases and death, are also subjected to a social determination.

The analysis of the determinants of mortality in a population is acomplex matter, because it is associated to a great number of frequentlyinterrelated biological, social, economic, and cultural features of theindividuals, the family and the community. It is obvious the advantage ofa theoretical frame, based upon an adequate social theory, to approach theproblem, allowing the elaboration of hypothesis leading to the attainmentof proper empirical referents.

Efforts to elaborate this theoretical framework on the genesis ofmortality have been limited(16). That is why the contribution of Beghin et.al. has a special significance, since it means a multidisciplinary approachto explain mortality in infancy and childhood. It is the result of a

(15) CICRED (1975). Seminar on Infant Mortality in Relation to the Level ofFertility. Bangkok. Thailand.

(16) See for example: Meegama. S.A. Socio-economic Determinants... op. cit. andMoley W.H. (1980). Social Determinants of Infant and Child Mortality: Someconsiderations for an Analytical Framework. Health and Mortality in Infancy and EarlyChildhood: Report of a Study Group. Cairo. The Population Council Regional Papers.

26

teamwork of demographers, sociologists and medical doctors specializedin epidemiology and nutrition, in the thesis that the problem is alsomultidisciplinary.

The authors assume that the mortality rate is the product of thefrequency of illness in a population (incidence) and the frequency of deathamong sick people (lethality). An analytic table is thereafter designedwhere, for each one of these two rates, it is specified the chain of eventsthat link variables of immediate action with the more far-fetched causes,such as social and economic determinants. In an example of application tothe infectious diseases, incidence appears linked to factors relative totransmission, characteristics of the agent and child susceptibility; this inturn is a function of its resistance and inmunity. At the same time,lethality appears determined, in addition to the characteristics of infection,by the opportunity and quality of the treatment, in relation withsocioeconomic determinants which affect the demand and supply ofmedical care services. Thus, the problem appears to be associated, at thelast instance, with the stage of development of the educational system, thehealth service, etc. Completed the design, the model can be elaboratedleaving only the most significant steps. The authors recognize that theframe of analysis has several questionable simplifications; that there aresome unsolved problems and that it requires improvement.

The contribution of this group is important, beyond any doubt. Ithas attempted to cross the barriers separating, several scientific disciplines.It is an effort to link biological variables directly connected to illness anddeath, with its determinants in the social and economic structures.Probably in view of previous experiences, the conventional biological andepidemiological steps appear more developed in the examples.

It is perhaps convenient to add that, in the background of anytheoretical framework of this type, there is always an ideological issue.The social determinants are to be framed within a more generalinterpretation of society. Some type of functionalist approach on theproblem can be used, or otherwise a materialistic, dialectic interpretation.This should always be an explicit matter.

In any case, the contribution of the Belgium group indicates apromising road and points out the advantages of an interdisciplinarycollaboration. The authors are planning to make a practical application ofthe method in a cooperative research to be carried out in Brazil.

ANALYSIS OF THE DETERMINANTS OF MORTALITY

Five studies that carry out a multivariate analysis are available, all ofthem of Latin American countries: Brazil, Costa Rica, Guatemala, Mexicoand Peru. Some of them are performed at the level of aggregates of

27

population, other at the individual levels. The independent variable isexpressed in conventional functions ( ^ standarized proportion of deadchildren) or as a ratio of expected and observed deaths. Independentvariables used are quite heterogenous in type and number. They refer ingeneral to indicators of fertility, family socio-economic status, householdsanitary conditions, work and marital condition of women, availability orutilization of medical care services. In Guatemala the proportion ofindigenous population is added and certain agriculture characteristics. Themethod of analysis is also variable: multiple classification analysis,multiple linear regression, etc.

It is known that comparative analysis of these studies runs intodifficulties because of the diversity of methods and information. In spite ofthem, the set of papers has contributed with findings of great interest.

The variable associated to child mortality in a most frequent andsignificant way is maternal education. The study of Guatemala is the onlyexception, apparently because of the rough character of the indicator(percentage of illiterate women in female population of 20-34 years). Theassociation is maintained in spite of controlling the effect of othervariables in the study. Of all the important effects, education shows thedifferences of mortality whose magnitude has more practical significance.In the study of Costa Rica, at a micro-level in spite of the lowproportion ofthe variance explained, the elasticity of response of child mortality tochanges in education, indicates a 6-7 % reduction in 2% for an increase ofone year in maternal education.

A finding of interest for the formulation of social policies is the onedescribed by Haines in Costa Rica. He has found "a much greaterelasticity of responses of child mortality to an additional year of educationfor women of less education than greater education (11 percent versus2 percent). Further, increases in socio-economic status had much greatereffect on child mortality among less well educated than among the bettereducated. This suggests that the greatest gains in child mortality reductionmight be achieved by raising educational attainment and the levels ofliving of the less well educated." This is another way of saying thatprogress in reducing child mortality can be greater if they are aimed tosocial groups actually less favored and of higher mortality.

Sawyer has comparatively analyzed in Brazil the effect of a numberof economic and social conditioning factors on child mortality within fourdifferent contexts: the advanced, industrialized Southern regions versusthe underdeveloped Northeast region, both subdivided in urban and ruralpopulation. With the exception of the rural Northeast, child mortality ofilliterate women is twice as high as that of better educated women, theeffect of other variables being controlled (table 3). In the South, maternaleducation almost cancelled rural/urban differentials. But it is also clearthat there are contextual variables related to the different regions, whichhave not been controlled in the analysis. In urban regions and for the

28

same level of maternal education, mortality is always higher in theNortheast than in the Southern region. These other determinantconditions, according to different evidences, would be "the overall lack ofadequate sanitary infrastructure, of preventive and curative. medicalservices, and worst, the overall level of poverty that predominates in theNortheast."

The Northeast rural region is an exception in respect to the restbecause child mortality does not show the gradient related to maternaleducation, on account of the predominant effect of the very adverse socio-economic condition which prevails in that region.

The interpretation of the association of child mortality with maternaleducation has been widely discussed(17). For certain authors, it is due tothe fact that education is a proxy of the level of living. On the other hand,Caldwell and others'2' think that in addition to that, education is animportant force by its own. He thinks that besides giving knowledge andskills, and to favour breaking with tradition, the main factor is thateducation changes the traditional balance of familiar relationship and so,the educated mother has more weight in the decision-making processrelated to the care of her children. Sawyer shows that, in Brazil, theeducation variable may reflect to a considerable degree, in addition toknowledge, levels of wealth and power, whether because it tends tocoincide with these aspects of socio-economic position or because, incontemporary society, education is increasingly important in determiningone's income(18).

The discussion has a practical interest. In many countries, it ispossibly more feasible to implement an equalitarian policy for socialbenefits such as education, rather than on the distribution of wealth. Ifeducation could have action per se on child mortality, it could be expectedthat such policies contribute to the reduction of actual levels of mortality.

The household sanitary conditions (provision of water and wastedisposal) is another factor that is outstanding and determinant in childmortality in several of the studies, although not in all of them. In Brazil, itis the second in importance, after maternal education, and has a moreevident association in urban population. Again, it is observed that, onequal terms of sanitary conditions and other variables, the regions present

( 17) See for example: University of Michigan. School of Public Health and WorldBank (1981). International Symposium on Literacy. Education and Health Development.Ann Arbor. Michigan. March 17-19.

(18) The social class is a variable of a greater explanatory power in the analysis ofinfant mortality because it belongs to a theoretical concept of higher level. Education isdiscriminated according to the social class as any other social and economic benefit. If theconcept of social class is used, education refers only to one of the mechanisms of influenceof social class on child mortality.

29

different mortalities, showing the effect of other contextual regionalvariables <19>.

An important subject is the evaluation of the impact in mortality thatthe application of adequate health care can have, using moderntechnologies for prevention and treatment of diseases, when those healthinterventions occur in the adverse socio-economic conditions prevailing inmany developing countries. Certainly, these health interventions aresocially determined by the same factors which explain underdevelopmentand high mortality. Unfortunately this is a point scarcely investigated inthe papers commented. In Peru, rough evaluation of the association withavailability of medical facilities did not reach significant results. On thecontrary, in Costa Rica it was found that the proportion of births havingreceived professional care was a powerful predictor of mortality at thedistrict level, accounting for 64 % of the variation on the mortality index.

In a rural population in Thailand, Kahnjanasthiti et.al. found thatperinatal mortality is lower when pre-natal care has been applied (17 and52 per 1000). Also, that mortality rates are lower if delivery care at homehas been made by midwives (6 per 1000) and not by traditional birthattendants (21 per 1000); in absence of all care, mortality rises to 55 perthousand. In this paper, nevertheless, the effect of other interveningvariables is no controlled.

The study of Castillo etal. in Chile is more specific and attempts todetermine what activities of the health system (out of a total of 59 indexeson resources and services provided by the 27 local health services) havegreater impact on child mortality and on the groups of avoidable causes ofdeath. Through a stepwise-regression analysis it was found that the mostsignificant were the proportion of births professionally assisted and thenumber of pédiatrie hours.

Riveron's et.al. paper summarizes the factors that, in the experienceof Cuba, appear as determinants of the important decline of childmortality in 1970-79 with a special mention on the improvements of thehealth system as to the care of the pregnant woman, birth and child. Thereis no attempt, however, to quantify the role which the health care has hadon this evolution.

Among other determinants, it has been found that children ofmigrant women from rural areas to urban ones, only have a moderatelygreater risk of death than resident women.

(19) In the concluding meeting discussion. Merrick presented a multivariate analysisin two urban samples of Brazil, showing that both education and access to water wereimportant mortality determinants. However, for migrant and low income mothers,mortality was less responsive, revealing that effectiveness of educational and environmen-tal policies in reducing mortality is limited by the persistence of other dimension ofpoverty for mothers who did not benefit. (Merrick T.. The Impact of Access of PipedWater on Infant Mortality in Urban Brazil. 1970 to 1976. IUSSP General Conference.Manila. December 1981).

30

Papers show that even within the restrictions derived from the typeof mortality estimations used and the limitations in the availability ofvariables, multivariate analysis can render important contributions to theknowledge of determinants in a given country. Their limitations are alsoevident. The definition of what is really being measured by the indicatorsand what is the interpretation of the results, is always a questionablematter because analysis is more the result of a better use of availableinformation than the product of a rational design. For example, the twovariables used in Guatemala in order to explore the influence of agrarianproduction conditions in early mortality lead to unexpected results, evenwith contrary signs. This forced the author to a reinterpretation of whatthe indicator was assumed to measure.

Sawyer's paper which compares the effects of several variables infour different socio-economic contexts in the same country, reaches animportant general conclusion. In the words of the author, "the basic pointto be made is that the same conditioning variables may have differenteffects in different contexts. There is no single 'most important' variable toexplain mortality wherever it may occur. Rather, the effects of eachvariable on mortality depend on the economic and social surroundings inwhich they occur."

Collected experience shows that in order that the analysis of thedeterminants on child mortality can make greater progress, it seemsnecessary to improve the theoretical framework to approach the problemand to collect information that permits a better elaboration of theintervening variables.

It is also fundamental to better describe the historical stage reachedby the studied population with information that necessarily does not enterthe multivariate analysis but helps to understand the research findings. Inthis sense a promising road could be the case-studies of countries, wherethe whole range of determining conditions of mortality may be properlyevaluated.

EVALUATION OF THE PROGRAM

Upon the initiation of the program a total of 24 demographic centreshad expressed to CICRED their interest in participating in a cooperativeprogram of research for the study of infant and child mortality incountries of the Third World. Only 14 of them, as observed in Table 4,could attend the initial meeting held in Chapel Hill, North Carolina,U.S.A. Budget restrictions prevented the attendance to this meeting ofcentres of less economical resources.

The Coordinator received a total of 27 papers, nine of which werematerials extracted from several researches made at the London Center forPopulation Studies. In relation to the objectives originally approved (as

31

detailed in the introduction), the summary and discussion of the paperspresented in the first part of this report, show that they made acontribution that may be considered significant to the knowledge of levelsand trends of infant and child mortality, and of its principal differentialsand determinants in the developing countries. The contribution is muchmore limited in the field of methods of data collection, methods ofestimation of mortality, and the study of the relationship between fertilityand mortality.

Considering the number of papers and their value for theimprovement of knowledge in the subject under study, it is fair to say thatthe project has had reasonable success. For various reasons, it is possibleto think that the project stimulated the production of some of these papers.

Nevertheless, this was not the main purpose of the program, definedas "strengthening the collaboration among population centres havingsimilar research interest... by eliminating needless duplication of effort, byspeeding the exchange of experience, by sharing supportive facilitieswhere possible, by providing the environment for a more stimulative andcritical intellectual interchange(20)."

With this purpose in mind, at the initial meeting, the participantsagreed to form working groups, each of them with a specific subject andalso entrust a centre to coordinate the work of each group and to elaboratea joint report. In this way, it was thought that collaboration among thecentres would be stimulated for the cooperative research of a subjectchosen by themselves. This did not occur in practice and the final papersare mainly the individual product of their authors. In three out of fivesubjects it was not possible to obtain an adequate contribution of thecentres.

One of the reasons for not having achieved a better process ofcommunication and joint work among the centres lies in the limitation offinancing. The CICRED project never intended to finance internationalcooperative studies. With limited resources it rather was intended toencourage cooperative research of subjects of mutual interest, but on thebasis that the centres could afford the investigators and resources. Inaddition to all this, a reduction of the initial budget allocated prevented thefinancing of the assistance of the center's representatives at the concludingmeeting. This meeting was important because it should provide thecentres the opportunity to discuss their own papers and draw up futurejoint plans. Representatives of only six centres could attend, two of whichhad no contributed papers.

Surely there are other factors which depend on the form in whichmuch of the research of population is carried out. In a competitive systemof grants for financing, it is not easy to stimulate a collaboration which

(20) CICRED (1979). A New Approach to Cooperative Research in the PopulationField.

32

TABLE 4. - Participation of demographic centers in the CICRED cooperative programon infant and child mortality in the third world

Present at Papers Present atCenter Originally initiating contri- concluding

interested meeting bution meeting

AFRICAInstitut de formation et de recherchedémographiques. Cameroon X XDemographic Research Unit, Institute ofStatistical Studies and Research, Egypt X X XDemographic Unit, University of Liberia X X XInstitut national de la statistique, Zaire XNORTH AMERICACarolina Population Center, Universityof North Carolina, USA X X XInternational Population Program,Cornell University, USA X X XPopulation Studies Center, Universityof Pennsylvania, USA X XLATIN AMERICACentro de Desenvolvimento ePlanejamiento Regional, UniversidadFederal de Minas Gerais, Brasil X X X XDepartamento de Salud Pública,Universidad de Chile X X XInstituto de Desarrollo de la Salud, Cuba X XCentro de Estudios Demográficosy de Desarrollo Urbano,El Colegio de México X X X XASIADemographic Research Centre, Facultyof Science, University of Baroda, India X XXMindanao Center for Population Studies,Xavier University, Philippines X X XPopulation Center, India PopulationProject, Lucknow, India XDemographic Research Unit, JadavpurUniversity, India XCentre for Mathematical Studies,Vaschuthacaud, India XFamily Planning CommunicationAction Research Centre, Universityof Kerala, India XCenter for Population and FamilyPlanning, Korea X XPopulation and Development StudiesCenter, Seoul National University, Korea XSocial Sciences Research Center,University of Punjab, Pakistan X

33

Present at Papers Present atCenter Originally initiating contri- concluding

interested meeting button meeting

EUROPEDepartement de démographie, UniversitéCatholique de Louvain, Belgique X X XOffice de la recherche scientifiqueet technique outre-mer, France X X XCentro de Estudos Demográficos,Portugal X X XCentre for Population Studies, LondonSchool of Hygiene and TropicalMedicine, United Kingdom X X

surpasses the'original projects. On the other hand, favorable factors arethe natural relations of language and culture which some demographiccenters of advanced countries maintain with demographic groups of theThird World.

As noted by Bui Dang Ha Doan in his evaluation of the program*2"the experience indicates several measures that could improve the results ofsimilar projects. The subjects selected for their study should be lessambitious and of a greater feasibility, in accordance with availableresources. It would be convenient that the coordination could dispose ofmore resources for the best practice of its functions. It is also clear thatfinancial aid is absolutely necessary so as to permit, for the success of theprogram, the assistance to joint meetings of the representatives of centres,in particular of those pertaining to developing countries.

The CICRED project tried to help solve an important contradiction.While countries of the so-called Third World suffer the burden ofexcessive mortality and offer an exciting field of study to analyze thecurrent transition of mortality, the greatest resources for research arefound in advanced countries. It is obvious the importance of joiningefforts in order to face this challenge, in which the researchers ofdeveloping countries can not be absent. They should contribute with theirknowledge of their own historical situation and their sense of urgency toreduce such excessive mortality. To this aim, the purposes of the CICREDprogram of collaboration for research among the centers of populationremain perfectly valid.

(21) Bui Dang Ha Doan, Tentative Appraisal of the CICRED Project of IntercentreCooperative Research "Infant and Childhood Mortality in the Third World", paperpresented at the Concluding Meeting of the Project.

SELECTEDCONTRIBUTIONS

LA MORTALITÉ INFANTILEET DE LA PETITE ENFANCEDANS LES PAYS AFRICAINS

LUSOPHONES :NIVEAUX ET TENDANCES

par

Carlos A. da COSTA CARVALHOCentro de Estudos Demografleos, Lisbon, Portugal

etNations Unies

INTRODUCTION

Le Cap-Vert, la Guinée-Bissau, le Saint Thomas et Prince, l'Angolaet le Mozambique, constituent l'ensemble des pays africains qui utilisent leportugais comme langue de communication.

Très diversifiés du point de vue de leur situation géographique, deleur étendue, de l'effectif et de la dynamique de leur population, et mêmede leur culture, ils font partie du nombre de pays pour lesquels l'absenced'un système valable d'enregistrement continu des faits d'état civil rendproblématique la détermination des niveaux et tendances des principalesvariables démographiques.

En effet, dans presque tous ces pays, l'état civil ne couvraitrécemment encore qu'une faible partie du territoire et ne fonctionnait, dumoins convenablement, qu'en milieu urbain.

L'étendue de certains d'entre eux, la dispersion de l'habitat, l'absencede voies de communication modernes, la couverture administrativedéficiente, le sous-développement généralisé contribuaient à cettesituation.

38

1. SOURCES DES DONNÉES ET MÉTHODOLOGIE

L'état civil, l'enquête par sondage et le recensement constituent lessources traditionnelles de mesure de la mortalité infantile et de la petiteenfance.

1.1. L'état civil

Dans le cas où on dispose d'un système valable d'enregistrement desnaissances et des décès et que, par ailleurs, on puisse connaître les effectifsmoyens de la population auxquels se rapportent les décès, il est aiséd'estimer les indices de mortalité entre 0 et 5 ans. Toutefois, la façon donton veut approcher le phénomène est évidemment tributaire du type declassement des événements dont on dispose (par âge et année de naissanceou par âge seulement).

Ainsi qu'on l'a précédemment remarqué, l'utilisation au niveaunational des données d'état civil est fortement à déconseiller, excepté pourle Cap-Vert - où la couverture des faits de population s'avère assezbonne depuis longtemps - et pour Saint Thomas et Prince.

En effet, dans le cas du Mozambique par exemple, le nombre moyende naissances et de décès déclarés à l'état civil entre 1969 et 1971 ne futque de 40 674 et 11 071 respectivement, chiffres qui, rapportés à l'effectifde la population recensée en 1970, conduiraient à des valeurs de tauxbruts de natalité et de mortalité invraisemblables (respectivement 5 et 1pour mille).

Or, selon les résultats obtenus à partir des structures dénombrées en1970 et sous l'hypothèse de la stabilité de la population, ces indices neseraient pas inférieurs à 45 et à 20 pour mille(I).

1.2. L'enquête par sondage

Visant à suppléer les carences d'un état civil ne couvrant qu'unefaible partie des événements(2), dès les années cinquante on assiste à laréalisation d'enquêtes par sondage un peu partout en Afrique.

(1) Costa Carvalho C.A., « Essai de détermination des niveaux de fécondité et demortalité de la population noire du Mozambique à partir des résultats du recensement de1970 », Département de démographie, Université catholique de Louvain, Working Papern°73, 1979, p. 37.

(2) Selon J. Vallin (« La mortalité infantile dans le monde, évolution depuis 1950 »,Population, 1975, n° 4-5, p. 803), le nombre de pays africains où la couverture desnaissances et des décès est supérieure à 90 % ne serait que de six pour le premier typed'événements et de trois pour le second.

39

Que ce soit à des fins multiples (structures, mouvement, etc.) ourestreintes (enquête fécondité, mortalité), qu'il s'agisse d'opérations à unseul passage ou à plusieurs passages, cette source de données permetd'estimer la mortalité infantile et juvénile, bien que les limitations desenquêtes, surtout à passage unique, soient suffisamment connues (risquesd'erreurs inhérents à toute observation rétrospective, notamment auniveau de la durée réelle d'une période passée, erreurs de dénombrement,etc.).

Malheureusement, cette méthode d'investigation n'a suscité que peud'intérêt dans les territoires alors sous administration portugaise et aucuneenquête d'envergure nationale n'a été conduite jusqu'à l'indépendance.

Cependant, un petit nombre d'enquêtes portant sur quelques ethniesde la Guinée-Bissau et de l'Angola ont été réalisées dans les annéescinquante ; mais, la plupart d'entre elles ne présentent qu'un intérêt trèslimité, en raison de l'insignifiance de leur univers et du nombre réduit derenseignements qu'on peut en déduire.

1.3. Le recensement

Entre 1940 et 1970, des recensements de nature statistique ont étéeffectués selon une périodicité décennale dans pratiquement tous ces pays.

Grâce aux questions pertinentes posées aux femmes en âge deprocréer, on dispose des proportions d'enfants décédés (sexes réunis) selonl'âge des femmes, pour le Cap-Vert (1960), la Guinée-Bissau (1950),l'Angola (1940) et le Mozambique (1940, 1950 et 1970).

L'application de méthodes indirectes d'analyse à ce type de donnéespermet d'obtenir une estimation du niveau de la mortalité aux jeunes âges.L'essentiel des estimations proposées dans cet article étant obtenues àpartir de ces renseignements et de l'application de ces méthodes, il asemblé nécessaire de rappeler brièvement les principes qui les orientent.

Il y a une vingtaine d'années, W. Brass a proposé une méthode detransformation des proportions d'enfants décédés en des probabilités demourir de la naissance au 1er, 2e, 3e et 5e anniversaire, grâce à l'applicationde coefficients correcteurs dépendant de la durée d'exposition au risque dedécéder, et donc de l'âge actuel et du calendrier de la fécondité desfemmes ; ces coefficients sont calculés à partir du rapport des paritésmoyennes à 15-19 ans et à 20-24 ans (?.I/P2), du rapport des paritésmoyennes à 20-24 ans et à 25-29 ans (P2/P3), de l'âge moyen de lastructure de la fécondité (m) et de l'âge médian de cette même structure(m').

Brass a en effet démontré la relation existante entre les proportionsd'enfants décédés parmi les femmes âgées de 15 à 19 ans et -¡c^, entre lesproportions d'enfants décédés parmi les femmes âgées de 20 à 24 ans et2qo, entre les proportions d'enfants décédés parmi les femmes âgées de 25

40

à 29 ans et 3 ^ , et entre ces proportions calculées pour les femmes âgées de30 à 34 ans et ¡¡q^ en prenant en considération la plus ou moins grandeprécocité de la fécondité.

Cette méthode se basant sur un seul schéma de mortalité (le standardgénéral de mortalité de Brass) et de fécondité (une fonction de 3e degré dutype f (x) = K (x - a) (a + 33 - x)2, Sullivan et Trussell ont dépassé ceslimitations en utilisant un très grand nombre de schémas de mortalité(issus des tables-types de Princeton) et de fécondité (des distributionsempiriques en ce qui concerne Sullivan et des distributions théoriquespour ce qui est de Trussell).

On doit rappeler que l'application de ces méthodes implique lasatisfaction de certaines conditions telles que la constance de la féconditéainsi que de la mortalité des enfants au cours des dernières années,l'absence de corrélation entre l'âge de la mère et la mortalité des enfants ouentre la mortalité des mères et la mortalité des enfants et, finalement, uneidentique proportion d'omissions pour les enfants déclarés décédés etsurvivants.

Parmi ces hypothèses, la dernière pose le plus souvent un problèmeen raison de la réticence des personnes à évoquer des événementsmalheureux, ce qui conduit le plus souvent à des mesures par défaut de lamortalité aux jeunes âges.

En définitive, si mortalité ou fécondité varient, les estimationsfournies par ces méthodes seront, évidemment, biaisées. Afin de tenircompte de ce fait, on peut recourir aux approches suggérées par W. Brass,F. Feeney et E.P. Kraly, et D.A. Norris, en particulier(3).

Les indices de mortalité utilisés pour mesurer la mortalité infantile etjuvénile sont respectivement ^ et 4q,.

Dans le cas où la source de données est le recensement, le premier deces indices est obtenu à partir de la valeur de gC calculée par la méthodede Trussell, en entrant avec la valeur correspondante de 15 dans la table1.2 du Manuel IV des Nations Unies(4).

Appliquant les probabilités de décéder entre l'instant de la naissanceet les premier et cinquième anniversaires à une cohorte fictive d'un effectifde 10 000 personnes à la naissance, on parvient à déterminer les valeursde 4q,.

(3) Voir à ce sujet. Wunsch G.. « La mortalité aux jeunes âges : l'apport desméthodes d'analyse démographique ». in La mortalité des enfants dans le monde et dansl'histoire, Département de Démographie. Université Catholique de Louvain. OrdinaEditions. 1980, p. 22.

(4) Nations Unies. « Méthodes permettant d'estimer les mesures démographiquesfondamentales à partir de données incomplètes ». ST/SOA/Série A/42. n° 42. New York.1969. p. 101.

41

Les événements déclarés à l'état civil n'étant pas classés par âge etannée de naissance, les « taux » de mortalité infantile relatifs à une année xont été calculés en rapportant les décès de moins d'un an à une sommepondérée de naissances survenues dans les années x et x-1, les « poids »utilisés étant de 0,67 et 0,33.

Puis, les probabilités de décéder entre 0 et 5 ans ont été obtenues entransformant les taux en quotients, selon la formule suivante :

1 0 / 2 8 < »

2. LES RÉSULTATS

L'essentiel de notre information provenant des recensements, c'est-à-dire, dans ce cas, d'opérations dont la qualité laisse souvent à désirer, lesestimations obtenues doivent être considérées avec réserve.

Et c'est particulièrement au niveau des proportions d'enfants décédésque le problème se pose avec le plus d'acuité ; en effet, leur évolutionselon l'âge des femmes s'effectue dans la plupart des cas de façon lente et,parfois, anormale (cf. Annexe : tableaux A.l et A.2).

Les estimations des indices retenus pour mesurer la mortalité auxjeunes âges sont présentées en annexe (cf. tableaux A.3 à A.7).

On y remarquera que l'essentiel de ces estimations provient desrecensements, le Mozambique étant particulièrement privilégié puisquel'information pertinente est disponible pour trois des quatre recensementsréalisés dans ce pays ; malheureusement, bien que la question sur la surviedes enfants ait été posée lors du recensement de 1960, ces données n'ontpas été publiées.

Par ailleurs, à notre connaissance, cette même information n'a pasencore fait l'objet d'une publication au recensement du Cap-Vert de 1970et la question pertinente a été abandonnée en Angola et en Guinée-Bissau.

Le pays pour lequel on dispose de la série la plus complète dedonnées d'état civil est le Cap-Vert (cf. Annexe : tableau A.3). La mortalitéinfantile aurait connu dans ce pays une diminution importante pendant lapériode d'observation, particulièrement entre 1940 et 1950 (taux moyenannuel de déclin de 2%) et entre 1962 et 1970 (taux moyen annuel dedéclin de 3 % ).

^ serait passé, entre 1950 et 1960. de 417 à 157 pour mille si on seréfère à la même source, et de 417 à 174 pour mille si on utilise dessources différentes, le taux moyen annuel de déclin de cet indice étant del'ordre de 10%.

(5) G. Wunsch et D.A. Cañedo in « La transformation des taux en quotients auxpremiers âges de la vie ». Département de Démographie. Université Catholique deLouvain. Document de Recherche n° 1. 1978. p. 8.

42

Pour 1960-1962, il est possible de vérifier la cohérence de cesestimations selon qu'on utilise l'une ou l'autre source. Elle seraitappréciable, particulièrement dans le cas du « taux » de mortalité infantile.

Les estimations incluses dans le tableau A.4 en annexe, concernentla Guinée-Bissau, pays pour lequel on ne dispose au niveau national quedes indices calculés à partir du recensement de 1950.

On a tenu à y inclure les estimations de la mortalité infantile relativesà la Circonscription administrative de Cacheu et à l'ethnie Fula de laCirconscription de Gabu, en raison des garanties de fiabilité qu'ellessemblent présenter.

Bien que les estimations de ^ concernent des entités adminitrativesdifférentes, elles sont suffisamment divergentes pour laisser penser à unesous-estimation au niveau national.

L'estimation de l'enquête semble être la plus proche de la réalité,compte tenu de la qualité de son organisation, ainsi que de l'importance deson univers : environ 2096 de la population féminine de 14 ans et plus aété enquêtée.

Le tableau A.5 en annexe présente les quelques estimations qui ontpu être effectuées à partir des données de l'état civil pour Saint Thomas etPrince.

En dépit du fait que ce petit archipel fait partie du nombre réduit depays africains où la couverture des faits d'état civil s'avère relativementbonne depuis longtemps, il a fallu attendre l'année 1966 pour disposer desdécès classés par âge (moins d'un an et d'un à cinq ans).

Entre 1966 et 1976, le «taux» de mortalité infantile est passé de86 pour mille à 54 pour mille, ce qui représente un taux moyen annuel dedéclin de presque 5 %.

On peut remarquer que la baisse de ^ est presque aussi importanteentre 1950 et 1960, sa valeur étant passée de 434 à 285 pour mille.

A la lumière des estimations concernant l'Angola (cf. Annexe :tableau A.6), le niveau de la mortalité de ce pays occuperait le premierrang parmi les pays africains lusophones, et devrait se situer parmi les plusélevés d'Afrique Noire.

Si, au niveau régional, la Province de Huila se singularise par unemortalité beaucoup moins importante que la moyenne, un tel fait peut êtreimputé à des déficiences au niveau de la collecte (sous-déclarationimportante des enfants décédés) plutôt qu'à d'autres raisons (par exemple,la salubrité de cette région).

On peut remarquer que, même au niveau du District de Luanda, oùse situait la capitale, le niveau de la mortalité était fort élevé ( égal à384 pour mille).

Les estimations incluses dans le tableau A.7 en annexe concernent leMozambique, pays pour lequel on dispose des proportions d'enfantsdécédés aux recensements de 1940, 1950 et 1970.

43

Tout d'abord, on doit noter l'évolution anormale de ces indices entre1940 et 1950, que ce soit au niveau national ou régional. Une telleanomalie peut s'expliquer, presque exclusivement, par le fait que lepremier recensement de nature statistique fut celui de 1940, la qualité desdonnées étant inférieure à celle de l'opération effectuée dix ans plus tard.

Si l'on ne considère que les seules estimations de 1950 et 1970, lamortalité aux jeunes âges n'aurait pas connu une baisse importante aucours de cette période, puisque le taux moyen annuel de déclin ne seraitque de l'ordre de 1 •% dans le cas de la mortalité infantile et de la petiteenfance.

Par contre, le District de L. Marques (région Sud), où se situait lacapitale, aurait connu un déclin de la mortalité aux jeunes âges bien plusimportant : entre 1950 et 1970 ^ serait passé de 154 à 106 pour mille et4q, de 90 à 53 pour mille.

Le remaniement des limites des frontières des régions Nord et Centrene permet pas d'effectuer une analyse de l'évolution au niveau régionalplus approfondie (cf. Annexe : tableau A.2).

3. CONCLUSION

L'état civil - source privilégiée pour la mesure de la mortalitéinfantile et de la petite enfance - ne couvrant qu'une faible étendue duterritoire de la plupart des pays d'Afrique d'expression portugaise, on a dûsouvent utiliser comme source des données les recensements pourestimer, à partir de méthodes indirectes d'analyse, la mortalité aux jeunesâges.

Or. ces méthodes - qui ne sont qu'un palliatif au manque dedonnées fiables - ayant été élaborées à partir de constatations empiriquesou de simulations, les hypothèses qui sont inhérentes à leur emploirendent parfois hasardeuse leur application.

A la lumière de nos estimations, on peut conclure à l'existence d'unetrès forte mortalité aux jeunes âges dans les pays africains lusophones.particulièrement en Angola et en Guinée-Bissau, la situation de SaintThomas et Prince et. dans une moindre mesure, du Cap-Vert, étant moinsalarmante.

BIBLIOGRAPHIE

COSTA CARVALHO C.A. Essai de détermination des niveaux de fécondité et de mortalité dela population noire du Mozambique à partir des résultats du recensement de 1970.Département de Démographie. Université Catholique de Louvain. Working Papern°73. 1979. p. 37.

VALLIN J. La mortalité infantile dans le monde, évolution depuis 1950. Population, 1975.n°4-5. p. 803.

44

WUNSCH G. La mortalité aux jeunes âges : l'apport des méthodes d'analyse démogra-phique. La mortalité des enfants dans le monde et dans l'histoire, Département deDémographie. Université Catholique de Louvain. Ordina Editions. 1980. p. 22.

WUNSCH G. et CAÑEDO D.A. La transformation des taux en quotients aux premiers âgesde la vie. Département de Démographie. Université Catholique de Louvain.Document de Recherche n" 1. 1978. p. 8.

NATIONS UNIES. Méthodes permettant d'estimer les mesures démographiques fondamen-tales à partir de données incomplètes. ST/SOA/Série A/42. Etudes démogra-phiques n° 42. New York. 1969. p. 101.

ANNEXE

TABLEAU ANNEXE A. 1 - Parités moyennes et proportions d'enfants décédés pourle Cap-Vert ( 1960). la Guinée-Bissau ( 1950) et l'Angola ( 1940).

Age des

femmes

CAP

Parités

VERT

Proportions

GUINEE

Parités

BISSAU (§)

Proportions

ANGOLA

Parités Proportions

15-1920-2425-2930-34

0.1 111.0652.4283.472

0.1850.1500.1590.175

0.4750.9231.5541.994

0.235 -0.272

-0.2690.288

0.0721.0552.0192.721

0.3050.3120.3410.366

Age des ANGOLA (Prov. Luanda) ANGOLA (Prov. Malange) ANGOLA (Prov. Benquela)

femmes Parités Proportions Parités Proportions Parités Proportions

15-1920-2425-2930-34

0.1241.2922.4033.210

0.3080.3350.3700.401

0.0651.0061.8212.471

0.3140.3150.3290.363

0.0591.1292.3223.182

0.2990.3090.3590.380

Age des ANGOLA (Prov. Bié) ANGOLA (Prov. Huila)

femmes Parités Proportions Parités Proportions

15-1920-2425-2930-34

0.0650.6141.0941.392

0.2170.2990.3340.353

0.0440.7991.7192.330

0.2150.1910.2070.246

(§) Concerne les seules femmes mariées.

Sources .-- Recenseamento Gérai da Populaçao-1960. Cabo Verde. Praia :- Censo da Populaçao nao Civilizada-1950. Guiñé Portuguesa. Bissau :- Censo Gérai de Populaçao-1940. Angola. Luanda.

45

TABLEAU ANNEXE A.2 - Parités moyennes et proportions d'enfants décédéspour le Mozambique (ensemble du pays, régions Nord. Centre et Sud).

MOZAMBIQUE

Age des

femmes

15-1920-2425-2930-34

Age des

femmes

Parités

0.3291,2342.1142.859

Parités

1940

Proportions

0.1790.2210.2250.237

1940

Proportions

Parités

0.1470.9051.8422.605

1950

Proportions

0.2580.2620.2640.264

NORD

Parités

1950

Proportions

Parités

0.2611.3782.4693.329

Parités

1970

Proportions

0.1960.1990.2070.219

1970

Proportions

15-1920-2425-2930-34

0.3371.0911.6722.461

0.2000.2230.2110.217

0.2990.9301.6132.124

0.3860.2870.2530.235

0.3011.3042.1922.877

0.1890.1880.1940.212

Age des

femmes Parités

1940

Proportions

CENTRE

1950

Parités Proportions Parités

1970

Proportions

15-1920-2425-2930-34

0,3381.4402,5723.492

0,1760.2490,2640,280

0,5041,4742,5363,280

0,2790,3160,3110,318

0,2761,5112,7923.807

0,2210,2230,2300,239

SUD

Age des

femmes

15-1920-2425-2930-34

Parités

0.2991.1711.9862.593

1940

Proportions

0.1420.1660.1720.185

Parités

0.1090.8701.7502.580

1950

Proportions

0.2530.2440.2570.257

Parités

0,2121,2712,3733.275

1970

Proportions

0,1570,1680,1770,193

46

Observations :1 - La région Nord comprenait les provinces et districts suivants :

1940 - la province de Niassa ;1950 - les districts de Lago. Cabo Delgado et Nampula ;1970 - les districts de Niassa, Cabo Delgado et Moçambique.

2 - La région Centre comprenait les provinces et districts suivants :1940 - les provinces de Manica e Sofala et de Zambézia ;1950 - les districts de Quelimane, Tete et Beira :1970 - les districts de Zambézia. Tete et Manica e Sofala.

3 - La région Sud comprenait les provinces et districts suivants :1940 - la province du Sul do Save :1950 - les districts de Lourenço Marques. Inhambane et Gaza ;1970 - les districts de Lourenço Marques, Inhambane et Gaza.

Sources :1) Recenseamento Gérai da Populaçao-1940, Moçambique. L. Marques ;2) Recenseamento Gérai da Populaçao-1950, Moçambique. L. Marques :4) Recenseamento Gérai da Populaçao-1970. Moçambique. L. Marques.

TABLEAU ANNEXE A.3 - Quelques indicateurs de mortalitéinfantile et juvénile au Cap-Vert.

Période Source Quotients de mortalité...

1940195019601960196219701977

Etat civilEtat civil et recensementEtat civil et recensementRecensementEtat civilEtat civilEtat civil

0.1790.143

_0.1180.1280,0940.091

0.063

0.4170.1570.174

Sources :Boletim Trimestrial de Estatistica, Cabo Verde ;Anuarios Estatisticos, Territorios Ultramarinos, Vol. II, Portugal, I.N.E.,Lisboa ;Recenseamento Geral da Populaçao - 1950, Cabo Verde ;Recenseamento Geral da Populaçao - 1960, Cabo Verde.

TABLEAU ANNEXE A.4 - Quelques indicateurs de mortalitéinfantile et juvénile en Guinée-Bissau.

Période Source Quotients de mortalité

4Q,

1950Oct.Sep.Juil.Dec.

1949 à19511954 à1956

Recensement(Etat civil (1)( Enquête (1)

Etat civil (2)

0.1910.2670.3240.258

0.117 0.286

Sources .-(I ) A. Carreira. "Movimento natural da populaçao nao civilizada da Circunscriçao

Administrativa de Cacheu". Bissau. 1956.

47

(2) E. Erito. "Taxas de natalidade e mortalidade dos Fulas do Gabu". BoletimCultural da Guiñé Portuguesa, Vol. XI. n° 44. Bissau. 1956.

- Censo da populaçao nao civilizada-1950. Guiñé,

TABLEAU ANNEXE A.5 - Quelques indicateurs de mortalitéinfantile et juvénile au Saint Thomas et Prince.

Période Source Quotients de mortalité

19501960196619701976

Etat civil et recensementEtat civil et recensementEtat civilEtat civilEtat civil

--

0.0860.0700.054

0.4340.285

Sources .-- Boletim Trimestrial de Estatistica. S. Tomé e Principe :- Anuarios Estatisticos. Territorios Ultramarinos. Vol. II. Portugal. I.N.E..

Lisboa.

TABLEAU ANNEXE A.6 - Quelques indicateurs de mortalitéinfantile et juvénile en Angola, en 1940.

Pays/

Région

AngolaProv. LuandaProv. M al angeProv. BenguelaProv. BiéProv. Huila

TABLEAU

Pays/

Région

MOZAMBIQUE

Nord

Centre

Sud

Source

RecensementRecensementRecensementRecensementRecensementRecensement

Quotients de mortalité

iQo

0.2350.2790.2290.2460.2250.159

4Qi

0.1480.1790.1440.1560.1410.094

0.3480,4080.3400.3640.3340.238

ANNEXE A.7 - Quelques indicateurs de mortalitéinfantile et juvénile au Mozambique.

Période

194019501970194019501970194019501970194019501970

Quotients de mortalité

0.1540.1740.1420.1380.1540.1370.1820.2090.155

.0.1210.1680.126

4Qi

0.0900.1040.0810.0780.0900.0770.1100.1290.0910.0650.1000.069

6q0

0.2300.2600.2110.206

- 0.2300.2030.2720.3110.2320.1780.2510.187

INFLUENCIA DEL SECTOR SALUDEN LOS NIVELES

DE LA MORTALIDAD INFANTILCHILENA

por

BERTA CASTILLO M., FRESIA SOLIS F.,GRACIELA MARDONES A.

Universidad de ChileFacultad de Medicina

Escuela de Salud Pühlica

INTRODUCCIÓN

Las características que adquiere la mortalidad infantil en unmomento determinado de su evolución, pueden ser el resultado de lainteracción de factores presentes en el contexto social, económico,cultural, político, ecológico, etc., de una sociedad M y por lo tanto, laidentificación de los factores que inciden en el nivel de la mortalidad delos menores de un año, es un proceso de alta complejidad y multivariado.

Se reconoce que los programas ejecutados por el sector salud, tienengran importancia en la reducción de la mortalidad infantil, sin embargo, laindividualización de las variables de salud más íntimamente relacionadascon el problema constituye un desafío, considerando que la potencialidadde ellas, va a depender tanto de la communidad receptora como de laorganización y cobertura del sistema de attención médica ß.

En el país, desde hace aproximadamente tres décadas diversosautores t3- 4>5- 6> 7> 8> 91 han explorado el nexo entre atención médica ymortalidad infantil por todas las causas, encontrando en su gran mayoría,que la asociación más alta se produce con la variable porcentaje denacimientos con atención profesional.

Estimamos que analizando las relaciones de la mortalidad infantilpor grupos de causas de muerte, sería posible identificar otras variables deatención médica responsables de la variación de los riesgos de muerte delos menores de un año.

Por lo tanto, los objetivos del trabajo se refieren a : medir el nivel dela mortalidad infantil por todas las causas, causas evitables y grupos de

50

causas evitables en los Servicios de Salud del país (l) ; cuantificar lasactividades y recursos de salud proporcionados a los menores de un año yfinalmente determinar las variables de atención médica más relevantespara explicar las diferencias de mortalidad infantil en el país, consideradacomo un todo y desagregada en grupos de causas evitables.

MATERIAL Y MÉTODO

En la construcción de las variables dependientes (tasas de mortalidadinfantil) se corrige el volumen de nacidos vivos según la omisióndiferencial estimada por los Servicios de Salud. A nivel nacional, seconsidera una integridad del 95 % t10'.

Se agrupan los decesos infantiles de acuerdo con la Octava Revisiónde la Clasificación Internacional de Enfermedades, Traumatismos yCausas de Defunción t>1], considerando las muertes evitables por : « buencontrol del embarazo y atención del parto » que comprende principal-mente defunciones originadas en el período perinatal ; « prevención,diagnóstico y tratamiento médico precoz», el 7596 de estas muertescorresponde a difteria, tos ferina, septicemia, y sarampión ; « buenascondiciones de saneamiento », principalmente enfermedades infecciosasintestinales ; « alimentación completa », defunciones por deficiencias en lanutrición y trastornos périnatales del aparato digestivo ; y dada suimportancia y naturaleza constituyen grupos aparte : « enfermedadesrespiratorias » y « accidentes, envenenamientos y violencias »(2).

(1) El Ministerio de Salud considera como divisiones administrativas 27 Serviciosde Salud en el país, creados por Decreto Ley n° 2763 de 1979.

(2) Causas evitables :Por « buen control del embarazo y atención del parto » : Tétanos (037), Sífilis

congénita (090), Enfermedades de la madre no relacionadas con el embarazo (760-761),Toxemias del embarazo (762), Infecciones maternas antes o intrapartum (763), Partodistócico (764-768), Afecciones a la placenta (770), Afecciones del cordón umbilical (771),Lesión ocurrida durante el nacimiento (772), Enfermedades hemolíticas del recién nacido(774-775) y Afecciones anóxicas e hipóxicas (776).

Por « prevención, diagnóstico y tratamiento médico precoz » : Difteria (032), Tosferina (033), Septicemia (038), Sarampión (055), Meningitis (320), Epilepsia (345),Enfermedades inflamatorias del oído (380-384), Enfermedades del aparato digestivo (520-577) y Enfermedades de la piel y del tejido celular subcutáneo (680-709).

Por « buenas condiciones de saneamiento » : Enfermedades infecciosas intestinales(001-009) y Hepatitis infecciosa (070).

Por « alimentación completa y acciones que prevengan la inmaturidad » :Avitaminosis y otras deficiencias nutricionales (260-269) e Inmaturidad (777).

« Enfermedades respiratorias » : Infecciones respiratorias agudas (460-466), In-fluenza (470-474), Neumonía (480-486), Bronquitis, Enfisema y Asma (490-493).

« Accidentes, envenenamientos y violencias (880-999) ».Causas no evitables : Varicela (052), Encefalitis vírica (065), Tumores (140-239),

Parálisis cerebral espasmódica infantil (343) y Anomalías congénitas (740-759).

51

Como variables independientes (indicadores de atención médica) seconsideran 59 actividades y recursos dirigidos a los menores de un año yque se refieren a : acciones de prevención, atención de la morbilidad ydesnutrición infantil y recursos de salud (véase tabla del anexo). Losindicadores de salud se elaboraron relacionando las cifras anuales de cadaServicio de Salud, con el número de nacidos vivos respectivo, yacorregido por omisión.

Sobre la base de esta información se diseña el estudio quecorresponde a un análisis de regresión lineal múltiple paso a paso I12-131,para obtener la mejor ecuación lineal de predicción de las variablesdependientes, a partir de un conjunto de variables independientes elegidaspreviamente para cada grupo de causas de muerte.

El modelo final en cada caso, incluye todas aquellas variables cuyoscoeficientes parciales, han probado ser significativamente diferentes decero, con un margen de error no superior al 10% (p < 0,10). Laproporción de la varianza explicada, se evalúa mediante el coeficiente dedeterminación (r2) y la potencia de las variables de salud en la explicaciónde la varianza de las tasas de mortalidad, queda determinada por el valorde la estadística « F ».

RESULTADOS

1. Descripción de las variables

1.1. Mortalidad infantil

En 1979, la mortalidad infantil en Chile es de 38,2 por mil nacidosvivos, registrando un amplio rango de variación entre los Servicios deSalud del país, desde 21,0%o para Santiago Oriente, hasta 68,7%o paraAisén.

En el mapa adjunto, se muestra la distribución de los 27 Servicios deSalud del país, clasificados en cinco grupos de mejor a peor nivel demortalidad infantil por causas evitables, según resultados de un estudiopreliminar ['4l

El grupo 1, constituido por las siete entidades de Santiago, la capital,además de Valparaíso-San Antonio y Arica el Servicio más septentrional,es el único que presenta tasas de mortalidad infantil por todas las causas ycausas evitables, inferiores a 30 por mil nacidos vivos ; los Serviciosagrupados en los tramos 2 al 4, registran tasas entre 30 y 60 por mil ;Aisén, en el grupo 5, posee tasas superiores a 60 por mil nacidos vivos.

Las muertes teóricamente evitables significan el 84,0 % del total dedefunciones infantiles. Dentro de este grupo, los mayores riesgos demuerte se vinculan con causas evitables por : « buen control del embarazo

y atención del parto » con una tasa de 11,4 por mil nacidos vivos ;« enfermedades respiratorias » (8,5 por mil) y por « prevención,diagnóstico y tratamiento médico precoz » (4,6 por mil). Las defuncionesde estos grupos representan el 63,996 del total de muertes infantiles y el76,1 % del total de decesos por causas evitables.

TABLA 1. - Distribución porcentual de las defunciones infantiles por grupos de causasde muerte evitables y tasas de mortalidad. Año 1979.

Causas de muerteDefunciones

TotalN° %

infantilesEvitables

Tasas de morta-lidad por milnacidos vivos

TOTAL 8825 100,0No evitables 1411 16,0Evitables 7414 84,0Enfermedades respiratorias (grupo A) . 1973 22,4Por buen control del embarazoy parto (grupo B) 2615 29,6

Por prevención, diagnóstico ytratamiento médico precoz (grupo C) 1053 11,9

Por buenas condiciones desaneamiento (grupo D) 872 9,9

Por alimentación completa (grupo E). . 394 4,5Por accidentes envenenamientosy violencias (grupo F) 507 5,7

100,026,6

35,3

14,2

11,85,3

38,26,1

32,18,5

11,4

4,6

3,71,7

6,8 2,2

1.2. Atención médica

1.2.1. Acciones de prevención

En el conjunto de las acciones de prevención, destacan por suvolumen, los controles de salud proporcionados por enfermeras ycontroles prenatales efectuados por matronas, que corresponden cadauno, en promedio, a cinco de estas actividades por nacido vivo. Estasacciones se dan en forma más o menos parecida en los Servicios, lo que setraduce en un coeficiente de variación relativamente bajo (18%) encontraposición a las actividades de prevención otorgadas en menormagnitud (visitas domiciliarias de enfermeras y auxiliares de enfermería,charlas educativas a madres con hijos menores de dos años, controles desalud efectuados por médico, matrona o auxiliar y controles prenatalesrealizados por médico) que presentan gran heterogeneidad a través delpais. (Véase tabla 2)

En relación a las inmunizaciones, con la tercera dosis de la vacunatriple y la antisarampionosa, el promedio es de 941 y 932 inoculacionespor mil nacidos vivos, respectivamente; sin embargo, en cinco Servicios,se proporcionan entre 700 y 800 vacunas triple por mil nacidos vivos.

,'Irjliiqiie

* - Antofagasta

quimbo

Vina del Mar-Quil locáj^S.in Fclípe-I.os Andes

Valparaíso-San Antonio! R- Metropolitana

[•*." l.ib.C.II. O'llim'.ins

. ' M - ' " " 1 ••

Talcaliuanol

Concepción-Arauco

i U,iiu]uíhiie-Ch ¡ \

53

Servicios de Salud : Mortalidad Infantil(por mil nacidos vivos)

nor todaslas causas

GRUPO 1 26,2

por causasevitables

20,7

Arica, Valparaíso-San Antonio, Santiago :Norte, Sur, Oriente, Occidente, CentralNor-Occidente, Sur Oriente

I GRUPO 2 40,9 36,2

Antofagasta, Magallanes, Viña-Quillota,San Felipe—Los Andes, Maule, Libertador,Bdo. O'Higgins,_Concepción-Arauco, Iqui-que, Coquimbo, Nuble

GRUPO 3 53,2 46,1

Osorno, Bio—Bio, Atacama, Llanquihue-Chiloé-Palena, Valdivia

I GRUPO 4 53,2

Talcahuano, Araucania

Antarticachilena

I GRUPO 5

Aisén

68,7

PROMEDIO PAIS 38,2

47,3

60,0

32,1

Polo sur

En 1979, el Ministerio de Salud distribuyó aproximadamente, 26millones de kilos de leche, siete millones (27,0%) se destinó a menores deun año y nodrizas. Esta cantidad relacionada con los nacimientos,representa 2,5 kilos mensuales por nacido vivo, que se divide en 1.5 y 1,0

54

. - 3

CD• O

a• a_3

co

•s»3O

rvi

&r-

3

o

igCu

•>

1!•8co

i•8

•2

1

<00<

i l .3¡I!

o13

<N O

» t ^V Ï r* CT\ Ooo ^- «i —

oo

r^ c? Ooo'o o• < 1 -

o t r^

<

o3<cotüQVituOfe8

sras

SII-i

ría

ic

•a

1M

uxi

Sw<

•aoo

•8

Soc

_«

o

[atr

Toó"

3 wi ^M r^

O OTT

• ¿ a ^•5= 5,2oó Z¡ S ^ tü

«lias » £ o5 " 2 2

S38 si

Q otü o"

u

Z 2tü

<tü

aCO

O<

i 5 PC 3

m

Q |

00 O

ΠO

lij «

SQ

IUScoO

§coUc.Q<

(N - -2 " "co

Q

D

[T | y^ go OtÜ ^°

o «

ZDUQcoaPÎou

Qû<

a

MO

RB

l

co< ¡g

^ « i3 S^ O P cd

Z -Otîi SO - C 3

55

S I-S

o•a

o r rnrs — —

- — (NO (7s OO

- ~ — OO OS

oo — r- —

en

IUUJQ

2 a s

C/3

IuuQ

Iua.

s saOu

^ -cd -4¿ •—• -cd -¿i 2—» '-5 to O -S « —s

i &8 5 sé I

8 1:!•§•gJS

I

i +

H

•§

•Scd "c3

3 ó c?

56

kilo de leche respectivamente, durante los dos primeros semestres de vida.En 10 Servicios, la leche distribuida a menores de seis meses y nodrizas,no sobrepasa 1,5 kilo por nacido vivo. En relación a niños de 6 a 11meses, en 16 Servicios se proporciona menos de 1,0 kilo mensual pornacido vivo.

/. 2.2. Atención de la morbilidad

De 4.636.014 atenciones proporcionadas a los menores de 15 años,una quinta parte de ellas, correspondió a los menores de un año. De éstas,(939,031) los médicos otorgan el 87 96, 9 % las auxiliares de enfermería y4% las enfermeras, ajustándose en general, al esquema del programamaterno-infantil, es decir, la mayor responsabilidad reside en el médico yen aquellos lugares en que este recurso es insuficiente, esta actividad essuplida por la enfermera o auxiliar.

Las consultas médicas a menores de un año, registran la menorvariación entre los Servicios (14,8%); las enfermeras y auxiliares,participan en esta actividad en forma muy variable, desde 8,4 consultas deenfermeras por mil nacidos vivos (Servicio de Iquique) hasta 1652,6(Santiago Nor-Occidente). En los Servicios de Santiago Oriente, Central ySur, las auxiliares no proporcionan consultas de morbilidad, en cambio, elServicio de Llanquihue-Chiloé-Palena, otorga 1391,6 consultas deauxiliares por mil nacidos vivos.

1.2.3. Atención de la desnutrición infantil

Las consultas por desnutrición a los menores de 15 años sonproporcionadas por nutricionistas, enfermeras y médicos, clasificando alos niños en eutróficos y desnutridos. Según las Tablas de Peso (Kg.) de M.Sempé(15) específicas para cada sexo, se determina el grado dedesnutrición : leve, moderado o avanzado.

Delas891.085 consultas proporcionadas a la población menor de 15años, una quinta parte corresponde a menores de un año. El mayor pesode la atención a este segmento de la población, recae en las nutricionistas(68,0%) y enfermeras (23,5%). La acción del médico es reducida,proporciona en promedio 64 consultas a desnutridos menores de un añopor mil nacidos vivos.

La relación consultas de desnutrición a menores de un año ynacimientos indica una gran heterogeneidad entre los Servicios, tanto alconsiderar al profesional que la otorga como el grado de desnutrición; asi,la nutricionista registra 76,6 consultas por mil nacidos vivos (Servicio deMagallanes) como mínimo y 1181,2 como índice máximo (Nuble). Enrelación al grado de desnutrición, el 78,8 % de las consultas a menores de

57

un año son calificadas de leves, 16,5 96 de moderadas y 4,7% deavanzadas.

1.2.4. Recursos de salud

A nivel nacional, existen 9,7 horas anuales pediátricas y obstétricas-ginecológicas contratadas por nacido vivo. La relación entre ambosíndices es una hora obstétrica-ginecológica por 1,7 hora pediátrica. Elrecurso hora médica es el que presenta mayor heterogeneidad entre losServicios; cabe destacar, que siendo el promedio 6,1 hora médicapediátrica por nacido vivo, en la mitad -de los Servicios, se dispone a losumo de 4,7 horas anuales.

De los 7.969 profesionales de colaboración médica registrados en1979, 34,996 son enfermeras, 24,696 matronas y 8,696 nutricionistas; el31,996 restante, se refiere a: asistentes sociales, tecnólogos médicos,kinesiólogos, etc.

En el país, en promedio, el 90,496 de los partos reciben atenciónmédica, variando de 75,296 a 99,1 96.

Se contabilizaron 12.277 camas obstétricas y pediátricas, de lascuales, el 6096 corresponden a pediatría. Sin embargo, el grado deutilización del recurso cama, es más homogéneo entre los Servicios(coeficiente de variación de 996) que la disponibilidad de él (17 96).

2. RELACIÓN ENTRE ATENCIÓN MÉDICA YMORTALIDAD INFANTIL

La metodología usada, permitió asociar a cada variable dependienteun modelo de regresión de valor predictivo que contiene un conjuntoseleccionado de variables de atención médica, ordenadas de mayor amenor potencia en la explicación de la variabilidad de la mortalidadinfantil.

De las 59 variables independientes consideradas, solamente 19 serelacionaron a un nivel estadísticamente significativo con los diferentesgrupos de causas evitables. Este conjunto de 19 variables se usó en laobtención de modelos de regresión para la tasa de mortalidad infantil totaly para la tasa de mortalidad infantil para todas las causas evitables. En latabla 3 se presenta una expresión simplificada de las ecuaciones obtenidasy el valor del coeficiente de determinación (r2) y en la tabla 4, se muestra elvalor de la estadística F y la probabilidad asociada a ella para las variablesde atención médica correlacionadas en las distintas variables dependientes.

58

TABLA 3. - Ecuaciones de regresión lineal múltiple y coeficientes de determinación (r2)para la mortalidad infantil total, por causas evitables y por grupos de causas evitables.

Mortalidad infantil Ecuación simplificada Q J

Y, : por todas las causasY2: por todas las causas evitablesY3: por enfermedades respiratorias

Y, = - X37 - X39 + X9 + X24

Y2 = - X37 - X39 + X9 + X24

Y3 = - X39 - X37 - X59 + X24- X , - X 3

Y4: evitable por buen control delembarazo y atención del parto

Y5: evitable por prevención, diagn.y trat, médico precozY6: evitable por buenas condiciones

de saneamientoY7: evitable por alimentación

completaYa: por accidentes, envenenamientos

y violencias Y8 = X24 - X15

Y4 — X3 3 X3B X42 + X5 0 X16

Y5 = X69 + X47 — X17 — X41

Y6 = X 9 - X 3 7 - X15

Y7 = X23 — X45 — X34

83,7

86,1

86,1

80,2

63,0

66,7

57,7

29,3

(I) En la tabla del Anexo se identifican las 59 variables de atención médica(X, a X59).

TABLA 4. - Valor de la estadística F y probabilidad asociada a ella, para las variablesde atención médico correlacionadas con las distintas variables dependientes.

Mortal.

Variables

infantil

F

total

P

Mortal, infantil porcausas evitables

Variables F P

Mortal, infantil porenfermedades respiratorias

Variables F p

X37

x39x9X24

18,9917,6915,883,91

0,00020,00040,00060.0608

X37

x39x9x24

25,4720,3416,625,97

0,00010,00020,00050,0230

x37x59Y

x,X3

39,8327,8816,4316,404,943,80

0,00010,00010,00060,00060,03790,654

Mort, infantil evitablepor buena atención del

embarazo y parto

Mort, infantil evitable porprevención, diagn. y trat.

médico precoz

Mon. infantil evitable porbuenas condiciones de

saneamiento

Variables Variables Variables

x33x3BX 4 2X 5 0

X16

17,1711.5711,215.364.60

0.00050.00270,00310.03080,0439

X„x47x17X41

21,7514,786,143,16

0,00010,00090.02140.0894

X9

X37

x,5

18.277.843,25

0,00030.01020,0844

Mortal, infantil evitablepor alimentación completa

Mortal, infantil por accidentes,envenenamientos y violencias

Variables

X 2 3

X 4 S

X34

F

26.0820.65

5,75

P

0.00010,00010.0249

Variables

x 2 4x 1 5

F

4,873.27

P

0,03710.0830

59

2.1. Atención médica y mortalidad infantil por grupos de causasevitables.

Horas médicas pediátricas y atención profesional del parto.En el estudio de muertes por grupos de causas, es de interés destacar

las variables de atención médica de mayor potencia para explicar lasdiferencias de mortalidad debida a enfermedades respiratorias y evitablespor « buen control del embarazo y atención del parto », puesto que en elmomento actual los mayores riesgos de muerte del menor de un año,están vinculados a estas causas

Como se expresó en la introducción del trabajo, diversos autores hancoincidido en encontrar la atención profesional del parto, como la variablede salud de mayor relevancia asociada a la mortalidad infantil por todaslas causas ; al subdividir por grupos de causas de muerte, el efecto de laacción profesional en el parto unido a las horas médicas pediátricas, tienemayor impacto en la reducción de la mortalidad por enfermedadesrespiratorias. Este resultado, podria estar indicando una acción educativahacia el recién nacido, pues se supondría que la madre durante su estadahospitalaria recibiría las indicaciones que hacen referencia al uso posteriorde los servicios pediátricos, lo que ayudaría a controlar el problema de lasenfermedades respiratorias, especialmente neumonías en el períodopostneonatal.

Que la atención profesional del parto esté más ligada a la prevenciónde la muerte por « enfermedades respiratorias », variable que ha ampliadosu cobertura de 63,0% en 1957M a 90,4% en 1979t16) unido al aumentode horas médicas pediátricas contratadas t'7), explicaría en parte, eldescenso sistemático en el último decenio de la mortalidad infantil porpatología respiratoria que, de una tasa de 30,6 por mil en 1968 ^ hallegado a 8,5 por mil en 1979M.

índice ocupacional de camas obstétricas y horas médicas obstétricasginecológicas contratadas.

Para el grupo de causas de muerte evitables por « buen control delembarazo y atención del parto », que reúne mayoritariamente enfermeda-des périnatales, las variables de mayor potencia para explicar los cambiosde mortalidad por este grupo de causas, son el índice ocupacional decamas obstétricas y las horas médicas obstétricas-ginecológicas contrata-das.

Esta combinación, pareciera ser mejor indicador de la calidad en laprestación de atención médica en el período perinatal, que sólo laasistencia profesional del parto, porque reflejaría el mejor manejo de lassituaciones de alto riesgo obstétrico para la madre y tratamiento concuidado médico y de enfermería para la mujer y el recién nacido, alocupar cama hospitalaria. La realidad muestra que este grupo de causas,continúan siendo el principal riesgo de muerte para los menores de un

60

año, variando las tasas de 16,8 por mil nacidos vivos en 1970 (17,2 pormü en 1975 ti »I) a 11,2 por mil en 1979.

Posible explicación a que el descenso no sea mayor, habría quebuscarla en los indicadores de salud. Un alto índice ocupacional,significaría que las camas probablemente estén mejor equipadas, es decir,cuenten simultáneamente con recursos de profesionales, laboratorios,banco de sangre, servicio de rayos X, etc. que redundaría en un mejor usode la dotación de camas disponibles y en consecuencia, se observaría unmenor nivel de mortalidad infantil. En efecto, de los datos de lasactividades de salud para 1979, los Servicios de Santiago, además deValparaíso-San Antonio, son los que registran las menores tasas demortalidad por este grupo de causas y al mismo tiempo, muestran el másalto índice ocupacional de camas obstétricas (entre 78 y 95 por ciento); porel contrario, Osorno, Atacama, Cautín y Valdivia, con altos niveles demortalidad perinatal, registran los índices más bajos, de 54,8 a 71,3 camasocupadas por cien camas disponibles.

Por otra parte, se ha expresado que la acción del médico, se reflejaríaen una prevención del alto riesgo obstétrico, lo que supone que tendríagran importancia en su labor, la destinación de un buen porcentaje de lashoras contratadas al control prenatal.

Al analizar las estadísticas, se encuentra que de 1.280.133 controlesprenatales efectuados en 1979, sólo el 7,296 los atiende el médico y el92,8 % los proporcionan las matronas. La escasa participación del médicoen esta actividad, puede deberse a que el recurso horas médicas obstétricascontratadas sea insuficiente y no pueda dedicarse a este controlpreventivo.

Cabe preguntarse, si en una labor tan importante, ambosprofesionales están capacitados por igual en la realización del examenfísico y obstétrico que implica habilidad diagnóstica y terapéutica o habríaque revisar las normas sobre encomendación de funciones para mejorar lacalidad de esta atención.

Controles de SaludLa variable controles de salud, está presente en cuatro de las seis

ecuaciones de regresión para los grupos de causas de muerte evitables.

Los controles efectuados por médicos o enfermeras a menores de unaño, muestran asociación inversa en todos los casos. Los controles deauxiliares y matronas, tienen un efecto mixto, es decir, colaboran en ladisminución de las defunciones por causas respiratorias y actúan ensentido contrario, cuando la atención es proporcionada a los menores deveintiocho dias en el rubro evitable por « prevención, disgnóstico ytratamiento médico precoz ».

Estos resultados, estarían evaluando la importancia del tipo deprofesional que supervisa el normal crecimiento y desarrollo del niño, que

61

implica anamnesis, examen físico, control pondoestatural y psicomotor eindicaciones de alimentación, vacuna y educación de la madre.

De 1.685.503 controles de salud practicados a los menores de un añoen 1979, el 67,8 96 los realizan las enfermeras, el 17,096 las auxiliares, el10,5 % los médicos y el 4,7 96 las matronas; estas últimas sólo atienden alos menores de veintiocho días.

Si se plantea que el niño, aunque aparentemente sano, es unindividuo potencialmente enfermo, la atención que recibe debe provenirde una persona formada en habilidades diagnósticas y que sea capaz dediscriminar en caso de encontrar signología sospechosa. Es comprensibleque, cuando el recurso médico es inexistente o insuficiente, otrosprofesionales traten de suplir la carencia. Esta situación se entiende comode emergencia y no como rutina establecida. Si en un momentodeterminado fue necesario adoptar esta política, pareciera convenienterevisarla y situar a los profesionales en las labores para las cuales estánmejor capacitados, puesto que, el control de salud, sobretodo al menor deveintiocho días, es el punto inicial para prevenir las enfermedades y porconsiguiente, la muerte por causas evitables.

Consultas de morbilidadLas consultas de morbilidad, aparecen relacionadas en forma directa

con las causas de muerte evitables por « buen control del embarazo yatención del parto », cuando esta actividad es proporcionada porenfermeras a menores de veintiocho días. Este resultado podría explicarse,por una parte, por la carencia o insuficiencia del recurso médico queobliga a delegar esta acción y por otra, por la calidad de la prestación,puesto que, es el médico el único con formación básica para otorgarla.

Visitas domiciliariasLas visitas domiciliarias a recién nacidos, están asociadas en forma

inversa con tres grupos de causas de muerte evitables, indicando que lamortalidad disminuye cuando se incrementa esta actividad. En el país, lasenfermeras realizan 121 visitas domiciliarias anuales por mil nacidosvivos, variando de 3,4 (mínimo) a 370,5 (como máximo). La participaciónde las auxiliares, es de 33 visitas anuales en promedio, por mil nacidosvivos.

Pese a lo reducido de su volumen, la relación indica que tienenimpacto en la reducción de la mortalidad infantil asociada a defuncionesevitables por « buen control del embarazo y atención del parto », « buenascondiciones de saneamiento » y « accidentes », es decir, la visita decarácter ecológico-pronóstica al hogar, cumple con sus objetivos.

Otra interpretación sería que mayor número de visitas domiciliariassea consecuencia de mayor disponibilidad de recursos, o también, deconcentración de la población en áreas urbanas que implica mejor accesoa los centros de salud y ser estas últimas variables las que, en forma masimportante, inciden en el descenso de la mortalidad.

62

Camas pediátricasLas camas pediátricas, muestran asociación positiva con la

mortalidad por « enfermedades respiratorias » y « accidentes ». En amboscasos, esta relación prodría ser consecuencia de que aquellos Servicios demás alta mortalidad por estas causas, corresponden a zonas con mayorporcentaje de ruralidad, donde el acceso a los centros de salud, es difícil,por lo tanto, existe el recurso cama, pero no se ocupa en todas susposibilidades porque el enfermo acude sólo en casos de extrema gravedado bien, que las camas estén insuficientemente equipadas. Los datosobservados corroboran esta apreciación : los Servicios de Santiagomuestran entre 16,6 y 42,0 camas pediátricas por mil nacidos vivos,mientras que Valdivia y Aisén de mayor mortalidad registran 42,8 y 49,7respectivamente.

Consultas por desnutriciónLas consultas médicas proporcionadas a desnutridos moderados,

muestran asociación directa con la mortalidad evitable por « buenascondiciones de saneamiento » ; probablemente, este resultado refleja laestrecha relación que existe entre el factor desnutrición y las muertes pordiarreas, y ser consecuencia, de que los Servicios tengan mayor númerode consultas porque existe un mayor númenrde desnutridos, por lo tanto,la mortalidad es más elevada.

Entrega de lecheDada la magnitud del programa, se esperaba una estrecha

vinculación entre esta variable y mortalidad evitable por « alimentacióncompleta » ; efectivamente, es la variable de mayor potencia en laexplicación de la varianza, sin embargo, su asociación es positiva.

Se puede argumentar que diversos factores socioeconómicos yculturales estarían también incidiendo en la obtención de este resultado,por una parte, el hecho empírico observado, que en aquellos Servicios connivel más bajo de mortalidad infantil evitable por « alimentacióncompleta » (Santiago Oriente, Central, Sur Oriente) se retira menos leche(aproximadamente un kilo mensual por nacido vivo) y en los Servicios demás alta mortalidad por esta causa (Iquique, Talcahuano, Concepción,Arauco), el promedio de leche retirada es mayor (alrededor de 1,9 kilosmensuales por nacido vivo). Por otra parte, existen antecedentes de unautilización inadecuada de la leche por las familias beneficiarías t19J ytambién se ha discutido el valor de la leche complementaria enpoblaciones en que la prevalencia de deficiencia de lactosa pudiera seralta120'.

63

2.2. Atención médica y mortalidad infantil por todas las causas y causasevitables.

A nivel nacional, la mortalidad infantil por todas las causas y causasevitables, aparecen asociadas con los mismos indicadores de atenciónmédica : horas médicas pediátricas, atención profesional del parto,consultas médicas a desnutridos moderados y camas pediátricas,manteniéndose el orden de importancia y el sentido de las relaciones. Laprincipal diferencia es la magnitud del coeficiente de determinación,donde la variación de la mortalidad infantil por causas evitables, resultaser mejor explicada (r2 = 86,1%).

Resta un 14,0% de la variación de la mortalidad evitable que nosería controlada por las variables estudiadas del sector salud.

Las horas médicas anuales pediátricas contratadas y la atenciónprofesional del parto, han resultado ser las variables de salud másimportantes para explicar las diferencias de mortalidad infantil entre los '27 servicios. Creemos que este resultado sintetiza los hallazgos obtenidos yexplicados en detalle en cada grupo de causas de muerte evitables.

DISCUSIÓN Y COMENTARIOS

Si bien es cierto que las estadísticas de salud y de hechos vitales hanexperimentado un evidente progreso en el tiempo, continúan presentandolimitaciones en relación a calidad, exactitud e integridad [211 que se trataronde atenuar en lo posible, para evitar problemas en el análisis de tasas eindicadores así como también, en la comparación de niveles entre lasdistintas unidades geográficas.

En el caso de los nacimientos, el Ministerio de Salud aplica un factorde corrección de 8,9 % de omisión a los nacimientos totales, derivado deun estudio realizado en 1966 I22l En esta oportunidad, se actualizó laintegridad del registro de nacimientos, ajustándose a la omisión estimadapor el Instituto Nacional de Estadística del 5 % para el total nacional í23iAdemás como la subenumeración es diferencial según lugar de residencia,se calculó la omisión para cada uno de los 27 Servicios de Salud I10!.

Respecto de las defunciones, a pesar que diferentes estudiosmuestran omisión de la inscripción en el Registro Civil I241 y errores en lacertificación de la causa de muerte[3], esta información no se corrigió porno tener antecedentes suficientes y confiables para estimar la magnitud delproblema.

Por otra parte, al clasificar las muertes infantiles, se observó que lasdefunciones por causas desconocidas presentaban un peso relativo de17,7%, teniendo al mismo tiempo, una alta variabilidad entre losServicios de 2,6 a 38,1 (141 por lo tanto, se optó por prorratearlas en forma

64

proporcional a la importancia relativa de cada grupo de causas, para evitardistorsión en el análisis de niveles de mortalidad.

Para la construcción de los índices de salud, se adoptó el criterio derelacionar las actividades y recursos con el total de nacimientosregistrados en cada Servicio del país. El numerador registra sólo lasprestaciones otorgadas a los niños beneficiarios del Ministerio de Salud yel denominador, reúne todo el universo de nacidos vivos en situación dedemandar atención médica, que puede ser satisfecha en institucionesdependientes o no del Ministerio. Cabe destacar que en el momentoactual, los establecimientos integrados al Sistema Nacional de Servicios deSalud realizan aproximadamente dos tercios de la atención ambulatoriadel país y 9096 de las hospitalizaciones, siendo gratuita para toda lapoblación la atención inicial de urgencia, la atención primaria maternainfantil, incluyendo la planificación familiar, la entrega de leche y ladetección del cáncer cérvico-uterino, como también las vacunaciones y elcontrol de las enfermedades infecciosas M.

Se ha coincidido con otros autores I3> 4> ^ en señalar importantesdiferencias geográficas en la mortalidad infantil, encontrándose que en losúltimos 35 años sistemáticamente las mismas zonas son las que mantienenlas tasas más altas y más bajas. En efecto, los Servicios de Salud del áreaMetropolitana poseen las tasas menores, destacándose Santiago Orientecon 21,0 defunciones por mil nacidos vivos y alejándose de la capitalhacia el sur, los indicadores son más altos, especialmente en la penúltimaregión, Aisén con 68,7 por mil.

Las muertes teóricamente evitables significan el 84,0 % del total dedefunciones infantiles, estando los mayores riesgos de muerte vinculadoscon las causas evitables por « buen control del embarazo y atención delparto » y « enfermedades respiratorias » que constituyen el 52 % del totalde decesos infantiles y el 61,996 del total de muertes infantiles evitables.

En el macronivel de análisis se ha planteado como factorescondicionantes o contribuyentes al descenso de la mortalidad infantil loscambios en la conducta reproductiva de la población, medido empírica-mente por : reducción de la natalidad (36,3 nacidos vivos por milhabitantes en 1965 y 22,3 en 1979 '25' ; concentración casi en un 70% enlos nacimientos de orden uno y dos y acumulación de los partos enmadres jóvenes, 60 % de los nacidos vivos provienen de mujeres entre 20y 29 años. Los hechos anteriores, determinan una mayor proporción denacimientos de hijos deseados que probablemente recibirán una mejoratención y cuidado materno disminuyendo el riesgo de muerte de estosniños [5. 26, 27, 28].

La disminución de la desnutrición en la población controlada ^2i\particularmente intensa en los grados medios y avanzado, puedeconsiderarse como otro hecho satisfactorio y favorable.

Tampoco debe desconocerse el progreso que se ha producido en elpaís en otros sectores y que puede tener efecto en la población infantil. Es

65

el caso del alfabetismo (90% en 1970), concentración de la población enáreas urbanas (85% de población urbana en 1979), mayor disponibilidadde agua potable y alcantarillado, mejorando las condiciones sanitarias delas viviendas M.

También se ha señalado que la mortalidad infantil tiende a ser dos omás veces mayor en aquellos sectores geográficos de Chile de bajo nivelde vida que en aquellos de elevado nivel ; usando el salario medio comoindicador de situación socioeconómica se encontró un coeficiente decorrelación de -0,46 M.

De los resultados obtenidos, creemos que la importancia del estudioreside en que ha permitido cuantificar el peso de la atención médica(superior al 80%) en la reducción de la mortalidad infantil por todas lascausas y por causas evitables. Además, la potencia de las variables haseñalado que las horas médicas pediátricas anuales contratadas y laatención profesional del parto, son las actividades de salud fundamentalespara explicar la variación de la mortalidad infantil entre Servicios deSalud.

También, de los resultados se desprende que el sector salud tiene unaacción limitada para evitar muertes por « alimentación completa » y« accidentes » (r2 = 57,7 y 29,3..% respectivamente) causas que seríanmejor explicadas por factores asociados al nivel de vida.

Por otra parte, de las variables que mostraron una asociación menoral actuar conjuntamente otros indicadores, se destaca por su volumen laprincipal actividad del programa materno-infantil : los controles prenata-les, que en promedio, significan 5,1 control de matrona por nacido vivo y0,4 por médico. El enorme esfuerzo que significa la cantidad deprestaciones registradas, tiene un efecto débil en la reducción de lamortalidad por causas périnatales, lo que pone de manifiesto la necesidadde estudiar en profundidad la calidad de esta actividad.

Otro hallazgo significativo, se relaciona con las variables controles desalud y atención de la morbilidad, que muestra una relación directa conalgunos grupos de causas de muerte evitables, cuando estas acciones sonefectuadas por auxiliares de enfermería. Este hecho está revelando poruna parte, falta de recursos, que en algunos Servicios es agudo y por otra,que habría que revisar las normas sobre encomendación de funciones ycapacitar al personal paramédico para evitar el deterioro de la atenciónmédica proporcionada a los menores de un año.

En resumen, el presente estudio proporciona una evaluación de lasvariables de atención médica que permite asignar prioridades en laprogramación, especialmente en la destinación de recursos, revisar lacalidad de las acciones que se están otorgando, individualizar lasactividades mas relacionadas en los grupos de causas que provocan losmayores riesgos de muerte del menor de un año, detectar los Servicios deSalud menos eficientes para reducir la mortalidad y ampliar el marco dereferencia para el estudio de los determinantes de la mortalidad infantil.

66

REFERENCIAS BIBLIOGRÁFICAS

[I] CICRED. - Infant and Child Mortality in the Third World. Proceedings of theinitiating meeting. Specific report of group n° 2. Carolina Population CenterEEUU. Sept. 3-6th 1979.

[2] ROMERO M.I., MAJLUF N. y Cols. - « Nivel de Salud y Atención PediátricaPreventiva. Une aplicación de Ingeniería de Sistemas». Cuad. Med.-Sociales.Vol. XVII, N° 2 : 7-13, Stgo.-Chile, Junio 1976.

[3] BEHMH. - Mortalidad infantil y nivel de vida. Ed. U. de Chile, Stgo. 1962.[4] LIVINGSTONE M., RACZINSKI D. - « Distribución Geográfica de la Salud del

Preescolar ». Salud Pública y Bienestar Social. Ceplan. U. Católica, Stgo-Chile,1976, pp 179-217.

[5] TAUCHER E. - Mortalidad Infantil en Chile. Tendencias, Diferenciales y Causas.Celade, Stgo. Julio 1979.

[6] DE LA FUENTE M., GONZÁLEZ J. y Cois. - « Nivel de Vida y Salud. Análisis porRegiones y Concepto de Pobreza Extrema (PEX) ». Cuad. Med. Sociales.Vol. XVII, N° 4 : 13-20, Stgo.-Chile, Diciembre 1976.

[7] KAEMPFFER A.M., MEDINA E. - Perspectivas en la Salud del Niño Latinoamericano.Mimeo 9055. Depto. Salud Pública y Medicina Social. U. de Chile. 1977.

[8] MEDINA E. - Elementos que Condicionan la Eficacia del Sistema de Salud. RelaciónEntre las Tendencias de los Problemas Sanitarios en Chile, la Atención Médica y laEstructura del Sistema. (Inédito).

[9] Me. CORMICK M., SHAPIRO S., DADAKIS S. - "The Relationship Between InfantMortality Rates and Medical Care and Socioeconomic Variables, Chile 1960-1970",International J'. of Epidemiol. Vol. 8 N° 2. Great Britain.

[10] MARDONES G., CASTILLO B., SOLÍS F. - Chile.- Integridad del Registro de NacidosVivos Correspondiente a los 27 Servicios de Salud del País. Anos 1969-1978.(Inédito).

[II] ORGANIZACIÓN MUNDIAL DE LA SALUD. - Clasificación Internacional de Enferme-dades. Octava revisión, Ginebra 1968.

[12] BARR A. y Cols. - Statistical Analysis System User's guide. S.A.A., Institute,Inc. 1979.

[13] NIE N. y Cols. - Statistical Package for the Social Sciences (S.P.S.S.) Cap. 20. MacGrawHill. 1975.

[14] SOLIS F., CASTILLO B., MARDONES G. - Chile : Mortalidad Infantil por Grupos deCausas Evitables en los 27 Servicios de Salud del País. Año 1979. Rev. Med. deChile, Abril 1982. (En prensa).

[15] MINISTERIO DE SALUD. - Normas sobre Acciones de Fomento y Protección enPediatría. Stgo. Septiembre 1976.

[16] CASTILLO B., SOLÍS F., MARDONES G. - Chile: Atención Médica y MortalidadInfantil en los 27 Servicios de Salud del País. Año 1979. (Inédito).

[17] DE KADT E. y Cois. - « Políticas y programas de salud, 1964-73 ». Salud Pública yBienestar Social. Ceplan. U. Católica. Stgo-Chile 1976, pp. 111-150.

[18] ACUÑA C, WOLFF R. - Mortalidad Materna ven la Niñez en las 25 Areas Pesmib.1970-1975. Ministerio de Salud Pública. Oct. 1976.

[19] UNDURRAGA O., MARGOZZINI J., MARÍN G. - « Aprovechamiento y Conservaciónde la Leche Semidescremada Entregada por el Servicio Nacional de Salud a susBeneficiarios ». Rev. Cliil. de Pediatría, 40 : 1039, 1969.

67

[20] PAIGE D.M., BAYLESST.M., GRAHAM G.G. - "Milk Programs: Helpful or Harmfulto Negro Children ?" Am. J. of Public Health. 62: 1486, 1972.

[21] MARCHANT L. - « Las Estadísticas de Salud en Chile ». Salud Publica y BienestarSocial, Ceplan. U. Católica. Stgo.-Chile, 1976, pp. 305-321.

[22] GUTIÉRREZ H. - Integridad del Registro de Nacidos Vivos en Chile. Facultad deMedicina. U. de Chile. 1968.

[23] INSTITUTO NACIONAL DE ESTADÍSTICAS. Demografía. Años 1972-73. INE. 1979.

[24] LEGARRETA A., ALDEA A., LÓPEZ I. - « Omisión del Registro de Defunciones deNiños Ocurridas en Maternidades, Stgo. Chile ». Boletín OPS. Vol. 75. N° 4 :308-314, Octubre 1973.

[25] MINISTERIO DE SALUD. Anuario de Nacimientos. 1965 y 1979.

[26] MEDINA E., KAEMPFFER A.M. - Elementos de Salud Pública. 3e Ed. Stgo., Edit.Andres Bello, 1978.

[27] KAEMPFFER A.M., MEDINA E. - « Morbilidad y Atención Médica Infantil en elGran Santiago». Rev. Chil. Pediatría. 51 : 355, 1980.

[28] VARGAS N., CARRETERO A., PENA C. - Variación de Algunos Factores de Riesgo dela Mortalidad Infantil en Chile entre 1975 y 1980. Resultado en 1 JornadasNacionales de Salud Pública. Stgo. Chile. Octubre 1981.

[29] MEDINA E., KAEMPFFER A.M. - « Morbilidad y Atención Médica en el GranSantiago». Rev. Med. Chile. 107 : 155, 1979.

[30] KAEMPFFER A.M., MEDINA E. - « Perspectivas de la Salud del Niño Latinoameri-cano ». Pediatría. 22 : 524, 1979.

Agradecimientos

Las autoras expresan su gratitud al prof. Claudio Silva, Jefe del Laboratorio deEstadística de la Universidad de Santiago, al Sr. Samuel Avila, del Centro de Computacióndel Hospital José Joaquín Aguirre, a la Dra. Erica Taucher de Celade y a las Dras. AnaMaría Kaempffer, María Ines Romero y al Sr. Francisco Cumsille, de la Escuela de SaludPública de la U. de Chile, por su valiosa ayuda en la elaboración del proyecto,procesamiento de la información y útiles sugerencias en el análisis y comentarios de losresultados.

68

x3x«x5x.x7x.x,

X,4

XXXXXXX

xXXXXXXX;xX;xX

xXxxX

X.

16

17

19

20

21

22

23

24

27

2»

29

3D

31

32

•33

34

37

38

3 9

ANEXO. - Listado de variables de atención médica

Controles de salud a menores de un año efectuados por médico.Controles de salud a menores de un año efectuados por enfermera.Controles de salud a menores de un año efectuados por aux. de enfermería.Consultas de morbilidad de menores de un año, atendidas por médico.Consultas de morbilidad de menores de un año, atendidas por enfermera.Consultas de morbilidad de menores de un año, atendidas por auxiliar.Consultas de desnutridos menores de un año, atendidas por médico.Consultas de desnutridos leves menores de un año, atend. por médico.Consultas de desnutridos moderados menor, de un año, atend. por médico.Consultas de desnutridos avanzados menor, de un año, atend. por médico.Consultas de desnutridos menores de un año, atend. por nutricionista.Consultas de desnutridos leves menores de un año, atend. por nutricionista.Consultas de desnutridos moderados menor, de un año, atend. por nutricionista.Consultas de desnutridos avanzados menor, de un año, atend. por nutricionista.Visitas domiciliarias a recién nacido, efectuadas por enfermera.Visitas domiciliarias a recién nacido, efectuadas por auxil. enfermería.Madres de niños menores de 2 años que asisten a charlas educativas.Controles prenatales efectuados por médico.Controles prenatales efectuados por matrona.Consultas de morbilidad a menores de 28 ds., atendidas por médico.Tercera dosis de vacuna triple, proporcionada a niños de 5 a 11 ms.Vacuna antisarampionosa, proporcionada a niños de 8 a 11 meses.Kilos de leche entregados mensualmente a menores de 6 meses y nodrizas.Número de camas pediátricas.Número de camas obstétricas-ginecológicas.Egresos hospitalarios obstétricos-ginecológicos.Egresos hospitalarios pediátricos.Consultas de desnutridos menores de un año, atendidas por enfermera.Consultas de desnutridos leves menor, de un año, atend. por enfermera.Consultas de desnutridos moderados menor, de un año, atend. por enfermera.Consultas de desnutridos avanzados menor, de un año, atend. por enfermera.índice ocupacional de camas pediátricas.índice ocupacional de camas obstétricas-ginecológicas.Número de nutricionistas.Número de enfermeras.Número de matronas.Horas médicas pediátricas contratadas (incluye becarios).Horas médicas obstétricas y ginecológicas contratadas (incl. becarios).Porcentaje de nacimientos con atención profesional.Controles de salud a menores de 28 ds., efectuados por médico.

69

X41 : Controles de salud a niños de 28 ds. a 1 lms., efectuados por médico.X42 : Controles de salud a menores de 28 ds., efectuados por enfermera.X4j : Controles de salud a niños de 28 ds. a 11 ms., efectuados por enfermera.X44 : Controles de salud a menores de un año, efect. por médico o enfermera.X4B : Controles de salud a menores de 28 ds., efect. por médico o enfermera.X46 : Controles de salud a niños de 28 ds. à 1 lms., efect. por méd. o enfermera.X47 : Controles de salud a menores de 28 ds., efect. por aux. de enfermería.X„ : Controles de salud a niños de 28 ds. a 11 ms., efect. por aux. de enfermería.X49 : Consultas de morbilidad de niños de 28 ds. a 11 ms., efect. por médico.X50 : Consultas de morbilidad de menores de 28 ds., efect. por enfermera.X51 : Consultas de morbilidad de niños de 28 ds. a 11 ms., efect. por enfermera.Xj2 : Consultas de morbilidad de menores de 28 ds., efect. por aux. de enfermería.X53 : Consultas de morbilidad de menores de un año, efect. por méd. o enfermera.XM : Consultas de morbilidad de menores de 28 ds., efect. por méd. o enfermera.Xss : Consultas de morbilidad de niños de 28 ds. a 11 ms., efect. por méd. o enf.X96 : Consultas de morbilidad de niños de 28 ds. a 11 ms., efect. por aux. enferm.X97 : Kilos de leche entregados mensualmente a menores de un año y nodrizas.X58 : Kilos de leche entregados mensualmente a niños de 6 ms. a 11 ms. y nodrizas.Xs9 : Controles de salud a menores de 28 ds., efectuados por matrona.

COMMUNITY VARIATIONS ININFANT AND CHILD MORTALITY

IN PERU: A SOCIALEPIDEMIOLOGICAL STUDY

by

Barry EDMONSTON and Nancy ANDESInternational Population Program

Cornell University, IthacaNew York

INTRODUCTION

Recent decades have witnessed an important change in our thinkingabout health and disease. There has been a striking movement towardconsideration of disease within a broader ecological framework and,likewise, an alteration in describing disease treatment solely in terms oftechnology. This change in our thinking about health has occurred duringa period when development goals have been no longer defined only interms of economic progress. Today, national health strategies usuallyinclude reflection of the social, economic, and political aspect- as well astechnology - and development plans have been increasingly broadened torecognize the health component. Consequently, health and developmentare now typically treated as interrelated, overlapping phenomenon, withoften mutual goals.

Such developments in the concept of health, and by implication inhealth planning and policy, result in new directions for health research.Although we surely need continued work on specific disease processesand on clinical medical treatment, a broader scope of health necessitates aconcern with various environmental, family, public programs, andmedical facilities, and their effect on health. Thus, we should be interestedin influences on health, not because they are specifically health programs,bu rather if they in fact have a health effect. More traditional medical

72

research has tended to stress clinical or narrow public (e.g., immuniza-tions) interventions; within an ecological framework, we require awholistic concept of the major disease influences.

For overall summary, four general influences on disease may bedistinguished. First, there are biological characteristics of the individual,including age, sex, various genetic characteristics, and idiosyncraticsusceptibilities. These are qualities on which neither medical treatmentnor public programs have much effect. Second, there are diverseindividual characteristics, both voluntary and largely involuntary, thatimpinge on health. These include voluntary habits such as smoking andfood choices as well as the general context of making these choices. Publicprograms, mainly in the area of education, can and have been tried toinfluence personal habits with the aim of improving self-identified healthgoals. Third, there are environmental conditions forming the larger basisfor individual health. The two major components of the environment arethe family, which offers a fundamental economic and nutritional context,and the community, which provides a basis for a diversity of publicservices (e.g., food, water, sanitation, and economic conditions). It isprimarily in this area of interest that recent enlarged health programs havebegun to work. Fourth and finally, there are direct medical and healthinterventions that have formed and will continue to be a crucial aspect ofindividual health.

This brief review suggests that our research attention should belimited to a broad set of conditions affecting health, along with selectedstudies emphasizing specific disease conditions and their treatment. Inaddition, our review warns us that there are aspects of health beyond ourinfluence, whether attempted as our most desirable social goals or aimedfor by individual aspirations.

The research reported here does not offer a complete specification ofhealth determinants, but does consider a selected set of variables withinthe framework outlined. We begin by taking into account known age-sexvariations in mortality by developing a standardized mortality ratio. Wewill then see that Peruvian community mortality varies by someimportant health influences. In particular, we will look at an overallmeasure of community development, a measure of family social andeconomic status (average years of female education), two indicators ofspecific health intervention (medical facilities and public sanitation), anddifferences by region and altitude.

The assumption underlying this inquiry is that there existsconsiderable variation in community mortality, and it is possible andimportant to understand even partially the broad factors related to thesevariations. The hope of this research is that it might clarify discussion ofwhat factors are subject to improving the health of children.

73

DATA

1. Peru Fertility Survey (PFS)

Data reported in this paper are derived from the household interviewand community inventory of the Peru Fertility Survey. The Peru FertilitySurvey (PFS) was carried out between July 1977 and June 1978 by theDemography Section, Census Bureau, National Statistical Office, underthe direction of Mr. A. Raul Garcia. It should be mentioned that the PFSis associated with an international series of fertility surveys, the WorldFertility Survey, which has been a pioneering effort to better understandthe levels, determinants, and factors associated with fertility.

It should go without saying that this research would not be possiblewithout the efforts of the statisticians, demographers, and workers on thePFS. It is an exceptionally fine achievement, and we sincerely appreciatethe work that went into this survey, and the permission from Peruvianauthorities to work with these data.

The PFS was a three-stage area/multi-probability sample. The firststage included the selection of 124 primary sampling units. In 93 of theseunits, a community survey was taken, with questions about communityfacilities, health conditions, and economic aspects. The second stageyielded a systematic selection of 8979 households, and the third stageproduced interviews with 5640 ever-married women, 15-49 years of age.

Data collected in the- PFS appear to be of excellent quality forstudying retrospective infant and child mortality. Comparisons of fertilityfrom WFS-associated surveys indicated generally complete coverage ofbirths, as reported by the maternity histories.

The omission of births is probably low in the PFS. More importantto our concerns here is that there is no evidence of omission of infantdeaths. The retrospectively reported infant mortality rates for the period1968-1977 is 102 per 1000, which is close to and in temporal agreementwith the figure of 92 per 1000 reported in the 1978 United NationsDemographic Yearbook. In more intensive checking than can be reportedhere, we have found the mortality data in the PFS to be consistent, andgenerally free of suspected omissions.

The problem of temporal displacement of reported events is not ofgreat concern for this research since we are not interested in the trendsover time. But while the data of the event, birth or death, is not of specialinterest, the PFS does present a problem in the age of death. The PFSreports age of death in age groups, coded 1 to 7. In order to assign an exactmonth of death, we imputed an age of death in the following manner. Weassumed a uniform distribution of deaths within an age group for thosechildren dying in that age group. If, however, the interview was

74

conducted at a date that would limit the date of death (right censorship ofthe date of the event), the uniform distribution was constrained of theperiod up to the interview date. Then within the possible duration ofdeath, a random number for a uniform distribution was generated. In thisways, all 25,504 children reported in the PFS were, if dead, assigned anage of death and, by addition to the date of birth, were given a time ofdeath.

For example, if a mother reported one of her children was nowdead, and the age of death was coded "4", indicating death during 6-11months, a random number of 6, 7, 8, 9, 10, or 11 was generated. Unlessthe interview was conducted at a date to make impossible later months,the random number was imputed as the child's age of death. Thisimputation seems reasonable, and should not bias any of the resultsreported here.

2. Mortality Ratios

The basic data for this study are mortality ratios, by age and sex, forthe 124 communities in the PFS, Peru. Mortality ratios were calculated toexpress to community's mortality relative to the Peruvian nationalaverage. First, all children living during 1968-1977 were included tocalculate by sex: (a) the probability of living from birth to age 1, (b) theprobability of living from age 1 to age 5, and (c) the probability of living tothe date of interview. The probabilities for (a) and (b) from the PFS dataare:

Birth to Age I Age I to Age 5

Males .8948 .9362Females .9000 .9437

and the probabilities for (c) include separate figures by sex for each of theten years prior to the survey.

Second, all children in each community were aggregated by sex andyear of birth. The number actually alive at age 1, age 5, and the surveywere compared to how many would be expected to be alive with nationalprobabilities. The actual mortality ratio is defined as:

Mortality Ratio = B o r n ' A l i v e

Born - Expected

Suppose 100 females are born in a community, and we therefore expect(100) x (.9000) = 90 to survive to age 1. If 95 actually survive, the

75

mortality conditions are better than nationally and we have a mortalityratio of (100-95)/(l 00-90) = 0.500. Note that the mortality ratios centeron a national average, by definition of 1.0, and evidence worse conditionswhen greater than 1 and better conditions when less than 1.

For the mortality ratios until the date of the interview, the numberexpected alive is the sum of all the age-sex groups who would be expectedto be alive. The Lima area illustrates this calculation. There were 2417children of both sexes reported during 1968 to 1977, and 2262 wereactually alive at the time of the interview. Based on the date and sex ofbirth, the national probabilities produce an expected 2071 alive at the dateof interview. The overall mortality ratio for Lima is therefore:

2 4 1 7 ' 2 2 6 2 = 0.452417-2071

This shows that, taking the age and sex distribution into account, Limahas about 45 percent of the expected proportion of number dead, givennational averages.

The mortality ratios reported here are essentially an adaption ofdirect standardization, a familiar demographic technique. Its usefulnesshere is that community mortality levels can be compared without concernthat they might arise because of differential age or sex structure.

RESULTS

The main findings of our epidemiological inquiry are contained inthree sections. First, the time series in mortality ratios show that therehave been no major fluctuations in the variation of community mortality.Second, examination of community mortality for selected factorsrevealing differential mortality by (a) medical facilities, (b) publicsanitation levels, (c) region, and (d) altitude. Third, we present a measureof community development which is then incorporated in a multivariateanalysis to explain variations in community mortality.

1. Mortality Variations

The average mortality ratio for both sexes, from birth to age 10, is1.08 for the 124 communities. The ratio ranges from 0 (there was acommunity with a small number'of women in which no women reporteda death during the past ten years) to 2.4, with a standard deviation of 0.51.Table 1 shows the mean mortality ratio for each year, 1968 to 1976 in thetop panel. There have been minor fluctuations, but no year of exceedinglyhigh or low national mortality. The standard deviation has also not varied

76

greatly on an annual basis, suggesting stability in the distribution.Individual community data are not shown here: if they were shown, theywould indicate comparable mortality levels for most communities overtime. That is, generally high mortality communities tend to have higherthan expected mortality on an annual basis. Because of this, we concludethat higher mortality communities suffer from persistent disadvantage,and probably not from a peculiar disaster.

The next panels of Table 1 show additional community variations inmortality. Metropolitan Lima has relatively low mortality overall, lessthan one-half the nationally expected level. Communities in the coast andjungle (selva) areas are slightly below national levels. Mountain (siena)communities, comprising about one-half the total communities, havemortality levels about one-third above the national averages.

A strong rural-urban difference in mortality exists. Predominantelyrural communities have relatively high mortality, and urban areas areslightly below the national averages.

TABLE I. - Mortality Ratios for Both Sexes, Birth to Age Tenfor 124 Peruvian Communities

StandardVariable Number Mean Deviation

Peru, by year:196819691970197119721973197419751976Overall

Region :LimaCoastMountainJungle

Residence:RuralUrban

Altitude, in meters:Less than 10001000 to 19992000 to 27992800 to 34993500 to 39994000 and more

124124124124124124124124124124

1396420

5173

5291323225

1.161.041.090.951.140.960.991.041.061.08

0.450.831.290.92

1.310.91

0.891.290.801.291.301.50

1.301.151.240.911.231.091.121.161.330.51

0.480.490.33

0.460.49

0.410.560.390.460.490.31

77

Finally, an interesting mortality pattern by altitude exists in Peru.Mortality is somewhat lower in the less than 1000 meters and in the 2000to 2999 meters regions. Mortality is higher elsewhere, especially in thecommunities at 4000 meters and over.

We begin our empirical investigation by noting these differences inmortality, and by wondering to what extent do they result fromdifferences in community development, levels of health resources, and thefamily situation? Since communities surely differ in these other variablesthat are important to health, how much variation in mortality persistsafter taking these variables into account? We continue by describing therelationship of mortality to the additional variables.

2. Variations in Mortality Ratios

A series of charts show the variation in the mortality ratios forselected variables. These charts display each category according to theproportion of communities with high, moderate, and low mortality. Thereare three lines in each chart, lying between 0 and 100 percent. The bottomline represents the percentage of communities with high mortality, a ratiogreater than 1.5. The proportion between the bottom line and the middleline are moderately high mortality communities, ratios between 1.0 and1.5. And the next group, between the middle line and the top line,represents moderately low mortality with ratios between 0.5 and 1.0.Finally, the communities between the top line and 100 are low mortalitycommunities, with ratios less than 0.5.

Figure 1 shows variations by region. Lima has exceptionally lowmortality, the coast and jungle areas have similar levels, and the mountain

Number

Figure 1. Proportion of Communities by Mortality Level,According to Regions of Peru

78

area has higher mortality. Note that the mountain areas differ mainlybecause of a greater proportion of communities with high, greater than1.5, mortality ratios.

Rural areas have higher mortality than urban areas, as seen inFigure 2. The primary difference stems from the greater proportion ofrural communities with high mortality, and the expense of fewercommunities with low and moderately low mortality.

Lou

Moderately low

Moderately high

High

Urban

(73)

Rural

(51) Number

Figure 2. Proportion of Communities by Mortality Level,According to Rural and Urban Location

100

None

(40)

Low

Moderately low

Less than

(21) (2 8)

All

(35) Number

Fip.ure 3. Proportion of Communities by Mortality Level,According to Percent of Households with Water

79

Figure 3 displays variations according to water supply. In theprevious section, the mean showed an erratic pattern, although we seehere a tendency for the category of high mortality to diminish withimprovements in water. On the other hand, there is an increase inmoderately high mortality in communities with some but not completewater supply.

Sewer facilities, shown in Figure 4, also show a diminution of highmortality with at least 50 percent sewers. There does appear to be loweroverall mortality when there is complete sewers for the community.

Figure 5 presents mortality variations for medical facilities. As with

Low

Moderately low

Moderately high

75

50

25

i i

r

into^ ^ ^ 17483

-

High

None Less than More than50% 50%

(58; (15) (19)

All

(32) Number

Figure 4. Proportion of Communities by Mortality Level,According to Percent of Households with Sewers

(56) (35) (31)

Low

Moderately low

Moderately high

High

Hospital

(48)(59)

Figure 5. Proportion of Communities by Mortality Level,According to Presence of Medical Facilities

Number

80

the mean, we see no striking pattern here. There appears to be nopronounced relationship with mortality nor a progressive decrease inmortality with enlarged facilities. The next section discusses thisrelationship in more detail.

Finally, mortality variations with altitude are shown in Figure 6. Wenotice an increase in high mortality ratios with altitude, starting about2000 meters, that becomes more dominant while there are decreases inthe proportions in the other categories. We note, in particular, that40 percent of communities at 3500-3999 meters have high mortality and80 at 4000 meters or above have high mortality.

3. Community Variables

Community conditions are reflected in three scales created byGuttman scalogram analysis (described in more detail in Appendix A).These scales were constructed to indicate three aspects of the community:first, the type of medical facilities available; second, the extent ofsanitation and public services; and third, an additive measure of generalcommunity development. These variables were all taken from the PFS-Peru community survey.

The variables comprising the medical facility scale are hospitals,clinics, medical centers, dispensaries, pharmacies, and medical posts. Thesanitation scale includes sewers, water, and electricity. The communitydevelopment scale includes (a) a transportation scale, composed of thetypes of roads accessing the community (asphalt, semiasphalt, true road,rough road, and bridle path); (b) a commerce scale, comprised of cinemas,

100

75 -

50 -

25 -

Moderately

1

low

\\

1

Low 1 f 1

/ Moderately high >

y High-' i i

yinto

17383

Less than1000

(52)

1000-1999

f9)

2000-2799

(13)

2800-3499

(23)

3500-3999

(22)

4000or more

(5) Number

Figure 6. Proportion of Communities by Mortality Level,According to Altitude in Meters of the Community

81

banks, markets, transient markets, security services, stores, and schools;and (c) a communication scale with television, telephone, radio,newspapers, and mail service variables. The overall communitydevelopment scale for each community is the sum of each of the threescales, ranging from a low scale value of 1 to a high scale value of 10.

We used the coefficient of scalability, a measure of the unidimensio-nality and cumulativeness of a scale, to evaluate the selection of Guttmanscales. A coefficient above 0.6 indicates that the index is a generally goodindicator of the underlying continuum. The scales used in this work havecoefficients well above the recommended 0.6 value. The coefficient ofscalability for sanitation is 0.83, 0.95 for communication, 0.83 forcommerce, and 1.00 for transportation.

The community survey was not taken in 31 communities withpopulations over 12,500 in the coast and mountain regions, and over2,500 in the jungle. Values were imputed to these communities for thethree community scales and for altitude. Based on the knowledge thatthese communities are urban and more developed, the maximum valuefor each scale was given to each community. The value for altitude wasassigned by locating each community on a map, and noting the altitude inmeters.

There were 13 communities which had community data on mostvariables, but were missing selected variables on the commerce orcommunication scales.

Values for these variables were assigned in two ways. First, tencommunities missing the code for cinema were also communities thatseemed to have few other facilities: the cinema code was assumed to bezero. Second, three communities had missing values for some variables ofthe commerce and communication scales. Inspection of the variablesavailable and the other scale were used to assign a value for thecommunity development scale.

4. Multivariate Analysis

We have seen that there are substantial variations in mortality bysome important variables. We have developed a broad, "fat" measure ofcommunity development from the ENF community survey. We now turnto a multivariate analysis of the relative importance of these variables. Inparticular, we will use multiple classification analysis (MCA) for thismultivariate analysis (Appendix B describes MCA in greater detail). Thereare three reasons for employing MCA at this point. First, it is important toconsider the impact of each independent variable in the case ofintercorrelated independent variables. While there is no unique solution tothe question of the exact influence attributable to an independent variablethat is intercorrelated with other independent variables, there are valuable

82

conclusions to be drawn from the form of the relationships. A secondpurpose of MCA is that it does not assume the effects of independentvariables are linear. That is, the categories of an independent variable canvary among each other in a way that is more realistic for studyingcommunity mortality. Third, MCA is used to see how the effect of avariable changes when other, possibly correlated variables, are includedin multivariate analysis.

Five independent variables are included in the MCA to explainoverall mortality up to age 10, mortality from birth to age 1, and mortalityfrom age 1 to age 5. The five variables are: (a) the Guttman scale forcommunity development, varying from low transportation, communica-tion, and urban facilities (scaled 1) to a high level (coded 10), (b) averageyears of female education, a measure of family social and economicresources, (c) medical facilities, varying from none to some medicalfacilities, (d) public sanitation and services, varying from none to thepresence of water, sewer, and electricity, and (e) altitude in meters, ameasure of an important geographical aspect of mortality in the Andeanregion.

a) Overall Infant and Childhood Mortality

Table 2 displays the MCA for mortality, for both sexes, for the 1968-1977 period. On the left-hand side are the five independent variables andthe categories of the variables. The second column shows the number ofcommunities in each category. The third column presents the mean foreach category, and the grand mean of 1.08 for all 124 communities. Thefourth column shows the unadjusted deviations, which are the categorymeans minus the grand mean; a positive unadjusted deviation indicateshigher mortality for that category. In the fifth column are the adjusteddeviations, the estimates for that category after holding constant the othereffects in the MCA model. The adjusted coefficients are, in other words,the net effect of that category on mortality after adjustment for the othervariables.

As a statistical preface, please note that the R2 of 0.46 indicates thatthis model, as well as the remaining two, accounts for a high proportion,almost one-half, of community variation in mortality. Looking briefly ateach independent variable, we see that the 36 communities with thelowest 3 levels of community development have higher mortality, aftercontrolling for other factors, and higher levels of development areassociated with lower mortality. Note, though, the important point thatcommunity development is not dominant, and that each incremental levelof development is not associated with a steady decrease in mortality. Itseems that community development has a general high-versus-low effect,rather than a progressive influence on mortality.

Average female education shows a striking relationship to mortality.

83

TABLE 2. - Multiple Classification Analysis of Mortality,Both Sexes, from Birth to Age Ten, for Peruvian Communities, 1968-1977

Independent Variable

Community Development:Low 1Low 2Low 3Low 4Low 5Low 6Low 7Low 8Low 9High 10

Average Female Education*0 - .99 years1 - 1.99 years2 - 2.99 years3 - 3.99 years4 + years

Medical Facilities:None, 0None, 1None, 2Some, 3

Public Sanitation:None, 0None, 1None, 2All, 3

Altitude, in meters:Less than 10001000 to 19992000 to 27992800 to 34993500 to 39994000 and more

Grand Mean

Multiple R2

Sample Size

SampleSize

1412106

111010101031

1721281741

28242448

33161263

529

132322

5

CategoryMean

1.02.98

1.291.281.151.381.311.151.01.83

1.531.341.131.01.76

.961.251.37.92

1.081.311.111.02

.891.28.80

1.291.301.50

1.08

.460124

UnadjustedDeviations

-.06-.10.21.20.07.30.23.07

-.01-.25

.45

.26

.05-.07-.32

-.12.17.29

-.16

.00

.23

.03-.06

-.19.20

-.28.21.22.42

AdjustedDeviations

.02

.25

.24-.01-.13-.05-.04-.22-.05-.02

.38

.28

.02-.05-.29

-.20.13.09

-.01

-.02-.02-.11.04

-.12.05

-.15.06.17.56

"Statistically significant at the .05 level.

even after controlling for other factors. Those communities with poorlyeducated women have about 40 percent more deaths while thosecommunities with more educated women have approximately 30 percentfewer deaths. Average female education is taken here to be a generalindicator of family economic resources, of formal knowledge, and of

84

ability to gain access to health resources. It appears that female education,as a general indicator of family resources, is highly associated withdifferential community mortality. We do not suggest here the exactlinkage of this relationship, but do note that this relationship persistsstrongly in the presence of other variables which might be thought todiminish the relationship.

Neither the presence or degree of medical facilities or publicsanitation, as measured here, seem to be powerful determinants ofcommunity. We wish to be very careful to add caution to this conclusion.First, the community survey merely ascertains the existence of a facility,and not how many people are served by it or its quality. Thus, acommunity may have a water supply, but we do not know if it is potablewater, for how long the water has been available, or if it is generallyaccessible to the residents. Thus, we should not expect these variables tobe superb predictors of mortality at the community level. Second,communities that have adequate medical and sanitation facilities are oftenmore developed and have higher average female education, both of whichare strong predictors. Third, health facilities are sometimes located inareas of the greatest need and hence, in studies such as this one, look as ifthe presence of a medical clinic is associated with higher mortality. (Its isindeed that case that the county containing the renowned Mayo Clinic hashigher mortality rates since high-risk patients attend the Mayo Clinic andmany of them die there.) So while expressing disappointment that thisspatial analysis does not show a persuasive influence for medical facilitiesor sanitation, it is clear that they are poorly favored in their empiricalmeasurement.

Altitude shows moderate levels of mortality below 3500 meters, aslightly higher level between 3500 and 3999 meters, and much higherlevels in the 5 communities at 4000 meters and over. It is important tonote that the 5 high altitude communities display mortality ratios thatbecome more deviant after controlling for other factors. They are not highmortality because of low community development or low familyresources, but apparently because of high altitude directly or some otherfactor associated with high altitude.

b) Infant Mortality

Variations in infant mortality, from birth to age 1, are shown inTable 3. They show similar patterns to those seen in the previousdiscussion. Community development shows a less uniform pattern,suggesting that this variable is less important in the first year of life.Indeed, the adjusted deviations fail to demonstrate that levels ofcommunity development have an easily-interpreted influence on morta-lity.

Average female education continues to show a predominant, clear

85

TABLE 3. - Multiple Classification Analysis of Mortality,Both Sexes, from Birth to Age One, for Peruvian Communities, 1968-1977

Independent Variable

Community Development:Low 1Low 2Low 3Low 4Low 5Low 6Low 7Low 8Low 9High 10

Average Female Education"0 - .99 years1-1.99 years2 - 2.99 years3 - 3.99 years4 + years

Medical Facilities:None, 0None, 1None, 2Some, 3

Public Sanitation:None, 0None, 1None, 2All, 3

Altitude, in Meters:Less than 10001000 to 19992000 to 27992800 to 34993500 to 39994000 and more

Grand Mean

Multiple R2

Sample Size

SampleSize

1412106

111010101031

1721281741

28242448

33161263

529

1323225

CategoryMean

.951.021.211.251.191.231.251.161.17.83

1.441.29.15

1.07.75

.931.291.30.93

1.091.29.98

1.02

.901.18.73

1.190.38

.45

1.07

.396124

UnadjustedDeviations

-.12-.05.14.18.12.21.18.09.10

-.24

.37

.22

.08

.00-.32

-.14.22.23

-.14

.02

.22-.09-.05

-.17.11

-.30.12.31.38

AdjustedDeviations

-.16.20.14

-.17-.10-.12-.03-.12.14.06

.37

.26

.04

.03-.33

-.16.18.15

-.07

.05-.10-.25.05

-.11-.05-.13-.01.24.57

"Statistically significant at the .05 level

effect. There is a progressive decline in infant mortality with higher levelsof family resources.

As before, medical facilities and sanitations show some variations,but they are not significant nor consistent. Even where a particular

86

category shows a particularly large adjusted effect, there seems to be noeasy explanation.

Altitude indicates higher mortality in communities at 3500 metersand above. Once again, there are genuine altitude effects since they persistafter controlling for other variables.

c) Childhood Mortality

A third group analyzed is mortality variations for children ages 1 to5. Table 4 displays the MCA results for this age group. As might beexpected, community factors grow in importance as the child leavesinfancy, and we see a much more pronounced effect of communitydevelopment on mortality for ages 1 to 5. All 36 communities in thelowest 3 levels of community development have comparatively highmortality, while there is an irregular, lower pattern for higher levels ofcommunity development.

Average female education continues to show its crucial effect, eventhough community development emerges as more important in this agegroup. There is a. considerable discrepancy between communities withpoorly educated women and more advantaged communities, even aftercontrolling for other variables.

Medical facilities and public sanitation display the same variedinfluence: there is no consistent pattern that yields easy interpretation. Weshould be cautious about the empirical results.

Finally, altitude shows that there is higher mortality in communitiesover 2800 meters, which is particularly pronounced for communities4 000 meters and over. Very high altitude not only affects infancy, butapparently is documented here as increasing the risk of mortality in theearly childhood years.

DISCUSSION

The interpretation of data collected in a multiple purpose surveywithout a specific focus on community mortality must be treated withsome caution. No specific hypotheses are tested in this paper for purposesof drawing rigorous statistical conclusions. Any attempt at interpretationmust therefore be tentative and somewhat speculative. Nevertheless,community mortality data are rare, the ENF-Peru is a valuable source ofmortality information, and it is possible to draw some conclusions.

The consistent, strong association of average female education andchildhood mortality collaborates other research on this topic. Laurell andher co-workers at the Cindad Universitaria in Mexico (Laurell et al.,1977) report that social processes involving economic activities were

87

TABLE 4. - Multiple Classification Analysis of Mortality,Both Sexes, from Age One to Age Five, for Peruvian Communities, 1968-1977

Independent Variable

Community Development:Low 1Low 2Low 3Low 4Low 5Low 6Low 7Low 8Low 9High 10

Average Female Education*0 - .99 years1-1.99 years2 - 2.99 years3 - 3.99 years4 + years

Medical Facilities:None, 0None, 1None, 2Some, 3

Public Sanitation:None, 0None, 1None, 2All, 3

Altitude, in meters:Less than 10001000 to 19992000 to 27992800 to 34993500 to 39994000 and more

Grand Mean

Multiple R2

Sample Size

SampleSize

1412106

111010101031

1721281741

28242448

33161263

529

1323225

CategoryMean

1.09.90

1.681.021.031.291.35.97.73.73

1.481.341.20.86.63

.921.281.32.80

1.001.351.29.90

.731.59.72

1.461.051.62

1.02

.312124

UnadjustedDeviations

.07-.12.66.00.01.27.33

-.05-.29-.29

.46

.32

.18-.16-.41

-.10.26.30

-.22

-.02.33.27

-.12

-.29.57

-.30.44.03.60

AdjustedDeviations

.55.64.84.01

-.19-.34-.06-.79-.63-.11

.35

.46

.14-.05-.46

-.45.05

-.27.37

-.13.27.20

-.04

-.26.52

-.25-.27.15.51

'Statistically significant at the .05 level

among the most important conditions for disease in two Mexican villages.Frisancho et al. (1976) document the connection between family socio-economic status and childhood mortality in an urban Peruvianpopulation. The work reported here indicates that family socioeconomicstatus is evidently reflected in broad community conditions as well: the

88

social processes that are so significant within the family are also closelyconnected at the community level. We may thus see that improving thesocioeconomic conditions of families decreases general childhoodmortality.

We have also seen that there is a general association betweencommunity development and mortality. It is not a clear, progressiveassociation, but does indicate an approximate improvement in mortalitywith greater community development. The association is not perfectbecause, among other things, community development is not necessarilylinear and is not necessarily positive for mortality. Two examplesillustrate this: several studies, such as Dutt and Baker's (1978) work insouthern Peru, note that migrants move often report medical symptoms.Particularly among migrants from high altitudes to rapidly growing low-altitude communities, we might see higher morbidity and mortalityassociated with community development. And second, Laurell et al.(1977) found that a more developed village with unstable cash cropsactually has higher morbidity than a less developed village withsubsistence agriculture. The nature of the social process, then, conditionsany empirical measure of community development. Health conditions areclosely linked to the community development process, but the relationshipis intricate.

Finally, we see in this research that altitude emerges with apersistent relationship with community mortality. This is important sinceprevious research, such as Mazess's (1965) study of neonatal mortality inPeru, report such an association, but have not included multivariateanalysis. It has been suggested by Mazess that mortality at higher altitudesmight stem from poorer health facilities or lower family socioeconomicconditions. We see in the results here that higher mortality persists athigher altitudes even after taking these other variables into account. Thereis thus the strong suggestion that higher mortality may be the direct effectof altitude.

As a final note, this inquiry concludes that traditional demographicsurveys may provide a valuable addition to epidemiological informationon childhood mortality. There have been relatively few uses of fertilitysurveys for this purpose (Edmonston, 1979), although other researchers(Aguirre, 1966 ; Minchaca, 1960) have also stressed the need for broadernational investigations of an epidemiological nature. With the variety offertility surveys now on hand and with the small incremental cost ofobtaining community inventory data, this type of epidemiological studypromises to be a useful addition to the study of childhood mortality.

89

APPENDIX A

GUTTMAN SCALES

In the well-known Guttman model of scalogram analysis, dichotomous items areconceived as points on a scale, where individual responses on the items represent whetherthe individual's point lies on one side or another for the item. If there is an affirmativeresponse to an item, then the Guttman model indicates that the individual's point liesabove the item, and conversely, the individual's point would lie below the item if therewere a negative response to an item.

For example, assume a Guttman scale analysis for individual ownership of threeitems: a watch, a camera, and an automobile. If an individual owns a camera, then theindividual response is above this item, with the implication that the individual would alsoown a watch. If an individual did not own a watch, then the individual response liesbelow this item, with the assumption that the individual would also not own a camera orautomobile. Note in the case of these items that there are four possible individualresponses: ownership of no items (0), or ownership of one, two, or three items (1,2, 3).

Guttman scales are stochastic and cumulative. The stochastic basis of Guttmanscales rests on the notion that there is a probabilistic relationship between the underlyingscale position (the individual's point) and the scale items. Moreover, there is anassumption that the items are cumulative, that affirmation of one item implies affirmationof the items below.

An attraction of Guttman scales in that the researcher is able to test some of thestatistical assumptions of the model and to ascertain if an underlying cumulative scale isappropriate.

APPENDIX B

MULTIPLE CLASSIFICATION ANALYSIS (MCA)

The multivariate model used in this work is a multiple classification model (MCA),Yy = Y + A, + Bj + eu

where Y¡j is the actual mean for the i, j cell, Y is the overall mean for the total sample, A¡ isthe net effect for the /th category, B¡ is the net effect for theyth category, and e¡j is the errorterm which by definition is equal to the actual cell mean minus the predicted cell mean.The MCA model is additive and assumes no interaction. In other words, the predicted cellmean is the sum of the specific categorical effects plus the overall mean. The predicted cellmean, for example, of category 2 on variable A and category 3 on variable B would be:

Y2J = Y + A2 + B3.Notice in this model that the categorical net effects are expressed as deviations from

the grand mean. We obtain such a solution by first requiring that the general constant forthe MCA model equals the grand mean. For ease of expression, we also specify that theweighted sum of the categorical effects for each independent variable is zero. Thus, wheren¡ represents the number of observations for each category of variable A, we require thatZn¡A¡ = 0. This is a useful approach, since it is convenient to interpret a set of categoricalnet effects in which any particular net effect reflects deviations from the grand mean.

90

The assumption of no interaction is an important one. If the assumption issatisfactory, it is clearly a useful one for purposes of statistical interpretation. It hasobvious economy if we can state that the predicted value of the dependent variable issimply the sum of each categorical net effect of the independent variable. How reasonableis it to assume no interaction? First of all, a researcher can test for the amount ofinteraction if worried about its existence. In general, we assume no interaction in thisresearch and made no special effort to incorporate interaction in the statistical models.Moreover, it is unlikely that interaction, if it exists, would add much to the larger, additiveeffects in the MCA model.

REFERENCES

AGUIRRE A., 1966. « Algunas Bases Epidemiológicas de la Mortalidad Infantil », RevistaColombiana de Obstetricia y Ginecología 17 : 349-357.

CHIDAMBARAM V.C., CLELAND J.G., and VUAY VERMA, 1980. "Some Aspects of WFSData Quality: A Preliminary Assessment", Comparative Studies, Number 16.London: World Fertility Survey.

DUTT James S. and BAKER Paul T., 1978. "Environment, Migration and Health inSouthern Peru", Social Science and Medicine. 12: 29<38.

EDMONSTON Barry, 1979. Population Research in Latin America and the Caribbean. AnnArbor, Michigan: UMI Publications.

FRISANCHO A.R., KLAYMAN J.E., and MATOS Jorge, 1976. "Symbiotic Relationship ofHigh Fertility, High Childhood Mortality, and Socio-Economic Status in an UrbanPeruvian Population", Human Biology 48: 101-111.

LAURELL Asa Cristina, et al., 1977. "Disease and Rural Development: A SociologicalAnalysis of Morbidity in Two Mexican Villages", International Journal of HealthServices 7: 401-423.

MAZEES Richard B., 1965. "Neonatal Mortality and Altitude in Peru", American Journalof Physical Anthropology 23: 209-213.

MINCHACA Francisco J., 1960. "La Mortalidad Infantil en Latinoamérica", Estadística 19:1-28.

PERÚ, Instituto Nacional de Estadística, 1977. "La Mortalidad en los Primeros Años deVida, 1967-1968", Boletín de Análisis Demográfico 17.

SPECTON R.M., 1972. "Mortality Characteristics of a High Altitude Peruvian Population",American Journal of Physical Anthropology 37.

WEISS-ALTANER E.R., 1975. "Producción de Salud y Mortalidad", Demografía y Econo-mía 9: 53-64.

WORLD HEALTH ORGANIZATION, 1980. Sixth Report on the World Health Situation, 1973-1977. Part II: Review by Country and Areas. Geneva: World Health Organization.

SOME FACTORS ASSOCIATEDWITH INFANT MORTALITY

IN MEXICO

by

Irma Olaya GARCIA y GARMA

El Colegio de México

1. MORTALITY LEVELS AND TRENDS IN MEXICOOver the last few decades Mexico has experienced a marked decline

in mortality. In 1940 official figures recorded a crude mortality rateof 23.4 deaths for every 1,000 inhabitants, and a life expectancy of41.5 years(", while in 1978 the figures registered at 8.5 and 64 yearsrespectively^).

The reduction in infant mortality has contributed significantlytowards the drop in mortality levels and consequently towards the gainsin life expectancy. In 1940 mortality in the first year of life constituted126 deaths for every 1,000 live births^), while in 1978 the figures came toapproximately 60 per 1,000(4). Even if we allow for the recognizeddeficiencies in the registration of births and deaths of children under theage of one and take these figures cautiously (5>, there is no doubt that theinfant mortality rate has steadily fallen in the last 40 years.

( 1 ) Dirección General de Estadística. SIC : Anuario Estadístico de los EstadosUnidos mexicanos (various years).

(2) Consejo Nacional de Población. México Demográfico. Breviario 1979.(3) Secretaría de Industria y Comercio. Dirección General de Estadística. Anuarios

Estadísticos de los Estados Unidos Mexicanos.(4) Some of the figures for infant mortality in México for recent dates which are

given in the following publications: 70.0 (1972-74). Secretaria de Programación yPresupuesto. Encuesta Mexicana de Fecundidad. Primer Informe Nacional 1979. p. 174 :66.0 (1976-77) Population Reference Bureau. World Population Data Sheet -, 56.5 (1979).Consejo Nacional de Población. México Demográfico. Breviario 1979.

(5) Cordero. Eduardo « La Subestimación de la Mortalidad Infantil en México »Demografía y Economía. Vo. II. Núm. 1. 1968. pp. 44-62.

92

While such gains are in no way insignificant, we cannot, however,be satisfied when we consider the infant mortality rate in Mexico,especially if we take into account that within the last decade variouscountries, developed and undergoing development, have achieved muchgreater triumphs than Mexico in their fight against infant mortality.

This is clearly illustrated by the following data on infant mortalityrates in various countries for the years 1966 and 1976.

As can be seen from Table 1, Sweden, Norway, Finland andDenmark, which already had low mortality rates in 1966, were able toachieve reductions of between 28 and 40% ten years later. Spain, France,Hongkong and Singapore began with slightly higher figures, and in thesame decade they effected reductions of between 43 and 70 %. On theother hand, the figures for Panama, Costa Rica and Cuba in 1966 werenot as high as those attained by Mexico. Nevertheless, by the end of thedecade which separates the two observations they had achieved muchgreater reductions than those registered in Mexico.

TABLE 1. - Infant mortality rates for 1966 and 1976 in some countries of the world,and the percent decrease obtained in the decade.

SwedenNorwayFinlandDenmarkFranceHongkongSingaporeSpainPuerto RicoCubaPanamaMexicoCosta Rica

1966

12.614.615.016.921.7

• 23,925.836.036.737.645.062.965.1

1976

8.310.59.9

10.210.413.711.610.720.9*23.035.654.733.6

Percent decreaseobtained in the decade

34.128.134.039.652.142.755.070.342.138.820.913.048.4

"Corresponding to 1975Source: United Nations. Demographic Yearbook 1970 and 1977.

2. RELATIONS BETWEEN MORTALITYAND SOCIOECONOMIC FACTORS

At the beginning of this century the inverse relation betweensocioeconomic levels of development and general mortality in differentcountries of the world was evident. In the course of time this associationhas weakened considerably, since the rapid decline in mortality in

93

developing countries is largely due to technological advances in theprevention and control of disease, and to the spread of medical and publichealth services which, to a great extent, are not related to the economicconditions of the region^6'. The preceeding statement is valid at thenational level, but there is evidence to show that in all countriesdifferentials in general mortality continue to exist, and that they varyaccording to the socioeconomic levels*?). In other words, the low incomesectors of the population are at a disadvantage as regards access to medicaland health services, and the quality of these services.

Just as the relationship between general mortality and economicdevelopment has gradually become much less significant, we can expectthe same to eventually occur with regards to infant mortality. However, atthe present time there is a consensus on the connection between infantmortality and socioeconomic factors.

Besides the economic and social factors, others such as sex, order ofbirth, age of the mother, intergenetic interval and so on, have also beenassociated with infant mortality. Anxious to explore, be it only partially,how much weight this second group of factors carries with relation to thesocioeconomic elements, this study attempts to evaluate the impact whichcertain characteristics associated with the child and his environment haveon the baby's chances of surviving his first year of life.

Given that, in Mexico, one baby in every 15 dies before completinghis first year of life, that the decline in infant mortality levels has beenslight despite the economic development which the country hasexperienced, and that one of the main concerns of the government is thatof preserving the life of its people, the identification of those factors whichcause infant mortality is fundamental for the defining of policies andprograms aimed at attacking, and consequently reducing, infant mortality.

3. SOURCES OF DATA

This study uses the information obtained by the Mexican FertilitySurvey which constitutes part of the World Fertility Survey. The projectwas financed by the Mexican Government and the United Nations Fundfor Population Activities (UNFPA). The field work was carried out fromJuly 18, 1976 to March 5, 1977 by the Department of Statistics (DirecciónGeneral de Estadística), with the cooperation of the Institute of Social

(6) Naciones Unidas. Boletín de Población Número 6, Naciones Unidas (1983),pp. 10-1 i.

(7) In this connection see Aaron Antonovosky «Social Class Life Expectancy andOverall Mortality». The Milbank Memorial Fund Quarterly, Vol. 45. Nos. 1-2. 1967.

94

Research, the National Autonomous University of Mexico (Instituto deInvestigaciones Sociales de la Universidad Autónoma de México,(ISUNAM), and assessment was given by the Center for Economic andDemographic Studies (Centro de Estudios Económicos y Demográficos(CEED) of the Colegio de México, the National Population Council(Consejo Nacional de Población (CONAPO)) and the Information Systemfor Economic and Social Planning (Sistema de Información para laPlanificación Económica y Social (SIPES)).

This study was the first of its kind to be made on a national level.Among its objectives was that of obtaining the information necessary foranalyzing fertility levels and trends in the country. It also aimed atfurthering the scientific study of fertility, with particular attention to itsexplanatory aspects(8). 13,200 household interviews were carried out, andfrom the information a subsample was obtained in which all women from20 to 49 years of age and women between the ages of 15 and 19 whowere married or who had had a liveborn child, were eligible forindividual interviews. Under these conditions information was obtainedon 7,310 women. Despite the fact that this study did not set outspecifically to provide an analysis on mortality, the information it containson pregnancy histories as well as on the socioeconomic characteristics ofthe women interviewed helps us to approach our objectives.

The information obtained from the Mexican Fertility Survey isconsidered to be free of inconsistencies and errors in reported answers; wecan therefore rely on the fact that the information handled in this studyreflects the reality of the country(9).

4. METHODOLOGY

Infant mortality can be defined as the number of deaths of childrenunder the age of one per 1,000 live births. For purposes of analysis, thismortality is frequently divided into neonatal mortality, which occurs inthe first month after birth, and postneonatal mortality, which occursduring the rest of the first year of life. Generally speaking, endogenousfactors are related to neonatal mortality, while exogenous factors areconnected to postneonatal mortality.

(8) Coordinación General del Sistema Nacional de Información EncuestaMexicana de Fecundidad. Informe Metodológico. Secretaria de Programación yPresupuesto. México. Octubre 1978, p. 2.

(9) For the evaluation of the survey see Ordorica and Potter, "An Evaluation of theDemographic Data Collected in the Mexican Fertility Survey" Draft for Circulation at theIUSSP Seminar on the Analysis of Maternity Histories, London, April 1980. ThePopulation Council Working Paper. N° 4.

95

Neonatal mortality is largely associated with genetic factors such asinherent weakness in the mother or in the foetus, or injuries incurredduring gestation or birth; in other words, it is more or less related tocauses of death which, at the present level of medical knowledge, aredifficult to control. On the other hand, it is estimated that postneonatalmortality is basically connected with non-biological factors, includingdisease, and therefore it includes such causes of death which could havebeen prevented by modern medicine and public health campaigns.Although it is probable that endogenous and exogenous causes acttogether in not a few deaths, the great changes which have ocurred ininfant mortality have been mainly due to decreases in postneonatalmortality.

In the analysis of the levels, trends and differentials of infantmortality, the measurement most commonly used is the infant mortalityrate. This rate can be directly calculated by dividing the number of deathsoccurring in children under the age of one year by the total number of livebirths in the same period. It can also be estimated by using life-tablefunctions, or by applying one of tfie indirect methods which derive therate from the proportion of deceased children according to the age groupsof the women (10>. In these cases the result is a measurement of thephenomenon at the aggregated level.

On the other hand, when data is analyzed on the individual level,and such is the case for our purposes, infant mortality is frequentlymeasured by obtaining the proportion of infant deaths from the totalnumber of live births corresponding to each woman. This measure is notinadequate, but neither is it really satisfactory. This becomes evident, forexample, if we consider that a woman who has only one child which dieswould have the same infant mortality measure as another woman whosefive liveborn children all later died. Even when the proportions are equal,one would be inclined to speculate that infant mortality is higher in thesecond case. On the other hand, a quotient which covers all infant deathsindiscriminately, as is the case in the proportion mentioned above, doesnot permit us to use certain individual characteristics as explanotoryvariables which could be important in the study. Among these

(10) For example: Brass, William, Seminario sobre métodos para medir variablesdemográficas (fecundidad y mortalidad) CELADE, San José, Costa Rica, 1973. Sullivan,Jeramieah, "Models for the Estimation of Dying Between Birth and Exact Ages of EarlyChildhood", in Population Studies, March 1972, pp. 79-97. Trussell, James, "AReestimation of the multiplying Factors for the Brass Technique for DeterminingChildhood Survivorship Rates", in Population Studies, March, 1975, pp. 97-107. FeeneyGriffith, « Estimación de tasas de mortalidad infantil a partir de información desobrevivencia de hijos clasificados por edad de la madre », CELADE, Santiago de Chile,1977.Mina, V. Alejandro, « Estimaciones de los niveles, tendencias y diferenciales de lamortalidad infantil y en los primeros años de vida en México, 1940-1977, Demografía yEconomía Vol. XV, Num. I (45) El Colegio de México.

96

characteristics we can mention birth order, the age of the mother atchildbirth and birth interval.

In our endeavours to carry out a more efficient microanalysis ofinfant mortality, using a "representation" of the phenomenon whichcould be more sensitive to possible determinants and thus enable us toidentify more clearly the factors associated with infant mortality, wedecided to examine each live birth individually. It is important toemphasize that the analytical unit used in this approach is neither aproportion nor a rate, but each live birth reported in the survey. In viewof the fact that we have no antecedents for the method of analysisproposed, we have attempted to apply it in this study, and invite thoseinterested in the subject to judge its validity.

The dependent variable. - In order to obtain the dependentvariable in the analysis, all live births were separated into two groups: onegroup consisted of children who died before reaching the age of one, thatis, infant deaths; the other group consisted of children who survived atleast up to their first birthday. The first group was then separated into2 sub-groups: those who died during the first four weeks of life (neonatalmortality) and those who died between the first and eleventh month of life(postneonatal mortality). For lack of a more appropriate name we havecalled this index the "survival condition".

In accordance with the above mentioned separation, each livebornchild recorded in the survey was assigned with the number 0, 1 or 2,according to his survival condition, in the following manner:

(0) If he died during the first month of live (neonatal mortality)(1) If he died between the first and eleventh month of life

(postneonatal mortality)(0) + (1) Total number of infant deaths

(2) If alive at the time of the survey and over the age of 12 months*,or if he had lived at least up to the age of one year (regardless) ofwhether he later died).

This survival condition is, precisely, our dependent variable.

Independent variables. - It is not easy to enumerate all the possiblefactors which could be mentioned as possible determinants of infantmortality: weight at birth, sex, race, nutritional deficiencies, access tomedical services, residential area, age of mother, order of birth - tomention just a few of them. Work is complicated by the fact that not allthe necessary information is available, and because often the variablesused are correlated to each other to a high degree.

*A11 reported liveborn children whose date of birth occurred within 12 months ofthe survey, were excluded from the study.

97

Since the survey used in this study was not directed towardsresearch on infant mortality determinants, we have found it necessary tolimit ourselves to exploring the connections between the survivalcondition and the following variables: the mother's age at childbirth, theintergenetic interval(I '), lactation, previous reproductive losses, themother's level of education, and the mother's residence (rural-urban).

The procedure followed in the study can be summarized in thefollowing stages:

1. From all women pregnancy history, the five years (1971-1975)birth cohort reported in the survey is selected for the study.Cross-tables are used to make a preliminary analysis of theassociations between the survival condition and the independentvariables. This is done by calculating the percentages of 0, 1 and2 from the total number of live births found in each column, thatis, in each of the different codes, values or characteristics whichthe independent variables take.

2. The independent variables which in the cross-tables showed themost marked changes in the survival condition are selected. Onthe basis of the changes observed from one column to another thevariables are transformed into dummy variables.

3. With all the possible combinations of the independent variablevalues, multidimensional cross classifications are used to obtainthe total number of live births according to their survivalcondition. The number of observations found in each case isrecorded in the respective cell.

4. With the results obtained in stage number 3, the proportions ofsurvival up to one month and up to one year are calculated.

5. A multiple regression analysis is made using the survivalproportions as a dependent variable and the values of theindependent (dummy) variables. The number of cases in theregression equation is given by the total number of possiblecombinations of the independent variable values.

Later the study concentrates on the survival condition of the lastliveborn child of each woman interviewed. In the regression this allowsus to use instead of dichotomies, the original values of the independentvariables.

(11) The intergenetic interval is the period which elapses between two successivebirths. The intervals between marriage and the first live birth (protogenetic interval) werenot included in the analysis.

98

5. RESULTS

Table No. 2 shows the relative distribution of the survival conditionfor both the total number of live births between 1940 and 1975 and thefive years birth cohort - July 1970-June 1975. It can be seen that infantdeaths represent 8.6% of the total number of live births (1940-75), that is86 per 1,000, while for the 1970-75 cohort they represent 6.9 % or 69 per1,000.

These infant mortality indicators derived from the survival conditionshow higher values than the rates corresponding to Mexico for 1966 and1976 which appear in Table 1. However, besides that the observationperiods are not the same, it is a well known fact that this indicator isunderestimated in Mexican statistics'12). On the other hand, the result forthe 36 years period (1940-75) is quite similar to the figure of 85.7 deathsfor every 1,000 births, which was obtained by applying Jorge L. Somoza'smethod of mortality estimation to the same information " 3).

In table 2 we can also observe that neonatal mortality indicators arehigher than postneonatal mortality indicators. This is related to the factthat 51 96 of the total number of infant deaths occur in the first month oflife, and the remaining 49 % are distributed throughout the following 11months (see Chart 1 for the relative distribution of the total number ofinfant deaths according to the child's age (months survived) at death).

Results of the cross-classification of the survival conditionwith each one of the independent variables.

Sex of the child

In most countries it can be seen that approximately 105 males areborn for every 100 females, and that male mortality is for all ages, greaterthan female mortality. In Mexico this is generally the case for both theseobservations. In the survey a greater number of liveborn children of themale sex was reported than of the female sex, while, as regards thesurvival condition, the advantage that females held over males withrespect to survival up to the first month and first year of life was clearly

(12) Eduardo Cordero, Evaluación de la mortalidad infantil en la RepúblicaMexicana 1930-1970, prologue to «Evaluación y Análisis», Serie III, N° 1 1975. AlsoOrdorica y Potter, op. cit. p. 28.

(13) Mina, Alejandro, op. cit. p. 9. The applied method can be found in Jorge L.Somoza, « Estimaciones de la mortalidad al comienzo de la vida en Colombia basadas eninformación de la encuesta nacional de fecundidad 1976 ». Celade, Santiago de Chile,1979.

99

oON

IÎ

«

7 1U O-J o

H

¡3

1'1o«j

•a

'E3

en

S Z

+o

I §11= ZfcE

I

100

Chart 1. Relative Distribution of the Total Number of Infant Deaths Accordingto the Child's age (months survived at death). Birth Cohort 1970-75.

55

50

Monthssurvivedat death

0]23456789101 1

Numberof cases Percent

274363032342233222114410

51.56.85.66.06.44.16.24.14.02.60.81.9

Total 532 100.0

| | Neonatal

Postneonatal

0 1 2 3 4 5 6 7

Source : 1976 Mexican Fert i l i ty Survey

10 IIMonths

seen (Table 3). If we consider the total number of births, the differencebetween the sexes is more marked in neonatal mortality than inpostneonatal mortality.

Birth order

Several studies, including one carried out by Puffer and Serrano,found that birth order is an important determinant in infant mortality"4'.Generally, children of high orders of births have less chances of survivalthan those of low birth order. Table 4 records the survival conditionaccording to birth order; as in the previous case, the results correspond tothe 1970-75 birth cohort.

(14) Puffer. R. Y Serrano. C. El peso al nacer, la edad materna y el orden denacimiento: Importantes determinantes de la mortalidad infantil. UPS. Publicacióncientífica N° 294, 1975.

101

X

•aÜCO

S 8

i SO C/3

UJ

03<

1 2.l-s st/ï O ^

<uz

È:3

• — • • « > ro • O ^

cfiZ CAZ

32

ïths

nt d

e¡

rtal

i

oiu

nata

l

o

moi

a

neon

í

s iO ^

a "3 o

'ivor

sye

ar

c «£ z o

CL

o

Ïc tí•a-5

o °

102

uOQ<

Os -o

1g•a

5^

o

S Z

iths

1

fan

1o

tal

I

OU

I

1o

IOH

153

lei

o

1v

103

Even taking these results with caution, for occasionally the numberof cases in each cell is very small, we can deduce from the Mexican data(on the total number of infant deaths, as well as the total figures forneonatal and post-neonatal mortality), that the relation between birthorder and infant mortality, if it does exist, is only to be found in birthsbeyond the ninth order. However, if we study the variations in thesurvival condition, taking into account not only birth order but also themother's age at the time of the baby's birth (see Table A1 in the appendix),two important facts can be seen:

1. First and second children born to women over the age of 20 havemore chances of survival than children born to women under theage of 20.

2. In high orders of birth, infant mortality is greater in the cases ofyoung women than in women more advanced in years. This ofcourse, is related to the frequency with which the mother has herchildren and, consequently, with the intervals between births.

The interval between previous date of birth and observed date of birth

A short interval between two pregnancies, and hence a rapidsuccession of births, will be associated with a shorter period of breastfeeding (when this is used) and the probability of the mother's having todjvide her attention between two or more babies. In the first case, whenthe child is breastfed and then taken off the mother's milk because of thearrival of another baby, the changeover to artificial milk, which is perhapslacking in protein or contaminated by bacteria, makes the child vulnerableto infection. In the second case, the small children are exposed to differentkinds of accidents which adequate attention would otherwise avoid.

The survival condition for intergenetic intervals last-birth-to this isshown on Table 5(15). What is strikingly evident is that the greatest risk ofinfant death occurs when the interval between two successive births lastsfor less than a year. Next in importance, although the risk of death isconsiderably lower, is the interval which lasts from 12 to 18 months.From then onwards, and up to a period of 48 months, as the intervalincreases so the risk normally decreases.

It would seem that the inverse relation between the intergeneticinterval and infant mortality becomes modified when the time spanbetween two live births is greater than 4 years. Although it is possible to

(15) Since births following a short interval disproportionately consist of prematurebirths all cases with intergenetic interval shorter than eight months were excluded. Itshould be mentioned also that, as first order births are excluded when this variable is used,the number of live births which appears in the tabulations is lower than that reported inthe tables in which the other independent variables appear.

104

f!n¡1.•" oo

•5 53

ass

fi

ctor

.2

o

l

v iLU_Jça

co

•s

'•3cou

Ta

rviv

.

03

f

•a

8

3en

+ • wi • oo • m

rs i Tf • * 6 J

Tf V^ „ ^

O v oo=o "*

— • * TT OO ( í , VO _ oo rs o I

•a Tj ""• _«s Op?ri» ¿om« o1"

+o

1 I

£ Z

o

1S S

3 8

III

105

associate long birth intervals with higher infant mortality, the decrease inthe number of events in each cell obliges us to take these figures withcaution. Moreover, it clearly illustrates the rareness of these cases and, forthe same reason, their lack of importance in the Mexican context.

The existence of previous reproductive losses

MacMahon, in his study on infant mortality in the United States,noted that mothers who had undergone a previous pregnancy whichterminated in foetal death would tend to suffer a further loss(16). Shapiro,for his part, found that in the cases of mothers whose last pregnancy hadterminated in foetal death or infant death there was a certain tendency forthe anomaly to repeat itself(17). In our study we have considered, in eachcase, previous foetal deaths (abortions and stillbirths) as well as previousinfant mortality, in order to observe their influence on infant mortality.Similarly to the findings made in the United States, Mexican data showthat the existence of previous losses is related to infant mortality.

In Table 6, it can be seen that lower infant mortality indicators wererecorded in cases where mothers had not previously suffered reproductivelosses. On the other hand, if the mother has a history of infant deaths orpregnancies which terminated in abortion or stillbirth, the probabilities ofinfant death occurring are greater.

The figures in the table show that, as the number of previousreproductive losses increases, there is a corresponding increase in infantmortality. The figures for neonatal deaths in children whose mothers havehad 3 or more losses are more than double than in cases where mothershave suffered no loss, and if the analysis is continued to include 4 or morelosses the figures are almost quadrupled.

Lactation

Maternal milk provides all that is necessary for adequatenourishment in the earlier months of life, and, naturally, it has morenutritious value than milk formulas or substitute foods. Lactation alsohelps to space births, for in a great many cases, it prolongs postnatalamenorrhoea which protects mothers against further pregnancy'18'.

(16) MacMahon, Brian, Mary Grace Kovar, and Jacob J. Feldman, InfantMortality Rates: Socioeconomic Factors (Washington, D.C.): Department of Health,Education and Welfare Publication N° (HSM) 72-1045 1972.

(17) Shapiro, S., L.J. Ross and H.S. Levine, "Relationship of Selected PrenatalFactors to Pregnancy Outcome and Congenital Anomalies", American Journal of PublicHealth, Vol. 55 (1965), pp. 268-282.

(18) H. Leridon, Human Ferlililv, The Basic Components, The University ofChicago Press, 1977, p. 83.

106

-S g

•a 3

ï lI 3c .2

i g§

ilBQ 3< tO

i I

co

•a

3C/3

SRZ

+o

S 2

I

i1oO)

o

î5S =.

lili

V

• - 3 ?»

107

| |O k.UOQ

1•a

O o o

+o

•S 3

1 I

108

Moreover it eliminates the risk of infections caused by lack of hygiene inthe preparation of bottles or other ustensils used in artificial feeding.

In developing countries malnutrition has been found to be one of themain causes of child mortality in the first years of life(19). Furthermore,infectious and parasitical diseases always occupy the first two places in thecause of infant deaths in Mexico(20). Taking into account the associationbetween the absence of breast-feeding and both malnutrition andinfectious diseases, it can be assumed that breast feeding and its duration(up to at least 6 months) are related to lower infant mortality.

In the Mexican Fertility Survey the women interviewed were askedif they had breast-fed their last and penultimate child, and for how long.According to the results recorded in the First National Report, 85% ofever married woman who had had at least 2 liveborn children answeredaffirmatively. The average duration of breast-feeding reported was 9.7months, and the modal was 12 months.

We can infer from the above that it is the rule rather than theexception for Mexican children to receive maternal milk in the firstmonths of life. It should be noted that the tabulations in the report alsoshow an inverse relation between the lactation period and the mother'slevel of education, and also between the lactation period and the size ofthe residential area(21 \ Since the development process is related toincreases in educational and urban levels, if the. inverse relation persistswe can anticipate a decrease in the average lactation period as the countrybecomes more and more modern.

As stated previously, information on breast-feeding was notobtained for each liveborn child, and so we had to assume that the repliesgiven concerning the last and penultimate child were valid for themother's previously born children; from that information each livebornchild was classified as breast-fed or non breast-fed. The exception were allchildren who died during the first month of life (index = 0). For thesecases, if the related information was not available, were classified underabsence of breast-feeding since a child can only be breastfed if it survives.We are aware that in some cases this assumption may be unfounded, butwe believe it to be correct in the majority of cases. Moreover, since we aredealing only with births occurred in a five years period the number ofliveborn children reviewed in this study received maternal milk. Asregards the influence that breast feeding has on infant survival, it was

(19) Puffer, Ruth Rice and Carlos V. Serrano, Patterns of Mortality in Childhood:Report of the inter-American Investigation of Mortality in Childhood, Washington, D.C.:Pan American Health Organization (1973), p. 345-355. This publication gives the resultsof a 5 years study concluded in 1972 on infant mortality in underdeveloped countries.

(20) Dirección General de Estadística, S.I.C., Anuarios Estadísticos (various years).(21) Secretaría de Programación y Presupuesto, Encuesta Mexicana de Fecundidad,

1er. Informe Nacional. Vol. Il, pp. 447-455.

109

found that the percentages for neonatal and postneonatal mortality, andconsequently for infant mortality, are higher for children who have notbeen breast-fed that for those who have. His information is consistentwith Alan Berg's findings in which he reports that the absence of maternalmilk is one the major causes of malnutrition and mortality in children(22).

Rural-urban residence

It is well known that the rural-urban differentials found in infantmortality levels are, in some cases, the result of socioeconomicdifferences, and that as countries advance in their development, thesedifferentials tend to diminish. In the comparison of rural and urban infantmortality levels in industrialized countries, recent data shows that adefinite pattern does not exist: some countries register rural overmortality,in others urban overmortality prevails, while others have similar figuresfor both areas. What is evident, however, is that when the values are notuniform the existing differences are not great(23).

On the other hand, in less developed countries, infant mortalitylevels are lower in the urban zones than in the rural. The advantageousposition held by the urban areas is associated, among other factors, with agreater variety of better medical, hospital and public health facilities aswell as higher levels of education, greater earnings and better, sanitaryconditions.

In our analysis of the survey, we have selected a population limit of20 000 in order to differentiate the urban from the rural zones. Accordingto this criterion, 48 % of the women interviewed lived in urban zones and52% were from places with a population of less than 20 000. Thepercentages for the total number of live births considered in the study are4496 and 5696 for urban and rural zones respectively. These figures arerelated to the fact that higher levels of fertility are recorded in the ruralzones of Mexico, in comparison with the urban zones.

Table 8 shows the survival condition for both rural and urban zones,and Chart No. 2 gives the percentages of neonatal and postneonatal infantdeaths corresponding to rural and urban zones.

Even if we consider the argument that, as a result of migration, thereis the possibility that some of the births and deaths of the interviewedwomen's children could have occurred in districts of a different size tothose in which they were actually reported by the mothers, and even if we

(22) Berg, Alan, The Nutrition Factor: Its Role in National Development. TheBrookings Institution. Washington. D.C.

(23) United Nations. Factores Determinantes y Consecuencias de las TendenciasDemográficas, Vol. I. United Nations ST/SUA/SER.A/50. pp. 140-142. Also Demogra-phic Yearbook various years.

(24) Puffer R. et. al.. Op. Cit.

110

i 81

!1S V

I 5S <2e :!O TI

i l3

en

303

2 feo ]3

il *i l 3OS•ses

1

c,o•3co

>

M I—

!o <Oj

r-2 -*r-

sez

1I3«

a z

i

ï1

el ¡gl o § |

Il l

Chart 2. Infant Mortality in Mexico. Percentages of Neonatal and PostneonatalInfant Deaths Corresponding to Rural and Urban Zones. Birth Cohort 1970-75.

°/oo 80BU

/U

60

50

30

20

10

n

—

—

.

—

—

••;• :• ' : • !

[ | Neonatal

Ey.y.VJ Postneonatal

-

-

-

infn~IIILU

181*3

Rural Urban

take into account the distorting effect that this would have on the results,the differentials shown in both the table and chart are so marked andconsistent that they confirm that infant mortality in Mexico is higher inthe rural zones than in the urban. This differential is related to the fact thatwithin the country there are thousands of small communities, many ofwhich have low levels of education, unsatisfactory sanitary conditionsand deficient facilities for preventing and treating illnesses. It is moreimportant to draw attention to the fact that in the course of time urbanzones have recorded more gains in infant survivorship than rural zones,and, consequently, the rural-urban differential has become moreaccentuated. Thus in this survey for the 1952-56 cohort, infant mortalitywas lower in urban zones than in rural zones by 18.9 96 ; ten years later(1962-66 cohort) there was a differential of 27.8 %, and in the most recentperiod studied (1972-76 cohort) a differential of 28.2%.

The age of the mother at childbirth

The mother's age is one of the factors most frequently associatedwith infant mortality; it has been observed that both neonatal andpostneonatal mortality generally increase as the mother gets older(24).

We can see from Table 9 that the highest risks of infant mortalityoccur in children born to mothers under the age of 20. The proportions ofinfant deaths occurring in children born to mothers older than 20 years ofage show a declining tendency; the lowest values, together with a certainstability, are to be found when the mother is between the ages of 25 and

112

3 *

o-g

a 8 »

cou

73

3CO

-tj- O i

CT\ oo

OS'O ^

D.

O

S Z

+o

JS

3o

i

tCO

I!

£ Z

73

O

1ac

I

a3tO>

3t/3

. 2 U.

§ ¿ o |3 S S

Ir; 3£ Z» Z

113

34 years, while a slight rising tendency is recorded in women of the age of34 years onwards. For a clear picture of this see the results shown onChart 3.

Chart 3. Survival Condition for the Age of the Mother at Childbirth.

Birth Cohort 1970-1975

10

—__— . -=- - - -

—*•-..—'

into180S3

Less than 20 20-24 25-29 30-34 35-39 40-49 Years

Mother's age at childbirth

Neonatal Postneonatal

Source : Table 9

The mother's education

Many studies carried out in countries of varying levels ofdevelopment have shown that there is a very important connectionbetween the mother's level of education and infant mortality. Forexample, a study carried out in the United States reported an infantmortality rate of 35 per 1 000 in cases where mothers had receivedelementary education, compared with a rate of 20 per I 000 in cases ofmothers who had attended college for 4 or more years(25). Caldwell in hisanalysis on mortality in Nigeria mentions various studies which establishan inverse schooling-infant mortality relation, and he concludes that inNigeria the proportion of infant deaths decreases as education increases.His data also show that the rural-urban differential decreases when themother's education has been controlled(26).

(25) Chease, Helen C. and Freída G. Nelson, "Education of Mother, Medical Careand Condition of Infant", in Helen C. Chease. ed. "A Study of Risks. Medical Care andInfant Mortality, American Journal of Public Health", Vol. 63, (September 1973).Supplement, 0.34.

(26) Caldwell J.C. Education as a factor in mortality decline. An examination ofNigerian data. World Health Organization Meeting on Socioeconomic Determinants andConsequences of Mortality. Mexico City. 19-25 June, 1979.

114

53oQuw

co

oo•a

3CO

O loO ^

o. °°

Î 8 Z

3 I

•a

I3a

s 1

o

3aco

CU

a 3

A" Z

115

11umx:5

ëo

1 I•ÏII

I'•o

c

1•g3

en

—

rs —

-il

o—

°oCJ f""> O OOON <~~ o OO

o

3nc

1 I

s ¡a "si§•?!•!s? i, S So § g S

S a i s

co"3

.22

§

se z I

116

In the Mexican Fertility Survey, 18 % of the women interviewedstated that they could not read and write. As was to be expected, thepercentage in illiteracy rises as age increases; thus, 9.8 of women betweenthe ages of 20 and 24 are illiterate, while for women in the 45 - 49 agegroup the figure reaches 32.6 %. It was also found that illiteracy decreasesas the number of inhabitants in the locality increases, (in populations ofless than 2,500 there is 31 % illiteracy while in populations of 500,000and over the figure is only 8 % ).

With respect to the level of education, 20.2 % were reported never tohave attended school; 43.7% had not completed primary school; 15.5%had completed primary school and 20.6% had studied beyond primarylevel(27). Although the level of schooling is low, the difference recordedfor women's age groups reflect the social changes which Mexico hasexperienced in the last few decades.

Table 10 gives the survival condition for the different levels ofmother's education. We can see from the table that decreases in infantmortality are generally constant as we go from one column to the next.The previous figures reveal just how significant the connection betweeneducation and infant mortality is. In this respect we can surmise thatwomen who have received some education will not only be aware of theneed for sanitary precautions and have some knowledge of nutrition, butthat they will seek medical attention more actively than women withoutany schooling.

With regard to the rural-urban differential and its connection withthe mother's education we can see fromfirst column-table 11 that thefigures for infant mortality are inferior in the urban localities to figurescorresponding to the rural localities. Thus the rural minus urbandifferences are: 2.2 for infant mortality, and 1.1 for its neonatal andpostneonatal components. However, when one controls the mother'seducation these differences decrease, in lower levels of schooling andincrease in the highest. It was noted, for example, that for mothers withno schooling at all, the infant mortality levels are quite similar in rural andurban zones.

Although there are great differences in the medical facilitiesavailable in urban and rural zones, the previous findings clearly indicatethat it is education, rather than rural-urban residence, which has greatersignificance with respect to neonatal and postneonatal mortality.

From our analysis of the previous tabulations we can sum up withthe following main points: it was found that infant mortality risks are

(27) The percentages relating to women's schooling were calcutated according tothe results of the Mexican Fertility Survey which were published in Volume II of the FirstNational Report (Primer Informe Nacional). Coordinación General del Sistema Nacionalde Información, Secretaria de Programación y Presupuesto. Mexico, Marzo 1979, pp. 53-64.

117

e8' oP S

¡«IC vj

iI 8c e

1 §

tí E ^

J ï2

PQO

ëeS

99 °°;VO »S

ol

^ 5 »s^ O VO * ^

r- 3¡ m Su

oI

_ es 2 — ro „ es —

a? a? è(?z a?z a?

3 2_ c

(rba

n

_>

nuiui

rai

3

iff

¿>

orta

l i

E

1 1n t-.C 3

C

E

1= 3

Q Z Q £

cE

1!=Û

118

higher when: the child is male, lives in rural districts, is not breast-fed,and when the interval between successive births is less than a year. Therisks of infant death are also higher when the mother is very young, hassuffered previous reproductive losses or has received no education. Thislast factor is significantly related to infant survivorship: on the one hand,the risk of infant death lessens as the number of years the mother hasspent at school increases; on the other hand, the rural urban differentialdecreases when the mother's level of schooling is controlled.

Statistical analysis

Up to this point, we have analyzed the relations between variables,but we have not evaluated the influence which each variable has on infantmortality. It is therefore necessary to make an analysis which clearlyestablishes to what extent each one of the independent variables affects thedependent variable. In order to do this without losing consistency with thelevel of analysis maintained up to the present, we applied a multipleregression model with dummy variables to the data.

We are aware that, by limiting the values of our independentvariables to only two possibilities, we will lose valuable information;consequently, when we study the final results we will have to bear inmind the different categories which are included within the groupings.

In accordance with the principal changes recorded in the crossedtabulations, the independent variables in the study (nominal andcontinuous) have been converted into dummy variables. This transforma-tion was done by giving them values of 0 or 1, depending on thefollowing characteristics:

/. The mother's age at the child's birth (Variable X2, X3)X2 X3

0 0 If the mother was less than 25 years old at the birthof the liveborn child

Values1 0 If she was between 25 and 34 years of age0 1 If she was 35 years old or more

//. The interval between the birth under observation and previous birth(Variable X4)

0 Interval of less than 12 monthsX4 =

1 Interval of more than 12 months

119

///. Breast-feeding Practice (Variable X5)0 Not breast-fed

X4 =1 Breast-fed

IV. The existence of previous reproductive losses (Variable X6)0 No previous reproductive losses

X6 =1 Any previous reproductive losses

V. The mother's education (Variable X7)0 Illiterate and up to three years of schooling

X7 =1 Four or more years of schooling

Once the independent variables had been transformed into dummyvariables, all the possible combinations of their values were determined.Given that for the age of the mother at childbirth there were 3 possibilities(0 0, 1 0 and 0 1) and that for schooling, previous losses, lactation and theintergenetic interval there were 2 possibilités (0 and 1), the total number ofcombinations came to 48. The matrix of the dummy variables wascomposed of 48 lines and 6 columns (see Table in the Appendix).Calculation of the Infant, Neonatal and Postneonatql Survival Rate.

For each one of the forty eight combinations, we calculated the totalnumber of live births, from which we then found the number of babiessurviving at the age of 12 months, the number of deaths occurringbetween 1 and 11 months, and the number of deaths for children under1 month.

This was done by means of a simultaneous cross-classification of thefive independent variables. From these figures we calculated for each linethe infant survival rate (IS), the neonatal survival rate (NS) and thepostneonatal survival rate (PS) in the following way:

number of survivors up to one year . C2IS =

NS =

PS =

total number of live births CO + CI + C2

number of survivors up to one month Cl + C2

total number of live births CO + Cl + C2

number of survivors between 1 and 11 months C2 - CO

total number of live births CO + Cl + C2

The IS, NS and PS vectors with the infant, neonatal andpostneonatal survival rates for each line are shown in Table A. 12 in theAppendix.

120

The following model was applied to each vector of the infant,neonatal and postneonatal rates.

Y = B, + B2X2 + B3X3 + BAXli + B5X5 + B6X6 + B7X7 + ewhere:Y = Survival rate (infant, neonatal or postneonatal) in one line.B, = Level of Y according to the situation: mothers less than 24 years,

intergenetic interval of less than 12 months, without lactation,without previous reproductive losses and without schooling.

B2 = The effect on Y of the mother's age when this is between 25 and34 years.

B3 = The effect on Y of the mother's age, when this is above 35 years.B4 = The effect on Y of the births interval of more than 12 months.B5 = The effect on Y of breast-feeding.B6 = The effect on Y of the existence of previous reproductive losses.B7 = The effect on Y of mother's education.

Regression results

From table 12, which gives the regression results, it can be seen thatthere are 3 equations for each survival condition: the first contains the5 independent variables; the second does not include breast-feeding andthe third does not include intergenetic interval. As the intergenetic intervaldoes not exist for first births, the regression results when this X s variableis incorporated are only applied to births which are second or higher inbirth order.

While bearing in mind the possible deviations derived from eitherthe data or the conversion of these into dummy variables, the group ofvariables explains in each case a good percentage of the variance in theinfant survival condition. The constants represent the level of the survivalrates for the situation: mothers less than 20 years of age, intergeneticinterval of less than 12 months, non use of breast-feeding, absence ofreproductive losses and mothers without schooling.

It can be seen in the regression equation, the highest coefficients, andconsequently, the highest contribution in R2, correspond to the variablesX5 (breast-feeding), X4 (intergenetic interval), X6 (previous reproductivelosses), and X7 (the mother's level of education). This means that greatersurvivorship, and therefore lower infant mortality, is associated with thefollowing condition: intergenetic interval of over 12 months, breast-feeding, absence of reproductive losses and mother's educational level.

When neonatal and postneonatal survival are analyzed separately,although the results appear to be quite consistent with regard to the

121

— 1 ~

•° g; ~2 ~ .2

u

eti

c

ac

£c

oen

ce

o

3•OU

i 00

s-I

o

X

qo

qo

Xo

qO

X

qo

X• *

qo

Xqo

oö

Xqo

XTÍ-

qd

X 5 X^ Osqo q

o

qo

X•oOsO

Ö

Xo

o

X

qo

X Xoo

XOs

Xoooqo

qo

qö

XOsqö

X

o+

Xoo

O

+

I

.c'S•o

o

X 52 5tqoIX*

o+ .

?5

X £q -S«S

" o s r i (*i ^fro«Ct- O T T °° r~ <rt

ni i— r oo y h> l—

c"ö öd d .i d o| =3 II II

a« „« „s« „ V] K VI U. OÍ V3

I

Sí

o y so

S 5 5 == í ':í •-S É 5

Ï. ü J i

122

importance of the variables, it can be seen that the coefficients inpostneonatal survival are, in general, higher than those for neonatalsurvival, and that they also explain a large percentage of the variance.

Finally it can also be seen from the regression that the coefficients ofmother's age at childbirth and educational level besides being small,statistically, are not significant. With respect to the mother's schooling, itis noteworthy that the cross-tables presented previously clearly showedthat as the mother's level of education rose, infant survivorship increased.It is very probable that because this analysis only allows for twopossibilities the influence of the mother's level of schooling on infantmortality cannot clearly be perceived.

From the above information we can conclude that the optimumconditions for a child to complete his first year of life are to be foundwhen the mother is over the age of 24 and under the age of 35, theinterval between births is longer than 12 months, the child is breast-fed,the mother has not suffered previous reproductive losses and has receivedsome education. Inversely, motherhood occurring at ages bordering onthe limits which mark the beginning or end of the reproductive years,short intervals between births, the absence of lactation, reproductivelosses and mother's lack of schooling, are -factors which are associatedwith a greater risk of the child's dying before he reaches his first birthday.

The above observations are valid for the whole of the first year ofthe infant's life, but the influence of the variables studied is much greaterin the postneonatal period than in the neonatal period.

Analysis of infant mortality of the last liveborn child

In order to make an analysis which would use rather than theconversion into dummy variables, the different values of the independentvariables and which, at the same time, would incorporate new elements inthe regression, we selected as the dependent variable the survivalcondition of each woman's last liveborn child(28). Besides the independentvariables included in the previous analysis: mother's age at childbirth,number of previous reproductive losses, number of years of schoolattendance and months of lactation, the following independent variableswere added: birth order, size of locality, piped water supply andavailability of hygienic facilities with toilet(29) (all of them with theiroriginal values).

(28) Only the last order births which occurred during the 5 years previous to thesurvey, that is from July 1970 to June 1975, were taken into account.

(29) The variables of availability of water and hygienic services with toilet aredichotomous.

123

The regression results are as follows:R2 = 0.1774

Dependent variables

Birth orderNo. of previous reproductivelossesMother's age at childbirthSize of localityNo. of years of mother'seducationDuration of lactationAvailability of waterAvailability of hygienic facilitiesConstantNo. of cases included: 2 769

unstandardized

-0.005*

-0.014-0.000*

0.053

0.0890.0320.0260.0241.852

standardized

-0.076*

-0.062-0.016*

0.087

0.2190.3250.0610.063

•Statistically not significant at 9596 Test F.

We can deduce from the above figures that the number of years ofmother's school attendance, the size of the locality, the duration oflactation, drinking water supply, and availability of hygienic facilities withtoilet, are positively related to infant survival, while the number ofprevious reproductive losses and birth order are negatively related toinfant survival. On the other hand, the mother's age at childbirth equationdoes not appear statistically significant in the regression.

Despite the fact that the variables taken together only explain 17 %of the variance, the proceeding analysis, besides confirming the previousfindings, clearly established the importance that education and theavailability of water and toilet in the dwelling have in infant mortality.

Conclusions and recommendations

The most important conclusions which we have obtained as a resultof our research are: Infant mortality is, to a large extent, a function offactors that it has not been possible to incorporate in the analysis.

Among the variables included in this study we have found that thegreatest risks of death for infants occur when the interval between twosuccessive births is less than a year. Generally speaking, as theintergenesic interval increases the risk of mortality decreases. However,this inverse relation is converted into a direct relation when the time spanbetween 2 live births is greater than 4 years.

124

The lowest infant mortality indicators are to be found when themother has not suffered previous reproductive losses. On the other hand,if the mother has a past record of infant deaths or pregnancies which haveterminated in abortion or stillbirths, the probabilities of infant deathoccurring are greater. It was also observed that an increase in previouslosses is accompanied by an increase in infant mortality.

The percentages for infant mortality are higher in children whowere not breastfed than in those who were.

Infant mortality is greater in rural localities than in urban localities,and in the course of time the breach, rather than narrowing, hasbroadened even more.

Greatest risks of infant death, together with a marked predominanceof neonatal mortality over postneonatal mortality occur in children bornto mothers under the age of 15. The proportions of infant deaths show adeclining tendency when mothers are 15 upwards at childbirth; theyreach the lowest values and certain stability when the mother is betweenthe ages of 25 and 34, and subsequently show a slight rising tendency.

The mother's level of education has a significant influence on infantmortality. First of all, it was observed that declines in infant mortality areconstant as the mother's level of education increased. Moreover, whenone controls for the mother's schooling, the rural-urban differentialdecreases, especially in the lowest levels of schooling. Since there is agreat difference in the medical facilities available in urban and rural zones,the previous finding clearly indicates the relevance of the mother'seducation, with respect to mortality risks in early childhood.

It was also observed that the availability of drinking water and toiletare positively related to infant survivorship.

Final considerations

Some deaths which occur in the first year of life are inevitable, butmany others could be prevented by modern medicine, hygienicconditions, and appropriate guidance for couples. These days, high levelsof mortality in early childhood cannot be justified, and there is a great dealwhich can be done in order to achieve considerable decreases in thisdemographic variable.

Among the urgent tasks to be carried out for attaining this goal, wecan mention the following:

To extend and improve health services mother-child care andimmunization programs, especially in the small communities.

In this respect it should be pointed out that the right to health is aconstitutional right of all Mexicans.

125

To launch campaigns for educating women on the importance of themother's age at childbirth, the intervals between births and breast-feeding.

To give priority to programs aimed at eliminating illiteracy andraising the cultural and socioeconomic level of the country, with specialemphasis on the low income groups on the social scale.

To provide programs aimed at providing drinking water anddrainage in all areas where such indispensable services are non-existent.

To sum up, the fight against infant and child mortality in ourcountry is an urgent task, and in order to carry it out successfully, wemust intensify our efforts in different directions. While extensive publichealth campaigns and mother-child care and guidance are necessary theyalone are not sufficient. Social and economic changes are indispensable ifthe fight is to be won.

126

io.

3

of-S

al

pë|

Hi

. uT3

HI

•Oí-o.

a c

tü

oööö — d — ö-d«6od-^

— o — d — o — c î o d — d d c i

^ oo O *N O O^ —^ r*i —->

o — o — o —o — o — o — o — o —o —

o o — — o o — — o o — — o o — — o o

o o o o — — — — o o o o — — — — o o

5000000000000

oooooooooooooooo — —

127

'— ooooo . oo ©I IoO ooooo . oo © n

-vi I inox I n o «©OOOOO©©©©©© — © — ©O©©©©© © — © © — ©

© o — — o © — — - © o — — — © — — © o — - ; i © ~ — ' © — _ :

o\0\*nt-^a^o\t^-ooa\' r-í — e-5 ' « -J o— O O O O O O O ' O — O ' O — ©

(N OO CN OO —

—. OO f N OO —

o — © — o — o — © — o — o — © " - © — © — o — © — o — © — ©

— —©© — — OO — — ©© — — © © — — O © « — O O — — OO —

!

o© — ©©©©• . © o © © — —•—. — • © © © © — —• —• —

©©©©©©• -OOOOOOOO'

- O O O O O C

oo * o — n n • * v) \o r-oo

128

C/3

S

<-3[lj Oí

sai3 3ai c

abuO

5<

£••5-5H — .o

mt- rv is

S - E °

«fi

^ =

A-aí-o —

.1ñ

I

u

ÚOOWOODait£0

o ' © o — © ö o " — ' o © ö — ö © ö — o

© — © — o — o — © — © — © — o — o — © — o — o —

©o oo — — o o oo o© o © —• —

-S¡2

© © © © — — — — < > — — o © o o —

MORTALIDAD INFANTILEN CUBA :

SU COMPORTAMIENTOEN EL DECENIO 1970-1979

por

Raúl RIVERÓN-CORTEGUERA*José A. GUTIÉRREZ-MUÑIZ**

Francisco VALDÉS-LAZO***Instituto de Desarrollo de la Salud

La Habana, Cuba

La mortalidad infantil revela las condiciones de salud de unapoblación y su importancia es de características tales, que a menudo esusada como indicator de desarrollo socioeconómico. La misma es unindicador que no solamente revela el desarrollo de la Salud Pública, sinoque, por ser sensible a diferentes factores como la nutrición, lascondiciones higiéníco-epidemiológicas, la atención médica maternoinfantil, la vivienda, etc., constituye un indicador de nivel de vida de lapoblación y del desarrollo del país.

Antes de 1959 y durante los primeros años de la década del 60 losdatos existentes sobre la mortalidad infantil, eran poco confiables. Existíaun subregistro marcado tanto en las cifras de defunciones como en la denacidos vivos. Durante este período se consideraba nacido vivo a losrecién nacidos que sobrevivían las 24 horas y en las áreas rurales se

* Especialista de II grado en Administración de Salud y I Grado en Pediatría,Miembro del Grupo Nacional de Pediatría del Ministerio de Salud Pública.

* * Especialista de II grado en Pediatría. Jefe del Departamento de Investigacionesde Pediatría del Instituto de Desarrollo de la Salud.

* * * Especialista de I grado en Pediatría. Jefe del Departamento Materno Infantil delMinisterio de Salud Pública. Secretario del Grupo Nacional de Pediatría.

130

desconocían tanto las defunciones como los nacimientos. No fue hasta1965, que se pone en vigor la definición de « nacido vivo » propuesta porla Organización Mundial de la Salud cuando se alcanza una coberturaaceptable de personal de estadísticas, y nuestros datos comienzan aconsiderarse aceptables[1.2,3,4,5]

Durante la década de 1970 a 1979, nuestro Ministerio ha trabajadoarduamente en el perfeccionamiento del Sistema Nacional de Salud y sinduda uno de los elementos, donde más se ha trabajado has sido en laatención materno infantil. Entre los factores que han contribuido amejorar la atención de la madre y el niño en esta etapa y por tanto areducir la mortalidad infantil podemos enumerar los siguientes.

Mejoría del estado nutricional de la población al existir unadistribución equitativa de los alimentos y estar al alcance de la poblaciónla posibilidad de adquirirlos.

Los niveles de educación alcanzados por nuestra población, que sinduda repercuten de manera directa en la salud y muy particularmente entodo, lo relacionado con la madre y el niño.

Las amplias posibilidades de trabajo que ofrece nuestra sociedadsocialista, al no existir el desempleo como fenómeno social, lo cual brindauna estabilidad económica a la familia, que repercute en la salud.

Mejoramiento de las condiciones de vivienda, al desaparecer losbarrios insalubres e iniciarse la construcción de viviendas en áreasurbanas y de nuevas comunidades en las áreas rurales.

Desarrollo de la educación sanitaria a las madres en consultasprenatales y de puericultura en los policlínicos ; en los debates de salud dela Federación de Mujeres Cubanas ; en audiencias sanitarias de losComités de Defensa de la Revolución y por los medios masivos decomunicación.

Presencia de la madre acompañante en todas las salas de Pediatría dehospitales del país. Esto es de particular importancia, tanto para el niño,por las ventajas de una mejor relación madre-hijo, como por la enseñanzadesde el punto de vista sanitario que la misma adquiere M.

Incremento del nivel immunitario de nuestra población, con especialinterés en la población infantil, aplicándose el BCG, la vacuna antipolio-mielítica, la vacuna Triple (Toxoide-diftérico-tetánico con vacuna antiper-tussis) ; la vacuna antisarampiónica, el toxoide diftérico-tetánico y eltoxoide tetánico, entre otras V\

Aumento de los recursos humanos profesionales y técnicos, tanto encantidad como en calidad. Se ha incrementado el número de especialistasde pediatría y ginecobstetricia y su distribución se ha realizado atendiendoa una distribución proporcional. Las enfermeras y demás técnicos medioshan experimentado un aumento sustancial en relación con el mejora-miento de los servicios que se ofrecen a la población M.

131

Mejoría de la calidad de los servicios pediátricos al desaparecer elhacinamiento e incrementarse el personal.

Implantación de la reanimación del recién nacido en los salones departo de todas las unidades que los realizan en el país &• 9'-

Modernización de los servicios de Neonatología de los hospitales dematernidad (servicios cerrados) y de los hospitales pediátricos (serviciosabiertos) con incubadoras de cuidados especiales y con servo-control,cunas térmicas, monitores, equipos de fototerapia y de presión positiva,uso de microtécnicas y de gasometría en todas las provincias.Conjuntamente con esto, se iniciaron los cursos posbásicos de enfermeríaespecializada en Neonatología y la ubicación de pediatras especializadosen Neonatologiat'0- ' '1.

Se elaboraron las Normas de Pediatría y de Obstetricia, lo cualpermitió uniformar en toda la Isla el diagnóstico y tratamiento de lasprincipales entidades nosológicas y la organización de los servicios deambas especialidades[12-I3].

Se trabajó tenazmente para reducir la morbimortalidad porenfermedades diarreicas agudas, para lo cual se realizaron múltiplesactividades t'4-15> l6l Entre ellas tenemos :- Fueron independizados los servicios de enfermedades diarreicasagudas del resto de las enfermedades infantiles y se crearon lascondiciones de un servicio de enfermedades infectocontagiosas.- Se destinó el 30% de las camas pediátricas de todo el país a lasenfermedades diarreicas agudas.- Se priorizaron la consulta y el ingreso de todo niño con enfermedaddiarreica aguda que tuviera un ligero signo de desnutrición.- Se dispuso que el paciente con enfermedad diarreica aguda encualquier parte del pais tanto en zona urbana como rural fuera atendidopor el personal de mayor calificación científica disponible, para garantizarsu óptima atención.- Se eliminaron las salas de hidratación en los servicios de urgencia y seingresa a todo niño que requiera hidratación.- Se dispuso el control de focos de enfermedades diarreicas agudas y elestablecimiento en todos los hospitales pediátricos o con servicios dePediatría, de las comisiones de gastroenteritis que analizaban mensual-mente, con la dirección de los hospitales, todo lo referente a los Servicios,material utilizado, medicamentos, venoclisis, etc.- Se normaron los criterios sobre el uso de medicamentos en las salas deenfermedades diarreicas agudas.

Se llevó a cabo una tarea de perfeccionamiento del trabajo en laObstetricia t|7í en todo el territorio nacional, lo cual puede expresarse enlos siguientes elementos :

- Coordinación entre el obstetra y el pediatra en la atención del bi-nomio madre-hijo.

132

- Captación precoz de la embarazada en la consulta prenatal demanera de ir educándola en su futura función de madre y a la vez detectarfactores de riesgo.

- Detección del riesgo obstétrico con priorización del verdaderoalto riesgo.

- Incremento del parto institucional y de las consultas por parto.- Establecimiento de las normas de control higiénico epidemioló-

gico de los salones de partos, de manera que los mismos tuvieran condi-ciones similares a las de los salones quirúrgicos.

- Realizar la reanimación del recién nacido en el interior del salónde partos o en un cubículo habilitado al efecto.

- Modernización de la atención al parto, con la aplicación demodernas técnicas de diagnóstico como los equipos de ultrasonido, elmonitoreo del parto, desde su ingreso a preparto, la introducción del usode la gasometría y de los equipos de microtécnicat'8!.

Se crearon los hogares maternos en las zonas más apartadas del paísen su inicio con la finalidad de incrementar el parto institucional. En laactual década, es un instrumento de educación sanitaria y a su vez deprevención de la prematuridad ['9l

Se incorporaron al programa de atención a la mujer, la divulgaciónde los métodos anticoncepcionales existentes en el país : DIU (anillos, asade Lippe, T. de Cobre) diafragmas, condón y los anticonceptivos oralespara la regulación de la fecundidad '2°J

Muchos otros elementos han sido motivo de trabajo durante todosestos años de programa para reducir la mortalidad infantil en nuestro país,que harían muy extenso este trabajo, por lo que hemos creído convenienteanalizar los resultados obtenidos en el transcurso de los años 1970 a 1979.

MORTALIDAD INFANTIL

Las defunciones en menores de 1 año en relación con el total demuertes por todas las edades alcanzaron en 1970 el 17,196 con 9173defunciones. Esta cifra se elevó en 1971 para posteriormente iniciar undescenso continuado hasta 1979 en que se registraron en todo el territorionacional 2 773 defunciones en menores de 1 año que representaron el5,1% de todas las muertes ocurridas ese año '2 ' I (Tabla I).

Al analizar la mortalidad infantil (Tabla II) lo haremos teniendo encuenta sus componentes : neonatal precoz (0-6 días), neonatal tardío (7 a27 días) y el postneonatal (28 días-11 meses).

El período neonatal precoz es el componente de la mortalidadinfantil donde más difícil resulta reducir su morbimortalidad. Está muy

133

TABLA I. - Porcentaje de participación de las defunciones menores de 1 añoen relación al total. 1970-1979

DEFUNCIONES

L Ñ O S

1970197119721973197419751976197719781979

Todas lasedades

53 76152 29949 44751 02652 90950 96152 91455 96755 94954 837

Menoresde 1 año

9 1739 3557 1296 6875 9465 2994 3704 22533132 773

DE PARTICIPACIÓN EN RELACIÓNAL TOTAL DE DEFUNCIONES

17,117,914,413,111,210,48,37,56,05,1

Fuente : Informe Anual 1979 del Ministerio de Salud Pública.

ligado al trabajo del ginecobstetra durante el embarazo y en el momentodel parto. La mortalidad neonatal precoz requiere de inversiones yrecursos considerables para su reducción, ya que está muy vinculada conlos fenómenos que se producen alrededor del parto como la anoxia, lainmaturidad y el bajo peso al nacer, las malformaciones congénitas y lasenfermedades de la madre que influyen sobre la salud del feto.

En el período analizado la mortalidad neonatal precoz redujo su tasade 17,7 en 1970 a 11,3 defunciones por 1 000 nacidos vivos, lo querepresenta una disminución del 36,2%.

El componente neonatal tardío es otra parte de la mortalidad infantilque resulta difícil reducir, ya que este período de 7 a 27 días está muyligado al desarrollo existente en los servicios de Neonatología. Lasinfecciones cruzadas, la sepsis y las malformaciones congénitas quesobrevivieron los primeros 7 días principalmente, entran a formar parte,como causas principales de muerte en este corto período. En la década de1970-1979 la mortalidad neonatal tardía se redujo de 6,1 en 1970 a 2,0por 1 000 nacidos vivos en 1979, para una rebaja del 67,2%.

Los componentes neonatal precoz y tardío constituyen la mortalidadneonatal que se plantea dependiente fundamentalmente de factoresrelacionados con el trabajo obstétrico.

La mortalidad postneonatal es el componente de la mortalidadinfantil que es expresión de los fenómenos que tienen relación con elambiente del lactante. Está muy vinculado con las enfermedadesdiarreicas y respiratorias agudas y con problemas nutricionales.

En nuestro país la mortalidad postneonatal disminuyó en la décadade 15,0 en 1970 a 6,1 por cada I 000 nacidos vivos en 1979. Esto,representó una reducción del 59,3%.

134

ia

,ro ô5 o\

o

<

o

<_OCO

SOzw

COÜJCOUJ

[AS

-l

Q

CO<

5CN

m<

Só

VI

C

*<co<

Op¿u

*<CO<h-

Ooíu

Z

*<VI<

Oc¿tüSDZ

<co<H

Oc¿uDZ

XI•3CL

T3J3

•a

o

c

s•a

es <

o go E2 3<~ S& ;•

H rs\ O N O N O N O N (^^ CT\

135

El total de defunciones menores de 1 año se redujeron de 9173 en1970 a 2773 en 1979. La tasa se redujo de 38,8 por 1 000 nacidos vivos, a19,4, para una disminución del 50,0% en el decenio.

El esfuerzo realizado por todo el equipo materno infantil, así comopor la Dirección de nuestro Ministerio de Salud Pública, ha sido enorme,pero los resultados han sido muy alentadores, ya que la disminución del50 % en la mortalidad infantil en un decenio, cuando las cifras no eranmuy altas, constituye un verdadero ejemplo de lo que puede lograr un paíscuando trabaja en una dirección determinada.

MORTALIDAD INFANTIL POR PROVINCIAS

Al analizar la mortalidad infantil en las diferentes provinciastenemos que exponer que hasta 1976, en que se establece la nuevadivisión político administrativa del país, sólo existían en Cuba6 provincias. A partir de esta fecha es que se establecen 14 provincias y elMunicipio Especial de la Isla de la Juventud. Con la nueva divisiónpolítico administrativa, las antiguas provincias mantienen un mayornúmero de recursos tanto humanos como materiales (camas, instituciones,equipos, etc.) en relación con las nuevas provincias. Durante los últimosaños de la década se han priorizado las provincias nuevas con la finalidad

Gráfico I. Natalidad, Cuba 1960-1979

Tasa por1000 habitantes40

I I I I I I I M i l I T M i l i

I I I I M M I II 9 6 0 1965 1970 1975

F u e n t e : I n f o r m e Anual 1979 d e l MINSAP

1979

Años

136

sa

•a t^c i-2 ^c T— o•o r-.•g s

pI

<

T cT cT oo" oo" rrÇ ey\ -*r TT OS

oo r^ o ' ~* oC r i ^ oC vo r

oo" rn r «T o" oo'MfNNnN

i^ooni^nw^x

aa a

.2.« § ö o >•Si" 5

v ; < ^u « M ¿ï^ r ^ l • • • Z3 t j ^ t a W * * ^ ^

älll^iHi^lgll^M •*> »t i3 ^5 ¿ï ^ i r C wl * ^ n G ^* irt " J

ö

I(/i

1w

<,S2

3J! 3 | g

«II

w= g gS 2 § •§¡ 1 2 S

9 °"

Mc

137

de ir borrando las diferencias, principalmente en recursos humanos ycamas.

Las provincias que alcanzaron las tasas más bajas al finalizar eldecenio fueron: Isla de la Juventud [14,7]; Matanzas [14,9]; Villa Clara[15,6]; Ciudad de La Habana [16,2]; Cienfuegos [17,3]; Camaguey [18,3];Holguín [18,5]; Provincia Habana [19,0] y Santiago de Cuba [19,5], queregistraron cifras inferiores a 20,0 por cada 1000 nacidos vivos.

Las provincias con mayor tasa de mortalidad infantil fueron: LasTunas [26,1]; Guantánamo [24,1]; Granma [23,9]; Sancti Spiritus [20,5];Pinar del Río [20,4] y Ciego de Avila [20,2]. Resumiendo, 8 provincias y elMunicipio Especial de la Isla de la Juventud quedaron por debajo de 20,0y 6 provincias quedaron por encima, de estas últimas 5 son nuevasprovincias y 1 sola es de las antiguas provincias. (Tabla III).

Independientemente de las tasas alcanzadas, las provincias quemayor porcentaje de reducción obtuvieron en el decenio fueron la Isla dela Juventud (71,496); Matanzas (61,4%); Ciudad de la Habana (59,5%);Camaguey (57,5%); Villa Clara (55,3%); Cienfuegos (54,6%) y Holguín(51,7%). Las provincias que menos porcentaje de reducción obtuvieronfueron Las Tunas (28,5 96); Pinar del Rio (39,1 %); Guantánamo (42,3 %);Sancti Spiritus (42,7%); Santiago de Cuba (46,3%); Granma (47,7%);Habana (48,5 % ) y Ciego de Avila (49,9 %)&A

Como elemento que se debe destacar es que en el decenio todas lasprovincias disminuyeron considerablemente su tasa de mortalidadinfantil, lo que demuestra que ei programa elaborado para reducirlacumplió su cometido ampliamente.

PRINCIPALES CAUSAS DE MUERTE

Se analizaron las primeras 20 causas de muerte según lista A de la IXClasificación Internacional de Enfermedades [23], de manera de poderprecisar con mayor detalle las causas que producen las defunciones en losniños menores de 1 año y tratar de tomar medidas para disminuirlas. Nose incluye el año 1979 por estar en el procesamiento de los certificados dedefunción, y no estar disponible el dato.

Como puede observarse en la Tabla IV, las afecciones anóxicas ehipóxicas no clasificadas se han mantenido durante todo el periodo en elprimer lugar de las causas de muerte, aunque han experimentado unareducción en su tasa de 6,6 en 1970 a 4,2 por 1000 nacidos vivos en 1978,disminuyendo la mortalidad en el 36,3%.

Las anomalías congénitas (A 126-130) que al inicio de la décadaocupaban el 6to. lugar como causa de muerte, se han colocado en 2do.lugar producto de la reducción que han experimentado las demás causas,

138

O VO OO VO fNf f r-í —T —T —T

(N fNO* O* o"o~ O* O" O" (N

oo n h \ o ^o o o o o o o o w->

•s• ^ •**" rn rñ —T c*í —" O O O O O O

o

10)_cc•o

1

1 —i fN — O O O O O O O O O O O O

O O O O O O O O CT\

o' o' o' o' o' o' o' o' o* o' o' o' o'

5— oC r-00 —

l a

— o\ oo oo rm m — fN <

*o >ûvi > - n cN rN — fN •—• o O v i r ^o* o' o" o' o' o o" o" o' o' o' o" — oo'

O\ fN '—' —• ^ r~lr rN — — — ° ° ^ "Oo" o* o' o' o' o' o' — vo'

i1i

oOs

— rN — •—• — l o OO

o o o' o' o — oo'

c-I

<1/1

« -S S S S g.2 a a u r, .

= J& O

139

ya que el aumento ha sido de 3,6 en 1970 a 3,8 por 1000 nacidos vivos en1978, lo que representa un incremento del 5,5%, la influenza yneumonías (A089, 090, 091 y 092) que ocupaban la 3ra. causa de muerte

' al inicio del período analizado, se mantienen en el mismo lugar a pesar dehaber reducido su tasa de 5,3 en 1970 a 3,0 por 1000 nacidos vivos en1978, para una disminución del 43,496. La cuarta causa de muerte loconstituyen otras causas de morbi-mortalidad perinatal (A 13 5) que se hamantenido en el mismo lugar a pesar de reducir su mortalidad de 5,3 en1970 a 2,6 por 1000 nacidos vivos en 1978, para una disminución del50,9%. En quinto lugar se mantienen las lesiones al nacer y partosdistócicos (Al 31), la cual desciende de 2,0 en 1970 a 1,8 por 1000 nacidosvivos en 1978, para una disminución del 1096.

Las enfermedades diarreicas agudas merecen ser destacadas, ya queocupaban en 1970 el 2do. lugar como causa de muerte en menores de 1año con una tasa de 5,6 por 1000 nacidos vivos y al finalizar el año 1978habían descendido al 6to. lugar con una tasa de 1,6, para experimentaruna reducción de su mortalidad en el 71,4%. La Sepsis (A021) que en1970 ocupaba el 5to. lugar como causa de muerte con una tasa de 5,0 por1000 nacidos vivos, en 1978 pasó a ocupar el 7mo. lugar con tasa de 1,2registrando un descenso del 76 %.

La reduction de la mortalidad en menores de 1 año en los rubros deenfermedades diarreicas agudas y sepsis, es sin duda, una expresión delmejoramiento del trabajo pediátrico y uno de los principales logrosobtenidos en el decenio que recién concluimos.

El resto de las causas de muerte aparecen en detalles en la Tabla IV,donde se observan ligeros aumentos y reducciones en el transcurso deestos 9 años, en algunas enfermedades las tasas son tan bajas que apenas siresulta útil su análisis, pero consideramos incluirlas para conocer el ordenen que aparecen como causas de mortalidad infantil.

CONSULTAS A EMBARAZADAS

El gráfico II muestra el número de consultas por parto que se hanbrindado a mujeres embarazadas durante la década que acaba de concluir.En 1970 se dieron 7 consultas de Obstetricia por cada parto, esta cifra havenido incrementándose progresivamente hasta alcanzar en 1979 unacifra de 11,0 consultas por parto. Este número de consultas, unido a laelevación de su calidad, han mejorado considerablemente la atenciónobstétrica y por ende ha producido una reducción considerable de lamortalidad perinatal. Hay que tener en cuenta además que en los casos dealto riesgo obstétrico, la cifra de 11 consultas/parto se incrementaconsiderablemente. Esto permite que en 1979 se haya alcanzado el 98,3 %de los partos en instituciones, siendo uno de los indicadores más alto delContinente Americano.

140

Gráfico I I . Consultas por parto. Cuba 1970-1979

12.0

10.0

8.0

6.0

4.0

2.0

/

i

EU17883

1970 1975 1979

Anos

Fuente : Informe Anual 1979 del HINSAP

CONTROLES A NIÑOS SANOS

En el gráfico III puede observarse el incremento del promedio deconsultas a niños sanos menores de 1 año que se ha alcanzado. A finalesde la década del 60 se implantó la consulta de puericultura, la cual fueincrementándose progresivamente hasta que en 1970 se llegó a alcanzar1.8 consulta por niño. Con la puesta en marcha del Programa dereducción de mortalidad infantil, esta cifra se fue elevando constante-mente hasta 1976 en que disminuyó ligeramente, para posteriormenteelevarse a 6,7 consultas por cada niño sano menor de 1 año. El programade atención integral al niño plantea dar como máximo 7 consultas porniño como promedio anualmente.

Consideramos que esta consulta a niños sanos es sumanenteimportante, ya que permite una evaluación del crecimiento y desarrollodel niño periódicamente ; administrar las inmunizaciones fundamentalesdel niño en su primer año de vida, así como para orientar la alimentacióny las medidas sanitarias fundamentales a tener en cuenta en estas edades.

141

GraFii'o I I I . Controles promedió a niños sanos menores de I año.Cuba 1970-1979

Controles a niñosmenores de 1 año10.0

8JO

6.0

2.0(>r

Jf

i

/>

&

1970 1975 1979

AnosFuente : Informe Anual 1979 del MINSAP

NATALIDAD

Como parte integral del Programa materno infantil, se incorporaronactividades de planeamiento familiar, teniendo en cuenta que elespaciamiento de los embarazos y la posibilidad de limitar el número dehijos respondía a los intereses y des.eos de la pareja. Conjuntamente conoíros factores estas actividades, no podían considerarse ajenas a unprograma materno infantil de gran proyección.

Las actividades de planeamiento al ser partes integrantes delprograma, se realizaron por el mismo personal y de conjunto con laconsulta de ginecobstetricia en todas las Unidades del Sistema de Salud.Los medios más modernos de anticoncepción han sido puestos adisposición de la pareja en forma gratuita, como los demás servicios desalud y sin que formen parte de un programa especial.

En los primeros años de la década del 60, hubo un aumento bruscode la natalidad llegando a alcanzar una tasa de 34,0 por cada 1000habitantes, en 1965. A partir de esta fecha ha ido decreciendopaulatinamente y de forma más acentuada en los últimos años hastaregistrar en 1979 la cifra de 14,7 por cada 1000 habitantes. No hay dudasque estos resultados son básicamente el producto de los altos niveles deeducación, salud y elevación de la calidad de la vida de la población.

142

CONCLUSIONES

Se pone de manifiesto la reducción de la mortalidad infantil en un50,0% en el decenio 1970-1979.

Se analizan los diferentes elementos, que de una manera u otra haninfluido en la disminución de la mortalidad infantil.

Se disminuye la mortalidad neonatal precoz en un 36,296 ; laneonatal tardía en un 65,2% y la postneonatal en un 59,3% en estos10 años.

Las provincias con menor mortalidad fueron Matanzas I'4-9], VillaClara H 5.fi] y Ciudad de la Habana H6-2' y las que exhibieron una tasa mayorfueron : Las Tunas I26-'], Guantánamo t24-1! y Granma con 23,9defunciones menores de 1 año por cada 1000 nacidos vivos.

Las enfermedades diarreicas pasan del 2do. lugar al 6to. con unareducción de un 71,4 % en su mortalidad.

Las anomalías congénitas pasaron a ocupar el 2do. lugar como causade muerte como consecuencia de la reducción de las demás causas.

Las enfermedades respiratorias agudas reducen su mortalidad en un43,4% en el decenio.

Se alcanzó la cifra de 11 consultas por parto y 6,7 controles de saluda niños sanos menores de 1 año, como promedio anual.

La natalidad se redujo de 27,7 a 14,7 por 1000 habitantes en losúltimos 10 años.

BIBLIOGRAFÍA

[1] RIVERON-CORTEGUERA, R. y Cois. « Salud Materno infantil : situación actual y pers-pectivas ». Rev. Cub. Pediatría 50 : 407-423, 1978.

[2] RIVERON-CORTEGUERA R., FERRER GRACIA, H. y VALDES LAZO F. « Avances en Pe-diatría y Atención Infantil en Cuba 1959-1974 ». Bol. Of San. Panam. 80 : 187-204, 1976.

[3] RIVERON-CORTEGUERA R., DUEÑAS GOMEZ E. y PEREA CORRAL J. « Mortalidad Infan-til en Cuba », publicado en Ecología en Clínica Pediátrica, Biblioteca PediátricaNo. 13, pp. 94-102, Ed. Médica-Panamericana, Buenos Aires, Argentina, 1975.

[4] RIVERON-CORTEGUERA R-, VALDES LAZO F. y Ríos MASSABOT E. « Mortalidad Infantilen Cuba : análisis de un programa para su reducción ». Bol. Med. Hospital Infantilde México 35 : 981-992, 1978.

[5] AZCUY P., RIVERON-CORTEGUERA R., ULLOA FRANCO F., HERNANDEZ HERNANDEZM. y DOSIL M. « Programa de Reducción de la Mortalidad Infantil en Oriente Sur ».Santiago de Cuba, Diciembre, 1969.

143

[6] GUERRERO RODRIGUEZ T. « La madre acompañante en el Hôpital Pediátrico de CentroHabana ». Trabajo para optar por el título de Especialista de I Grado en Pediatría,Junio de 1971.

[7] WERTHEIN L. y RIVERON-CORTEGUERA R. « Resultados obtenidos en el control de al-gunas enfermedades prevenibles por vacunas ». Comunicación en Carteles presen-tada a la XIX Jornada Nacional de Pediatría, Ciudad de la Habana, Diciembrede 1977.

[8] RIVERON-CORTEGUERA R. Recursos humanos para la atención materno infantil.Rev. Cub. Ped. 47 : 329-334, 1975.

[9] FARNOT U., DUEÑAS GOMEZ E. y RIVERON-CORTEGUERA R. « Mortalidad Perinatal enCuba». Rev. Cub. Obst. Ginec, 2 : 135-146, 1976.

[10] ROJAS-OCHOA F. y RIVERON-CORTEGUERA R. « Mortalidad según el peso al naceren los Servicios de Neonatología de Cuba 1968-1974». Rev. Cub. Adm. Salud2:205-210, 1976.

[11] DUEÑAS GOMEZ E. y RIVERON-CORTEGUERA R. « La neonatología en Cuba ».Bol. Of. Sanit. Panam. 86 : 406-419, 1979.

[12] MINISTERIO DE SALUD PUBLICA. GRUPO NACIONAL DE PEDIATRÍA. Normas de Pedia-tría, 2da. Edición, Instituto Cubano del Libro. Ciudad de la Habana, Enero 1976.

[13] MINISTERIO DE SALUD PUBLICA. GRUPO NACIONAL DE GINECOBSTETRICIA. Normasde Obstetricia, 2da. Edición, Instituto Cubano del Libro. Ciudad de la Habana,Septiembre 1975.

[14] MINISTERIO DE SALUD PUBLICA. «Estudio sobre las enfermedades diarreicasagudas». Rev. Cub. Ped. 35: 161-193, 1963.

[15] RIVERON-CORTEGUERA R., VALDES LAZO F. y PEREA CORRAL J. « Morbimortalidadpor enfermedades diarreicas agudas en Cuba 1962-1973 ». Rev. Cub. Ped.48: 7-15, 1976.

[16] RIVERON-CORTEGUERA R., CORDOVA VARGAS L. y VALDES LAZO F. « EnfermedadesDiarreicas Agudas en Cuba». Rev. Cub. Ped. 5 1 : 181-193, 1979.

[17] RODRIGUEZ-CASTRO R. Organización de la atención obstétrica en Cuba. Rev. Cub.Adm. Salud 2 : 129-140, 1976.

[18] RIVERON-CORTEGUERA R. « La Salud y la Mujer en Cuba ». Rev. Cub. Adm. Salud2: 219-238, 1976.

[19] CASTELL MORENO J. Hogares Maternos. Folleto 17 Años de Ginecología, Obstetriciay Neonatología, Ciudad de la Habana, Enero, 1976.

[20] ALVAREZ VASQUEZ L. y ALVAREZ LAJONCHERE C. « Cuba : Fecundidad, Diferencia-les, Contracepción y Abortos en Zonas Seleccionadas ». Simposio CUBA-OMS So-bre. Reproducción Humana y Regulación de la Fecundidad. Ciudad de la Habana,30 Enero-1 Febrero, 1978. Publicado por Ed. ESP AXS Barcelona, 1978.

[21] MINISTERIO DE SALUD PUBLICA. Informe Anual 1979, Ciudad de la Habana,Marzo, 1980.

[22] COMITÉ ESTATAL D E ESTADÍSTICAS. Anuario Demográfico 1977. Ciudad de laHabana, Febrero 1980.

[23] ORGANIZACIÓN MUNDIAL DE LA SALUD. IX Clasificación Internacional de Enferme-dades. Ginebra, 1975.

CHILD MORTALITYIN DIFFERENT CONTEXTS

IN BRAZIL:VARIATION IN THE EFFECTS

OF SOCIO-ECONOMIC VARIABLES

by

Diana Oya SAWYER (1)and Elidimar Sergio SOARES (2)

Federal University of Minas GéraisBrazil

INTRODUCTION

Most of the natural history of the morbid process that individualsmay be subjected to is well known and the medical and sanitarytechnology needed to prevent and cure the most lethal diseases in theThird World has been known for some time (3). Still, millions of childrendie before completing the first five years of life.

Numerous studies have been undertaken of the reasons for thesepremature deaths and of means to avoid them. In order to understand theprocess that leads a person from a state of health to lack of health and todeath, it is not sufficient to focus exclusively on the interplay betweenindividuals' physiological status and pathogenic factors in their environ-ment. It is also necessary to take into account factors of a political, socio-economic or cultural nature that aggravate or attenuate harmfulinteractions.

( 1 ) Associate Professor at CEDEPLAR, Federal University of Minas Gérais. Brazil,and Visiting Professor in the Amazon Research and Training Program. Center for LatinAmerican Studies. University of Florida.

(2) Systems Analyst at the Computation Center of the Federal University of MinasGerais.

(3) The authors express their appreciation to Donald R. Sawyer for his commentsand editing and to Thomas Merrick for providing helpful information.

146

In the Third World, it is common to find an uneven distribution ofexpensive, privately owned or quasi-private health services. Adequatemedical attention is most available to those strata which can afford it, andexisting services are designed primarily to care for their needs, whichdiffer fundamentally from those of the less privileged. Especially in urbanareas, there is a tendency for adequate nutrition, sanitation and livingconditions, which play an important role in resisting disease, to depend onone's ability to pay for them. Health increasingly becomes a commoditythat has to be purchased. The access a person has to materials andinformation needed to provide health for himself and for his family is akey factor in avoiding premature deaths. Such access is related to positionin the social class structure and to social networks, as well as beingaffected by political forces. It depends, ultimately, on the way theproductive structure and society are organized.

One of the ways in which we can begin to discover how socio-economic factors lead to premature deaths is by using empirical data ondifferentials in infant mortality by various socio-economic variables.Numerous studies on mortality in the Third World indicate large gapsbetween the highest and lowest categories of the socio-economic variablesused. This reflects not only the prevailing inequality of wealth but alsoinequality of the provision of and access to health care systems amongdifferent sectors of these populations.

The situation in Brazil is not different, and may be particularlyinstructive because of its marked socio-economic contrasts betweensectors covered by a single census. There are considerable differentials inchildhood mortality among regions in Brazil and, within regions, amongdifferent socio-economic strata (Carvalho and Sawyer 1978, Carvalho andWood 1978, Sawyer 1981). The Northeast and the South of Brazil standin sharp contrast to each other in socio-economic and demographic terms.The states from Rio de Janeiro and Sao Paulo to the south arecharacterized by rapid industrial growth and development, as well aspossessing most of the health-related infrastructure and services. Internalmigration in Brazil has to a large extent been directed toward these states.They have the lowest rates of infant and child mortality. The Northeast,on the other hand, has the highest mortality rates in Brazil. The statesfrom Bahia north to Maranhäo are characterized by economic stagnationand by out-migration (IBGE 1977).

This paper interprets the effects of a number of economic and socialconditioning factors on child mortality within four different contexts:Urban South, Rural South, Urban Northeast and Rural Northeast. TheSouth comprises the states of Rio de Janeiro, Guanabara (the formerFederal District, since incorporated in the state of Rio de Janeiro), SäoPaulo, Paraná, Santa Catarina and Rio Grande do Sul. The Northeastincludes the states of Maranhäo, Piaui, Ceará, Rio Grande do Norte,Paraiba, Pernambuco, Alagoas, Sergipe and Bahia. Definitions of urban

147

and rural are those given by the census and are based on administrativecriteria rather than population size.

There is no attempt to cover all of the possible conditioning factors,but only to show some of the variation in relationships between variablesin different settings within the same country or region. The basic point tobe made is that the same conditioning variables may have different effectsin different contexts, There is no single "most important" variable toexplain mortality wherever it may occur. Rather, the effects of eachvariable on mortality depend on the economic and social surroundings inwhich they occur.

THE DATA

The data used in this study are at the individual level, from a one-percent public use sample of the Brazilian demographic census of 1970.Data on households were matched with individual data.

Multiple Classification Analysis (MCA) was chosen for the analysis.It provides for multivariate analysis of categorical data, without theassumption of independence among the so-called independent variables.

The dependent variable, abbreviated PCD, was the proportion ofchildren who had died among children ever born to women age 20-29,who were in conjugal union of any type and who had at least one childever born. This proportion is an approximation of mortality amongchildren up to five years of age. It was standardized for length of exposureto risk of death among children of different ages using techniquesdeveloped by Trussel and Preston (1981).

Among the independent variables, the following were considered as"factors" in the MCA analysis: water and sewerage of the household,previous residence of mother, mother's education and mother's worksituation. Control variables, called "covariates", were mother's length ofresidence in the municipality of present residence and per capita familyincome.

It can be seen that the independent variables are intended to beindicators of conditions regarding sanitation, migration, social relationsand the economic situation of the household and the mother. They do notexhaust, by any means, the possibilities; much of the socio-economic anddemographic context remains beyond the set of variables that could beincluded. The independent variables included in the analysis are describedand justified as follows:

148

1. Water and Sewerage of the Household

Access to adequate water supply and sewerage facilities is a majorcontributing factor to lower levels of infant mortality in some segments ofBrazil's population (Vetter and Simöes 1981, Merrick 1981). The waterthat a family uses for drinking, cooking, washing and waste disposal is awell-known potential means of transmission of many infectious andparasitic diseases, especially those of an intestinal nature. In Brazil,gastroenteritic diseases are one of the principal causes of infant deaths(Carvalho and Sawyer 1978, Sawyer 1980). Even in cities where publicsystems of water treatment and waste disposal are well developed, parts ofthe population, especially in outlying or poor neighborhoods, may not becovered. These neighborhoods tend to absorb poor migrants, due to rapidpopulation growth of cities and high rents in central areas.

The water and sewerage variable used in this analysis wasconstructed from information on the two types of installations. Watersupply was considered adequate if the household had piped water insideor outside the house, or if there was a well or spring with insideplumbing. Wells or springs without inside plumbing or any other sourceof water were considered inadequate. It is true that wells or springswithout plumbing are not necessarily risks to health, but when combinedwith inadequate sewerage there is a greater possibility of contamination ofthe water. Sewerage installations were considered adequate if connectedto a central system or to a sanitary cesspool. Simple privies, other meansof waste disposal or no installation were considered inadequate.

It should be remembered that the presence of plumbing is not initself an adequate indicator of circumstances with regard to contamina-tion, since the method of treatment and distribution of water have to betaken into account, as well as the degree of contamination of the watersources when piped water is not present in the household. However, thisinformation was not available.

Using the above-mentioned criteria for adequate and inadequateinstallations, the combined water and sewerage variable was given threepossible values: both adequate, one inadequate and both inadequate.

2. Previous Residence of Mother

Migrants to large cities in Brazil have higher levels of mortality thannon-migrants (Simöes, 1979). This is probably due to some extent to theunfavorable social and economic conditions which urban settlers faceupon arrival. It must be remembered, however, that one of the problemsof analyzing the effects of the migration variable as normally defined onthe proportion of children who have died is that the deaths might haveoccurred before the mother migrated. To the extent this occurs, mortalitycannot be attributed to conditions at the place of present residence.

149

The purpose of including this migration variable is to attempt toisolate the effect of migration on child mortality from other conditionsprevailing at the present place of residence. If the effect of this variablepersists even after other variables are controlled, it is quite possible thatmortality is due to conditions at the previous place of residence.

The variable chosen for use as a factor was urban or rural previousresidence of the mother. Categories were: a) rural, b) urban and c) non-migrant. The length of residence in the municipality of present residencewas included as a control variable, in order to further evaluate the effect ofmigratory status of the mother on child mortality.

3. Mother's Education

In recent years, studies in various parts of the world have shownthat mother's education is one of the most important variables forexplaining mortality differentials (Caldwell 1980, Behm 1980). Interpreta-tion of the meaning of this variable, however, is still controversial. It maybe, on the one hand, that mother's education is a particulary accurateindicator or proxy of the general socio-economic position of the family. Itappears early in the life cycle, before other socio-economic characteristicshave developed clearly, and is very stable over time, as well as being easyto measure and scale. This may make it a better general socio-economicindicator than variables such as income or occupation, at least as they arecommonly used. On the other hand, education of mothers may be, in andof itself, an important determinant of lower child mortality by providingthem with knowledge of how to care for children, in both preventive andcurative ways, and how to seek appropriate medical attention. It would bevery difficult to separate these two dimensions without furtherinformation. The minimum that is attempted in this paper is to examinethe extent to which mother's education can be regarded as universallyimportant by evaluating its effects in different contexts, as well as byseeing to what extent the effects are attenuated or accentuated when otherconditions are controlled.

The education variable was constructed using two types ofinformation from the census, one giving the level of schooling(elementary, secondary, college, etc.) and the other the grade (year)attended within each level. The derived categories are: a) no education, b)incomplete elementary, c) complete elementary and d) at least one year ofsecondary.

4. Mother's Work Situation

Because of monetarization of the economy and higher costs of living,many women in Brazil have begun to work outside the home in order to

150

supplement family income. Some studies have concluded that infantmortality is lower among women that work because family income ishigher (Schultz 1980, Merrick 1981). It should be remembered that theactual effect of mother's work on child mortality depends on the structureof the family, particularly on the availability of other family members totake care of children in the mother's absence, or on the availability ofadequate child care centers. Working conditions of the mother, hoursspent working and commuting and salary levels are also relevant. Effectsmay not be uniformily positive. The categories of the variable used herewere: a) working mother and b) non-working mother. No distinctionswere made among different types of work.

These four independent variables were the factors analyzed todetermine their effects on the proportion of children who have died forwomen in different situations. The effects of each factor were analyzedafter adjustment for the effects of the other factors. The two controlvariables used were Family Income and Length of Residence in theMunicipality of Present Residence. The former serves as an indicator ofthe family's economic situation. The latter is assumed to be an indicator ofadaptation to the context in which the family lives, including knowledgeabout and access to local health-care services.

It should be pointed out that, as is generally the case in studies whichuse this type of analysis, there may be differences between the conditionsexisting at the time of the census and those which prevailed at the time thechildren died. The effects of this problem were reduced by restricting theanalysis to women in the 20 to 29 age group, i.e., those for which the leasttime had elapsed between childbearing and the census.

RESULTS

The frequency distribution of the variables shown in Table 1confirms the great disparities between the regions studied. While in theUrban South 41.67% of the households had adequate water and adequatesewerage (which by itself is certainly not a high figure), the figure for theUrban Northeast was only 14.68%. The rural areas were worst off,showing proportions of households with adequate sanitary installations of3.38% for the Rural South and 0.29% for the Rural Northeast. In theSouthern states, 62.98% of urban women and 53.91 % of rural womenwere migrants. In the Northeast the corresponding figures were 39.51 %and 19.86%. The difference is due in part to inter-regional migration tometropolitan areas in the South. Another contrast lies in the proportion ofmothers whith no formal education: 18.23% in the Urban Southcompared to 75.99% in the Rural Northeast.

151

TABLE 1. - Frequency distribution of the variables in the model, by context.Brazil 1970(96)

FACTORS

A) Water and sewerageof household

1. Both inadequate2. One inadequate3. Both adequate

B) Mother's previous residence1. Rural2. Urban3. Non migrant

C) Mother's education (*)1. No education2. Incomplete elementary3. Complete elementary4. More than elementary

D) Mother's work situation (**)I. Working2. Non-working

Number of cases

URBANSOUTH

100.0038.6519.6941.67

100.0016.9945.9937.0299.8318.2327.2235.2019.1999.9810.9089.09

10078

RURALSOUTH

100.0086.729.903.38

100.0045.92

7.9946.0899.9341.4540.0116.102.37

100.002.98

97.02

5566

URBANNORTHEAST

100.0072.4312.8914.68

100.0011.5427.9760.4899.8741.4926.5618.6813.1399.89

9.0590.84

4454

RURALNORTHEAST

100.0099.140.570.29

100.0016.243.62

80.1499.9275.9920.19

3.080.65

99.996.46

93.53

7231

Source.- Calculated from 1 % sample tape of 1970 demographic census.(*) There were 17 women in Urban South, 4 in Rural South, 6 in Urban Northeast

and 6 in Rural Northeast, whose educational level was not reported.(**) There were 2 women in Urban South, 5 in Urban Northeast and 1 in Rural

Northeast, whose work situation was not reported.

These socio-economic and demographic contrasts are reflected inregional differentials of child mortality. The proportion of children whohave died (PCD) was 58.14 per thousand for the Urban South, 67.98 forthe Rural South, 116.97 for the Urban Northeast and 109.31 for the RuralNortheast. It is interesting to note that even though the Rural Northeasthad worse sanitary and economic conditions than the Urban Northeast,its child mortality levels were lower. In order to understand thesedifferentials more fully, it is useful to examine mortality differentials bycategories of the variables, in conjunction with information from othersources.

Table 2 presents the results in terms of adjusted values of PCD,obtained through MCA, for each category of the independent variables orfactors. Table 3 shows the relative difference between the values ofadjusted PCD of the lowest category and the highest category of thevariables. (Some of the categories of the variables are not rankable fromlow to high, but this ordering was maintained according to the expecteddirection of the differentials). These results, as well as those obtained by

152

TABLE 2. - Proportion of children who have died (per thousand) adjustedfor factors and covariates, by context. Brazil 1970

FACTORS

A) Water and sewerageof household

1. Both inadequate2. One inadequate3. Both adequate

B) Mother's previous residence1. Rural2. Urban3. Non migrant

C) Mother's education1. No education2. Incomplete elementary3. Complete elementary4. More than elementary

D) Mother's work situation1. Working2. Non-working

Grand meanNumber of cases

URBANSOUTH

70.8957.7046.52

62.0762.5550.85

83.1466.3548.8240.31

64.9457.31

58.1410078

RURALSOUTH

69.0559.0066.89

67.6674.3167.20

87.7859.6542.6436.47

72.2467.85

67.985566

URBANNORTHEAST

127.9292.5984.35

117.60120.20115.36

140.97116.43101.3165.31

123.81116.41

116.974454

RURALNORTHEAST

109.81(*)(*)

126.92116.54105.50

111.43101.07111.99100.13

112.87109.06

109.317231

Source: Calculated from 1 % sample tape of 1970 demographic census.(*) Less than 45 cases.

TABLE 3. - Relative difference of the adjusted values ofpcd, between the lowestand highest categories of the variables, by context.

Brazil 1970(90)

FACTORS

A) Water and sewerageof household

Both adequate/both inadequateB) Mother's previous residence

Non migrant/ruralC) Mother's education

More than elementary/no educationD) Mother's work situation

Non working/working

URBANSOUTH

-34.38

-18.08

-51.52

-11.75

RURALSOUTH

- 3.13

- 0.68

-58.45

- 6.05

URBANNORTHEAST

- 34.08

- 1.90

-53.67

- 5.98

RURALNORTHEAST

(*)

-16.88

-10.14

- 3.38

Source: Calculated from 1 % sample tape of 1970 demographic census.(') Not calculated because of the small numbers of cases in "both adequate" (N = 21).

153

an additive model of analysis of variance that indicates the statisticalsignificance of the factors, can now be described for each of the fourcontexts.

Urban South. In the Urban South, with the exception of themother's work situation, all the variables were statistically significant inexplaining the variance of PCD, with values close to 0.000 for probability(p) associated with the f ratio in the analysis of variance. The values ofPCD associated with different categories of the variables reveal theirrespective impacts in the reduction of the child mortality. For thehousehold's water and sewerage, PCD varied from 70.89 per thousand for"both adequate" to 46.52 for "both inadequate". When both installationswere adequate, the average value of PCD was 34.38 % less than whenboth were inadequate. More striking was the difference between values ofPCD for mothers with at feast some secondary education and those withno education at all. The PCD value was 51.5296 less in favor of womenwith higher educational levels. Even though migratory status of themother had been shown to be significant by the analysis of variance, itsinfluence was not as marked as with the other variables. The differencebetween non migrant women and those coming from rural areas was18.08% less in favor of non-migrants. The PCD values for non-workingmothers were 11.75% less than for working mothers.

Rural South. In the Rural South, the only significant variablethat was revealed by the analysis of variance was mother's education(p - 0.000). The PCD value for women with some secondary educationwas 36.47, while for women with no education it was 87.78, a differenceof 58.45% less for the more educated women. The other variablesshowed relative differences between the highest and lowest categories ofless than 6.10%.

Urban Northeast. For the Urban Northeast, there were twosignificant factors, water and sewerage of the household (p = 0.000) andmother's education (p = 0.000). For mothers living in households withadequate sanitary installations, the PCD value was 84.35, while for thoseliving in households with both sanitary installations inadequate the valuewas 127.9. The relative difference was 34.08%. The difference betweenthe most educated women and those with no education was 53.67 %, withPCD values of 65.31 and 140.97, respectively.

Rural Northeast. Up to this point we have seen a pattern wheresanitary conditions of the household and education of the mother areimportant factors in explaining variations in child mortality. This patternis not maintained, however, in the Rural Northeast. The only variablediscernible in the analysis of variance was the place of mother's previousresidence (p = 0.003). The non-migrant women had an average PCDvalue 16.88 % less than migrants of rural origin. The household's sanitaryinstallations, although presenting a high differential among its categories.

154

did not have much significance (p = 0.106) because of the small numberof mothers that lived in households with any installation that could beconsidered adequate.

For the urban areas of the South and of the Northeast, the mostimportant variables were mother's education and water and sewerage ofthe household. The PCD values for the most educated women were on theorder of 50 96 less than those for women with no education. For womenwho lived in households with adequate water and sewerage installations,PCD rates were about 35% less than for those with inadequateinstallations. One can note, however, that in the Northeast women whosehouseholds had adequate sanitary installations had a value of PCD (84.35)higher than those in Southern households with inadequate sanitaryconditions (70.89). Also, those women who had at least some secondaryeducation in the Northeast had a value of PCD (65.31) only slightly lowerthan those who had not completed elementary school (66.35). Also, forwomen with complete elementary schooling in the Northeast, the PCDvalues were 100.3 96 higher that the corresponding women in the South.

These contextual differences among the socio-economic variablespersisted in spite of controls for the other factors and covariates. The datathus indicate that there were other conditions that contributed to higherlevels of mortality in the Northeast, conditions that would evencompensate for advantages of the more privileged economic strata. Theseconditions could include lack of adequate medical resources and anenvironment favoring dissemination of diseases. Published data (IBGE1979a, IBGE 1979b) are indicative of such disadvantages in the Northeast.Its hospital/population ratio was only 0.39 per 10,000 inhabitants, whilein the South this ratio was 0.61. There were 467 persons for each hospitalbed in the Northeast and three doctors for each 10,000 inhabitants. In theSouth these ratios were 189 and 11/10,000. The proportion of householdswhere there were more than four persons for each bedroom was 9.36 %in urban areas of the Northeast, while in the South the comparable figurewas 6.82 %. The proportion of Northeastern households with mud, wattleor thatch walls and earth floors was 10.7296, while in the South it wasonly 0.66 96. These are only some of the indicators of the overall lack ofadequate sanitary infrastructure, of preventive and curative medicalresources, and worst, of the overall level of poverty that predominates inthe urban areas of Northeast.

We can now compare urban and rural areas in the same region. Inthe South, mother's education almost cancelled rural/urban differentials.In the rural areas, mothers whith some education tended to have evenlower child mortality than their counterparts in urban areas. The childmortality for the categories of the other variables always showed atendency to higher values in the rural areas.

In the Northeast, child mortality in urban areas was, with someexceptions, higher than in rural areas. Among those exceptions, rural

155

previous residence of the mother was a factor of higher mortality in therural areas. Mothers who had at least completed elementary school hadlower child mortality in urban areas. This suggests that environmentalhazards in urban areas are greater than in the rural areas, and that only themost privileged families can overcome them.

The work situation of the mother was not an important variable inany of the contexts under study. Adjusted values of PCD indicated thatthere was a tendency towards higher child mortality among workingmothers, inverting the tendency to lower rates for the entire South and forthe Urban Northeast, when compared to values prior to adjustment forfactors and covariates. Tables 4 and 5 show values of PCD prior toadjustment and relative differences between adjusted values and non-adjusted values. For the three contexts just mentioned, the adjusted valuesfor working women were up to 28 % higher than non-adjusted values,correcting the underestimation of child mortality for those women. Thismatter certainly deserves further study.

Still examining Tables 4 and 5, one can see that there is a tendencyto underestimate PCD values of the more educated women when they are

TABLE 4. - Proportion of children who have died (per thousand)without adjustment for factors and covariates, by context. Brazil 1970.

FACTORS

A) Water and sewerageof household

I. Both inadequate2. One inadequate3. Both adequate

B) Mother's previous residence1. Rural2. Urban3. Non migrant

C) Mother's education1. No education2. Incomplete elementary3. Complete elementary4. More than elementary

D) Mother's work situation1. Working2. Non-working

Grand meanNumber of cases

URBANSOUTH

77.0058.0040.71

73.4161.1047.45

88.6569.1946.8034.81

55.8258.44

58.1410078

RURALSOUTH

70.1553.3255.15

71.1569.5464.55

87.8659.5542.3239.14

57.4068.31

67.985566

URBANNORTHEAST

134.9782.3858.52

130.13110.04117.66

147.03119.6996.6146.67

96.37119.17

116.974454

RURALNORTHEAST

109.79(*)C)

127.13113.41105.51

111.80100.34109.9087.31

112.54109.08

109.317231

Source: Calculated from(*) Less than 45 cases

sample tape of 1970 demographic census.

156

TABLE 5. - Relative differences ( % ) between adjusted and non adjusted values of PCD(adjusted/non adjusted), by context. Brazil-1970.

FACTORS

A) Water and sewerageofhousehod1. Both inadequate2. One inadequate3. Both adequate

B) Mother's previous residence1. Rural2. Urban3. Non migrant

C) Mother's education1. No education2. Incomplete elementary3. Complete elementary4. More than elementary

D) Mother's work situation1. Working2. Non-working

URBANSOUTH

- 7.94- 0.52

14.27

-15.452.377.17

- 6.22- 4.10

4.3215.80

16.34- 1.93

RURALSOUTH

-1.5710.6521.29

-4.916.864.11

-0.090.170.76

-6.97

25.82-0.67

URBANNORTHEAST

-5.2212.3944.14

-9.638.88

-1.95

- 1.44-2.72• 4.8639.94

28.47-2.32

RURALNORTHEAST

0.10(*)(*)

-0.172.76

-0.01

-0.330.731.90

14.68

0.29-0.02

Source: Calculated from(*) Less than 45 cases

sample tape of 1970 demographic census.

not adjusted for other variables, especially for women who lived in urbanareas, as well as for women that lived in households with adequatesanitary installations. Another bias that could occur when non-adjustedvalues are used is overestimation of child mortality of mothers withprevious residence in rural areas.

CONCLUSIONS

It can be seen in the foregoing analysis that the effects of thevariables used as factors are far from uniform from one context toanother. Within different regions of the same country and within differentsettings of the same region, the effects vary considerably with regard totheir statistical significance and their magnitude. Of course, some of thisvariation could be due to problems of sampling, quality of the data,definition of the dependent and independent variables, internal composi-tion of open-ended categories and other technical problems. There isreason to believe, however, that there is more to it than this, and that

157

relationships between variables should be interpreted, not as universal,but in relation to the specific historical circumstances in which they areobserved.

One of the most interesting variables to discuss in this regard ismother's education, which has attracted considerable attention as a factorwhich seems to be most important in explaining statistical variation inchild mortality rates around the world, as well as having an apparentlystrong causal relationship. This interpretation is quite plausible. It haspolicy implications which are almost universally acceptable. Unfortuna-tely, it may not be entirely correct, at least in all the diverse socio-economic situations which exist around the world, or even within acountry like Brazil.

Education does make a considerable difference in Southern Braziland in the urban areas of the Northeast, where, as was seen above,women with at least one year of secondary education have child mortalityrates less than half of those for women with no education. It is quitereasonable to believe that women with at least some secondary educationknow more about proper nutrition, dangers of contamination from wateror other unhealthy aspects of their habitat, vaccination and otherpreventive measures, as well as what sort of attention to give or to seekonce children become ill. They almost certainly know more about what isneeded and how to get it than do their less educated counterparts. Indeed,the fact that health-related goods and services are not at all evenlydistributed in Brazil makes such knowledge more important than it mightbe where societal mechanisms provide health protection for allindividuals, whether or not they happen to know much about preventingand curing disease.

At the same time, however, suspicions arise that such knowledgealone is not the key factor. Where socio-economic conditions are worst, inthe Rural Northeast, mother's education has very little effect on childmortality, rates of which are uniformly high for all socio-economiccategories. In this case, the explanation may be that it does not do muchgood to know what is best if the risks to health are severe and the meansto deal with them are simply not available. The question of materialavailability should not be restricted to the Rural Northeast, but can berelated back to the contexts in which mother's education does make aconsiderable difference. There, the question is not one of whether or notphysical means exist in the area, but rather to whether or not thepopulation at large has access to them. It would seem, in the Braziliansituation, that one's general socio-economic position is crucial indetermining his access to health services and infrastructure. Since only afew socio-economic variables were included as "factors" in the presentanalysis, little can be said about the relative importance of education ascompared to variables such as income and occupation. It is quite clear,nevertheless, that education does not simply indicate how much time was

158

spent in classrooms, but also provides information on the way in whichindividuals are inserted in the social and economic structure. Husbands'and wives' education are closely associated. Parents with more educationare much more likely to come from families of better position and tomuster the material conditions to control their environment and protectthemselves and their children against disease and death, given that suchprotection is possible in the context in which they live. Thus, in additionto knowledge, the education variable may reflect, to a considerabledegree, levels of wealth and power, either because it tends to coincidewith these aspects of socio-economic position, or because in contempo-rary society education is increasingly important in determining one'sincome.

With regard to water and sewerage, it can be seen that they do havean impact, especially in urban areas, where high population densityincreases the chances of contamination. Independently of the othervariables, there was considerably lower mortality for households withadequate installations in both the South and the Northeast. However, thereduction was only about half of that associated with higher levels ofeducation. If education is taken as an indicator of position in the socialstructure rather than of more specific knowledge of how to protect thehealth of one's children, this could mean that expansion of water supplyand sewerage networks would need to be accompanied by more profoundstructural change.

Although migration has been considered responsible for a good partof the high mortality in cities, the census data analyzed here show that ithad a minor impact on mortality differentials. What matters more thanmigratory status is insertion in the social structure. The only context inwhich migration had a clear effect on mortality rates was the RuralNortheast. Since this is an area of extensive net out-migration, theindividuals who migrate to it may be those in the most unfortunatecircumstances, and the death of children may itself be an indication of theseverity of the problems which led them to undertake such migration.

The mother's work situation did not have much effect in any of thecontexts. This may be due to the fact that the variable itself is quiteambiguous. On the one hand, mother's work may indicate upward socialmobility and higher family income, as well as fewer children, all of whichfavor lower child mortality. On the other hand, mothers may be forced towork in order to supplement insufficient family income, or even toprovide the sole or principal source of family income. Their work mayresult in less care for their children than they would otherwise provide.Much more needs to be known about particular contexts beforemeaningful interpretations can be made of the effect of mothers' work.

In sum, while enormous tasks remain in the Northeast in terms ofproviding better control of the environment, adequate housing andnutrition, public health facilities and preventive and curative medical care,

159

there is also much to be done in the South. Even in the more developedareas, a considerable part of the population does not have access to themeans needed to avoid disease and death of young children, eitherthrough prevention - better housing, water supply, sewerage, nutrition,vaccination, etc. - or through curative medicine - health posts, clinics,nurses, doctors, hospitals. In defining what needs to be done and what thepriorities should be, great care needs to be taken in using conventionalapproaches and demographic data. These may suggest that technicalsolutions are indicated - sanitary engineering, more schooling, healtheducation, etc. - while persistent high rates of child mortality are moredeeply rooted in the social and economic structure.

REFERENCES

BEHM H., 1980. - Socioeconomic determinants of mortality in Latin America.Proceedings of the Meeting on Socioeconomic Determinants and Consequences ofMortality, Mexico City, June, 1979. Geneva: World Health Organization.

CALDWELL J.C.. 1980. - Education as a factor in mortality decline: an examination ofNigerian data. Proceedings of the Meeting on Socioeconomic Determinants andConsequences of Mortality, Mexico City. June, 1979. Geneva: World HealthOrganization.

CARVALHO J.A.M. and SAWYER, D. 0.. 1978. - Os diferenciáis de mortalidade no Brasil.Anais do I Encontró de Estudos Populacionais, Campos do Jordäo, AssociaçàoBrasileira de Estudos Populacionais, p. 231-259.

CARVALHO J.A.M. and WOOD. C. H.. 1978. - Mortality, income distribution and rural-urban residence in Brazil. Population and Development Review 4: 405-20.

F.I.B.G.E.. 1977. - Geografía do Brasil. 5 vol. Rio de Janeiro : IBGE.F.i.B.G.E. 1979a. - Indicadores Sociais : Retatorio 1979. Rio de Janeiro : IBGE.F.I.B.G.E.. 1979b. - Perfil Estatislico de Criancas e Mâes no Brasil. Rio de Janeiro : IBGE,

UNICEF.MERRICK T.W., 1981. - The impact of access to piped water on infant mortality in urban

Brazil, 1970 to 1976. Contributed paper for session on Determinants of Late Foetal.Infant and Child Mortality Changes. XIX General Conference of the InternationalUnion for the Scientific Study of Population, Manila. December, 1981.

SAWYER D.O.. 1980. - Mortality-Fertility Relationships Through Historical Socio-Economic Changes: The Case Of Sao Paulo, Brazil. Doctoral Dissertation presentedto Harvard University.

SAWYER D.O.. 1981. - The effect of urbanization and industrialization on mortality indeveloping countries: the case of Brazil. Proceedings of the XIX General Conferenceof the International Union for the Scientific Study of Population, Manila. December.1981.

SIMÖES C.C.S.. 1979. - O quadro da mortalidade por classes de renda : un esludo dediferenciáis ñas regiöes metropolitanas (núcleo e periferia). Master Dissertationpresented to COPPE. Universidade Federal do Rio de Janeiro.

160

SCHULTZ T.P., 1980. - Interpretation of relations among mortality, economics of thehousehold, and the health environment. Proceedings of the Meeting onSocioeconomic Determinants and Consequences of Mortality, Mexico City, June,1979. Geneva: World Health Organization.

TRUSSELL J. and PRESTON S., 1981. - Estimating the covariates of childhood mortalityfrom retrospective reports of mothers. Paper presented at Annual Meeting of thePopulation Association of America, Washington, D.C.

VETTER D.M. and SIMÔES C.C.S., 1981. - « Acesso a infra-estrutura de saneamentobásico e mortalidade ». Separata da Revista Brasileira de Estatistica. 42 (165):17-35.

LA MORTALITEAUX JEUNES AGES :

UN ESSAI D'APPROCHEEXPLICATIVE

INTERDISCIPLINAIRE

par

Département de Démographie(Université Catholique de Louvain, Belgique)

Unité d'Epidémiologie(Université Catholique de Louvain, Belgique)

Unité de Nutrition(Institut de Médecine Tropicale, Anvers, Belgique)

Unité de Santé Publique(Institut de Médecine Tropicale, Anvers, Belgique)

1. INTRODUCTION

Les facteurs actuellement envisagés en matière de déterminants etd'explication de la mortalité des enfants ne sont pas seulement nombreux,mais relèvent aussi de disciplines différentes. Le résultat en est unensemble de démarches explicatives partielles et disparates, et sommetoute assez incohérent. Le but de notre projet et de ce document est deprocéder à une réflexion pluridisciplinaire en la matière et de proposerune méthode de travail que nous croyons prometteuse. On ne trouvera icini données, ni analyse sophistiquée, ni modèle raffiné et testé. Maisreprenons l'historique de notre projet avant d'en présenter les objectifs etde préciser le plan de ce document.

C'est lors d'un colloque organisé par le département de démographiede l'Université de Louvain, en 1979, sur le thème de la mortalité desenfants dans le Tiers-Monde(1) qu'a surgi la nécessité d'une collaboration

(1) Département de Démographie, Chaire Quetelet 1979, La mortalité des enfantsdans le Tiers-Monde : orientations et méthodes de recherche, Ordina Editions. Liège. 1980.226 pages.

162

entre diverses disciplines pour une approche causale « décloisonnée » etplus complète- du phénomène. C'est ainsi que des nutritionnistes, desépidémiologues, des spécialistes de santé publique, des sociologues et desdémographes ont tenté un triple pari : constituer un groupe de recherche,trouver un financement, et travailler de façon relativement continue à ladétermination des causes médicales, socio-économiques et culturelles de lamortalité aux jeunes âges. C'est un bilan provisoire de ces travaux quenous nous proposons de présenter ici(2).

L'objectif général du projet est de nature méthodologique, tout aumoins dans une première étape. Il existe un grand nombre d'étudesdémographiques ou médicales sur les causes de la mortalité des enfants,ou sur les liaisons entre cette mortalité et la malnutrition, mais bien peuvisent à dégager l'enchaînement des facteurs individuels, sociaux,économiques et culturels, en d'autres termes les différents mécanismes quifont que tel groupe est plus frappé qu'un autre, ou que dans un groupe,certains enfants sont plus gravement touchés que d'autres.

Notre première tâche a été de construire un cadre analytiquedégageant toutes les variables pouvant intervenir. Comme nous le verronsultérieurement, il s'agit d'un cadre de référence global qui, partant dudécès, progresse jusqu'au système social et politique. Outre ses avantagesdidactiques, ce type de démarche a fourni un langage commun auxmembres des diverses disciplines concernées - médecine, démographie,nutrition, sociologie - et a permis de repérer le niveau d'interventionspécifique à chacune d'elles. De plus, un tel cadre analytique devraitfaciliter la construction de modèles de natures différentes, d'explication oud'intervention, en rationalisant notamment le choix de variables adaptéesà chaque situation concrète (données disponibles, type de collecte possible,etc.). Enfin nous essaierons de montrer les intérêts et les limites de notreapproche.

En définitive, nous présenterons :- la démarche générale ;- le cadre analytique ;- un exemple d'utilisation de ce cadre pour la modélisation ;- les intérêts, limites et problèmes de l'approche envisagée.

2. DEMARCHE GENERALE

Les recherches sur la mortalité des enfants sont essentiellementeffectuées dans le cadre de trois disciplines : la démographie, la médecineet l'épidémiologie. Mais chacune d'elles aborde le problème de manièredifférente.

(2) Ce texte a été écrit vers la mi-l 981. mais les travaux ont depuis lors progressé.

163

L'objet de la démographie est la mesure la plus fine et la plus exactepossible du phénomène : sa méthode consiste à rapporter les décèseffectivement survenus à la seule population susceptible de les subir, et àmesurer les variations de l'incidence des décès selon les milieux sociaux,économiques ou écologiques. De ce fait, la démographie est bien plusanalytique et descriptive que synthétique et explicative ; et comme elleenvisage la mortalité comme un phénomène touchant des populations oudes collectivités, son approche est fondamentalement statistique.

Les méthodes de recherches en médecine - l'étude des cas cliniqueset l'expérimentation - répondent à sa préoccupation centrale, qui est desoigner des individus. La médecine se penche avant tout sur lesmécanismes biologiques et physiologiques qui conduisent un individu dela santé à la maladie et de la maladie à la mort. Elle accorde certes unelarge place aux interactions entre l'hôte et son environnement immédiat,mais sa vocation l'amène à étudier beaucoup moins les phénomènescollectifs qui font l'objet de l'épidémiologie et de la démographie.

L'épidémiologie utilise une méthodologie beaucoup plus proche decelle de la démographie. Elle étudie en effet des groupes d'individus et voitles problèmes de santé comme des phénomènes collectifs. La diminutionde l'importance relative des maladies infectieuses dans les paysindustrialisés et la responsabilité croissante de l'épidémiologie dansl'élaboration et l'évaluation des programmes de santé publique l'ontconduite à une approche multicausale des phénomènes morbides. Elle acependant tendance à se limiter aux causes relativement immédiates de cesderniers, et à considérer les variables sociales et économiques, au moinsen pratique, plus comme des facteurs d'identification de risque quecomme des causes véritables.

Chacune de ces trois disciplines : démographie, médecine, épidémio-logie, est donc régie par un cadre de référence qui, le plus généralement,est implicite. De plus, chacune d'elles ne couvre qu'un (ou quelques-unsseulement) des niveaux d'explication du phénomène étudié, ici lamortalité chez les jeunes enfants. C'est parce que le champ d'explicationde chacune d'elles est limité, et parce que leurs cadres de référencerespectifs ne sont pas assez explicites, qu'il est difficile d'effectuer unerecherche réalisant l'intégration des explications partielles en un tableaucohérent et global.

Entre le niveau biologique et physiologique exploré par la médecine,et le niveau collectif où se situent la démographie et l'épidémiologie,s'étend une vaste zone d'ombre, une sorte de « boîte noire » dans laquelleagissent différents mécanismes. Parmi ceux-ci, figure toute la gamme descomportements dont on sait l'importance, mais dont l'impact est encorelargement inconnu et relativement peu étudié. S'il est difficile de rendretransparente la « boîte noire » en conceptualisant les interactions desdifférents niveaux, il ne l'est pas moins d'établir une distinction entrecause, facteur de risque, déterminant, ou facteur associé. Par exemple.

164

une des causes de décès infantiles liés à la grossesse et à l'accouchement estla prématurité ; mais la prématurité doit-elle être comprise comme unecause proprement dite, ou comme un facteur de risque ? Force est dereconnaître que cette distinction est bien rarement opérée dans lalittérature. Or, il est essentiel d'établir une distinction entre les loisscientifiques (ou à défaut les généralisations empiriques) d'une part et lesénoncés de leurs conditions d'application de l'autre.

La prise de conscience de ces insuffisances nous a conduit à adopterune démarche qui, pour nous, est préalable à toute opération de terrain,ou même à toute tentative de construction d'un modèle explicatif oupragmatique de la mortalité des enfants dans une situation particulière.Cette démarche consiste à construire un cadre de référence qui recouvreles différents niveaux d'explication du phénomène, en partant des décès eten progressant de proche en proche, depuis les causes médicales jusqu'auxdéterminismes socio-économiques globaux, en enchaînant de façoncohérente des événements, des actions et des comportements.

Ce qui est désigné dans ce texte comme le « cadre analytiquegénéral » est le produit provisoire de cette démarche. Il ne doit pas êtrecompris comme un modèle d'explication utile dans un contexe précis,mais comme un « analytical framework » à partir duquel des modèlesparticuliers peuvent être élaborés. On peut le comparer à une carteroutière où toutes les routes possibles sont représentées : sur cette basechacun peut établir l'itinéraire de son choix en tenant compte de sesobjectifs, de ses ressources et de ses contraintes. Dans cette constructiondeductive, l'accent pourra être mis, selon les besoins, sur telle hypothèseou tel fondement important.

Les avantages de cette démarche sont donc notamment d'obliger àrendre explicites tous les présupposés du cadre analytique, de fournir unlangage commun qui permet de retirer le maximum de profit de lacontribution de chaque discipline, et aussi de permettre la construction demodèles particuliers à partir du cadre général.

3. LE CADRE ANALYTIQUE

Méthodologie

Le cadre analytique décrit la chaîne d'événements et de circonstancesmenant au décès. Le schéma se construit en partant des décès. Ondécompose tout d'abord la mortalité en ses « causes immédiates » les plusfréquentes : maladies transmissibles, malnutrition (comme cause directe),causes liées à la grossesse et à l'accouchement, accidents et « autrescauses ». Cette première étape constitue une décomposition en sommes(figure 1).

165

Matrans

ladies

Tiissibles

1 Décès de 0 à 5 ans

Causes liées âla grossesse età 1'accouchement

1 Accidents Malnutrition | Autrescauses

Figure IConstruction du cadre analytique : première décomposition

Ces décès par « causes immédiates » dépendent d'une part de lafréquence de ces maladies, l'incidence qui en indique le nombre dans lapopulation, et d'autre part de la fréquence des décès parmi les malades oulétalité. Ces deux notions, rarement utilisées en démographie, traduisent,en médecine, deux voies d'intervention si l'on veut diminuer la mortalité :soit supprimer la maladie, soit éviter le décès lorsqu'on a contracté lamaladie. Dans cette deuxième décomposition, c'est le produit del'incidence et de la létalité qui donne le taux de mortalité (figure 2).

Les étapes ultérieures de la construction du cadre analytiquedétaillent, le plus précisément possible, les chaînes d'événements quiinfluencent l'incidence (figure 3) ou la létalité (figure 3').

Une partie seulement de la démarche est illustrée ici, celle quidéveloppe de proche en proche des relations explicatives influençant letraitement. Les diverses variables retenues sont décomposées, soit ensomme, soit en produit. A côté de ces deux manières de décomposer unevariable, il existe un autre type de relation entre les cases d'un niveaudonné et celles du niveau immédiatement inférieur, que nous appelons« relation d'influence ». Dans bon nombre de cas, cette relation est unehypothèse causale stricte : par exemple, l'état nutritionnel influencedirectement la résistance de l'hôte, la perception de la maladie est cause dela décision d'utiliser le service de santé. Mais, pour nous en tenir à ce

Décès de 0 à 5 ans

1

Maladies transmissibles

i

Incidence Létalité

Figure 2

Construction du cadre analytique : deuxième décomposition

166

l - l

l - l

OJ4-1

1-4

3

On

* |CCJ

4.))-lOE

V•a

QJ

'ECO

ccd

4-1

a

•anji-iraE

licë

1

i r j

4-1

|

QJT3

1 T

aux

ad

ié

I

ef - i

<u

Dour

t

Q)

j a

wto

*BO)

Crt) j4-1

167

0)uc(den

(0

S

at

<oJ=

-a

101 H

3CD

J JU

COr-t

01GCOen| J

eua

JJ

c01Eai

t d

uH

ai Gu -^•<-t 0> ent - i

ai ai

"tr3aiCOeo

'u101

J_ai

fa

JJT"

C

£ai 3a"

i c i

• a

LOIJ J

! _ !

n)

eu3t r

iO>

ct-iuC

•f-(

C

i-i

3

_

aiT3euCai

eis

¿i

toaiJJc

rai

ont

u

CO01

erocoo

••-iXI

euu3j j

>3O

3

eu>

J J

uCl

U-l

14-1

OJ

c0)Eau

«

eneu

T 3

euU-J

O

01TD

01 C

u -H.^ o> enl_l01U)

ai-i

3J Jü3MJJen(dj - i

H

ai

lotJ JdE

ai

H ai>" S .

5 ,°

3?:

O eo

ai13

aiT5CaEaia

ucCJEClJJ

duJ J

01

ucwCO

rdCe0u

onai

en• H

eu

ai

atu

t u

O

i—(

01-a

JJ

3• 3

n

cisi

a

ser-

cnai

iser

nins

en

n

ces

>

gJ J

CJ

ai

«no

en

rce

sou

en01

tau

168

dernier exemple, l'acceptabilité ou l'accessibilité des services de soins sont,au contraire, des conditions d'application (des facteurs conditionnants) dela relation perception-décision. Sans être des « causes » de la décisiond'utilisation, elles vont néanmoins l'influencer.

Observations

a) Les variables d'identification ne se retrouvent pas dans le schémadans la mesure où elles ne « causent » pas ouvertement le décès. Il s'agitpar exemple des caractétristiques des parents (profession, revenu,instruction, etc.), des caractéristiques de l'enfant (âge, sexe, rang denaissance, etc.), des variables culturelles (religion, tradition, etc.), etc.Celles-ci permettent la constitution de groupes humains homogènes face àun des risques étudiés. De plus, elles peuvent être dans certains casutilisées comme variables dans un modèle (ou des modèles) appliqué à unesituation particulière.

b) Les liaisons entre variables situées au même niveau (ou presque aumême niveau) n'ont pas été retenues. Cet écartement des « liaisonshorizontales », qui est temporaire, présente l'avantage d'une simplificationconsidérable tant du discours que de la représentation graphique du cadreanalytique. Dans notre expérience, cet avantage compense très largementles inconvénients, nous y reviendrons au point 5.

c) Des éléments qui ne sont pas repris de façon constante dans lesdiverses disciplines en présence ont été introduits dans le cadre général, àune place bien définie de la chaîne d'événements : c'est le cas des variablesliées à l'utilisation des services de santé, ou encore de celles qui ont traitaux comportements.

d) Plusieurs cases de ce schéma représentent des concepts qui ne sontpas mesurables dans l'état actuel de nos connaissances, ou dans lesconditions habituelles de collecte de données (par exemple la virulenced'un agent infectieux). Elles peuvent ne pas être reprises lors de laconstruction de modèles dérivés du schéma, ou lors de la traduction desdifférentes cases en indicateurs.

e) La poursuite de proche en proche du processus de construction ducadre analytique aboutit, à première vue, à un schéma pyramidal, maiscertains facteurs obéissent à des déterminants communs. Le schéma serétrécit à la base jusqu'au système socio-politique, et sous sa formeachevée, il prendra plus ou moins l'aspect d'un losange.

169

4. UTILISATION DU CADRE ANALYTIQUEPOUR L'ELABORATION D'UN MODELE EXPLICATIFDE LA MORTALITE PAR MALADIES DIARRHEIQUES

Prenons donc comme exemple le cas de la mortalité par maladiesdiarrhéiques, sous-ensemble des maladies transmissibles, dont on connaîtl'importance dans la mortalité des enfants des pays en développement.Nous allons montrer comment, à partir du cadre analytique, on peutrapidement construire a priori un modèle explicatif simple et opérationnel.

Construction du modèle

Partant du cadre analytique précédemment présenté, la figure 4illustre l'itinéraire choisi pour notre problème. Nous avons volontaire-ment, dans cette première étape, repris la plupart des variables du cadre,mais nous avons distingué celles qui sont mesurables (cases en trait pleinsur la figure 4) de celles qui ne le sont guère (cases en pointillés) ou quipeuvent être considérées comme une constante quand on travaille à unmoment donné (modèle statique) dans une région donnée. Les variablesnon mesurables dans une collecte classique de données sont « la virulencede l'agent » et les « facteurs personnels », tandis que « les facteurs propresà l'agent » peuvent être pris comme une constante.

De ce cadre analytique, il ressort que certaines variables intervien-nent plusieurs fois : c'est le cas de l'état nutritionnel des enfants qui a unimpact direct aussi bien sur l'incidence que sur la létalité ; c'est aussi le casde la durée d'allaitement qui a un impact direct sur l'incidence comme ilen a évidemment un sur l'état nutritionnel. Nous avons volontairementpeu développé la partie traitement, qui nécessite à notre avis une collectede données tout à fait spécifique et qui devrait faire l'objet d'uneproblématique d'action plutôt que d'une problématique d'explication.

Le modèle qui en résulte est présenté en figure 5. Les relations entreles différentes variables du modèle sont indiquées par une flèche et par unsigne positif ou négatif. La direction de la flèche indique la « variableexpliquée », tandis que le signe positif signifie qu'un accroissement de la« variable explicative » entraîne un accroissement de la « variableexpliquée », et vice-versa pour le signe négatif. Ce modèle pourrait êtretesté par un traitement statistique du type path-analysis.

Choix des indicateurs

L'étape suivante consiste à déterminer les indicateurs qui vontpermettre de recueillir les données pour les variables retenues dans le

170

iê i

01

- t

o- .. 1

pro

igen

!

s™ iS 1

41CJCniLJU>

I

1•H

^ i-(

U V3 C

c cOXJ «H

171

ci01J->

(0

(O

0 )10)

J-l

3Q

t-(

0)

c014J

(0B

+,—I

EV4J

cM

ique

CO

c10)

oc0 )

c• 1-t

W0)uc01T3

•H

1 In

c

[su

o• H4 J

to

• •H

autoí>

172

modèle. Un exemple est présenté dans le tableau suivant, pour quelquesvariables.

VARIABLES DONNEES A COLLECTER

Y Mortalité

XI Incidence

X2 Létalité

X3 Hygiène du milieuEauHygiène alimentaire

X4 VaccinationsX5 Etat nutritionnel

X6 Action médicale curativeX7 Action familiale

X8 Durée de l'allaitementetc.

Nombre d'enfants morts de diarrhée (0 à 2 ans)

Nombre total d'enfants de 0 à 2 ans

Nombre de malades

Nombre d'enfants

Nombre de décédés

Nombre de maladesIndice bactériologique de l'eauPourcentage de population qui a accès à l'eau potableLatrines, EgoutsElimination d'ordures% enfants vaccinésTaille - Poids - Pli cutané -Symptômes de malnutritionRéhydratation I.V. et médicaments adéquatsRéhydratation per osRégime adéquatNombre de mois d'allaitement seul

5. INTERETS, ACQUIS ET PROBLEMES

Le travail jusqu'alors réalisé présente des caractéristiques qui lerendent, sinon original, du moins nouveau par rapport à ce que chacund'entre nous faisait jusqu'ici. L'approche inter-disciplinaire résulte bienplus des apports mutuels et des interactions entre les disciplines que d'unesimple juxtaposition de ces dernières. L'étude a débordé largement lescadres de la démographie, de la nutrition et de l'épidémiologie, et il est trèsprobable qu'aucun des groupes n'aurait pu mener seul la recherche.

Cette expérience confirme et renforce celle des nutritionnistes, quiavaient déjà utilisé une approche similaire. Leur association avec deséconomistes, des agronomes et des médecins a été bénéfique pour choisirles variables à récolter en vue de poser un diagnostic nutritionnel pour unpays donné, et économiser ainsi temps et ressources.

Parmi les caractéristiques de notre recherche qui méritent d'êtresoulignées, se trouve en premier lieu l'approfondissement de l'analyse.

173

Comme on l'a vu, dans la formulation des chaînes d'événements reliant lavariable dépendante à ses causes éloignées telles que les facteurs sociaux etéconomiques, le souci a été constant d'éviter la rupture de continuité.Notre groupe s'était en effet fixé pour règle de ne laisser aucun chaînonintermédiaire ni aucun facteur causal ou associé inexploré, de décomposerau maximum les chaînes en tous leurs maillons, de rendre explicites, dansla mesure du possible, toutes les propositions, hypothèses et définitions.Le processus a été assez lent, mais avant d'adapter le cadre analytique àune question concrète, nous voulions connaître globalement le problèmeet décomposer les différentes chaînes causales dans le détail.

La méthodologie utilisée avait été inaugurée avec succès par lesnutritionnistes : notre groupe l'a reprise pour l'étude des causes de lamortalité des enfants en s'efforçant d'élargir la problématique et de lapréciser. Rappelons quelques caractéristiques de cette approche : le refusdes liaisons horizontales, la distinction entre les relations d'influence et dedécomposition, et surtout le mode de construction « descendante » ducadre analytique (de la variable dépendante vers les causes de plus en pluséloignées). Cette manière de procéder permet d'arrêter l'analyse àn'importe quel niveau choisi : l'expression du phénomène sera incom-plète, mais cohérente.

Les acquis

Quels sont les acquis provisoires du travail réalisé ? Ils sont assezvariés, et certains d'entre eux demandent confirmation. Citons en premierlieu la découverte de nombreuses incohérences, absences de définition ouerreurs de logique relevées non seulement dans notre propre langage,mais aussi dans la littérature. Les malentendus liés à l'utilisationinadéquate des termes de « prématurité » et de « dysmaturité », lesdéfinitions imprécises du « niveau socio-économique » ou de « l'utilisationdes services de santé » en sont quelques exemples. Cette confusion,souvent terminologique mais parfois même conceptuelle, nous a rendusconscients de la nécessité d'une grande rigueur dans les définitions et dansl'emploi des termes. Elle nous a aussi montré les limites des façonsclassiques d'appréhender les causes de la mortalité de la petite enfance, etles faiblesses dans ce domaine de la classification internationale des causesde décès.

Un autre acquis, de fond celui-ci, est l'amélioration certaine de notreconnaissance des mécanismes et des causes de la mortalité et de lamalnutrition, grâce sans doute à la revue de la littérature, grâce surtout àl'apport mutuel des différentes disciplines et aux échanges d'informationsentre les différents membres du groupe, et à l'examen systématiquementcritique des concepts intuitifs : ce fut le cas pour l'utilisation des servicesde santé, ou pour celui des comportements, répétés à plusieurs reprisesdans le cadre analytique. Ces derniers, faut-il le rappeler, sont pertinents à

174

la fois comme causes directes et comme facteurs d'efficacité des mesuresde prévention et de traitement.

Enfin, en plus des acquis méthodologiques énoncés plus haut - etqui devront bien entendu être contrôlés - ce travail a permis de dégagerdès à présent des thèmes de recherche pour l'avenir, des hypothèses àvérifier, des synthèses à faire. La méthodologie appliquée devrait en effetpouvoir s'étendre également à d'autres sujets comme l'étude de lamortalité périnatale, celle de la mortalité des jeunes enfants dans les paysriches, celle de l'utilisation des services de santé, celle aussi de la fécondité,etc.

Les problèmes

Signalons en premier lieu les pièges sémantiques, et parfois logiques :ceux que nous créent l'intitulé des variables, ou leur catégorisation, ouencore l'introduction dans le cadre analytique de certaines cases serapportant à des facteurs non mesurables. Au large éventail des définitionsemployées dans la littérature s'ajoute l'imprécision des termes - etsouvent des concepts - qui servent à exprimer les relations entre cesvariables, tels que « cause », « association », « influence », « détermina-tion », etc. D'autre part, la connaissance des liens qui unissent lesdifférents éléments du cadre analytique est insuffisante ; cela apparaîtclairement lors de la construction de ce dernier : quand ces liens sont-ilscausaux, et quand ne représentent-ils que de simples associations ?

Le fait d'avoir ignoré volontairement les relations horizontales dansla construction du cadre général crée un double problème :1) celui des interactions entre deux variables situées à des niveauxsemblables, comme par exemple, entre la malnutrition et les infections ouencore entre l'utilisation du service de santé et la qualité de ce dernier.Notre réponse actuelle - qui est que le cadre analytique est statique -n'est que partiellement satisfaisante : dès lors que la période d'observationn'est pas instantanée, le modèle n'est plus entièrement statique. Dans le casd'un modèle dynamique, ces relations horizontales ne pourraient plus dutout être mises à l'écart ;2) celui des variables qui paraîtraient devoir intervenir à plusieursendroits du cadre analytique. L'éducation de la mère, par exemple, peutagir au niveau de la perception de la maladie, dans le choix des alimentsde l'enfant, dans la décision de consulter le service de santé. Un termegénérique comme « l'éducation de la mère » recouvre en fait un ensemblede facteurs qui demandent à être précisés, de façon à ne rester pertinentsqu'à un seul endroit du cadre analytique.

Un autre problème provient du passage constant de notionsindividuelles à des notions agrégées (et vice-versa) dans les définitions quenous donnons aux variables. Le cadre analytique contient des cases telles

175

que « incidence », qui est une notion collective, et « quantité ingéréed'aliments » qui est une mesure individuelle. La distinction entre ces deuxtypes de variables est importante, car elle va influencer la collecteultérieure des données, le type d'analyse et son interprétation. Dans l'étatactuel d'avancement du projet, nous n'avons pas résolu la question desavoir s'il y a là un problème de concepts, ou tout simplement unproblème de construction.

La place du facteur temps compte au nombre des difficultésrencontrées. Même si le cadre analytique se rapporte à une périoded'observation brève, il n'est cependant jamais tout à fait statique : il y atoujours un intervalle de temps entre le moment où les facteurs causaux semanifestent et le décès. Certains de ces facteurs, en outre, peuvent semodifier pendant la période considérée. Le passage ultérieur du cadreanalytique général, statique ou partiellement statique, à un modèledynamique, présentera donc des problèmes supplémentaires.

Enfin, il faudrait encore mentionner les nouvelles difficultés quiapparaîtront sans aucun doute lors du passage du cadre analytique à desmodèles particuliers. Jusqu'ici, notre groupe a procédé à une démarcheanalytique poussée, sans préjuger des types de modèles que l'on pourraiten dériver. Or, tôt ou tard, de tels modèles devront être construits, dont lechoix dépendra évidemment de l'usage auquel on les destine : modèles desimulation, modèles « génétiques » (donnant des états successifs avec laprobabilité de chacun d'eux), modèles agrégatifs, etc.

ANNEXELISTE DES PARTICIPANTS AU GROUPE

I.M.T. : Institut de Médecine Tropicale, 25, Kronenburgstraat, 3000 Anvers (Belgique)U.C.L. : - Département de Démographie, Université Catholique de Louvain, 1 Place

Montesquieu, 1348 Louvain-la-Neuve (Belgique)- Faculté de médecine, U.C.L., 1200 Bruxelles

D.D. : Département de DémographieU.N. : Unité de NutritionU.E. : Unité d'EpidémiologieU.R.E.S.P. : Unité de Recherche et d'Enseignement en Santé Publique

Françoise BARTIAUX, sociologue et démographe, D.D. - U.C.L.Yvan BEGHIN, médecin et nutritionniste, U.N. - I.M.T.Irène BORLEE, pédiatre et épidémiologue, U.E. - U.C.L.Paul-Marie BOULANGER, sociologue, D.D. - U.C.L.Godelieve MASUY-STROOBANT, démographe, D.D. - U.C.L.Denis NziTA (janvier à juillet 1981 ), démographe, D.D. - U.C.L.Mpembele SALA-DIAKANDA (1980), démographe, D.D. - U.C.L.Dominique TABUTIN, démographe, D.D. - U.C.L.Marc VANDERVEKEN, médecin et épidémiologue. U.E. - U.C.L.Wim Van LERBERGHE, médecin de Santé Publique, U.R.E.S.P. - I.M.T.Jacques VUYLSTEKE. pédiatre et nutritionniste, U.N. - I.M.T.

176

RESUMEL'explication des niveaux et des variations de la mortalité aux jeunes âges suppose

la collaboration de disciplines différentes (démographie, médecine, épidémiologie,sociologie, etc.). Pour que cette collaboration soit possible et fructueuse, compte tenu desdifférences de points de vue, de vocabulaire et de méthodes entre ces disciplines, il importede construire un cadre analytique et conceptuel qui permette l'intégration de cesdifférentes approches. Le présent article expose la méthode adoptée en Belgique par ungroupe interdisciplinaire de recherches sur la mortalité aux jeunes âges pour construire untel cadre analytique. Elle consiste essentiellement à partir des décès et à remonter parétapes successives vers les causes médicales et épidémiologiques d'abord, vers lesdéterminations économiques, sociales et culturelles ensuite. On montre aussi comment desmodèles particuliers, exhaustifs mais" plus opérationnalisables et plus adaptés à tel ou telproblème concret, peuvent être générés à partir de ce cadre général.

SUMMARYThe explanation of the levels of child mortality and its variations requires a

collaboration between several scientific disciplines (e. g. demography, medicine,epidemiology, sociology, ...) differing by scope, methodology and level of conceptualisa-tion. In order to make such a collaboration possible, it is necessary to integrate them in acommon analytical framework. This paper presents the method used by an interdiscipli-nary research team in Belgium for the construction of such a framework. It consistedmainly in taking as departure the crude death rate, to disentangle the various medical andepidemiological causes of death and for each of them to settle down by successive stagesthe whole path of social, economic, cultural and behavioral causes. The paper shows howit is possible to use such a general framework in building more specific causal models.

SUMMARIES

Infant and Child Mortality in Costa Rica*

This paper examines differentials in childhood mortality in Costa Rica for theapproximate period 1968-1973. Costa Rica is an interesting case study because of theexistence of significant differentials in child mortality in a nation undergoing rapid fertilitydecline along with economic development. The period chosen is related to the use of the1973 census of Costa Rica, which provided data on children ever born and childrensurviving.

Two levels of analysis are used. The first constructs a model of mortality for smallgeographic units (cantons, which are political subdivisions having an average of about20,000 persons each). The dependent variable is the proportion of children dying beforereaching age 2 (2q0) and is constructed using estimated fertility histories of women aged20-34 and the proportion of children surviving for each canton. This variable is thenrelated to several socio-economic, geographic, and demographic independent variables insingle equation and simultaneous equation regression models with cantons as the units ofanalysis. The second level of analysis uses the individual woman as the unit of analysisand relates a mortality index for her (i.e. the ratio of actual to expected child deaths) tocharacteristics of the woman, her family, and the ambient level of infant mortality in hercanton.

The canton model of mortality includes the following independent variables:fertility, level of urbanization, proportion of women with more than five years ofschooling, proportion of dwellings with sanitary facilities, estimated average income perhead, proportion of births with medical attention, and altitude of the canton. Of these, theproportion of births with medical attention clearly has the most powerful influence (aloneexplaining about two thirds of total variation), followed by fertility, altitude, andeducation. Sanitation, income per head, and urbanization do not have significant effects inthe multivariate model, although the zero-order coefficients are strong and significant forall variables. The sign of the urbanization variable changes from negative in the zero-ordercase to positive in the multivariate model, suggesting if anything that cities had highermortality when other factors are taken into account. The model was tried with slightlydiffering specifications and a two stage least squares simultaneous model is estimatedtaking into account of the possibility of joint causality between fertility and childmortality. The results of these exercises do not differ significantly from those presentedabove.

'Differential Infam and Child Mortality In Costa Rica: 1968-1973. By Michael R. Haines(Dept. of Economics, Wayne State University, Detroit, MI, USA, and International PopulationProgram, Cornell University, Ithaca, NY, USA) and Roger C. Avery (International PopulationProgram, Cornell University, Ithaca, NY, USA) [Published in Population Studies. Vol. 36, N° 1(March. 1982), pp. 31-43]

178

The individual level model of mortality uses the ratio of actual to expected childdeaths as the dependent variable and employs weighted least squares estimation. Theweights used are the number of children ever born to each woman, normalized to leavethe total number of cases unchanged. The rationale for weighting is to bring the level ofanalysis closer to that of the child rather than the mother. Three models are presented inthe paper: one having only the characteristics of the woman and her family; the secondhaving, in addition, the level of child mortality in the canton; and the third including thecharacteristics of the woman and her family as well as other characteristics of the cantonbut not its child mortality level. The independent variables include the age and educationof the woman and the sanitary facilities and dwelling quality of the house. All these arefound to have substantial effects on the level of child mortality in the family. The effectsare in'the predicted direction: that is, more educated women in homes of good qualitywith sewers or septic systems had much lower levels of child mortality. (In other modelsincome was used but failed to have significant effect). The inclusion of regional effects onindividual mortality by the inclusion of canton infant mortality rates decreases the size ofthe coefficients of some of the variables, indicating that at least some of their effect is dueto regional characteristics. Sanitation is particularly affected. The coeffecients of housingquality actually increase, however, indicating that regionally it is particula'rly important tohave the wealth that a good quality house indicates.

There are several general conclusions. Differentials in child mortality in Costa Ricawere relatively large during this period, and they can be measured adequately along manydimensions only with census data. Both individual characteristics, education and housingfor example, and community characteristics, such as medical care at childbirth andambient child mortality levels, can have a substantial effect on differentials in childmortality. Finally, income in and of itself does not appear to be very important; it is suchthings as housing and education, some of which are purchased with money directly andsome of which are furnished through the government, which most greatly affect childmortality in Costa Rica.

Infant and Child Mortality in Guatemala*

This paper deals with the change in differential childhood mortality among smallgeographic areas of Guatemala between 1959-64 and 1968-73. The time periods aredictated by use of five percent samples of the 1964 and the 1973 censuses of Guatemala toestimate mortality rates. The paper has two parts, the first dealing with methodology andthe basic estimates, and the second dealing with multivariate analyses of the data for smallgeographic areas.

The first part treats the methodology necessary to estimate mortality from censusdata. The 1973 Guatemala census asked questions on children ever born and childrensurviving and so the standard indirect estimation techniques can be applied. For thiscensus, however, the Preston-Palloni technique of backward projection of the agestructure of surviving children is used. For 1973, the age structure of surviving ownchildren (i.e. children matched to their mothers) is adjusted for non-own children (i.e.children unable to be matched) in order to obtain the proper age distribution to reversesurvive to equal total children ever born. The reverse projection program then searches

'"Differentials in Infant and Child Mortality and their Change Over Time:Guatemala, ¡959-1973. " By Michael R. Haines (Department of Economics, Wayne State University,Detroit, MI, USA and International Population Program, Cornell University, Ithaca, NY, USA), RogerC. Avery (International Population Program, Cornell University, Ithaca, NY, USA), and Michael A.Strong (Population Studies Center, University of Pennsylvania, Philadelphia, PA, USA)

179

iteratively in a one parameter model life table system to obtain a life table that will justproduce the correct equivalence between children ever born and survivors by age.However, the 1964 census only asked a question on children ever born and not onchildren surviving, thus this must be estimated. The procedure used is to estimate anoverall level of mortality most likely to have prevailed in 1959 (roughly the time period towhich the 1964 census would have applied) and take an overall life table at that level.Then, taking the proportion of surviving total children to surviving own (matched)children in the 1973 census and applying it to the 1964 census age distribution ofsurviving own children, the total surviving children are estimated and the life table foundin the same manner as in 1973. We also compare estimates of children ever born tocohorts of women by 1964 using both the 1964 and 1973 census, where the estimates forthe 1973 census were made by own children methods. We found the 1964 census hadsubstantially underestimated children ever born. Further investigation revealed that thisundercount was greater among the less well educated and among Ladinos. Correctionsare applied to account for these differentials in'order to obtain more accurate estimates ofdifferential child mortality. (An implication of this finding is that questions on childrensurviving should be asked in censuses if only to improve the count of children ever born.)-

Several assumptions are necessary to estimate these mortality rates. First, the agedistribution of children who have left the household are assumed to be the same as thosepresent. This paper uses rates estimated ortly for younger women (aged 20-34), for whomthe problem of children absent for reasons other than mortality is not too serious.(Simulations confirm this result). A second assumption is that the extent of differentialunderreporting of children ever born in 1964 can be adequately described for geographicsubregions by the proportions of mothers with more than six years of schooling and theproportions of mothers who were indigenous. Third, it is assumed that the ratio of totalsurviving children to own (matched) children in 1973 is adequate to estimate therelationship in 1964.

A life table was thus estimated for each geographic subregion for 1959-64 and1968-73. The subregions are aggregations of municipios, grouped to create aggregationsof about the same population which are also geographically contiguous and do not crossprovincial boundaries. For Guatemala City, groups of zones are used for the samepurpose. The goal is to reduce sampling errors while still allowing multivariate analysisand retaining homogeneity within subregions.

The results confirm the major differentials for 1968-73 found by Behm and Vargas,but the level of mortality is found to be higher, despite the use of the same data. Ruralareas, indigenous populations, and the less educated in both rural and urban areas hadsignificantly higher than average child mortality rates. In addition, there was significantregional variation with the lowest mortality in Guatemala City. The South region, an areaof substantial export agriculture, had particularly high rates of mortality. The direction ofthe differentials for 1959-64 (from the 1964 census) were substantially the same as thosefor 1968-73, but they were much smaller. In particular, Guatemala City had mortalityrates much closer to the national average than in 1968-73.

The second part of the paper uses the mortality rates estimated for the subregionsand employs OLS regression to relate them to various socio-economic, geographic, anddemographic characteristics of the subregions. For 1968-73. the regressions indicate thatmost of the included variables had significant effects on child mortality. More education, ahigher level of urbanization, better quality housing, a larger proportion of non-indigenouspopulation, and higher income are associated with better child survival, while highfertility, small farms, and export agriculture tend to produce higher child mortality. Theregressions for 1959-64 reveal much less in terms of relationships between a similar set ofindependent variables and child mortality across regions. For both periods, theindependent variables are highly intercorrelated, since they are all affected by thedevelopment process. As a consequence, a number of alternative specifications were tried.The sizes of the relationships in the multivariate models are smaller than the zero-orderrelationships and sometimes insignificant, but generally in the same direction.

180

A major result is that differentials in child mortality for Guatemala widenedbetween 1959-64 and 1968-73. Factors causing the mortality decline during this periodwere unequally spread by geographic region, rural-urban residence, and ethnic andeducational groups. Guatemala City seems to have benefited most from the decline.Commercialization of agriculture also seems to have worsened child mortality in thoseaffected areas relative to the rest of the country.

Health Problems in Perinatal Period and Infancy in a Rural District in Thailand*

The study consisted in the follow-up of all pregnancies and new-borns up to 28days after birth during 15 months in 1977-1978 in Bang Pa-In, a rural district with 36,000inhabitants, 66 km. from Bangkok, the capital of Thailand. The infant death rate was 42per 1,000 births in Bang Pa-In, and 21 per 1,000 births in Bangkok. It is noteworthy that,in the Mekong River basin area, approximately 160 km. from Bangkok, the infant deathrate was 64 per 1,000 births. In Bang Pa-In, approximately 50 % of infant deaths occurredduring the first week of life and 75 % during the 4 first weeks of life.

The distribution of deaths according to cause was basically different in Bang Pa-Inand in the capital, as shown hereafter:

Out of 100 infant deathsCause in Bang Pa-In - in Bangkok

(Ramathibodi Hospital)-

111

44

Most infant deaths in Bang Pa-In were preventable and reducible if maternal andchildcare had been sufficient. It is interesting to observe that home deliveries by traditionalbirth attendants had a perinatal death rate of 21 per 1,000, three times higher than homedelivery by midwives (6 per 1,000). Home self delivery or with the assistance of relativeswas the most risky (more than 60 per 1,000).

The nutritional status of newborns according to age was as follows:

% of normal growthby Thai standards

8L266.669.555.355.734.4

The critical age was during weaning (13 - 18 months of age) when the proteinenergy malnutrition started. It worsened through later ages. During the post neonatalperiod, the leading cause of death was diarrhoea, and the second, pneumonia. Theassociated cause of death by diarrhoea was protein energy malnutrition, which covered50% of them.

InfectionAsphyxia, birth traumaToxaemiaPre-term and complication

31199

12

Below13 -19 -25 -31 -37 -

121824303642

monthsmonthsmonthsmonthsmonthsmonths

"by Pensri Khanjanasthili and Vilai Benchaharn, Ramathibodi Hospital, Bangkok,Thailand.

181

The Effects of Fertility on Fetal, Infant and Child Mortality in Bangladesh*

The reciprocal relationship between fertility and infant-child mortality has been thesubject of numerous publications. Both biological and psycho-social-cultural factors haveassociated higher fertility with increased mortality. Many such studies suggest a consciouseffort of parents to replace children who have died. In developing countries whereprolonged post partum amenorrhea is a relatively effective contraceptive, the biologicaleffect of child mortality may directly increase fertility by shortening the birth interval. Theconverse relationship of excessive fertility as a determinant of child infant mortality hasalso been documented in both developing and developed countries.

In this Bangladeshi study population, pregnancy spacing influences the outcome ofpregnancy to a greater degree than any of the other selected maternal factors (pregnancyorder, age of mother, history of previous fetal and child losses). Fetal deaths in the secondtrimester of pregnancy have the strongest inverse relationship with length of theinterpregnahcy interval when compared to other outcomes such as stillbirth, neonatal,post neonatal and early childhood mortality. Both second trimester fetal mortality andpost neonatal mortality were highest with pregnancy intervals of less than 12 monthsfollowing a live birth that survived and was breastfed. Early fetal wastage and postneonatal mortality would each be expected to decline by more than 40 % if all secondorder pregnancies were preceded by intervals greater than 24 months.

Among first order pregnancies stillbirth mortality is 5096 higher and neonatalmortality is 25 96 higher than among two to three order pregnancies. Second trimesterfetal mortality, post neonatal and early childhood mortality are all less than 1096 higheramong first order pregnancies than among two to three order pregnancies. There is adirect linear relationship between fetal, infant and child mortality and maternal age. As innumerous other studies the maternal history of previous fetal, infant and child deaths isalso associated with significantly higher rates of fetal, infant and child deaths insubsequent pregnancies. Nevertheless, prolonged pregnancy spacing is associated withreductions in these unfavorable outcomes of pregnancy even among the pregnancies ofwomen with a history of such previous losses.

'Swenson, I. (University of North Carolina at Chapel Hill): "The Effects of Pregnancy Spacingon Neonatal and Postneonatal Mortality in Bangladesh". Journal of Tropical Pediatrics.

I M P R I M E R I E L O U I S - J E A NPublications scientifiques et littéraires05002 GAP - Tél . : (92)51.35.23Dépôt légal : 332 - Juin 1983