The causal model approach to nutritionalproblems: an effective tool for research andaction at the local levelR. Tonglet,1 Maheshe Mudosa,2 Masumbuko Badashonderana,3 1. Beghin,4 &P. Hennart5

Reported are the results of a case study from Kirotshe rural health district, Northem Kivu, Zaire, where aworkshop on the causal model approach to nutrition was organized in 1987. The model has since beenused in the field for research design, training of health professionals, nutrition intervention, and commu-nity development. The rationale behind this approach is reviewed, the experience accumulated fromKirotshe district is described, and the ways in which the causal model contributes to comprehensivehealth and nutrition care are discussed. The broad range of possible policy implications of this approachunderlines its usefulness for future action.

IntroductionSevere nutritional problems in the Kivu highlands ofZaire have been reported and analysed since the1950s (1-4). However, despite the implementationof coordinated health activities and the support givento community development, little progress has beenmade in controlling these problems, and the nutri-tional situation in the area still gives cause for con-cern (5).

On taking charge of the Kirotshe health district,Northern Kivu, in 1985, we observed that many heal-th personnel were aware of the severity of thenutritional problems in the area, but did not knowhow to integrate nutrition into health programmes.Instead, they focused most of their efforts on growthmonitoring at clinics for under-5-year-olds andexpressed little concem about the selection of appro-priate actions. Growth monitoring, however, is notlikely to be worthwhile unless attention is paid alsoto the health and nutrition interventions needed (6).

1 Project Manager, Centre Scientifique et M6dical de l'Universitede Bruxelles pour ses Activit6s de Coop6ration (CEMUBAC),and Research Associate, Ecole de Sante Publique, UniversiteLibre de Bruxelles, Brussels, Belgium.2 Nutrition Technician, Zone de Sant6 Rurale de Kirotshe, NordKivu, Zaire.3 Rural Development Technician, Zone de Sant6 Rurale deKirotshe, Nord Kivu, Zaire.4 Head, Department of Nutrition, Institut de Medecine Tropicale"Prince L6opold", Antwerp, Belgium.s Director, CEMUBAC, and Professor, Ecole de Sant6 Publique,Universit6 Libre de Bruxelles, route de Lennik, 808, B-1070Brussels, Belgium. Requests for reprints should be sent to Pro-fessor Hennart.Reprint No. 5335

Based on this and other observations, we discoveredthat many health staff working in the field required acomprehensive approach to health and nutrition care.Lack of knowledge about the causes of nutritionalproblems, the feasibility of nutrition interventions,and the means of effective communication within thecommunity are a major constraint in many healthand nutrition programmes (7, 8). In such pro-grammes, a fundamental issue is how to design a toolthat would be effective in generating and organiz-ing knowledge that can lead to action.

This article advocates such a tool, which wasinitially developed for community nutritional assess-ment. The conceptual and methodological bases ofthis approach are reviewed, the experience accumu-lated in the Kirotshe health district is analysed, andthe ways such an approach contributes to more com-prehensive health and nutrition programmes are dis-cussed.

The causal model approachConceptual backgroundThe causal model approach assumes that complexproblems require to be dealt with in a comprehen-sive, holistic manner. In the last two decades a num-ber of global models of hunger and malnutritionhave been proposed. In one of the earliest suchattempts, Call & Levinson substantiated claims for asystematic approach to nutrition programmes (9).It is beyond the scope of the present article to reviewcritically the various competing models, which al-though useful for analytical purposes, often are oflimited use for planning or decision-making. The

Bulletin of the World Health Organization, 70 (6): 715-723 (1992)) © World Health Organization 1992 715

R. Tonglet et al.

review by Field illustrates the common failures ofmost attempts made to apply such models at themacrolevel (10). As discussed by Jonsson, the dis-crepancy between the level of analysis and the levelof proposals for action is certainly a problem, andone which requires attention (7).

In the late 1970s Pradilla et al. provided an inno-vative altemative to modelling nutritional problemsusing a purely pragmatic approach (11). Startingfrom this proposal, Beghin et al. developed andimproved the method, initially focusing on nutrition(12, 13)a, b, followed by its further application toother fields such as health services utilizationc or thecontrol of infectious diseases (14, 15). This approachhas been endorsed by WHO (16) and UNICEF."

A causal model is a hierarchically structured setof hypotheses on the causes or mechanisms that leadto the problem under study. Such a model attemptsto provide an easily understandable analytical frame-work showing the complex relationships between alldeterminant factors underlying the problem. In thiscontext, causality is suspected when a logical linkbetween a factor and the problem of interest is iden-tified.

The method applies knowledge, insight, and thefield experience of a multidisciplinary panel of localexperts to the problem under study and attempts tomodel how particular individuals achieve consensuson their situation; this consensus, however, is subjectto re-examination and reformulation. The model isspecific only for a given population at a givenmoment, and for a specific purpose.

MethodologyThe methodology used in the causal model approachis described in ref. 16. By applying this method-ology, workshop participants, through a series of"brainstorming" sessions, can easily pinpoint thefew well-defined rules that are necessary to buildthe model.

The easiest way to start the process is to list allthe factors that are relevant to the dependent

a Wilson, I.M. et al. On the use of a conceptual model in theempirical research setting. Antwerp, Institute of Tropical Medi-cine, 1989 (Working Paper No. 23).b Perez, J.A. Breast-feeding and medicine: a reassessment ofhistorical trends and Third World needs. Antwerp, Institute ofTropical Medicine, 1989 (Working Paper No. 25).c da Sliveiria, V.C. et al. Development and uses of a con-ceptual model in the study of antenatal services utilizationby migrant women in Belgium. Antwerp, Institute of TropicalMedicine, 1988 (Working Paper No. 19).d Strategy for improved nutrition of children and women indeveloping countries. New York, UNICEF, 1990 (UNICEFPolicy Review E/ICEF/1990/L.6, 9 March 1990).

variable. It does not matter which factors - biologi-cal, sociological, political, economic, etc. - areincluded in the model, as long as the group agreeswith the listing. In a further step, causal factors arelinked together in either a logical sum or a logicalproduct. Causal chains are mapped as a network ofboxes that are broken down at successive levels. Thetechnique of model building is retrospective; itrelates the dependent variable to the proximate deter-minants, in a top-to-bottom approach, against theflow of causality. The mapping ignores horizontallinks and feedback loops, but these are most oftenimplicit. Repetitions of the same factor are permit-ted. Finally, the model is completed by ordering andcombining causal chains into a causal framework.

ResultsDevelopment of the modelThe Kirotshe workshop on the causal modelapproach to nutritional problems took place inSeptember 1987. The development of the model wasthe task of a multidisciplinary group of two inter-national consultants in nutrition and 24 local partici-pants (2 doctors, 3 nurses, 2 health workers, 3 nutri-tion technicians, 2 rural development technicians, 2agronomists, 2 veterinarians, 2 farmers, 3 teachers,and 3 local politicians). Over five consecutive days,this group met in the momings for brainstormingsessions devoted to model building, and in the after-noons to collect useful additional information ontopics addressed during the development process. Asecretary drafted the minutes of each meeting andcollected the successive developmental stages of themodel.

The group started with a basic model that includ-ed the following causal factors: breast-feeding, foodintake, feeding practices, and the health status of thechild (Fig. 1). According to this simple model, all thedeterminants of the nutritional status of the young

Fig. 1. Schematic representation of the basic model.

Nutritional statusof the child

Food intake Food utilization

Breast milk Intake of food other Feeding Health statusintake than breast milk practices of the child

(See Fig. 2a) (See Fig. 2b) (See Fig. 5)WHO 92753

716 WHO Bulletin OMS. Vol 70 1992

The causal model approach to nutritional problems

Fig. 2. Schematic representation of the food intake sub-model: a) breast milk intake; b) intake of food other thanbreast milk.

her child. In African rural communities, women areoften overwhelmed by their routine household tasks,and children are fully dependent on their mothers fortheir feeding because of the pattem of social roles.The submodel for mother's time availability (Fig. 3)emphasizes the potential benefits of time-savinginterventions that can free mothers from routinetasks. The workshop participants therefore agreedwith the conclusions of a previous workshop onwomen in poverty: for poor women in developingcountries saving time is development, for time savedfrom humdrum tasks is time to invest in human capi-tal (17).

Those participants who were involved in agri-culture and stock farming helped the group to ana-lyse the complexity of the food production submodelprocess, which is one of the principal factors that hasa bearing on a child's food intake. A food productionsubmodel was developed (Fig. 4), and causal chainswere identified for the following factors: soil, capi-tal, techniques, and manpower. These factors wereassembled in a logical sum (Who is doing what, howand where?). The nutritional status of the childappeared to be an indicator of the socioeconomicsituation as a whole. The relationships between foodavailability, soil preservation, and demographic pres-sure were strikingly similar to those modelled byWils et al. in their systemic analysis of the ecosys-tem of the Kivu Mountains (3).

Fig. 3. Schematic representationmother's availability of time.

(See Fig. 5) WHO 92755

child can be divided into two categories those thatinfluence the intake of food by the child, and thosethat influence the child's utilization of the food; intum, factors that affect food intake can be dividedinto two categories, etc. From this, the group built aseries of submodels to elaborate on and providefurther details about each factor under scrutiny.Examples that illustrate the development of thesesubmodels are discussed below.

The food intake submodel (Fig. 2) highlights thecrucial importance of the mother's time availabilityas a factor that determines the nutritional status of

WHO Bulletin OMS. Vol 70 1992

of the submodel for

717

R. Tonglet et al.

Fig. 4. Schematic representation of the submodel for foodproduction.

Foodproduction

Capital L Techniques Manpower

ItAvailability Quality

of soils of soils

Demographic Migrations Family Protectionpressure size against

erosion

WH2-O92757

Cropping Famer's Appropriateand education technology

farmingpatterns

The submodel for child health status (Fig. 5)calls for particular attention. Although the partici-pants were familiar with the distinction betweenhealth problems (needs), care-oriented behaviour(demand), and availability of health services (sup-ply), it became clear to them that needs, demand, andsupply are never matched; implementation of healthservices is not automatically followed by increasedattendance at health centres or by direct improve-ment of health status. The participants concludedthat it was more accurate to make a distinction be-tween "the possibility of using services" and "thedecision to use services". Such a conclusion wasdrawn also by da Silveiria et al.,e and Mosley &Chen support this view in advocating the inclusionof personal illness control in any analytical frame-work of health problems (18). This part of the exer-cise stressed the need, in assessing health status, tocombine sociological and anthropological methodswith commonly used epidemiological approaches(19, 20).

Uses of the modelSince September 1987 the causal model has beensuccessfully used for research design, education of

e See footnote c, p. 716.

health professionals, nutrition intervention, and com-munity development.

Research design. Applied research, mainly on riskassessment in young children and on the healthimpact evaluation of the health programme, occupiesa substantial part of the health and development ac-tivities in the Kirotshe district. The research agendais planned by the health district authorities, withthe collaboration of several funding agencies.

In designing a 1-year follow-up study that focus-ed on the morbidity risk associated with the nutri-tional status of young children, we found that the cau-sal model was extremely helpful for improving datacollection. For example, the submodel for mother'savailability of time (Fig. 3) stressed the need foraccurate identification of a child's effective caretakerduring the recall period. By including in the follow-up forn a set of appropriate questions, we were ableto calculate the potential risk of ill health associatedwith defective child care. In focusing on the distinc-tion between the possibility and willingness to usehealth services, the submodel for child health status(Fig. 5) emphasized the usefulness of careful investi-gation of attitudes towards health care and services.Although social scientists have underlined this keyquestion for some time (21, 22), its crucial importan-ce for daily work in Kirotshe only became apparentafter the workshop on the causal model approach.Therefore, in the follow-up form, we made a clear-cut distinction between a subject's and an observer'spoint of view. For example, a mother may consider

Fig. 5. Schematic representation of the submodel for childhealth status.

Health statusof the child

Personal Ill health Psychologicalillness control status

Illness control Use of health Personal Environmentalat the familial services factors factors

level

Possibility to Decision to Hygieneuse health use healthservices services

Availability Acceptability Professional Quality ofof health of health ability of health healthservices services personnel services

WHO 92758

718 WHO Bulletin OMS. Vol 70 1992

The causal model approach to nutritional problems

that her child does or does not need care, and anobserver may or may not agree with her. Indepen-dently of these viewpoints, the mother may expressher intention to attend the health centre or not, andmay subsequently make a decision that is or is notconsistent with her expressed intention. Reportedand observed attitudes require to be assessed separ-ately and should be tested for consistency.

The causal model also proved useful for analy-sing data on 1096 under-5-year-olds collected duringa 1-year follow-up study of the health impact evalua-tion of a water supply programme in Kirotshe (23).The analysis was carried out in keeping with thechild health status submodel (Fig. 5) and the need toidentify the users and non-users of facilities. Thosepeople who were able to attend public taps had todecide whether or not to use them. No significantassociation was observed between the incidence ofdiarrhoea and the use of a water supply facility whenthis was assessed using subjective information (e.g.,Do you use the tap? or Where are you drawing waterfrom?). Many individuals could have claimed falla-ciously that they were public tap users because of thesocial value attributed to "good" answers. However,very significant associations were observed betweenthe incidence of diarrhoea and the use of public tapswhen such use was determined through objectivewater-related variables (the quantity of water drawnper household per day or the distance from thehousehold to the tap). Causal factors previouslyidentified in the workshop (i.e., "the possibility ofusing" and the "decision to use") therefore proved tobe useful in a completely different context.

Education of health professionals. Many paramedi-cal students undergo 3 months of practical training inKirotshe and are requested to submit a workingpaper at the end of this period. In this context, thecausal model proved to be an effective pedagogictool. Students were told to focus less on the scholas-tic aspect of the exercise and concentrate instead onthe causal submodel most appropriate to the problemthey had chosen. Each student then collected obser-vational data on a set of households, and tried torefine the construction of the submodel. They werenot asked to make any innovative contributions,but rather to try to fornulate and address rel-evant questions and to enrich their own understand-ing of health problems.

One of the students, for example, concentratedon breast-feeding practices and improved the exist-ing submodel by noting that breast-feeding wasaffected by seasonal variations in the availability offood. This observation is in accordance with pre-vious studies by Vis et al. in the Kivu area (24, 25)and with a causal model of breast-feeding proposed

by Perez! Another student tried to assess the differ-ences in the feeding practices of children of two eth-nic groups in the district. The Banyarwanda ethnicgroup from the Mitumba mountains, who live mainlyon stock farming, almost never ate meat, but had abalanced diet of cereals, vegetales and tubers; theBahunde ethnic group, on the other hand, who prac-tise agriculture along the Kivu lakeshore, purchasedmore meat but lived on a diet of cassava flour. Theseobservations provided the link between the produc-tion and consumption of food: food habits. A thirdstudent, who focused on the relationships betweenhealth status and health services utilization discov-ered that the children who were more frequently illwere those who made least use of the health centres;these children also had the poorest social conditions.This highlighted the social stratification that existseven in an apparently homogeneous poor rural com-munity and prompted the following question: Howcan we improve the accessibility and the acceptabilityof the health services for the poorest (26)?

Using the causal model as a framework for col-lecting and connecting observational data, the stu-dents therefore succeeded in rediscovering by them-selves causal factors and causal chains that hadpreviously been identified in a different setting. Thestudents learned in an interesting and stimulatingway how the formulation of a conceptual model waseffective in organizing knowledge and initiating afruitful education process.

Nutrition intervention. Many health professionalswho operate at the local level are convinced thattaking care of a malnourished child only requires theprovision of extra food. Considering that thisapproach was inappropriate and inadequate in thecontext of the rural health district of Kirotshe, wetested an alternative strategy. Initially we concentra-ted on the feeding practices causal submodel, whichidentifies the following common characteristics ofthe usual children's diet in the Lake Kivu highlands:insufficient number of meals (1 or 2 per day); dullcassava flour diet; and fairly low consumption ofgrains (maize, sorghum) or legumes (beans, ground-nuts), which are the major source of proteins. Wetherefore planned a new strategy, based on a proposalmade previously by Beghin & Van Lerberghe (27).

In January 1989, 23 malnourished children fromthe village of Kirotshe-Mushindi (1475 inhabitants,medical census of 1987) were identified. All weresuffering from clinical malnutrition and exhibitedgrowth retardation, oedema, hair loss or depigmenta-tion, and skin changes. The following proposal was

f See footnote b, p. 716.

WHO Bulletin OMS. Vol 70 1992 719

R. Tonglet et al.

made: in return for providing the mothers with dailymedical supervision and nutritional counselling for13 weeks, they were asked to attend, in groups, dailyeducational sessions at the local clinic. During thefirst week the children received, if necessary, anti-malarial and antihelmintic drugs and iron and folicacid supplements. The nutritional counselling focu-sed on convincing the mothers to feed their childrenat least three times a day with balanced meals con-taining tuber flour, grains, and legumes. In addi-tion, regular home visits were made to their house-holds. No extra food was given or purchased. As aresult, the intervention was almost restricted tosocial support and "homing-in" (28), both of whichare time-consuming but inexpensive. The mothersattended the sessions regularly and better feedingpractices resulted. All 23 children gained weightand height, and their growth curves improveddramatically. Oedema disappeared after 2-4 weeks,and the children rapidly became more active andcheerful; all but one were clinically healthy after10 weeks (R. Tonglet et al., unpublished data, 1989).

Community development. The workshop stimulatedparticipants to generate initiatives for communitydevelopment. For example, one participant started aninteresting development process in his own village incollaboration with the local health and developmentcommittee. The district health authorities and the vil-lage population had together built a small maternityward near the health centre, but a large field remain-ed vacant near the building. Encouraged by the localrural development technician who previously hadattended the Kirotshe workshop, the committeedecided to grow soya beans, which were then madeavailable on the local market. The first crop wasplentiful and the committee brought it to the vil-lage's mill; however, because milling is expensivethe villagers could not afford to buy the flour. Thecommittee then decided to buy its own mill. Acooperative store was established, funds were raisedfrom its members, the U.S. Peace Corps provided themill, and the committee was finally able to grind itsown grain. After a 2-year follow-up study, we observ-ed that most of the village women enriched the usualweaning porridge with soya bean flour, and alsomade biscuits from maize, sorghum, and soya beanflour; the feeding patterns were slowly changing.Encouraged by these results, the committee is nowtrying to popularize improved chicken and rabbitproduction in the village.

DiscussionAt a meeting on child mortality held in Antwerp,Belgium, in 1985, Palloni underlined that causativeresearch is flooded by theories and mathematical

models but is short on conceptual models (29).Futhermore, Beghin stated that to explain better aphenomenon and the mechanisms leading to it, thefollowing are needed: a theory; a conceptual frame-work or a causal model - preferably derived fromthe theory but not verifiable in itself; and a statisticalor epidemiological model consisting of a set of test-able hypotheses (30). These objectives are rarely ifever achieved.

Moreover, the lack of an appropriate analyticalapproach often results in ineffective interventions.The factors and processes that affect or underlie thehealth of the community, such as those that deter-mine nutritional status, are usually complex. Healthprofessionals working in the field are not alwaystaught how to analyse and solve complex issues.This deficiency in the training of health personnel,lack of knowledge about causes, poor ability to inter-vene, and limited ability to communicate within thecommunity, are probably among the commonreasons for the failures observed when the impactof health and nutrition care is evaluated.

To correct this situation, health personnel need amethod that assists them to clarify complex situa-tions and to bridge the gap between analysis andintervention. Based on our experience in Kirotshe,building a causal model may be one of the most cost-effective methods for assessing health problems,identifying objectives for action, and evaluatinghealth interventions.

At the Kirotshe workshop, the participantsreached consensus over a structured set of causalfactors and causal chains, which were postulated tobe the major determinants of the nutritional status ofyoung children in the health district. The resultingcausal model is not theoretical, but rather a practical,conventional representation of the complex situationin the study area. Unlike models that are intended tobe used in many different situations, our model in itspresent form is suitable only for local application;also, it is not a general model for malnutrition nordoes it provide an etiological explanation of thenatural course of malnutrition. Instead, it is a com-prehensive set of causal hypotheses that reduces thechance of omitting relevant determinants and poten-tial confounders. Futhermore, it is not a substitute forepidemiological models, and makes no use of statis-tics, although it facilitates better-designed data col-lection and analysis; it makes explicit hypotheseslikely not to be evident and facilitates communica-tion between individuals from different disciplines. Itcould be asked, however, whether these causalhypotheses could not have been formulated as well aswithout the need to construct a formal model. Prob-ably they could, but certainly not in the localcontext of Kirotshe.

WHO Bulletin OMS. Vol 70 1992720

The causal model approach to nutritional problems

The model leads to better understanding amonghealth professionals of the causal factors and causalchains that affect nutritional status and providesthem with a framework for determining the pro-cesses whereby this status is influenced by pro-grammes and policies. Therefore, the model is ofvalue for research purposes and for communitydevelopment and is consistent with current methodsused in development planning; for example,target-oriented programme planning (TOPP).TOPP, however, is more than an approach: rather,it is a general and detailed planning method that con-siders causes and ranks objectives. The causal modeladopted in Kirotshe is more empirical, opening upa wide range of unexpected initiatives, and stimu-lating investments in health and development.

The place of the causal model approach in theprocess of generating knowledge is still a matter ofdebate. However, its scientific and operational meritsare evident for professionals in the field. The modelbases its legitimacy on the classic paradigm of exper-imental science: first construct a hypothesis, thenderive from it a model that is a simplified representa-tion of the reality, and finally test the goodness-of-fitof the model and its validity. In addition, we believethat it also satisfies the following basic criteria ofaction-research, as defined by Susman & Evered(31): it is future-oriented and has close affinities withthe planning process; it is collaborative; it impliessystem development; it generates theory grounded inaction; it is subject to re-examination and reformula-tion; and it is situational, i.e., can change with thesetting. With these criteria action-research can alsobe viewed as a cyclical process of identifying a prob-lem, selecting courses of action, evaluating interven-tions, and specifying leaming from action to reassessthe original problem. As a result of the Kirotsheworkshop, various individuals in the district wereinvolved in such a cyclical process. With a fairly lowinvestment - mainly human resources - weestablished, surprisingly successfully, the utilityand feasibility of a complex process of organizingknowledge, generating new hypotheses, and iden-tifying appropriate interventions.

In addition, the Kirotshe experience seems to bethe first example of the sustained use of the causalmodel approach over a period of years in the samesetting.

9 [Target-orientated programme planning]. Eschborn, GermanAgency for Technical Cooperation, 1987 (in German).

AcknowledgementsThe Kirotshe workshop was funded by the Centre Scienti-fique et Medical de l'Universit6 de Bruxelles pour ses Acti-vites de Coop6ration (CEMUBAC), a non profit organiza-tion for cooperation in development.

We thank Dr G. Stott for his useful editorial assist-ance.

Resume

L'approche causale en nutrition: un outilutile pour la recherche et l'action auniveau localDans les pays en d6veloppement ou les pro-blemes nutritionnels sont extremement preoccu-pants, les professionnels de la sant6 eprouventtres souvent des difficultes a organiser des soinsde sante et de nutrition int6gres; 1'exp6rience amontr6 qu'ils m6connaissent les causes des pro-blemes nutritionnels, sous-estiment les possibilit6seffectives d'intervention et n6gligent les moyensd'am6liorer la communication au sein de la com-munaut6 pour promouvoir de bonnes conditionsalimentaires et nutritionnelles. Dans ce contexte, ilest n6cessaire de pouvoir mettre a leur dispositionun outil qui les aide a recueillir et organiser lesconnaissances indispensables a I'action. L'appro-che causale, propos6e par l'OMS en 1988, noussemble adapt6e a cet objectif.

L'approche causale est fond6e sur la convic-tion qu'il est n6cessaire de gerer les problemescomplexes d'une maniere globale et qu'une bonnecompr6hension des causes et des m6canismesde ces problemes est un prealable indispensablea toute decision. La construction d'un modele cau-sal hypoth6tique, applicable a une situation don-n6e, est au cceur de cette m6thode. Le terme"modele" est employe ici au sens d'une represen-tation simplifiee d'un systeme ou d'un processus.Le modele est appel6 "causal" car il pr6sente unjeu d'hypotheses logiques mettant en relation demaniere hierarchique les diff6rents facteurs d6ter-minants du probleme etudi6. La methode faitappel aux connaissances, aux intuitions et al'experience d'une equipe pluridisciplinaired'experts locaux, qui s'efforcent de clarifier leurperception collective du probleme 6tudie, dans uncontexte particulier et a un moment donn6. Laconstruction d'un modele causal passe par a)I'identification de tous les facteurs biologiques,sociologiques, politiques, economiques ou autres,qui permettent de formuler des hypotheses cau-sales quant a la genese du probleme 6tudi6, b) lerep6rage des liens logiques qui unissent ces fac-

WHO Bulletin OMS. Vol 701992 721

R. Tonglet et al.

teurs entre eux et a la variable dependante, etc) I'organisation et la combinaison de ces chainescausales au sein d'un schema conceptuel mate-rialise par un graphique simplifie. Le modele d6finiau terme de cet exercice n'a rien de d6finitif. Leshypotheses causales seront verifiees ou non; denouvelles hypotheses pourront etre formulees etenrichir le cadre analytique initial; les resultats del'analyse causale devront etre adapt6s en fonctiondes modifications du contexte de depart.

En septembre 1987, la zone de sant6 ruralede Kirotshe (Nord-Kivu, Zaire) a organis6 un ate-lier afin de proceder a l'analyse causale des pro-blemes nutritionnels du jeune enfant dans lar6gion. Ce s6minaire, qui a r6uni 26 expertslocaux pendant une semaine, a permis d'identifierles sources d'information disponibles, d'elaborerune repr6sentation simple et facile a communi-quer de la realit6 complexe des problemes nutri-tionnels de la region, et de s6lectionner les fac-teurs qui se pretent a une intervention.

Au terme de cette analyse, la zone de santede Kirotshe disposait d'un outil de travail qui,depuis lors et de maniere continue, est apparuextremement utile pour a) am6liorer la collecte etle traitement des donnees dans le cadre de larecherche appliqu6e, b) contribuer a la formationdu personnel de sant6, c) planifier des interven-tions nutritionnelles pertinentes au niveau local etd) encourager de nouvelles initiatives en faveurdu d6veloppement communautaire. L'exp6rienceacquise a Kirotshe demontre que la m6thode pro-pos6e cr6e des conditions susceptibles de dyna-miser un programme de sante et de nutrition et demobiliser ses diff6rents partenaires.

11 serait utile de soumettre I'analyse causale ala critique 6pist6mologique, mais il suffit ici deconstater que cette methode est manifestementfond6e sur le paradigme classique de la scienceexperimentale: d'abord formuler une hypothese,ensuite batir un modele et donc proposer unerepr6sentation simplifiee de la realit6, enfin cher-cher a v6rifier la validit6 du modele et son ade-quation a la realite. II n'est donc guere etonnantde voir des professionnels de sante constater, demaniere tout a fait empirique, la valeur scientifiqueet op6rationnelle de l'approche causale, sur le ter-rain. Un modele causal, en effet, n'est pas un6chafaudage theorique ayant une valeur univer-selle, mais la repr6sentation conventionnelle etpratique d'une realite locale. L'analyse causale nese substitue pas aux modeles 6pid6miologiques etne fait pas usage de techniques statistiques, maiselle permet d'identifier de maniere exhaustive lesfacteurs determinants du probleme etudi6 et declarifier des hypotheses qui souvent sont impli-

cites mais dissimulees aux yeux des operateurs,sur le terrain. L'approche causale, en outre, initieun processus cyclique qui debute avec l'identifica-tion d'un probleme, se poursuit avec l'analyse decelui-ci et s'oriente ensuite vers la planification etl'evaluation d'interventions pertinentes. L'experien-ce de Kirotshe confirme le bien-fond6 des hypo-theses conceptuelles et methodologiques a labase de cette methode et illustre a quel pointl'approche causale est un outil utile pour larecherche-action au niveau local.

References1. Vis, H.L. General and specific patterns of marasmic

kwashiorkor in the Kivu area. In: McCance, R.A. &Widowson, E.M., ed. Caloric deficiencies and pro-tein deficiencies. London, Churchill, 1968,pp. 119-134.

2. Vis, H.L. et al. The health of mother and child inrural Central Africa. Studies in family planning, 6:437-441 (1975).

3. Wils, W. et al. Le Kivu Montagneux (surpopulation,sous-nutrition, erosion du sol). M6moires de l'Aca-d6mie Royale Belge des Sciences d'Outremer,21(3): 1-201 (1986).

4. Hennart, P. et al. Long-term follow-up of severeprotein-energy malnutrition in eastern Zaire. Journalof tropical pediatrics, 33: 10-12 (1987).

5. Tonglet, R. et al. Pattern of attained growth in0-to-5-year-old children from Kivu (Zaire). Ecologyof food and nutrition (in press).

6. Gerein, N. Is growth monitoring worthwhile? Healthpolicy and planning, 3(3): 181-194 (1988).

7. Jonsson, U. The causes of hunger. Food and nutri-tion bulletin, 3(2): 1-9 (1981).

8. Scrimshaw, N.S. Integrating nutrition into pro-grammes of primary health care. Food and nutritionbulletin, 10(4): 19-28 (1988).

9. Call, D.L. & Levinson, F.J. A systematic approachto nutrition intervention programs. In: Berg, A. et al.,ed. Nutrition, national development, and planning.Cambridge, MA, MIT Press, 1973, pp. 165-197.

10. Field, J.O. Multisectoral nutrition planning: a post-mortem. Food policy, 15-28 (February 1987).

11. Pradilla, A. et al. Interpretative models for selectionof nutrition priorities. Archivos latinoamericanos denutricion, 27(2) (suppl. 1): 89-107 (1977).

12. Beghin, I. Improving nutrition at the local level. Car-nets de l'Enfance, 35: 9-24 (1976).

13. Beghin, I. L'approche causale en nutrition. In:Lemonnier, D. & Ingelbeeck, Y., ed. La malnutritiondans les pays du Tiers-Monde. Paris, 1986, pp.615-628 (INSERM Serie Colloque n° 136).

14. Beghin, I. et al. Can the causal model approachcontribute to the study of the epidemiology and thecontrol of sleeping sickness? Annales de la Societ6Belge de M6decine Tropicale, 69(suppl. 1): 31-47(1989).

15. The Antwerp Trypanosomiasis Causal ModellingGroup. Constructing a causal model of African

722 WHO Bulletin OMS. Vol 70 1992

The causal model approach to nutritional problems

human trypanosomiasis. Annales de la Societ6Belge de M6decine Tropicale, 69(suppl. 1): 49-72(1989).

16. Beghin, I. et al. A guide to nutritional assessment.Geneva, World Health Organization, 1988.

17. Briscoe, J. Water supply and health in developingcountries: selective primary health care revisited.American journal of public health, 74: 1009-1013(1984).

18. Mosley, W.H. & Chen, L.C. An analytical frame-work for the study of child survival in developingcountries. In: Mosley, W.H. & Chen, L.C., ed. Childsurvival: strategies for research. Cambridge, Cam-bridge University Press, 1984, pp. 25-45.

19. Kroeger, A. Anthropological and sociomedicalhealth care research in developing countries.Social science and medicine, 17: 147-161 (1983).

20. Buzzard, S. Appropriate research for primary healthcare, an anthropologist's view. Social science andmedicine, 19: 237-277 (1984).

21. Kroeger, A. Health interview surveys in developingcountries: a review of the methods and the results.International journal of epidemiology, 12: 465-481(1983).

22. Igun, U.A. Stages in health-seeking: a descriptivemodel. Social science and medicine, 13(A):445-456 (1979).

23. Tonglet, R. et al. Can improvements in water sup-ply reduce childhood diarrhoea? Health policy andplanning, 7(3): 260-268 (1992).

24. Vis, H.L. et al. L'allaitement en zone rurale pauvre:I'alimentation maternelle et I'allaitement au Kivu,

Zafre. Carnets de l'Enfance, 55/56: 171-189 (1981).25. Vis, H.L. & Hennart, P. Exclusive and partial

breast-feeding and infant development in CentralAfrica. In: Ballabriga, A. & Rey, J., ed. Weaning:why, what and when? New York, Raven Press,1987, pp. 169-185. (Nestle Nutrition WorkshopSeries No. 10).

26. Chen, L.C. Primary health care in developing coun-tries: overcoming operational, technical, and socialbarriers. Lancet, 2: 1260-1265 (1986).

27. Beghin, I. & Van Lerberghe, W. La courbe de"gpoids-cible": un outil pratique pour le suivi desenfants malnourris pris en charge. In: Actes desQuatriemes Journ6es Scientifiques Internationalesdu Groupe d'Etudes et de Recherches sur la Malnu-trition (GERM), Spa, Belgium, 23-29 April 1989.Paris, Hopital Bichat, Ul INSERM Nutrition, 1989,pp. 1 1-27.

28. Keutsh, G. Homing-in on interventions in themalnutrition-infection complex. American journalof clinical nutrition, 33: 727-729 (1980).

29. Palloni, A. Theory, analytical frameworks andcausal approach in the study of mortality at youngages in developing countries. Annales de laSoci6t6 Belge de M6decine Tropicale, 67(suppl. 1):31-45 (1987).

30. Beghin, I. Comments on the paper by A. Palloni.Annales de la Societ6 Belge de Medecine Tropicale,67(suppl. 1): 47-50 (1987).

31. Susman, G.I. & Evered, R.D. An assessment of thescientific merits of action-research. Administrativesciences quarterly, 23: 582-603 (1978).

WHO Bulletin OMS. Vol 70 1992 723