Methods andResponseCharacteristics1980 National Natalityand Fetal MortalitySurveys

This report describes the methods

employed in the 1980 National Natality

Survey and the 1980 National Fetal

Mortality Survey. These surveys are based

on information obtained from mothers,

hospitals, attendants at delivery, and

providers of radiation examinations for a

sample of live births and a sample of fetal

deaths.

Data Evaluation andMethods ResearchSeries 2, No. 100

DHHS Publication No. (PHS) 86-1374

U.S. Department of Health and Human

Services

Public Health Service

National Center for Health Statistics

Hyattsville, Md.

September 1986

Copyright Information

All material appearing in this repai is in the public domain and may bereproduced or copied without permission; citation as to source, however, ISappreciated.

Suggested Citetiin

National Center for Health Statistics, K. G. Keppel, R. L. Heuser, P. J. Placek,et al.: Methods and response characteristics, 1980 National Natality and FetalMortality Surveys. Vita/ and Hea/th Statistics. Series 2, No. 100. DHHS Pub.No. (PHS) 86-1374. Public Health Service. Washington. U.S. GovernmentPrinting Office, Sept. 1986.

Library of Congress Catabging in Pdicatiin Date

Methods and response characteristics.(Series 2, Data evaluation and methods research ;

no. 100) (DHHS publication ; no, (PHS) 8&1374)By Kenneth G. Keppel and others.Bibliography: p,1. Childbirtt+United States--statistics. 2. Prenatal death-United States—

StatieOcs. 3. Prenatal car+United Statee-Evaluation. 4. Health survey%United State-Forms. 6. United States-Statistics, Medical. 7. United States—StatiWIcs, vital. 1. Keppel, Kenneth G. Il. National Center for Health Statistics(U.S.) Ill. Series: Vital and health statistics. Series 2, Data evaluation methodsresearch ; no. 100. IV. Series: DHHS publication ; no. (PHS) 661374. [DNLM:1. Data Collecfio~ethods. 2. Fetal Deatk-lJnited State+statistics.3. Health Surveys-United States. 4. Prenatal Care-United States.W2 A N146vb no. 100]RA409. U45 no. t 00 362.1 ‘0723 S 8&600090[RG530.3.U5] [304.630973021 ]ISBN 0-6408-0336-3

For sale by the Superintendent of Documents, U.S. Government Printing Oftice. Washington, D.C. 20402

National Center for Health Statistics

Manning Feinleib, M. D., Dr. P. H., Director

Robert A. Israel, Deputy Director

Jacob J. Feldman, Ph. D., Associate Director for Analysis and

Epidemiology

Garrie J. Losee, Associate Director for Data Processing andServices

Alvan O. Zarate, Ph. D., Assistant Director for International

Statistics

Peter L. Hurley, Acting Associate Director for Interview and

Examination Statistics

Stephen E. Nieberding, Associate Director for Management

Gail F. Fisher, Ph. D., Associate Director for Program Planning,

Evaluation, and Coordination

Monroe G. Sirken, Ph. D., Associate Director for Research andMethodology

Peter L. Hurley, Associate Director for Vital and Health CareStatistics

Alice Haywood, Information Ofjicer

Viial and Health Care Statistics Program

Peter L. Hurley, Associate Director

Gloria Kapantais, Assistant to the Director for Data Policy,

Planning, and Analysis

Dwision of Vial Statistics

John E. Patterson, Director

James A. Weed, Ph. D., Deputy Director

Robert L. Heuser, M. A., Chiej Natality Statistics Branch

Joseph D. Farrell, Chiej Computer Applications Staff

Mabel G. Smith, ChieJ Statistical Resources Branch

Data Processing and Services Program

Garrie J. Losee, Associate Direclor

Martin A. Baum, Ph. D., Deputy Associate Director

Dwision of Data Services

James C. Jacks, Ph. D., Director

Phillip R. Beattie, Deputy Director

Jacqueline P. Davis, formerly Chiej Data Collection Branch(presently Special Assistant to the Director of the O#ice ofInternational Statistics)

Office of Research and Methodology

Monroe G. Sirkcn, Ph. D., Associate Director

James T. Massey, Ph. D., Chief, Survey Design Staff

.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .“....Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .The 1980NationalNatalitySurvey(NNS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .The 1980National FetalMortalitySurvey(NFMS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Conductofthesurveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Availabilityofdataandfindings. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Surveyprocedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Samplingofcertificatesoflivebirth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Samplingofreportsoffetaldeath. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Sourcesofinformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Questionnaires . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Collectionofsurveydata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Dataprocessingandimputation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Samplingerror . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Nonsamplingerror . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..”..

Responsecharacteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Responseratesformothers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Consentstatementsfrommothers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . .Responseratesformedical sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....”..

Listofdetailedtables, ..,.. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . .

Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

List of text tables

A, Total number of live births in the United Statesin 1980 and number of live births included in the sample for the1980National NatalitySurvey(NNS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . -

B, Estimated number of fetal deaths in the United States in 1980 with gestation of 28 weeks or more, or delive~weight of 1,000 grams or more, and number of fetal deaths included in the sample for the 1980 National FetalMortalitySurvey(NFMS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

C, Number of potential sources ofinformation in the 1980 National Natality Survey (NNS) and the 1980 National FetalMortality Suney(NFMS) bythetype ofsource andmatital status of the mother . . . . . . . . . , . . . . . . . .

D. Number of married mothers included in the 1980 National Natality Survey and National Fetal Mortality Survey andpercentdistribution byresponsestatus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...

E. Number of hospitals and attendants at delivery inthe 1980 National Natality and Fetal Mortality Surveys and percentdistribution byresponse status, according tothematital status of the mother . . . . . . , . . . . , . . . . . . . .

F. Number of secondary radiation sources andseconda~ attendant-at-delive~ sources in the 1980 National Natality andFetal Mortality Surveys and percent distribution by response status, according to the marital status of the

mother . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ........G, 1980National Natality Suney(NNS) poststratification cell definitions, number oflivebitihs inthe NNS, number of

live births toresidents of the United Statesin 1980, and sampling weights . . . . . . . . . . . . . . . . . . . . .

1

11222

3334679

121417

18181819

21

23

32

3

4

6

7

8

9

13

Ill

H. 1980 National Fetal Mortality Survey (NFMS) weighting cell definitions, number of fetal deaths in the NFMS,estimated number of fetal deaths with gestation of 28 weeks or more, or delivery weight of 1,000 grams or more, toresidents of the United States in 1980, and sampling weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

J. Parameters used to approximate the relative standard errors for estimates based on the 1980 National Natality Survey(NNS)by domain ofstudy andsource of information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

K. Parameters used to approximate the relative standard errors for estimates based on the 1980 National Fetal MortalitySurvey (NFMS) by domain of study and source of information . . . . . . . . . . . . . . . . . . . . . . . . . . . Is

L. Response rates for hospitals and attendants at delivery in the 1980 National Natality and Fetal Mortality Surveys bymarital status of themother andtype of consent statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ]9

Symbols

. . . Data not available

. . . Category not applicable

Quantity zero

0.0 Quantity more than zero but less than0.05

z Quantity more than zero but less than500 where numbers are rounded tothousands

* Figure does not meet standard ofreliability or precision

# Figure suppressed to comply withconfidentiality requirements

Methods and ResponseCharacteristics,1980 National Natality andFetal Mortality Surveys

by Kenneth G. Keppel, Robert L. Heuser, Paul J. Placek,David P. Johnson, and Gloria A. Simpson, Division of VitalStatistics; Steven L. Botman, Office of Research andMethodology; Carla E. Weinberg, formerly with the Division ofData Services (presently with the Division of HealthExamination Statistics); and Arthur L. Whitaker, formerly withthe Division of Data Services (presently with the Division ofHealth Interview Statistics)

IntroductionThis report describes the methods employed in the 1980

National Natality Survey and the 1980 National Fetal MortalitySurvey. This introduction provides a brief overview of thenature and content of these surveys. The next section describesthe sampling of vital records, the collection and processingof survey data, the production of national estimates, and theapproximation of sampling errors. The third section examinesdifferentials in response rates for mothers, hospitals, and at-tendants at delivery included in the surveys.

13ackground

The vital registration system provides annual data on vitalevents that occur in the United States. Because vital recordsserve both legal and statistical purposes, they provide limitedsocial, demographic, health, and medical information. Thedata from vital records can be augmented, however, throughperiodic mail or telephone “followback” surveys. These sur-veys are referred to as “followback” because they obtain addi-tional information from sources named on the vital record.A followback survey is a cost-effective means of obtainingsupplementary information for a sample of vital events. Fromthe sample it is possible to make estimates of vital eventsaccording to characteristics not otherwise available on a na-tional basis, Periodic followback surveys respond to the chang-ing data needs of the public health community without requir-ing changes in the vital record forms.

Previous National Natality Surveys were conducted bythe National Center for Health Statistics (NCHS) based onlive births in 1963, 1964-64, 1967–69, and 1972. The focusof the 1963 survey was on the exposure of married mothersto radiation (particularly x rays) during pregnancy. In the1964-66 and 1967–69 surveys information was sought frommarried mothers concerning their social and demographic char-acteristics, and various health and prenatal characteristics.

NOTE Peer review was prnviclcd by Kenneth Harris, Office of Researchand Methodology,

Information collected in the 1972 survey included social,demographic, health, prenatal, labor, and delivery informationfrom married mothers, hospitals, and attendants at delivery.See the references at the end of this report for a list (refer-ences 1–22) of reports and papers based on these previousnatality surveys.

The 1980 National Natality Survey (NNS)

The 1980 NNS is based on a probability sample of 9,941live births that occurred in the United States during 1980.The live birth certificate represents the basic source of infor-mation. Based on information from the sample birth certifi-cates, questionnaires were mailed to mothers who were mar-ried. These mothers were asked to provide information onprenatal-health practices, prenatal care, previous pregnancies,and social and demographic characteristics of themselves andtheir husbands. Each mother was also asked to sign a consentstatement authorizing NCHS to obtain supplemental informa-tion from her medical records. If the mother did not respondafter two questionnaires were sent by mail, a telephone inter-viewer attempted to complete an abbreviated questionnaireand to obtain a consent statement. To ensure their privacy,unmarried mothers were not contacted.

Regardless of the mother’s marital status, questionnaireswere mailed to the hospitals and to the attendants at delivery(for example, physicians or nurse-midwives) named on thebirth certificates. A questionnaire was sent to the hospitalfor each sample birth that occurred either in a hospital oren route to a hospital. If the mother signed a consent statementauthorizing NCHS to obtain supplemental medical informa-tion, a copy was included with the questionnaire. The focusof the hospital questionnaire was on characteristics of laborand delivery, health characteristics of the mother and infant,information on prenatal care visits, and information on radia-tion examinations and treatments received by the mother duringthe 12months before delivery of the sample live birth.

A questionnaire was also mailed to the attendent at delivery

1

for each sample birth for which the attendant’s address wasnot the same as the address of the hospital. Again, if themother signed a consent statement, a copy was included.The questionnaire sent to attendants at delivery also containedquestions concerning prenatal care visits and exposure to radia-tion examinations and treatments before delivery.

If the hospital or attendant at delivery identified on thebirth certificate possessed little or no information concerningprenatal care, he or she was asked to provide the namesand addresses of other facilities or individuals that could supplythis information. Appropriate questionnaires were then sentto those facilities or indhiduals.

Finally, the mothers, hospitals, and attendants at deliverywere asked to provide the names and addresses of other medicalsources that provided radiation examinations and treatmentsto the mother before delivery. A special question~aire concern-ing the mother’s exposure to radiation was sent to these medicalsources.

Though the 1980 NNS incorporates selected aspects ofearlier natality followback surveys, it is different in severalrespects. 23 First, the number of births included in the sample

is greater than that in any previous NNS. Second, low-birth-weight infants (those less than 2,500 grams) were oversampledto facilitate studies of these births. Third, this is the firstfollowback survey for which extensive medical informationwas collected for births to unmarried mothers. Although un-married mothers were not contacted, information was obtainedfrom their hospitals, attendants at delivery, and providersof radiation examinations and treatments. Fourth, this is thefirst time that data have been collected on maternal alcoholconsumption, occupation and industry of both parents,hematocrit and hemoglobin values, blood pressure readings,tests for urine protein, and amniocentesis. This survey isalso the first source of national information on the exposureof mothers to ultrasound during pregnancy. Finally, this isthe first followback survey in which consent statements weresought from mothers in order to encourage medical sourcesto provide information.

The 1980 NNS is composed of information from birthcertificates and information from questionnaires sent to marriedmothers, hospitals, attendants at delivery, and providers ofradiation examinations and treatments. The survey representsan extensive source of information concerning specific mater-nal and child health conditions and obstetric practices forlive births in the United States.

The 1980 National Fetal MortalitySurvey (NFMS)

The 1980 NFMS is based on a probability sample of6,386 fetal deaths with gestation of 28 weeks or more, ordelivery weight of 1,000 grams or more, that occurred inthe United States during 1980. The report of fetal dedth repre-sents the basic source of information in this survey. Marriedmothers, hospitals, attendants at delivery, and providers ofradiation examinations and treatments were surveyed under

the same conditions as those described for the 1980 NNS.The same questionnaires were used for both surveys. Althoughsome questions pertained only to live births and others per-tained only to fetal deaths, instructions to skip inappropriatequestions were included in the questionnaires.

A national followback survey based on fetal death recordshad not been conducted previously. A pretest conducted priorto the 1980 survey indicated that it was feasible to collectsubstantially the same information about fetal deaths as aboutlive births .24The sampling design for the NFMS was developedso that the NFMS sample would be large enough to permitcomparisons between live births in the NNS and fetal deathsin the NFMS.

Conduct of the surveys

The 1980 NNS and NFMS is a major research effortof the National Center for Health Statistics (NCHS) requiringthe collaboration and cooperation of many other agencies.The content and methods of these surveys were planned anddirected by the Division of Vital Statistics of NCHS. Thesurveys were approved by health and vital statistics officialsin the State and independent registration areas that sampledvital records for inclusion in the surveys. The Division ofData Services of NCHS processed the vital records and col-lected the data. Seven other Public Health Service agenciesparticipated in planning the surveys and provided fundingthrough NCHS’S Reimbursable Work Program:

. The National Center for Devices and Radiological Health,Food and Drug Administration.

. The National Institute for Occupational Safety and Health,Centers for Disease Control.

. The Center for Health Promotion and Education, Centersfor Disease Control.

. The National Institute of Child Health and HumanDevelopment, National Institutes of Health.

. The National Institute on Drug Abuse; Alcohol, DrugAbuse, and Mental Health Administration.

. The National Institute on Alcohol Abuse and Alcoholism;Alcohol, Drug Abuse, and Mental Health Administration.

● The Bureau of Health Care Delivery and Assistance,Health Resources and Services Administration.

Availability of data and findings

A public-use data tape containing information collectedin the 1980 NNS and NFMS may be purchased from theNational Technical Information Service (NTIS), 5285 PortRoyal Road, Springfield, Virginia 22161. Data tapes for the1964-66, 1967-69, and 1972 National Natality Surveys arealso availabIe.25 Analyses of the 1980 NNS and NFMS dataare being published as NCHS reports and as journal articles,and are being presented at meetings of professionalorganizations.

Survey procedures

Sampling of certificates of live birth

The files of birth certificates in the 50 States, the Districtof Columbia, and the independent registration area of NewYork City constituted the sampling frame for the survey.Certificates were registered for an estimated 99.3 percent ofall live births that occurred in the United States in 1980.26

In each registration area a sequential file number is as-signed to each birth certificate received from the beginningto the end of each calendar year. Certificates with file numbersending with specified combinations of digits were drawn sothat 105 of every 10,000 certificates would be selected forthe sample. The certificates were drawn as soon as the fileof certificates for the most recent month of registration wascomplete, Copies of the sample certificates were sent monthlyto NCHS for further processing. Certificates of live birthreceived after March 27, 1981 were not selected; there isno reason to believe that any substantial number of birthcertificates were received after that date.

Some of the certificates of live birth selected for thesample in the registration areas were not included in the NNS.Births that resulted in adoption and births to nonresidentsof the United States were excluded. Births to unmarriedmothers in New Jersey and Idaho were excluded at the requestof the State registrars of vital statistics in those States.

The next stage of sample selection was designed to producean oversampling of low-birth-weight infants, Certificates withfile numbers ending with a subset of the initially specifieddigits were selected for the sample. In general, 25 of every105 certificates were selected in this way. From the remainingcertificates (80 of every 105 certificates), on]y those on whichthe birth weight of the infant was less than 2,500 grams(5 pounds 8 ounces) were selected for the sample. If thebirth weight on a certificate was not stated, it was assumedto be 2,500 grams or more for sampling purposes. The numberof low-birth-weight infants in the NNS sample was thereforeabout 4 times the number that would have been obtainedby simple random sampling.

Finally, it was necessary to exclude from the samplesome additional certificates for the States of Idaho andWashington. In Idaho permission was requested from marriedmothers to include the birth certificates of their infants inthe NNS. The State of Washington requested permission fromunmarried mothers to include the birth certificates of theirinfants in the NNS. If the mothers refused permission, thecertificates were excluded.

The total number of registered live births to residents

Table A. Total number of live birtha in the United States in 1980 andnumber of live births included in the sample for the 1980 Natiinal NataliiSurvey (NNS)

Classification of live births Number

Live births inthe United States... . . . . . . . . . . . . . . . 3,612,258

Live births selected for the NNS sample . . . . . . . . . . 10,615

Live births excluded from the NNS sample . . . . . . . . . . 674Attendsnt’s name or address illegible . . . . . . . . . 189Mother’s questionnaire returned by Post Office . . . . . . 485

Live births included in the NNS sample . . . . . . . . . . . . 9,941

of the United States during 1980 was 3,612,258 (table A).According to the procedures outlined above, 10,615 certificatesof live birth were selected for inclusion in the NNS sample.During the conduct of the survey, however, a shortage ofstaff dictated a reduction in the number of cases that requiredspecial handling. In order to send questionnaires to the at-tendants at delivery, it was necessary to obtain their namesand addresses. In some cases this information was illegibleon the vital record, and survey staff had difficulty identifyingthese individuals through medical and telephone directories.For married mothers, some of the addresses were incomplete“or out of date, and questionnaires mailed to these motherswere returned by the Post Office. Survey staff had difficultyobtaining new addresses for these mothers. The decision wasmade to exclude from the NNS 189 certificates on whichthe attendant’s name and address were illegible, as well as485 certificates for which the mother’s questionnaire was re-turned by the Post Office. Therefore, 9,941 certificates oflive birth form the NNS sample (table A). Hereafter all refer-ences to the NNS sample refer to these 9,941 certitlcatesof live birth. This represents approximately 1 in 363 livebirths to residents of the United States in 1980.

Although the NNS is based on a sample of live births,the terms “women” and “mothers” are frequently used insteadof “births” for convenience, particularly when the focus ison mothers.

Sampling of reports of fetal death

The designation for the document filed in the event ofa fetal death differs among the registration areas (for example,certificate of fetal death, certificate of stillbirth, report offetal death). Inas “reports of

this report these documents are referred tofetal death .“ Although the requirements

3

governing the registration of fetal deaths vary among the Table B. Estimated number of fetal deaths in the United States in 1980registration areas, all areas require the registration of fetal

with gestation of 28 weeks or more, or delivery weight of 1,000 grama or

deaths of at least 22 weeks gestation or 500 grams.more, and number of fetal deaths included hr the sample for the 1980National Fetal Mortalii Survey (NFMS)

The extent to which fetal deaths are not registered isunknown .27 However, underregistration is suspected to bemore of a problem near the lower limit for areas with aminimum gestation period requirement. The NFMS was lim-ited to fetal deaths of 28 weeks gestation or more, eventsthat were relatively well registered,

The files of reports of fetal deathin51 of the 52 registrationareas constituted the sampling frame for the NFMS. MichiganState law prohibited the use of reports of fetal death in Michiganfor purposes of conducting a followback survey.

In each registration area a sequential file number is assignedto each report of fetal death received from the beginning tothe end of each calendar year. Reports of fetal death withfile numbers ending with specified digits were drawn so that4 of every 10 reports would be selected for the sample. Copiesof the sample reports of fetal death were sent to NCHS ona monthly basis. Reports of fetal death for 1980 received afterMarch 27, 1981 were not selected, No substantial number ofqorts of fetal death for 1980were expected after that date.

The sample reports of fetal death were then examinedto determine whether they were within the scope of the NFMS;fetal deaths to nonresidents of the United States and inducedabortions were excluded. Reports of fetal death to unmarriedmothers in New Jersey and Idaho were also excluded at therequest of the State registrars of vital statistics in those States.

The NFMS was limited to “late” fetal deaths becausethey were more likely to have occurred in a hospital, makinggood medical information available. The following reportsof fetal death were selected for the sample: a) Those forwhich length of pregnancy or physician’s estimate of gestationwas stated as at least 28 weeks or 7 months; b) those lackingan explicit statement of gestational length for which the deliv-ery weight of the fetus was at least 1,000 grams or 2 pounds,3 ounces; and c) those lacking an explicit statement of bothgestational length and delivery weight for which the lengthof gestation based on the date of last menstrual period was196 days (28 weeks) or more. Although not strictly correctgiven these selection criteria, the fetal deaths included inthe NFMS can be referred to as “fetal deaths with gestationof 28 weeks or more, or delivery weight of 1,000 gramsor more. ”

Finally, it was necessary to exclude some reports of fetaldeath for the States of Idaho and Washington. In Idaho permis-sion was requested from married mothers, and in Washingtonpermission was requested from unmarried mothers to includetheir reports of fetal death in the NFMS. If the mother refusedpermission, the report of fetal death was not included.

In accordance with these procedures, 7,391 reports offetal death were selected for the NFMS sample. It was alsonecessary to exclude reports of fetal death from the NFMSbecause of illegible names on vital records, and because somemothers’ questionnaires were returned by the Post Office.A total of 1,005 reports of fetal death were excluded fromthe NFMS for these reasons (table B). Therefore, the 1980NFMS sample included 6,386 reports of fetal death. Hereafter

Classification of fetal deaths Number

Estimated number of fetal deaths in the United Statesl . . . . . 19,202

Fetal deaths selected for the NFMS sample . . . . . . . . . . . 7,391

Fetal deaths excluded from the NFMS sample . . . . . . . . . 1,005Attendant’s name or address illegible . . . . . . . . . . . . 512Mother’s questionnaire returned by Post Office . . . . . . . 493

Fetal deaths included in the NFMS sample . . . . . . . . . . . 6,366

1See tesl forderivationof thisestimate.

all references to the NFMS sample refer to these 6,386 reportsof fetal death.

The total number of reports of fetal death eligible forthe 1980 NFMS could not be determined from the vitalregistration system because of the way in which the gestationand birth weight criteria were applied in the selection ofthe NFMS sample. The total number of eligible reports offetal death in 1980 was estimated by multiplying the numberof reports of fetal death selected (7,391) by the reciprocalof the probability of selection and then adding the numberof eligible reports of fetal death in Michigan. This providedan estimate of 19,202 fetal deaths with gestation of 28 weeksor more, or delivery weight of 1,000 grams or more. Thesurvey therefore includes approximately 1 in 3 of the estimatednumber of late fetal deaths eligible for the NFMS in 1980.

While the NFMS represents a sample of fetal deaths,the terms “women” and “mothers” are used in discussingthe data.

Sources of information

Mtal records

Although there are some variations in the nature of theinformation collected on the State live birth certificates, thecontent of these forms generally corresponds to that of theU.S. Standard Certificate of Live Birth (appendix figure 1),Data items for which reporting varies among the registrationareas are shown in appendix table I. The birth certificatecontains information on the demographic and social character-istics of the parents, the woman’s pregnancy history, character-istics of the pregnancy and delivery, and characteristics ofthe newborn. This information was coded from the birth certifi-cates as part of the data record for each live birth in the1980 NNS sample. The complications of pregnancy and com-plications of labor and/or delivery items were coded onlyin terms of whether or not any complication was noted, Infor-mation concerning concurrent illnesses or conditions affectingthe pregnancy was not coded. If data items were missingor coded inappropriately, values were imputed according toprocedures described elsewhere in this report.

Reports of fetal death also vary among the registrationareas, but they generally correspond to the U.S. StandardReport of Fetal Death (appendix figure II). Data items forwhich reporting varies among the registration areas are shown

4

in appendix table II. The items on the report of fetal deathwhich deal with the characteristics of the parents, the mother’spregnancy history, and characteristics of the pregnancy anddelivery are comparable to those on the live birth certificate.The report of fetal death also contains information on theweight of the fetus, when the fetus died, and whether anautopsy was performed. This information was coded and im-puted as described for live births.

Mothers

The mother was a potential source of information aboutsocial and demographic characteristics, prenatal care, andhealth behavior not available from the vital record. In the1980 NNS and NFMS, questionnaires were sent only tomothers who were married. Consideration for the mother’sprivacy led to the decision not to contact those who werenot married. Although the vital records of most registrationareas included an item on the mother’s marital status, thevital records of 11 registration areas did not. In addition,marital status was not indicated on a small number of vitalrecords from registration areas that reported marital status.

When marital status was not reported, other informationon the vital record was used to infer the mother’s maritalstatus. Briefly, a live birth or fetal death was classified asoccurring to a married mother if the parents’ surnames werethe same, or if the child’s and father’s surnames were thesame and the mother’s current surname was missing. A livebirth or fetal death was classified as occurring to an unmarriedmother if the father’s name was missing, if the parents’ sur-names were different, or if the father’s and child’s surnameswere different and the mother’s current surname was missing.These rules are listed in the order in which they were applie~the first applicable rule was coded. Beginning in 1980 thisprocedure was used to infer marital status for all birth certifi-cates in the United States where marital status was not other-wise indicated.26

The names and addresses of the married mothers wereobtained from the vital records so that questionnaires couldbe mailed to them. Most vital records provide only the mother’smaiden name, so it was necessary to infer the mother’s currentlast name. Generally it was assumed that the name of a marriedmother consisted of her first name, her maiden last name,and the father’s surname.

Hospitals

For those live births and fetal deaths that occurred ina hospital or en route to a hospital, the hospital’s patientrecords represented another source of information about laborand delivery, the health of the mother and infant, and prenatalcare, The name and address of the hospital were extractedfrom the vital record so that a questionnaire could be sent.

Attendants at detivery

The person listed on the vital record as the attendantat delivery was also a potential source of additional informationabout the mother’s prenatal care. Only attendants at deliverywho were medical persons (for example, physicians, physi-cians’ assistants, or certified nurse-midwives) were includedas sources of prenatal care information. Vital records for

which the attendant at delivery could not be identified wereeliminated from the samples as described previously. In manycases it was necessary to obtain the attendant’s address frommedical directories or other sources. If the address for theattendant at delivery was the same as that of the hospital,it was assumed that the person was on the hospital staffand that he or she could not provide information about prenatalcare beyond that available from the hospital’s patient records.If the address for the attendant at delivery was different fromthat of the hospital, the name and address of the attendantat delivery were recorded so that a questionnaire could besent.

Secondary sources of information

Mothers, hospitals, and attendants at delivery were re-quested to identify other sources of specific medical informa-tion. The mothers were asked for the names and addressesof any medical sources that provided them with radiationexaminations or treatments during the year preceding their1980 deliveries. Likewise, the questionnaires sent to the hospi-tals and attendants at delivery requested the identity of medicalsources that could provide additional detail concerning themother’s exposure to radiation. If a questionnaire had notalready been sent to these sources, one that dealt exclusivelywith radiation examinations and treatments was sent. Thesesources are referred to as “secondary radiation sources.” Theterm “secondary” indicates that the source was identified byone of the other sources in the survey instead of being takenfrom the vital record.

The questionnaires for hospitals and physicians also re-quested the identity of other sources that might provide addi-

tional information about the mother’s prenatal care. If thesesources had not already been sent a questionnaire, a copyof the questionnaire intended for attendants at delivery wassent. These sources are referred to as “secondary attendant-at-delivery sources.”

Numbers of sources in the NNS and NFMS

The number of sources included in the surveys accord-ing to the marital status of the mother are presented intable C. There were 9,941 sample certificates of live birthincluded in the NNS; 7,825 of the mothers were marriedand 2,116 were unmarried. Of the 6,386 reports of fetaldeath included in the NFMS sample, 4,814 of the motherswere married and 1,572 were unmarried.

In addition to the vital records, there were about 45,000other potential sources of information in the two surveys.Questionnaires were sent to all of the married mothers inboth surveys. There were potential hospital sources for about

99 percent of the live births and fetal deaths. There werepotential attendant-at-delivery sources for about 80 percent

of the live births and 77 percent of the fetal deaths.

There were no secondary radiation or attendant-at-deliverysources for most live births and fetal deaths. Multiple second-ary radiation sources were permitted, and there were as manyas four for some cases. Only one secondary attendant-at-

delivery source was permitted for each case. The vital recordwas the only source of information in those cases involvingunmarried mothers who delivered at home without a medical

5

Table C. Number of potential eouroea of information in the 19S0 NationalNataStySurvey (NNS) and the 1S60 National Fetal Mortelii Survey(NFMS) by the type of aouree and marital atatua of the mother

Number of soumes

Source and marital status of the mother NNS NFMS

Vital records

Total . . . . . . . . . . . . . . . . . . . . . . . .

Married . . . . . . . . . . . . . . . . . . . . . .Unmarried. . . . . . . . . . . . . . . . . . . . .

Mothers

Married . . . . . . . . . . . . . . . . . . . . . .

Hospitals

Total . . . . . . . . . . . . . . . . . . . . . . . .

Married . . . . . . . . . . . . . . . . . . . . . .Unmarried. . . . . . . . . . . . . . . . . . . . .

Attendantsat delivery

Total . . . . . . . . . . . . . . . . . . . . . . . .

Married . . . . . . . . . . . . . . . . . . . . . .Unmarried. . . . . . . . . . . . . . . . . . . . .

Secondaryradiationsources

Total . . . . . . . . . . . . . . . . . . . . . . . .

Married . . . . . . . . . . . . . . . . . . . . . .Unmarried. . . . . . . . . . . . . . . . . . . . .

Seconda~ attendant-at-deliverysources

Total . . . . . . . . . . . . . . . . . . . . . . . .

Married . . . . . . . . . . . . . . . . . . . . . .Unmarried. . . . . . . . . . . . . . . . . . . . .

9,941

7,8252,116

7,825

9,855

7,7652,090

7,939

6,5801,359

1,443

1,42914

430

315115

6,386

4,8141,572

4,814

6,297

4,7621,535

4,926

3,9161,010

1,089

1,07118

375

270105

attendant. There were 9 such cases in the NNS and 20 intheNFMS.

Questionnaires

The same questionnaire forms were used for live birthsin the NNS and for fetal deaths in the NFMS. The samecover letters were used for medical sources in both surveys,but the cover letter for mothers in the NNS was differentfrom that for mothers inthe NFMS. The questionnaires andthe cover letters that accompanied the first mailing to eachsourceareshownin appendixfiguresIII throughXII.

Mother questionnaire

Thequestionnairethat wasmailedtomothers ,the’’M-CS”form, includes questions about medical care received beforedelivery, health-related practices before and during pregnancy,pregnancy history, marital history, wontedness of the preg-nancy, expectations for additional children, and methods offeeding the newborn. Instructions toskipinappropriate ques-tions, such as methods of feeding, are provided for womenwhohad a fetal death. Themother questionnaire also containsquestions about the socioeconomic characteristics of the motherand father, including education, occupation, income, and na-tional origin or descent. Also included are questions aboutwhether the mother received examinations and treatments

involving exposure to radiation during the year precedingher 1980 delivery, and space is provided for the mother tolist the names and addresses of the providers of these examina-tions and treatments.

A pretest of data collection methods for the 1980NNSand NFMS indicated that medical sources were more likelyto supply information when consent statements signed by the

24Therefore, a consent statement formpatient were provided.was included in the questionnaires sent to married mothers.The following consent statement was included at the endof the questionnaire for mothers to read and voluntarily sign:

1 have voluntarily participated in this national health surveyand hereby give my consent for the National Center for HealthStatistics to obtain supplemental medical information from healthrecords maintained on me by medical sources. I understandthat the National Center for Health Statistics will usethis infer-mation only for statistical purposes in health research, and noinformation which identifies either me or the medical sourcewill ever be released or published.

Mothers were initially sent an M–CS questionnaire by

maiI. If they dld not respond within 4 weeks, they weresent a second questionnaire, If the mother failed to respondto both questionnaires, an attempt was made to conduct aninterview by telephone using an M–CS Abbreviated TelephoneInterview. Some questions on the mother’s M-CS mail ques-tionnaire were excluded to reduce the time required for theinterview.

At the conclusion of these telephone interviews, the con-sent statement was read to the mother along with this question:“Do you agree with the consent statement which I just read?”The telephone interviewer recorded whether or not the motheragreed. If the mother agreed, the interviewer signed and datedthe form certifying that the consent statement had been readto the mother and that she had agreed. The signature ofthe interviewer represented the mother’s consent by proxy;thus the consent statements obtained by telephone are referredto as “proxy consent statements.”

Hospital questionnaire

The hospital questionnaire, the “H’ form, seeks informa-tion about the pregnancy, the labor and delivery episode,and the condition of the infant or fetus at delivery. It includesquestions about the induction and duration of labor, use ofanesthetics, type of delivery, underlying medical conditions,complications of pregnancy and labor, and postpartum sterili-zation. Questions about the use of electronic fetal monitoring,Apgar scores, length and weight, and congenital anomaliesand other conditions are included for live births and fetaldeaths as appropriate. The hospital questionnaire also containsall of the questions from the attendant-at-delivery questionnairedescribed below.

Attendant-atdetivety questionnaire

The questionnaire that was sent to the attendant at delivery,the “P form, obtained information on the medical care pro-vided to mothers before delivery. The questionnaire seeksinformation about the dates of prenatal visits, blood pressurereadings, tests for urine protein, hematocrit and hemoglobin

6

values, amniocentesis, and weight gain. This questionnairealso contains detailed questions about med]cal x ray,ultrasound, and nuclear medicine procedures, as well as queriesregarding short wave and microwave examinations and treat-ments, during the 12 months preceding delivery. In connectionwith each of these procedures, the respondent was asked toindicate whether the procedure had been performed at theirlocation or elsewhere. If elsewhere, space was allowed forthe name and address of the provider.

Radiation questionnaire

A questionnaire concerning radiation examinations andtreatments, the “X” form, was sent to the secondary radiationsources. This questionnaire also contains the questions aboutmedical x ray, ultrasound, nuclear medicine, short wave, andmicrowave procedures, as well as detailed questions aboutdental x rays.

Collection of survey data

Contacting mothers

Each married mother was sent a questionnaire (M–CS),an introductory cover letter, and a postage-paid return envelopeby first-class mail. These were mailed as soon as possibleafter the date of delivery to minimize problems of recall.The first mailing to mothers whose deliveries occurred inJanuary 1980 took place in June 1980. Mothers were instructedto complete the questionnaire and to return it in the accompany-ing postage-paid envelope. If there was no response after4 weeks, another questionnaire was sent with a followupcover letter and another return envelope.

If there was no response after an additional 4 weeks,an effort was made to contact the mother by telephone. Ifa telephone number was obtained, as many as seven attemptswere made to reach the mother over a 2-week period. About58 percent of the mothers who had not responded to twomailings were contacted by telephone. If the mother wascontacted, an attempt was made to conduct the M-CS Ab-breviated Telephone Interview. These interviews were con-ducted by trained interviewers who were sensitive to the factthat some of the mothers had recently experienced a fetaldeath. Only the mothers were interviewed; no proxy interviewswere accepted. Information was obtained by telephone from84 percent of the mothers who were contacted. Therefore,a telephone interview was conducted with about half of themothers who did not respond to the two mailed questionnaires,

If it was not possible to reach the mother by telephone,a questionnaire, cover letter, and return envelope were mailedfor the third time. The last of these mailings to motherswas sent in September 1981.

When blank questionnaires were returned by mothers whodid not wish to participate, no further attempts were madeto contact these mothers. In some cases the questionnairesmailed to mothers were returned by the Post Office as undeliv-erable. Early in the surveys, an attempt was made to obtainnew addresses for these mothers. It became necessary, how-ever, to eliminate this time-consuming process. After February1981, cases for which the mother’s questionnaire was returnedby the Post Office were eliminated from the surveys as dis-cussed under “Sampling of certificates of live birth” and “Sam-pling of reports of fetal death.”

When a mail questionnaire or an abbreviated telephoneinterview containing information was obtained, it was thor-oughly edited. The editors identified inappropriate and incon-

sistent responses and recoded or reformatted certain responseson the preceded forms for keypunching. If certain questionswere not answered completely, an attempt was made to recon-tact the mother by mail or by telephone to obtain the missinginformation. The editors also examined the consent statement.

If the mother gave her consent by signing the statement onthe M-CS questionnaire or by agreeing with the statementread to her as part of the M–CS Abbreviated Telephone Inter-view, copies of the consent statement were made so thatthey could be included with questionnaires sent to her medicalsources.

The distribution of married mothers included in the twosurveys by their response status is presented in table D. “Re-spondents” include only those who provided at least someof the information requested. Among married mothers includedin the NNS, 79.5 percent provided information. In the NFMS,74.5 percent of the married mothers responded. The responsesof 36.0 percent of the mothers in the NNS and 37.1 percentof the mothers in the NFMS were attributed to the first mailingof the questionnaire. The responses of an additional 17.1percent of the mothers in the NNS and 14.9 percent of themothers in the NFMS were attributed to the second mailing.Substantial proportions of the mothers provided informationin a telephone interview (23.2 and 20.1 percent, in the respec-tive surveys). The responses of 3.2 percent of the mothersin the NNS and 2.3 percent of the mothers in the NFMS

were attributed to the third mailing. In the NNS. 20.5 percentof the married mothers were nonrespondents, including 3.7

Tabte D. Number of married rnothere included in the 1980 National Natalii Survey and National Fetal Mortalii Survey end percent distribution byresponse status

Number of Response statusmarried

Respondentsmothers

in the All First Second Third Non-Survey suwey Total respondents mailing mailing Telephone mailing reqwrrdents

Percent distribution

NationalNatalltySurvey . . . . . . . . . . . . . . . 7,825 100.0 79.5 36.0 17.1 23.2 3.2 20.5

National Fetal Mortality Survey . . . . . . . . . . . . 4,814 100.0 74.5 37.1 14.9 20.1 2.3 25.5

7

percent of the mothers who refused or returned blank question-naires, 2, I percent for whom there were early Post Officereturns or who were deceased, and 14.7 percent for whomthere was no response of any kind. In the NFMS, 25.5 percentof the married mothers were nonrespondents, includ]ng 6.4percent of the mothers who refused, 2.7 percent for whomthere were early Post Office returns or who were deceased,and 16.5 percent for whom there was no response.

Contacting hospitals and attendants at delivery

Questionnaires were mailed to the hospitals and attendantsat delivery for both married and unmarried mothers. Theinitiation of mailings to these medical sources was determinedby the availability of a consent statement from the mother.Because there were no attempts to obtain consent statementsfrom unmarried mothers, mailings to medical sources for thesemothers were begun as soon as possible after the deliverieswere selected for inclusion in the surveys. These mailingswere sent from June 1980 to June 1981.

The initiation of mailings to medical sources for marriedmothers was determined by the mother’s response to the con-sent statement. If the mother returned a signed consent state-ment by mail, or if a proxy consent statement was obtainedthrough a telephone interview, mailings to the mother’s medi-cal sources were sent with a copy of the consent statementincluded. If the mother indicated that she did not wish tohave one or more of her medical sources contacted, her requestwas respected. If a consent statement was returned blank,mailings to medical sources were begun without a consentstatement. Finally, if there was no response from the mother3 weeks after the last attempt to contact her, mailings weresent to her medical sources without a consent statement. Mail-ings to medical sources for married mothers were sent fromJune 1980 to October1!381.

Medical sources were sent a hospital or attendant-at-delivery questionnaire (as appropriate), an introductory coverletter, a postage-paid return envelope, and a copy of the

mother’s consent statement, if available, by first-class mail.The recipients were requested to consult their records, tocomplete the questionnaire, and to return the questionnairewithin 1 week. If there was no response after 4 weeks, asecond set of questionnaire materials was mailed. If therewas no response after 2 additional weeks, an effort was madeto contact the hospital or the physician by telephone. Thetelephone calls encouraged these sources to complete andreturn the questionnaires that they had already received; therewas no attempt to obtain questionnaire information over thephone. If the source requested an additional copy of the ques-tionnaire, a third set of questionnaire materials was sent.The last of these mailings to medical sources was sent inDecember 1981.

Some questionnaires were returned by the Post Officeas undeliverable. In most cases it was possible to obtaina new address and to remail the questionnaire, but in someinstances it was learned that the medical facility was closedor that the attendant at delive~ had retired or moved. Blankquestionnaires were also returned by sources that refused toprovide information, by sources that refused to provide infor-mation without a consent statement (or without completionof their own form of consent statement), and by sourcesthat indicated that they had no records for the delivery inquestion.

Hospital and attendant-at-delivery questionnaires contain-ing information were thoroughly edited. As with the motherquestionnaires, the editors identified inappropriate and incon-sistent responses and recoded or reformatted certain responseson the preceded forms for keypunching. If specific questionswere not answered completely, another copy of the question-naire was mailed to the respondent with the required informa-tion noted.

The response rates for the medical sources included inthe two surveys according to the mother’s marital status are

.shown in table E. Medical sources that were not contacted

Table E, Number of hospitals and attendants at delivery in the 1980 Natiinal Netelii and Fetal Mortalii Surveys and percent distribution by responsestatus, according to the marital statua of the mother

Response statusNumber of

soursesRespondents

in the All First Second Non-Survey, source, and marital status of the mother survey Total respondents mailing mailing Telephonel respondents

National Natality Survey Percent distribution

Hospitals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9,a55 100.0 76.1 50.7 18.3Married . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7.1 23.9

7,765 100.0 77.6 52.8 18.6 6.2 22.4Unmarried . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2,090 100.0 70.8 43.1 17.1 10.6 29.2

Attendants at delivery . . . . . . . . . . . . . . . . . . . . . . . . . 7,939 100.0 61.6 41.1 14.5Marriad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

6.0 38.46,580 100.0 84.3 43.3 15.2

Unmarried . . . . . . . . . . . . . . . . . . . . . . . . . . . .5.8 35.7

1,359 100.0 48.6 30.5 11.3 6.8 51.4

National Fatal Mortality Survey

Hospitals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6,297 100.0 74.0 49.2 18.2 6.5 26.0Marriad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4,762 100.0 76.3 51.5 18.3 6.5 23.7Unmarried . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,535 100.0 66.7 42.0 18.0 6.7 33.3

Attendants atdelivery . . . . . . . . . . . . . . . . . . . . . . . . . 4,928 100.0 55.2 35.3 14.6 5.4 44.8Married . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,916 100.0 57.9 37.7 15.0 . .Unmarried . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5.2 42.1

1,010 100.0 44.6 25.6 13.0 5.9 55.4

lAmongmsdical sources, telephcmerespondents represent those whotid not return mail questionnaires until atter the telephone reminder.

8

Table F. Numkrof seconda~ratiation sources andsecon&~ atien&nt-at-dehve~ sources inthe1980 Na~mnalNatah~and Fetal Motiah~Suweysandpercent distribution by response status, accordirtg tothemantal atatusof the mother

ResDonse statusNumber of

sourcesRespondents

in the All First Second Non-Survey, source, and marital status of the mother survey Total respondents mailing mailing Telephone’ respondents

National Natality Survey Percent distribution

Secondary radiation sources . . . . . . . . . . . . . . . . . . . . . 1,443 100.0 79.8 50.8 22.9 8.2 20.2Married . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,429 100.0 79.9 51.0 22.9 6.0 20.1Unmarried . . . . . . . . . . . . . . . . . . . . . . . 14 100.0 71.4 28.6 21.4 21.4 28.6

Secondary attendant-at-delivey sources . . . . . . . . 430 100.0 52.8 25.6 16.0 11.2 47.2

Married . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315 100.0 51.7 28.3 14.0 9.5 48.3

Unmarried . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 100.0 55.7 18.3 21.7 15.7 44.3

National Fetal Mortality Survey

Secondaryradiationsources. . . . . . . . . . . . . . . . . . . . 1,089 100.0 78.2 48.3 23.6 6.3 21.8Married . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,071 100.0 78.3 48.4 23.7 6.3 21.7Unmarried . . . . . . . . . . . . . . . . . . . . . . . ie 100.0 72.2 44.4 16.7 11.1 27.8

Secondary attendant-at-delivery eources . . . . . . . . . . . . . 375 100.0 55.7 31.7 16.3 7.7 44.3

Married . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270 100.0 57.8 32.2 17.6 7.8 42.2Unmarried . . . . . . . . . . . . . . . . . . . . . . . 105 100.0 50.5 30.5 12.4 7.6 49.5

‘Among medical sourcas, telephona respondents represent those who did not return mail questionnaire until after the telephone reminder.

at the mother’s request are included in table E as nonre-spondents. In both surveys, more than 76percentofthe hospi-tals associated with the deliveries of married mothers returnedquestionnaires containing information. Among hospitals as-sociated with the deliveries of unmarried mothers, the responserates were 70.8 percent for the NNS and 66.7 percent forthe NFMS. The response rates for attendants at delivery whowere associated with the deliveries of married mothers were64.3 percent in the NNS and 57.9 percent in the NFMS.The differences in response rates by the marital status ofthe mother were greater for attendafits at delivery than forhospitals.

The response rates for the attendants at delivery are notan indication of the proportion of respondents to the prenatalcare questions, however. For most live births and fetal deathsthis information was also sought from hospitals and in somecwwsfromsecondary attendant-at-delivery sources.

Contactingsecondary medicalsources

The medical sources were requested to indicate whetherradiation procedures received by the mother had been per-formed at their own location. If the examination or treatmenthad been performed elsewhere, the name and address of theprovider (hospital, physician, clinic) was requested. If themedical source provided most of the information about theprocedure, there was no reason to query the other sourcenamed, If the medical source provided only thetypeofradia-tion examination or treatment and the date on which it wasperformed, and the provider named by the medical sourcewas not alreadya potential source for the mother in question,the providerwas added to the survey as a secondaryradiationsource, Ifthemother named anadditional source that providedradiation examinations or treatments, it was added to thesurvey as a secondary radiation source. In addition, eachhospital and attendant-at-delivety source was requested to

provide the name and address of other facilities or individualsthat could provide additional information about the mother’sprenatal care. The first mailing to these secondary radiationand attendant-at-delivery sources was sent as soon as possibleafter their names and addresses were obtained. Mailings tothese sources consisted of a radiation or attendant-at-deliveryquestionnaire, an introductory cover letter, a postage-paid re-turn envelope, and a consent statement from the mother ifavailable. The followup and editing procedures for these ques-tionnaires were the same as those described for hospital andattendant-at-delivery questionnaires.

The response rates for the secondary medical sourcesare shown in table F. Between 71 and 80 percent of thesecondary radiation sources provided information. The re-sponse rates for these sources are similiar to those for hospitals.The response rates for the secondary attendant-at-deliverysources ranged from 51 to 58 percent.

Data processing and imputation

Individual live births and fetal deaths are the units ofobservation in the 1980 NNS and NFMS data file. The datafor each event include information from a certificate of livebirth or a report of fetal death and information from applicablequestionnaires. The processing of data from the vital records,including the treatment of missing data, is described first.The sequential process of merging each set of questionnairedata with the vital record data is then described.

Processingof vital records

Information on certificates of live birth was classifiedand coded according to the rules set forth in “Vital StatisticsClassification and Coding Instructions for Live Birth Records,1980,” NCHS Instruction Manual, Part 3a. Information onreports of fetal death was classified and coded according to

9

rules set forth in “Vital Statistics Classification and Coding which to resolve the inconsistency. An analysis of the compdra-Instructions for Fetal Death Records, 1980,” NCHS in.struction bility of re@xting appears elsewhere .2xManual, Part 3b. These instruction manuals were modifiedto provide for the coding of a few additional data items inthe NNS and NFMS. The vital record data were keyed tomagnetic tape, and the keying was 100percent verified.

The subsequent processing of the vital record data wasdone by computer. The data were edited for items with missingor invalid values. Data were missing either because the infor-mation was not provided or because the item was not includedon the vital record used by the registration area. Values wereinvalid if they were not within a predetermined range foreach data item. Each missing or invalid value is referredto as an “item nonresponse.” Most of the variation in theproportion of item nonresponse among the vital records dataitems was due to differences in the number of areas thatreported some items (see appendix tables I and II). Therewere no vital records for which all values were missing orinvalid.

The item nonresponses were replaced by appropriatevalues through a “hot deck” imputation procedure. This proce- .dure involves the sequential processing of the vital recorddata file. When an item nonresponse was encountered, itwas replaced with a value for that data item from a previousrecord in the file. In the simplest case, an item nonresponseon the current record was replaced by the value for the samedata item from the immediately preceding record in the file.When place of delivery was not reported, for example, theplace of delivery code from the previous record was assignedto the current record.

A more complex procedure was used to impute mostdata items. This procedure involves the identification of oneor more data items on the vital record that are associatedwith the data item to be imputed; these data items are referredto as “predictor items.” The previous vital record in the filewith similar values for these predictor items was located,and the item nonresponse on the current record was replacedby the value from this previous record. For example, whena nonresponse for the father’s age was encountered, themother’s age was used as a predictor item. The previousrecord in the file with a mother in the same age categorywas located, and the father’s age from this previous recordwas assigned to the current record.

The records for live births were imputed separately fromthose for fetal deaths. Because birth certificates for infantswho weighed less than 2,500 grams were overrepresentedamong the live births, birth certificates for infants who weighedless than 2,500 grams were imputed separately from thosefor infants who weighed 2,500 grams or more.

The vital record data items were imputed before the ques-tionnaires were processed. Information was occasionally sup-plied by the mother or the hospital for a data item that hadbeen imputed on the vital record. In these instances, theimputed value on the vital record was replaced by the reportedvalue from the mother questionnaire or the hospital question-naire. However, if information reported on the vital recordwas inconsistent with information reported by another source,the values were not changedbecausethere was no basis upon

Processing of mother questionnaires

After the mother’s M-CS mail questionnaire and M-CSAbbreviated Telephone Interviews were manually edited, thedata were keyed to magnetic tape for further processing. Thesedata were edited to detect missing and invalid values whichwere treated as item nonresponses. For mothers who respondedto the M-CS Abbreviated Telephone Interview, responsesto the questions that were excluded from the telephone inter-view were also treated as item nonresponses. Each question-naire record was also examined to determine whether responsesto related questions were internally consistent. For example,a woman’s responses about the number of prenatal visits duringeach month of pregnancy were compared with her responseabout whether she had any prenatal care. In some cases theinconsistency was resolved by assuming that one set of re-sponses took precedence. In these cases the precedent re-sponses provided a basis for recoding the inconsistent re-sponses. In other cases the inconsistency could not be resolved,and the inconsistent responses were treated as item nonre-sponses. The questionnaire data for mothers who providedinformation were matched to and merged with the correspond-ing vital record data.

In these surveys the vital record was available for eachcase, but one or more of the potential sources of information(mothers, hospitals, attendants at delive~, secondary radiationsources, and secondary attendant-at-delivery sources) mightnot have responded. The failure of a potential source to respondis referred to as a “source nonresponse.”29 If there was nomother questionnaire information for a live birth or a fetaldeath to a married mother, it was designated as a sourcenonresponse. In the NNS, 20.5 percent of the questionnairesfor married mothers were source nonresponses as were 25.5percent of the questionnaires for married mothers in the NFMS.

The -item nonresponses and source nonresponses werethen imputed. In this process each source nonresponse wastreated as a series of item nonresponses. For some mothers’data items (mother’s age, for example), the value assignedwas drawn directly from the vital record. Mother questionnaireitems for which information was not available from the vitalrecord were imputed using the hot-deck procedures describedabove with data items from the vital record and from thequestionnaire as predictor items. The imputations were carriedout separately for fetal deaths, low-weight live births, andother live births. There was no imputation of mother question-naire data for unmarried mothers.

Processing of hospital andattendant-atdetiiery questionnaires

The hospital and attendant-at-delivery questionnaires werekeyed to separate magnetic tape files. These data were editedfor missing values, invalid values, and internal inconsistencies.Inconsistencies were either resolved or designated as itemnonresponses. The subsequent processing of the questionson radiation exposure is described under “Processing of radia-tion data.” The rest of the information on questionnaires

10

returned by hospitals, attendants at delivery, and secondaryattendants at delivery was combined to produce a single datarecord for each live birth and fetal death for which at leastone of these sources responded. There wasonly one hospitalquestionnaire for each case, but all of the questions aboutprenatal care on the hospital questionnaire were also on theattendant-at-delivery questionnaire. Prenatal care informationcould have been obtained from as many as three sourcesfor some cases (a hospital, an attendant at delivery, and asecondary attendant at delivery). When more than one sourceprovided prenatal care information, the information was com-bined into a single set of responses.

The combination of responses from two or more question-naires was accomplished in one of several ways. If any sourceindicated that hematocrit and hemoglobin values were ob-tained, that information was retained. If two or more sourcesprovided different values for the highest or lowest hematocritor hemoglobin values, the higher and lower values, respec-tively, were retained, If any source indicated that amniocentesiswas performed or that specific prenatal advice was given,that information was retained. If two or more sources providedinformation about the mother’s prepregnancy weight or herweight at time of delivery, an order of precedence was as-sumed: The hospital response was chosen over the responseof an attendant at delivery, which was chosen over the responseof a secondary attendant at delivery. If two or more sourcesprovided weight at first prenatal visitor weight at last prenatalvisit, the lowest and highest values were retained, and itwas assumed that they represented the earliest and latest visits,respectively, An indication of which source supplied eachprenatal care data item was included in the data record.

The data on prenatal visits from two or more questionnaireswere also combined. All visits that were reported on differentdates were retained, When more than one visit was reportedon the same date, information for only one visit was retained.Although the selection criteria were complex, the effect wasgenerally to retain blood pressure readings reported by at-tendant-at-delivery sources, and to retain any indication thatthe result of a urine protein test was positive. Data for amaximum of 30 prenatal care visits were retained for eachlive birth and fetal death.

The combined hospital and attendant-at-delivery data rec-ords were matched to and merged with the correspondhgvital record and mother data records. If the information ob-tained only from hospitals (questions l–l 2 and 19–35 onthe hospital questionnaire) was missing for a live birth ora fetal death that occurred in a hospital or en route to ahospital, this missing information was designated as a sourcenonresponse. The percent of source nonresponses for the dataobtained only from the hospital corresponds to the percentof nonrespondents for hospitals in table E. If there was noprenatal care information for a live birth or fetal death forwhich the vital record indicated that the mother had prenatalcare, this missing information was designated as a sourcenonresponse,

Source nonresponses were treated as a series of itemnonresponses, which were replaced by values from the vitalrecord whenever possible or were imputed in a manner similar

to that described for the mother questionnaires. DaVditemsfrom the vital record and from the hospital or attendant-at-

delivery questionnaires were used as predictor items in theimputation process. These imputations were carried out sepa-rately for fetal deaths, low-weight live births, and other livebirths.

Processingof radiationdata

Questions about the mother’s exposure to radiation wereincluded on the hospital, attendant-at-delivery, and radiationquestionnaires. Once the hospital and attendant-at-deliveryquestionnaires were edited and checked for consistency, theradiation data were separated for processing with the radiationquestionnaire data. The radiation questionnaires were alsokeyed to magnetic tape, edited, and checked for internal consis-tency. The radiation information from all of the sources as-sociated with each live birth and fetal death was combinedinto a single set of responses.

The process of combining responses from two or moresources was designed to retain any indication of exposureto radiation and to retain nonduplicated information aboutindividual radiation procedures. If any of the respondingsources reported that the mother had particular radiation proce-dures (medical x ray, dental x ray, ultrasound, nuclearmedicine, short wave, or microwave) during the 12 monthspreceding delivery, that information was retained. If two ormore sources provided duplicated information about a particu-lar x ray, ultrasound, or nuclear medicine procedure, theinformation was included only once. An indication of whichsource or sources supplied each radiation data item was in-cluded in the data record.

The combined radiation data records were merged withthe data records based on vital records, mother, hospital,and attendant-at-delivery questionnaires. Item nonresponsesamong the radiation data were imputed using the techniquesdescribed previously. The hospital, attendant-at-delivery, sec-ondary attendant-at-delivery, and secondary radiation sourcesthat did not respond were identified, and their radiation ques-tions were designated as source nonresponses. These sourcenonresponses were nottreated as a series of item nonresponses;the radiation data for each source nonresponse was imputedas a block. The mother’s age, race, and her responses tothe questions on radiation exposure were utilized as predictoritems. A hot-deck procedure was used whereby radiation infor-mation from another similar mother with a hospital, attendant-at-delivery, or radiation source was imputed to each sourcenonresponse as appropriate. For example, if the hospital sourcefor the case currently being processed did not respond, theradiation data from a previous similar mother with a hospitalsource was assigned to the current case. The same procedurewas followed for the attendant-at-delivery and radiation sourcenonresponses.

Secondary radiation sources were most often added tothe surveys based on information supplied by the mother.Consequently, there were very few secondary radiation sourcesfor mothers who did not return questionnaires and for unmar-ried mothers who were not sent questionnaires. In order tobetter estimate the extent of radiation exposure for these

11

mothers, secondary radiation source information was imputedfor nonrespondent mothers and for unmarried mothers withoutsecondary radiation sources. In the sequential processing ofthe data records, nonrespcmding mothers and unmarriedmothers were matched with the previous record for a marriedmother of similar age, race, and education. If the previousmother had no secondary radiation source, no secondary radia-tion source data were imputed to the present case. If theprevious mother had a secondary radiation source, the datafrom that source was assigned to the present case.

These imputation procedures were also carried out sepa-rately for fetal deaths, low-weight live births, and other livebirths.

Consistency between responses

One disadvantage of imputing data on an item-by-itembasis is that inconsistencies may be introduced. When twoor more data items are imputed for a particular source, thedata may be drawn from different cases. The predictor itemswere selected to avoid inconsistencies among directly relatedquestions. In some cases the same predictor items were used,and in other cases the response to one data item was employedas a predictor item in the imputation of a related data item.Some inconsistency among responses that are not directlyrelated are inevitable, as are inconsistencies among responsesfrom different sources.

Estimation

Probability sampling allows the data from the NNS andthe NFMS to be weighted to produce national estimates. Italso allows approximation of the sampling errors for theseestimates,

NNS estimation procedure

NNS sampling weights were prepared by a poststratifiedratio estimation procedure. The purpose of a ratio estimationprocedure is to use available independent information to reducethe variability of estimates. Independent information aboutall live births is available from the vital registration system.

This procedure was applied in each of the 50 poststratifica-tion cells, or weighting strata, shown in table G. These cellswere defined in terms of data items from birth certificates:Mother’s marital status and age, and child’s race, live-birthorder, and birth weight. Combinations of these cells formthe major domains of study in the NNS. According to thevital registration system for 1980, certificates for 42,129 of3,612,258 live births lacked response for one or more ofthe stratification variables. For the NNS estimation, certificatesfor these live births were distributed among the poststratifica-tion cells.

The sampling weight for each cell, Wi, is the ratio ofthe number of births to U.S. residents in 1980 to the numberof sample births in the NNS:

Wi = YilJ’i

where Yi = total number of births in the ith poststratificationcell based on the vital registration system, and

yi = total number of sample births in the ith poststratifi-cation cell.

These 50 ratios comprise the sampling weights for the NNS.Thus, the NNS-estimated number of live births for each

poststratification cell is consistent with the correspondingnumber of live births from the vital registration system. Al-though the estimated total number of births for a poststratifica-tion cell is not subject to sampling error, these totals aresubject to nonsampling error (for example, undercoverageof the vital registration system and errors in classifying livebirths whose birth certificate omitted responses for one ormore poststratification variables). However, the nonsamplingerror is considered negligible compared with the samplingen or associated with other NNS estimates.

The number of live births with a characteristic of interest,.f, is estimated from the NNS using the formula

where ~i = sampling weight assigned to sample births inthe ith poststratification cell, and

xi = total number of sample births with the characteris-tic in the ith poststratification cell.

For publication, NNS estimates of aggregates are roundedto the nearest thousand. Therefore, sums of detailed figuresin tables may not always equal the total. Rates and percentagesare calculated using the unrounded estimates,

NFMS estimation procedure

NFMS sampling weights are the products of adjustmentsfor the probability of sample selection, for nonresponse, andfor the lack of fetal death reports from Michigan. Unlikethe NNS estimation procedure, the NFMS estimation proceduredoes not include a poststratified ratio adjustment, The 22NFMS weighting strata shown in table H were defined interms of data items from reports of fetal death: Mother’smarital status and age, and fetus’s race.

The sampling weight in the ith weighting strata, wi, is

1 tl’iWi. _ Ni+ (ll~lp).— .P tij (n; /p)

where p = 2/5, the probability of selection for each samplefetal death,

n’i = number of sample fetal deaths eligible for theNFMS in the ith weighting stratum,

hi = number of sample fetal deaths in the ith weightingstratum, and

Ni = number of 1980 fetal deaths in the ith weightingstratum eIigible for the NFMS in Michigan asestimated from summary data provided by theMichigan Department of Public Health.

12

Tde G. 1960 National Nstality Survey (NNS) postetrstifioetion ceHdefirtitiin~ number of We births in the NNS, number of We births to residents of theUrdted States in 19S0, snd sampling weights

Cell definitionsNumber of Number of

Birth Marital Age of tive births in births in

Cell

Samplingweight status Race of mother in birth the AINS2 19803 weight

number in grams (M or UN)l child years order (Yi) M) (Wj

TotsI . . . . . . . . .

01 . . . . . . . . . .02 . . . . . . . . . .03 . . . . . . . . . .04 . . . . . . . . . .05 . . . . . . . . . .06 . . . . . . . . . .07 . . . . . . . . . .08 . . . . . . . . . .09 . . . . . . . . . .10 . . . . . . . . . .11 . . . . . . . . . .12 . . . . . . . . . .13 . . . . . . . . . .14 . . . . . . . . . .15, ..,,..,..16 . . . . . . . . . .17, . . . . . . . . .18 . . . . . . . . . .19 . . . . . . . . . .20 . . . . . . . . . .21 . . . . . . . . . .22 . . . . . . . . . .23 . . . . . . . . . .24 . . . . . . . . . .25 . . . . . . . . . .26 . . . . . . . . . .27 . . . . . . . . . .26 . . . . . . . . . .29 . . . . . . . . . .30 . . . . . . . . . .31 . . . . . . . . . .32 . . . . . . . . . .33 . . . . . . . . . .34 . . . . . . . . . .35 . . . . . . . . . .36 . . . . . . . . . .37 . . . . . . . . . .36 . . . . . . . . . .39 . . . . . . . . . .40 . ...,.,...41 . . . . . . . . . .42, ...,.....43 . . . . . . . . . .44 . . . . . . . . . .45 . . . . . . . . . .46 . . . . . . . . . .47 . . . . . . . . . .48 . . . . . . . . . .49 . . . . . . . . . .50 . . . . . . . . . .

. . .

2500>2500=2500>2500>2500>2500

>2500

. . .

MMMMMMMMMMMMUNUN

UN

UN

UNUNM

M

M

MM

M

M

MMM

M

M

M

MM

MM

M

MM

MM

M

UNUN

UN

UN

UNUNUN

UN

UN

. . .

White

White

WhiteWhiteWhiteWhiteWhiteWhite

White

Other

OtherOtherWhite

White

White

Other

OtherOtherWhiteWhite

White

WhiteWhite

White

White

WhiteWhiteWhite

White

White

White

WhiteWhiteWhite

Other

Other

OtherOther

OtherOther

Other

WhiteWhite

WhileWhite

WhiteOtherOthar

Other

Other

. . .

Tab&H. 19WNational Fetal Motiak~Suwey (NFMS)weighting cell definitions, number of fetal deaths inthe NFMS, estimated number of fetal d~thswith gestation of 28 weeks or more, or deliiery weight of 1,000 grams or more, to residents of the United States in 1980, and sampling weights

Cell definitionsNumber of Estimated number

Merital Age of fetal deaths of eligible fetal Sampling

status Race of mother in the NFMS deathsin 1980 weight

Cell number (M or UN)’ fetus in years (Y/) (y)’ (W/)

Total . . . . . . . . . . . . . . . . .

01 . . . . . . . . . . . . . . . M

02 . . . . . . . . . . . . . . . M

03 . . . . . . . . . . . . . . . M

04 . . . . . . . . . . . . . . . M

05 . . . . . . . . . . . . . . . M

06 . . . . . . . . . . . . . . . M

07 . . . . . . . . . . . . . . . M

08 . . . . . . . . . . . . . . . M

09 . . . . . . . . . . . . . . . M

lo . . . . . . . . . . . . . . . M

11 . . . . . . . . . . . . . . . M

12 . . . . . . . . . . . . . . . M

13 . . . . . . . . . . . . . . . UN14 . . . . . . . . . . . . . . . UN15 . . . . . . . . . . . . . . . UN

16 . . . . . . . . . . . . . . . UN

17 . . . . . . . . . . . . . . . UN

18 . . . . . . . . . . . . . . . UN

19 . . . . . . . . . . . . . . . UN20 . . . . . . . . . . . . . . . UN

21 . . . . . . . . . . . . . . . UN

22 . . . . . . . . . . . . . . . UN

. . .

White

WhiteWhiteWhite

White

White

White

OtherOther

Other

Other

OtherWhiteWhiteWhite

Whits

White

OtherOther

OtherOther

Other

. .

marital status, Because a minimum of 30 sample births support-ing an estimate is needed for the theory for large samples,estimates less than 10,890, the product of 30 and the averageNNS sampling weight, were excluded from these samples.A minimum number of sample births was similarlycalculatedfor NNS low-weight estimates (3,400). Each sample was thenused to calculate the parameters for the formula

RSE(X) = ~).

Three pairs of A and B parameters for this formula wereadequate to approximate the relative standard error for aggre-

gate NNS estimates. One pair is for estimates for low-weightbirths (less than 2,500 grams), one pair for estimates of eitherunmarried mothers or infants of races other than white, andone pair for all other estimates (all births, births of whiteinfants, and births to married mothers). The standard errorof an estimate can be obtained by multiplying the relativestandard error of the estimate by the estimate itself.

This procedure underestimates the standard error for cer-tain NNS estimates because data for some sources (sourcenonrespondents) were imputed in entirety. Therefore, the stand-ard errors for NNS estimates were adjusted based on theresponse rates for sources included in the survey. Becauseit is not practical to calculate the response rate correspondingto each estimate, a generalized procedure was adopted wherebya multiplicative adjustment factor based on response rates(k) was incorporated in the formulas for the relative standarderrors,

The A and B parameters in table J were resealed sothat they would be appropriate for most NNS estimates basedon mother, hospital, attendant-at-delivery, and radiation ques-tionnaires, For these estimates, the adjustment factor, k, is1. Some questions on the mother’s M-(X mail questionnaire(the M-CS form) were excluded from the mother’s M-CSAbbreviated Telephone Interview, however. The responsesto some of the questions that were excluded from the M-CSAbbreviated Telephone Interview could be inferred; responsesto the other questions were imputed. Estimates based on thelatter responses are therefore based on a larger proportionof imputed responses than estimates based on other questionson the M–CS form. The adjustment factor for estimates basedon the questions on the mother’s M

A, B = parameters from table J or K,

RSE(X) = relative standard error ofx, and

SE(x) = standard error ofx.

These formulas are not appropriate for estimates of the totalnumber of live births in a poststratification cell or in a combina-tion of such cells because these estimates have negligibleerror.

Example: Based on the NNS, it is estimated that 134,646married mothers under 20 years of age smoked cigarettesduring the 12months before their 1980delivery.

The standard error of this estimate is calculated as follows:

RSE(134,646) = 1V( –0.00010438) + (377.81921/ 134,646)

= 0.052

and

SE(134,646) = 134,646 “0.052 = 7,002.

Estimates based on fewer than 30 sample cases are consid-ered unreliable. Based on the average sampling weight, 30sample cases in the NNS correspond to about 3,400 low-weightbirths and 10,890 live births in all other domains; 30 samplecases in the NFMS correspond to about 90 fetal deaths. Theseformulas should not be used to approximate the standard errorfor smaller estimates in each respective NNS and NFMSdomain.

(2) Standard error for ratios or proportions where thedenominator is assumed to have negligible error—Where thedenominator of a ratio is the estimated total number of livebirths in a poststratification cell or in a combination of suchcells, the relative standard error of the ratio is the relativestandard error of the numerator. Thus, the approximate stand-ard errors may be calculated using the formulas

RSE(r) = RSE(x/y) = RSE(X)

and

SE(r) = r . RSE(X),

where r=

x =

Y=

RSE(r) =

SE(r) =

RSE(X) =

ratio or proportion,

numerator of the ratio,

denominator of the ratiowith negligible error,

relative standard error of r,

standard error of r, and

relative standard error ofx.

Example: An estimated 46.5 percent (134,646) of the marriedteenage mothers (289,677) smoked cigarettes during the 12months before delivery. The number of teenage marriedmothers is a combination of the poststratification cells specifiedin table G.

Therefore,

RSE(46.5) =

=

and

SE(46.5) =

RSE( 134,646/289,677) = RSE( 134,646)

0.052 (from the prior example)

46.5 “0.052 = 2.4.

Estimates based on fewer than 30 sample cases are consid-ered unreliable. Based on the average sampling weight, 30sample cases in the NNS correspond to about 3,400 low-weightbirths and 10,890 live births in all other domains; 30 samplecases in the NFMS correspond to about 90 fetal deaths. Theseformulas should not be used to approximate the standard errorfor smaller estimates in each respective NNS and NFMSdomain.