Persons using assistive technology might not be able to fully access information in this file. For assistance, please send e-mail to: firstname.lastname@example.org. Type 508 Accommodation in the subject line of e-mail.
National Infant Mortality Surveillance (NIMS) 1980
This report presents data from the National Infant Mortality Surveillance (NIMS) project. The NIMS project was a collaborative effort between the Public Health Service and states to address the issue of infant mortality. Factors that affected the risk of infant mortality for single-born infants included birthweight, race, sex, gestation, birth order, maternal age and education, and prenatal care. The most important predictor for infant survival was birthweight, with improved survival for both blacks and whites associated with increased birthweights. Overall, black infants had twice the mortality risk of white infants. The higher risk for blacks was related to higher prevalence of low birthweight and to higher mortality risks in both the neonatal and postneonatal periods. In general, the black-white differential exists regardless of other infant and maternal characteristics.
Findings from NIMS have been published in a special section of Public Health Reports (March-April 1987) and elsewhere and have been presented at several national meetings. Foreword
After two decades of consistent, gratifying improvements in levels of infant mortality in the United States, progress has slowed substantially during the 1980s. In fact, progress has lagged behind that of other developed countries; among the countries of the developed world, the United States now ranks 19th in infant survival. Even more disturbing is that essentially no progress has been made in reducing the twofold risk of infant mortality among blacks as compared with whites.
The nation is mounting new efforts to regain the momentum toward achieving our maternal and infant health objectives. To ensure the maximum use of limited resources, we need to assess current programs, evaluate new ones, and determine unmet needs. So that there will be a solid basis for making and justifying our decisions, this assessment requires good information that links infant health outcomes with the use of the programs.
Before the National Infant Mortality Surveillance (NIMS) project, the most recent national linkage of birth and infant death records had been done in 1960 by the National Center for Health Statistics (NCHS). Since that year, however, there had been no national benchmark to evaluate information on infant survival within specific birthweight categories and related to other key maternal and infant factors. The linkage of vital records offers a number of substantial opportunities other than providing infant mortality data. For example, such linkage provides the base upon which a state can overlay the characteristics of participation in a program such as Medicaid; Women, Infants and Children (WIC); prenatal outreach; and others to evaluate its coverage and impact. Since this archival effort to produce national data for 1980, NCHS has begun annual linkage of birth and infant death records, beginning with the birth cohort of 1983. The NIMS data provide a valuable baseline for the systematic and regular monitoring and evaluation of program effectiveness, while the new system will supply the data for continuous measurement of our progress toward achieving the Year 2000 Objectives for Mothers and Infants. James S. Marks, M.D., M.P.H. Deputy Director for Public Health Practice Center for Chronic Disease Prevention and Health Promotion Partial support for this project was provided by the National Institute of Child Health and Human Development, the Health Research and Services Administration, and the National Center for Health Statistics, CDC. Overview of the National Infant Mortality Surveillance (NIMS) Project
Carol J.R. Hogue, PhD, MPH
Lilo T. Strauss, MA James W. Buehler, MD
Jack C. Smith, MS Summary
A slowdown in the decline of infant mortality in the United States and a continuing high risk of death among black infants (twice that of white infants) prompted a consortium of Public Health Service agencies, in collaboration with all states, to develop a national data base of linked birth and infant death certificates for the 1980 birth cohort. This project, referred to as National Infant Mortality Surveillance (NIMS), provides neonatal, postneonatal, and infant mortality risks for blacks, whites, and all races in 12 categories of birthweights.* *Neonatal mortality risk = number of deaths of infants less than 28 days of age/1,000 live births; postneonatal mortality risk = number of deaths of infants ages 28 days up to 1 year/1,000 neonatal survivors; and infant mortality risk = number of deaths of infants less than 1 year of age/1,000 live births.
Tabulations were requested for infants born in single and multiple deliveries. For single-delivery births, tabulations included birthweight, age at death, race of infant, and each of these characteristics: infant's live-birth order, sex, gestation, type of delivery, and cause of death; and mother's age, education, prenatal care history, and number of prior fetal losses at greater than or equal to 20 weeks' gestation. An estimated 95% of eligible infant deaths were included in the NIMS tabulations. Analyses have focused on various components of infant mortality, including birthweight distribution of live births, neonatal mortality, and postneonatal mortality.
The most important predictor for infant survival is birthweight; survival increases exponentially as birthweight increases to its optimal level. The nearly twofold higher risk of infant mortality among blacks than among whites was related to a higher prevalence of low birthweights, to higher mortality risks in the neonatal period for infants with birthweights of greater than or equal to 3,000 g, and to higher mortality during the postneonatal period for all infants, regardless of birthweight. Moreover, the black-white gap persisted for infants with birthweight of greater than or equal to 2,500 g, regardless of other infant or maternal risk factors. INTRODUCTION
Each year, approximately 40,000 U.S. infants die before reaching their first birthday. The 1990 Objectives for the Nation call for an infant mortality rate of no more than 12 deaths/1,000 live-born infants of any racial group for an overall national infant mortality rate of no more than 9 deaths/1,000 live-born infants (1). In 1986, the infant mortality rate was 18.0/1,000 live-born black infants and 8.9/1,000 live-born white infants (2). It is thus unlikely that the United States will achieve the 1990 objective for black infants, especially since black infant mortality rates decreased only 15.9% from 1980 (3) to 1986; to meet the 1990 objective, the rate for these infants would have to be reduced by 33.3% within the 4 years that remain in the period.
Three observations should be kept in mind as the United States strives to attain the 1990 objective for infant mortality. First, success in improving survival for low-birthweight (i.e., less than 2,500 g) infants has not been accompanied by a parallel decrease in the frequency of low birthweight--the most important determinant of infant survival (4-11). Of all infants born in 1980 who died, more than 60% were of low birthweight (12). Second, other industrialized nations have already achieved lower infant mortality rates (11,13,14). Third, although mortality rates have decreased for both white and black infants, neonatal and postneonatal mortality rates for black infants continue to be approximately twice the rates for white infants (8,15,16).
Epidemiologic analyses of low birthweight and infant mortality can guide health planners in developing and assessing interventions to reduce infant mortality. By identifying problems in maternal and infant care, health planners can target high-risk groups for more intensive interventions. Calculation of mortality risks by birthweight and other characteristics--which is essential to such analyses--requires linkage of individual birth and death certificates. Despite its importance, such linkage had not been done nationally since that done for infants born in 1960 (17).
The National Center for Health Statistics (NCHS), CDC, has an ongoing project to produce an annual, national, linked birth and infant-death file beginning with the 1983 birth cohort. The 1983 and 1984 files were released in 1989; the birth-cohort files for 1985 and 1986 are scheduled for publication over the next few years (18). To provide a more immediate data source between 1960 and future national linkages and to identify issues to be addressed before a national linked file could be implemented, CDC implemented the 1980 NIMS project. The sections below present the history and rationale of the NIMS project, its design and methods, and basic analyses of differences in infant mortality for blacks and for whites.
More details on the design, methods, and results of basic analyses are provided in an unpublished document, the unabridged NIMS report. Numerous references are made to the unabridged report in this surveillance summary. MMWR subscribers who need the longer, more detailed NIMS report may request a copy by writing to the Pregnancy Epidemiology Branch, Division of Reproductive Health, Mailstop C06, Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control, Atlanta, GA 30333. HISTORY AND RATIONALE
In May 1983, CDC sponsored a meeting with representatives of the Association for Vital Records and Health Statistics, state directors of Maternal and Child Health, the American Academy of Pediatrics, the American College of Obstetricians and Gynecologists, NCHS, and the National Institute of Child Health and Human Development (NICHD). The participants agreed to assemble a national, birthweight-specific infant mortality data base by means of collecting data from all states with available linked birth and infant death records. CDC received partial funding from NICHD, NCHS, and the Health Resources and Services Administration (HRSA). Key to this project was the full cooperation that CDC received from all states and other vital registration reporting areas.
States provided necessary programming support, computer time, and professional and clerical time. To minimize the workload for approximately half the states that had linked certificates but did not have the data in machine-readable records, CDC requested tabular data from all the states rather than individually linked records on computer tape. This data-collection method enabled all the states to participate in the surveillance and allowed this national compilation to be done relatively quickly.
The planning group recommended collecting birthweight-specific infant mortality data for the cohort of infants born in 1980. Selection of that year allowed sufficient time for linkages to have been completed and allowed a 20-year follow-up from the most recent national linkage done by NCHS (for infants born in 1960).
Finally, the planning participants urged a prompt dissemination of the information. The mechanisms to accomplish this recommendation included a NIMS Conference (May 1-2, 1986), preliminary data analyses (12,19), a preliminary version of this NIMS report--with detailed tabulations and methods--distributed at the NIMS Conference and on request by mail, 12 scientific articles in a special issue of Public Health Reports (20), followed by others (21,22), and a public-use tape (to be released in 1990). The public-use tape and documentation will be available through the National Technical Information Service. DESIGN AND METHODS
Fifty-three vital statistics reporting areas participated in the NIMS project: 50 states, New York City, the District of Columbia, and Puerto Rico (subsequently referred to as "states"). The national-level tabulations do not include Puerto Rico. All states were able to link birth and death certificates for infants who were born alive in 1980 and who died within the first year of life (death in 1980 or 1981). We requested each state to provide data for all infant deaths for resident births in 1980, regardless of state of residence at death. State of residence was defined as state of mother's residence at the time she gave birth. The request was limited to data for live births involving birthweights of greater than or equal to 500 g. Assuming that all infants weighing less than 500 g died during the neonatal period, we added data for those births to the death data provided by the states; this sum gave us the total number of infant deaths.
Information on all resident births for each state serves as the denominator for the calculation of infant mortality risks. To reduce the reporting burden for each state, CDC generated denominators for states (except Maine and New Mexico, as described in Appendix II of the unabridged report) from the computer tape of 1980 natality records produced by NCHS. We also used the NCHS natality tape for information on live births of less than 500 g.
Because the NIMS data set is for deaths occurring within a birth cohort rather than for births and deaths occurring in a given year, we use the term mortality "risk" instead of "rate." The neonatal mortality risk (NMR) was defined as the number of neonatal deaths (survival of less than 28 days)/1,000 live births, the postneonatal mortality risk (PNMR) as the number of postneonatal deaths (survival 28 days up to 1 year)/1,000 neonatal survivors, and the infant mortality risk (IMR) as the number of infant deaths (survival less than 1 year)/1,000 live births.
Birthweights were divided into 250-g intervals for infants born weighing 500 g1,499 g, into 500-g intervals for those between 1,500 g and 4,499 g, into a less than 500-g interval for those between 227 g and 499 g, and into a greater than or equal to 4,500-g interval for infants born weighing 4,500 g-8,165 g. Birthweights of less than 227 g (8 oz), those of greater than 8,165 g (18 lb, 0 oz), and missing values were included with unknowns. All references to weight in this report refer to weight at birth.
States provided CDC with the number of infant deaths by birthweight and age at death (neonatal and postneonatal) for single deliveries, by race (blacks, whites, and all races), and by plurality. The NCHS algorithm was used to determine race of infant according to the race of both parents (23). Hispanics have emerged as a major ethnic group in the United States having distinct reproductive life histories (24-28). However, because states use a variety of methods to define Hispanics (24-26), a national study of Hispanics could not be done with the NIMS data. Puerto Ricans are described in a separate report (29), and information on Hispanics in California and New York City is provided in Appendix V of the unabridged report. Native Americans, Asians, and Pacific Islanders in selected states have also been analyzed (21 and CDC unpublished data).
We requested separate tabulations for infants born in single and in multiple deliveries. For single deliveries, we obtained tabulations by birthweight, age at death, race, and each of the following characteristics: infant's live-birth order, sex, gestation, type of delivery (vaginal or cesarean section), and the specific underlying cause of death (using the International Classification of Diseases, 9th Revision) (30); and mother's age, education, prenatal care history, and number of previous fetal losses at greater than or equal to 20 weeks' gestation. Type of delivery is not recorded on the NCHS natality tape; therefore, we have death data only for type of delivery.
Definitions used for the tabulation of these characteristics paralleled as much as possible those for natality statistics used by NCHS. Details of the definitions for each characteristic are provided in Appendix I, unabridged report. Occasionally, states had some difficulty using the definitions (see Appendix II and 31). Presentation of mortality risks for individual characteristics includes only those data for states that reported the given variable substantially as the definitions provided (see Appendix III.B, unabridged report). The number of states excluded from particular tables varied from two for infant's sex to 19 for gestational age of infants born in multiple deliveries (see Appendix III.B, Table A-IV, unabridged report).
Information on the number of prenatal care visits and the month prenatal care began was tabulated by birthweight rather than by gestational age. Although number and timing of prenatal care visits are confounded by duration of gestation, gestational age is one of the least reliable variables obtained from birth certificates and is missing for approximately 20% of births (32).
After states provide natality data to NCHS, final processing of data occurs, such as imputing unknown or extreme values for selected variables. For births in 1980, NCHS assigned a value by imputation for unknown race and maternal age. In obtaining data for deaths by these characteristics (race and maternal age) from individual states, we could not recreate this imputation procedure. However, we did use the NCHS method of assigning unknown plurality to the category of single delivery. In NIMS, unknown race was included only in tabulations for all races and was not distributed to other racial groups. We assigned unknown maternal age to known categories proportional to the distribution of known values. Unknown values of other maternal and infant characteristics were handled similarly (see Appendix III.A, unabridged report). Unknown race is rare, since race for both parents must be unknown before the child's race is classified that way. Exclusive of data for New Mexico (see Appendix II, unabridged report), NCHS imputed race for only 0.36% of births.
For calculation of mortality risks, we assigned infants with unknown birthweight (0.2% of births and 3.3% of infant deaths) to birthweight categories according to the proportion of births and deaths with known birthweight. Appendix III of the unabridged report includes details of the procedure used.
To promote uniformity in reporting and tabulating deaths, we prepared a detailed set of instructions for the states and provided extensive technical assistance via telephone (31). After receiving the data, we conducted editing checks that included visual examination and graphing of birthweight distributions, birthweight-specific neonatal mortality, and birthweight-specific postneonatal mortality for each state. States provided information regarding the death certificates that they could not link (33,34). To estimate both state-specific and total underreporting of infant mortality related to failure to link death certificates with corresponding birth certificates, we produced a synthetic cohort of infant deaths among infants born in 1980 (see Appendix IV.B, unabridged report, and 33,34). We obtained the expected number of deaths by selecting from NCHS annual mortality tapes those deaths in 1980 and 1981 that occurred among U.S. residents less than 1 year of age who were born in 1980 (see Appendix IV.B, unabridged report). The infant death certificate does not have the infant's residence at birth but does include the infant's residence at death and place of birth and death. Several estimates of unreported deaths by state of residence at birth can be obtained by using different assumptions of the correlation of each of these three items to residence at birth (see Appendix IV, unabridged report, and 33,34). One such analysis is presented in this overview.
Our analyses of U.S. infant mortality focus on three components: birthweight distribution of live births, neonatal mortality, and postneonatal mortality. We treat these components separately because interventions aimed at improving each one differ substantially. For example, reducing low birthweight requires identifying the risk factors for low birthweight that are present before pregnancy and during the prenatal period. On the other hand, reducing neonatal mortality focuses not only on raising birthweight but also on improving intrapartum and newborn care. The latter improvement entails regionalizing perinatal services and identifying infants who might be at high risk of mortality following their release from the hospital. Intervention to reduce postneonatal mortality focuses on improvement of well- and sick-child care and on intensive follow-up of infants at high risk of postneonatal mortality. We also examine the birthweight distribution of neonatal, postneonatal, and infant deaths to determine how deaths of lighter and heavier infants contribute to overall IMR.
The analysis presented here can focus attention on those groups of infants for whom improvement in survival would have the greatest overall impact on reducing infant mortality. Results are for single-delivery births only. Infants born in multiple deliveries are included in the report but are not represented in our results. This group, which accounted for 10% of deaths of both black and white infants (19), may be the subject of forthcoming analysis. RESULTS Birthweight Distribution
In 1980, there were 3,542,995 single deliveries (Table 1). Infants with birthweight of less than 1,500 g comprised 1.0% of all live births but comprised 2.1% of black live births. Likewise, infants in the intermediate low-birthweight category of 1,500 g-2,499 g comprised 5.0% of all births but 9.2% of black births. This birthweight discrepancy indicated that although blacks comprised 16.2% of single-delivery births, among lower-weight infants blacks comprised 39.8% of those weighing less than 500 g, 35.3% of those weighing 500 g-1,499 g, and 29.8% of those weighing 1,500 g-2,499 g. Within each birthweight category of births less than 2,500 g, the risk for black infants' being born with low birthweight was over twice that for whites. At the higher birthweight extreme, however, in 1.9% of all births, but in only 0.8% of black births, the infants weighed greater than or equal to 4,500 g. Birthweight-Specific Infant Mortality Neonatal mortality.
Neonatal mortality decreased sharply with increasing birthweight up to 4,000 g for both blacks and whites (Table 2). When we used data with unknown values distributed among known values, the NMR for single-delivery infants weighing less than 1,500 g at birth was 431.2 deaths/1,000 live births, whereas the risk was 2.1 deaths/1,000 live births for single-delivery infants weighing greater than or equal to 2,500 g. Infants weighing 500 g-999 g at birth had an almost 400-fold relative risk when compared with those weighing 3,000 g-3,999 g. Compared with whites, black infants weighing less than 3,000 g at birth experienced a lower birthweight-specific neonatal mortality, whereas the birthweight-specific NMR for heavier black infants was much higher (Table 2 and Figure 1). Postneonatal mortality.
Postneonatal mortality decreased with increasing birthweight up to 4,000 g, although the slope was not so steep as for neonatal mortality (Table 2 and Figure 2). Within all birthweight categories, blacks experienced higher postneonatal mortality than did whites. The overall relative risk of 2.1 was higher than all birthweight-specific relative risks because of the greater preponderance of black neonatal survivors in the lower birthweight ranges.
For neonatal survival, optimal birthweight for blacks was in the 3,000-g to 3,499-g category, while optimal birthweight for whites was in the 3,500-g to 3,999-g category. For postneonatal survival, the optimal categories were 500 g more for both blacks and whites. A comparison of these optimal groups showed that the IMR for blacks was 2.1 times that for whites. This risk is slightly higher than the overall relative risk of infant mortality for black infants (2.0). This overall risk reflects higher birthweight-specific neonatal mortality for black infants weighing greater than or equal to 3,000 g, higher postneonatal mortality for black infants of all birthweights, and the greater proportion of black infants weighing less than 3,500 g at birth. Optimal birthweights in Table 2 differ somewhat from those reported elsewhere (32) because this table included unknown gestational ages and birthweights; these data were distributed proportionately to known birthweights and gestational ages (see Appendix III.A, unabridged report). Birthweight Distribution of Deaths
Two-thirds of infant deaths occurred during the neonatal period (Table 3). Of those infants, more than half had birthweights of less than 1,500 g. Those infants, who comprised less than 1% of all live births, accounted for almost 40% of all infant deaths. Another two-fifths of infant deaths occurred among the 94% of infants with birthweights of greater than or equal to 2,500 g. In the postneonatal period, those with birthweights of greater than or equal to 2,500 g represented about three-fourths of deaths. Infants weighing less than 500 g at birth accounted for 10.7% of all deaths of black infants and 6.3% of all deaths of white infants with known birthweight.
States differ in the race-specific proportions of deaths of infants with birthweight of less than 500 g (see Appendix IV.C, unabridged report, and 35). This difference may be due to varying practices of recording such deliveries as live births or fetal deaths. If all recorded live-born infants weighing less than 500 g are excluded, the NMR for white infants with birthweights of less than 1,500 g is reduced 10.7%, from 441.9 to 394.4/1,000 live births. The corresponding NMR for black infants is reduced 15.1%, from 406.5 to 345.1/ 1,000 live births. Infant mortality is thereby reduced 7.2% overall, representing a 6.1% reduction for whites (from 9.3-8.7/1,000 live births) and a 10.2% reduction for blacks (from 18.9-17.0/1,000 live births). Postneonatal mortality is not affected, since all infants with birthweights of less than 500 g are assumed to have died during the neonatal period. Risk Factors for Infant Mortality Sex.
Regardless of race, males experienced higher birthweight-specific infant mortality than did females (Table 4). However, the female advantage was proportionately somewhat higher for infants with birthweights of greater than or equal to 1,500 g and for all white infants. Gestational age.
For each reported gestational age, infant mortality decreased with increasing birthweight. At virtually all gestational ages, blacks weighing less than 2,500 g had lower infant mortality than did whites. The exceptions were infants weighing less than 1,500 g whose gestational ages were reported as 40 and 42-45 weeks (Table 4)--gestational ages that most likely represented misclassified values that we discuss subsequently. White infants weighing greater than or equal to 2,500 g but born after a gestation of less than 37 weeks also had higher gestation-specific infant mortality. From 37 weeks on, however, black infants in this birthweight category experienced higher gestation-specific infant mortality. Several combinations of birthweight and gestational age undoubtedly included a large proportion of infants whose gestational age was incorrectly reported. Gestation- and birthweight-specific measures of IMR for these categories will be misleading because of this misclassification. These categories are noted in Table 4. Racial differences in birthweight- and gestation-specific infant mortality are explored in greater detail elsewhere (32). Live-birth order.
Second-born infants experienced lower infant mortality (10.4/1,000 live births) than infants of other birth orders (Table 4). However, among infants with birthweights of greater than or equal to 2,500 g, those who were first-born experienced the lowest infant mortality among both blacks and whites. Among those heavier infants, infant mortality increased steadily with increasing birth order except for lower mortality for black infants of birth order 6 or higher. For specific birth orders, IMRs were 58% to 77% higher for those blacks than for whites. Maternal age.
Infant mortality decreased with increasing maternal age through 30-34 years of age but increased thereafter (Table 4). Optimal maternal age was 25-29 years for black mothers and 30-34 years for whites. Differences in infant mortality by maternal age were most pronounced for infants weighing greater than or equal to 2,500 g. Heavier black infants experienced higher mortality than whites, regardless of their mothers' ages. IMRs associated with young maternal age primarily relate to a lower birthweight distribution, especially for neonatal mortality differentials (36). Maternal education.
Infant mortality declined with increasing level of maternal education for both races but did so more steeply for infants born to white women and, for both races, for infants born weighing greater than or equal to 2,500 g (Table 4). Heavier black infants also experienced higher mortality, regardless of their mothers' educational levels. Prenatal care.
Infants born to mothers who obtained prenatal care beginning in the first trimester experienced substantially lower infant mortality (Table 4). This trend was most pronounced for infants weighing greater than or equal to 2,500 g but was also present for those weighing between 1,500 g and 2,499 g. In part, because of differences in birthweight distributions for black infants whose mothers received late prenatal care, the IMR for those infants was lower than that for black infants whose mothers received prenatal care beginning during the first trimester. Black-white differences in birthweight-specific infant mortality persisted for infants of women obtaining prenatal care beginning in the first trimester. Completeness of NIMS Data
There were 2,104 fewer deaths in NIMS than were predicted by using the synthetic cohort derived from 1980 and 1981 mortality tapes (Table 5). This table reflects an estimated underreporting of 4.6%; however, the impact varied among the states. In seven states, NIMS reported more deaths based on residence at birth than the synthetic cohort number of deaths based on residence at death. There was exact agreement in two states, and 2,159 fewer deaths were reported in the NIMS data for 42 states. Eight states reported greater than 10% fewer deaths than were predicted from the synthetic cohort (see Appendix IV.B, Table VI1, unabridged report). These eight states included 44.8% of all estimated unreported deaths but only 15.5% of all deaths in the United States.
We asked states to report the number of death certificates that they were unable to link with birth certificates. Because state of residence at birth is not reported on death certificates, we could not estimate directly the effect of these unlinked certificates on state-specific data concerning resident IMRs. However, we could approximate the impact by examining unlinked certificates for deaths of infants who were likely to have resided in the given state at birth. The likelihood that an unlinked death certificate is for an infant who resided at birth in the given state is higher if the state of residence at death is also the state in which both birth and death occurred; of the 2,604 reported unlinked certificates, 1,202 met these criteria. With the number of these deaths used as the numerator and the number of expected deaths (from the synthetic cohort) as the denominator, reported unlinked certificates of six states were greater than or equal to 5% of expected resident deaths. Three of these six states were also among the eight with greater than 10% fewer linked deaths reported than expected from the synthetic cohort. This finding suggests that the chief problem with unlinked certificates may be one of in-state linkage. The other five states with greater than 10% discrepancy between linked and predicted infant deaths reported very few unlinked certificates of any kind. This finding suggests that 1) the states may need to revise their definition of what kind of death to link, 2) they are unaware of interstate linkage gaps or do not have the resources to improve interstate linkage efforts, or 3) the synthetic cohort estimate of expected deaths may be relatively invalid for those states. DISCUSSION
In 1980, there were 3,542,995 single-delivery and 69,912 multiple-delivery infants born in the United States. In 1980-1981, there were 43,217 deaths among these infants, for a total IMR of 12.0/1,000 live births. This number can be compared with the infant mortality rates reported by NCHS of 12.6 and 11.9 infant deaths/1,000 live births in 1980 and 1981, respectively (3). The difference in IMR as measured in the NIMS project and the infant mortality rates for 1980 and 1981 results in part from the inherent difference between risk and rate. The former is based on the experience of a birth cohort, whereas the latter is a period measure based on occurrences of deaths in a given year divided by births in that year. In times that have a fairly steady birth rate and a declining IMR, the period rate will overestimate the risk of mortality for that year's birth cohort. However, because two-thirds of the infant deaths occurred in the neonatal period, the 1980 infant mortality rate should closely approximate the IMR for the 1980 birth cohort. If a correction factor for the estimated underreporting of infant deaths in NIMS of 4.6% is added to the NIMS total, the resultant IMR for the 1980 birth cohort would be 12.5/1,000 live births, within 0.1/1,000 of the 1980 infant mortality rate.
Factors that increased the risk of infant mortality for single-born infants included low or very high birthweight, black race, male sex, short or long gestation, birth order (all but second), maternal age (younger or older), lower level of maternal education, and lack of prenatal care during the first trimester. Of these, the most important predictor for survival was birthweight: survival improved exponentially as birthweight increased to its optimum level (3,000 g-3,999 g for blacks and 3,500 g-4,499 g for whites, based on the neonatal and postneonatal mortality components of birthweight-specific infant mortality discussed above) (Table 2 and Figures 1, 2).
Overall, black infants had over twice the risk whites had of dying in their first year. This risk was related both to a higher prevalence of low birthweight and to higher mortality risks in the neonatal period for infants weighing greater than or equal to 3,000 g and in the postneonatal period for all infants, regardless of birthweight. These results parallel state-specific analyses of black-white differences in neonatal and infant mortality (15,16).
Proportionately, more than three times as many black as white newborns have birthweights of less than 500 g. If these infants are excluded, the relative risk of neonatal mortality for blacks compared with whites drops from 2.01 to 1.87. Some live births of infants weighing less than 500 g are classified as fetal deaths, but this tendency is probably no more frequent for white than for black deliveries. Thus the true risk of neonatal death for small black infants--and thus the true black-white gap--is probably even greater than is reported.
Characteristic-specific analyses of black-white differences in infant mortality reveal that regardless of other infant or maternal risk factors, black infants were up to twice as likely as white infants to die within their first year of life. This gap is most pronounced for infants with birthweights of greater than or equal to 2,500 g. The exceptions to this general finding were lower neonatal mortality among black infants weighing less than 3,000 g and lower infant mortality for black infants of less than 37 weeks' gestation (32).
Overall infant mortality risks are associated with maternal characteristics that reflect social-class differentials in access to and availability of health care. Prenatal factors affecting birthweight--as well as postnatal factors affecting infant care--contribute to increased infant mortality for socially disadvantaged infants, although the relative contributions of these components of infant mortality vary among different groups of infants. Part of the black-white gap is related to the relatively disadvantaged status of blacks in the United States. However, the widespread differences between black and white infant mortality across all maternal characteristics reported on birth certificates suggest that there may also be a further problem with access to effective health care for black infants and pregnant black women.
National black-white differences in infant mortality are explored in a series of articles (20-22) dealing with the composite effect of birthweight and gestation on the excess black infant mortality (32), regional variations in race-specific infant mortality (37), differences in cause-specific mortality (38,39), the effect of maternal age on black-white differences in infant mortality (22,36), and the change in race-specific infant mortality between the 1960 and the 1980 birth cohorts (40). With the exception of deaths resulting from congenital anomalies, all analyses point to the same conclusion: Far more effective strategies need to be developed and applied 1) to decrease the incidence of low birthweight, especially among black infants, 2) to increase neonatal survival for black infants weighing greater than or equal to 3,000 g, and 3) to increase postneonatal survival for all black infants.
Three potential sources of bias exist with limitations in the NIMS data: 1) underestimation of cohort mortality risk through failure to link data for deaths and births, 2) underascertainment of death through failure to file death certificates, and 3) lack of congruity between numerator data (from state tabulations) and denominator data (from the NCHS natality tape). The first two are inherent biases with vital records analyses; the third is unique to the NIMS methodology.
As to the extent of underestimated mortality relative to unlinked deaths, NIMS data include an estimated 95.4% of all reported infant deaths to U.S. resident infants born in 1980. The 4.6% estimated underreporting of known deaths is related in part to problems with state-level linkage. First, when vital events do not occur within the state of residence, linkage requires transferring certificates to the proper state for recording and linking. Since death certificates do not include information on place of residence at birth, appropriate transfer may be difficult. Second, there may not be a birth certificate for a reported infant death, especially for an early neonatal death of a very small infant. Third, when information necessary for the linkage is inaccurately recorded, linkage is less likely to occur. When linkage does not occur for any of these reasons, the death cannot be included in cohort mortality risks.
The extent of underestimating IMRs is not uniform throughout the states. The eight states that reported greater than 10% fewer deaths than expected through the synthetic cohort comprised nearly half of all the estimated underreporting. Also, postneonatal deaths, especially out-of-state events, are proportionately more underreported than earlier infant deaths (33,34). The same may be true of deaths of black infants overall (33). Because the extent of underreporting was small for the United States as a whole, these state-level biases probably do not affect national results in measurable ways. However, for specific states, the difference between NIMS estimates and true state-specific risks could be substantial. For this reason, persons examining state-level NIMS data should examine the estimated extent of underreporting (see Appendix IV.A-B, unabridged report, and 41).
Regarding the underestimation of cohort mortality resulting from unreported infant deaths, neither NIMS data nor NCHS data include deaths for which there were no death certificates. Among infants with birthweights of less than 1,500 g, unregistered deaths have been traced to problems in vital records registration (42); however, such underreporting has been lower in more recent reports (43,44). As states have become more aware of the need to track infants whose birthweights are very low, their reporting of infant deaths may have improved. Another registration problem is the failure to report deaths of live-born infants who die very soon after birth. These infants are sometimes reported as fetal deaths. The perinatal mortality risk that includes fetal deaths will include the risk for those infants; an examination of birthweight-specific perinatal mortality risks reveals patterns similar to those found for NIMS IMRs (40).
The third potential source of bias was the use of different data sources for births (NCHS computer tapes) and deaths (state reporting) of infants weighing greater than or equal to 500 g. NCHS birth reporting could be less complete than state-based death reporting because NCHS uses an earlier cutoff date for inclusion in its birth file than that used by some states for inclusion of death reports. On the other hand, NCHS reporting of births for a given state may be more complete because NCHS is not dependent on the interstate exchange of vital records. For the nine states for which we had access to both NCHS and state birth tapes, we investigated the differences between state and NCHS reports of births of infants weighing greater than or equal to 500 g as a proxy for estimating the differences in the two sources of data for births and deaths. We found that the differences between states' reported births and births reported by NCHS for those states were small (see Appendix IV.C, unabridged report).
We therefore concluded that when bias does occur, it probably results in underestimating the black-white infant mortality differential. Our conclusions regarding the black-white gap are thus somewhat conservative. Biases related to analyses of specific characteristics are discussed in the Appendices of the unabridged report.
The NIMS project represents phase I of a two-stage process to develop routine reporting of national birthweight-specific infant mortality data. Phase II is the annual microlevel tape, being produced by NCHS, that begins with the 1983 birth cohort. A report on the pilot stage of phase II is presented elsewhere (18). Besides compiling interim national data, NIMS has provided insight into developing and improving this ongoing system. We have identified states with particular linkage problems and have isolated problems with definitions of vital records variables (see Appendix II, unabridged report, and 31).
Through NIMS, we have explored the usefulness of linked birth and infant death record data for program planning and evaluation (45). The NIMS Conference brought together representatives from all states. Participants--including vital registrars, health statisticians, and maternal and child health directors--discussed problems of linkage as well as uses of linked vital records for program planning, targeting, and evaluation (45). The examination of birthweight distributions, birthweight-specific infant mortality, and birthweight distributions of infant deaths can provide insight into targeting programs more effectively and identifying problem areas for more intensive program evaluation.
Promoting health/preventing disease: objectives for the nation. Washington, DC: US Department of Health and Human Services, 1980:15-20.
2. NCHS. Advance report of final mortality statistics, 1986. Monthly Vital Statistics Report 37 (No. 6 suppl). Hyattsville, Maryland: US Department of Health and Human Services, Public Health Service, 1988:7.
3. NCHS. Advance report of final mortality statistics, 1981. Monthly Vital Statistics Report 33 (No. 3 suppl). Hyattsville, Maryland: US Department of Health and Human Services, Public Health Service, 1984:4.
4. Lee KS, Paneth N, Gartner LM, Pearlman MA, Gruss L. Neonatal mortality: an analysis of the recent trends in the United States. Am J Public Health 1980;70:15-21.
5. Williams RL, Chen PN. Identifying the sources of the recent decline in perinatal mortality rates in California. N Engl J Med 1982;306:207-14.
6. David RJ, Siegel E. Decline in neonatal mortality, 1968-1977: better babies or better care? Pediatrics 1983;71:531-40.
7. Kleinman JC, Kovar MG, Feldman JJ, Young CA. A comparison of 1960 and 1973-74, early neonatal mortality in selected states. Am J Epidemiol 1978;108:454-69.
8. Kleinman JC, Kessel SS. The recent decline in infant mortality. In: Health United States 1980. Hyattsville, Maryland: US Department of Health and Human Services, Public Health Service, 1980; DHHS publication no. (PHS)81-1232.
9. Kleinman JC. Trends and variations in birth weight. In: Health United States 1981. Hyattsville, Maryland: US Department of Health and Human Services, Public Health Service, 1987; DHHS publication no. (PHS)82-1232. 10. Lee KS, Paneth N, Gartner LM, Pearlman M. Very low birthweight rate: principal predictor of neonatal mortality in industrialized populations. J Pediatr 1980;97:759-64. 11. Guyer B, Wallach LA, Rosen SL. Birthweight standardized neonatal mortality rates and the prevention of low birth weight: how does Massachusetts compare to Sweden? N Engl J Med 1982;306:1230-3. 12. CDC. Years of potential life lost attributable to low birthweight--United States, 1980 birth cohort. MMWR 1986;35:188-90, 195. 13. Erickson D, Bjerkedal T. Fetal and infant mortality in Norway and the United States. JAMA 1982;247:987-91. 14. NCHS. Proceedings of the International Collaborative Effort on Perinatal and Infant Mortality, vol l. Hyattsville, Maryland: US Department of Health and Human Services, Public Health Service, 1985; DHHS publication no. (PHS)85-1252. 15. Binkin NJ, Williams RL, Hogue CJR, Chen PM. Reducing black neonatal mortality: will improvement in birth weight be enough? JAMA 1985;253:372-5. 16. Alexander GR, Tompkins ME, Altekruse JM, Hornung CA. Racial differences in the relation of birth weight and gestational age to neonatal mortality. Public Health Rep 1985;100:539-47. 17. Armstrong RJ. A study of infant mortality from linked records by birth weight, period of gestation, and other variables, United States. Vital Health Statistics 20 (12). Hyattsville, Maryland: NCHS, 1972; DHEW publication no. (HSM)72-1055. 18. NCHS. Linked birth and infant death data set. Hyattsville, Maryland: US Department of Health and Human Services, Public Health Service, 1989. 19. CDC. National birthweight-specific infant mortality surveillance: preliminary analysis--United States, 1980. MMWR 1986;35:269-73. 20. Public Health Rep 1987;102:126-223. 21. VanLandingham MJ, Buehler JW, Hogue CJR, Strauss LT. Birthweight-specific infant mortality for Native Americans compared with whites, six states, 1980. Am J Public Health 1988;78:499-503. 22. Friede A, Baldwin W, Rhodes PH, et al. Older maternal age and infant mortality in the United States. Obstet Gynecol 1988;72:152-8. 23. NCHS. Public use data tape documentation, 1980 natality detail. Hyattsville, Maryland: US Department of Health and Human Services, 1982. 24. Giachello AL, Bell R, Aday LA, Anderson RM. Use of the 1980 census for Hispanic health services research. Am J Public Health 1983;73:266-74. 25. Hayes-Bautista DE. On comparing studies of different Raza populations. Am J Public Health 1983;73:274-6. 26. Selby ML, Lee ES, Tuttle DM, Loe HD. Ethnic differentials in infant mortality in an urban county: the case of Mexican Americans reexamined. Presented at the 111th Annual Meeting of the American Public Health Association, Population and Family Planning Section, Dallas, Texas, Nov. 14, 1983. 27. Gee SC, Lee ES, Forthofer RN. Ethnic differentials in neonatal and postneonatal mortality: a birth cohort analysis by a binary variable multiple regression method. Soc Biol 1976;23:317-25. 28. Williams RL, Binkin NJ, Clingman EJ. Pregnancy outcomes among Spanish-surname women in California. Am J Public Health 1986;76:387-91. 29. Becerra J, Kristal AR, Perez N, Buehler JW, Smith JC. Infant mortality and childbearing patterns among Puerto Rican Hispanics. Presented at National Infant Mortality Surveillance Conference, Atlanta, Georgia, May 2, 1986. 30. International Classification of Diseases, 9th Revision, Clinical Modification, vol l. Ann Arbor, Michigan: Commission on Professional and Hospital Activities, 1978. 31. Strauss LT, Freedman MA, Gunter N, Powell-Griner E, Smith JC. Experiences with linked birth and infant death certificates from the NIMS project. Public Health Rep 1987;102:204-10. 32. Sappenfield WM, Buehler JW, Binkin NJ, Hogue CJR, Strauss LT, Smith JC. Differences in neonatal and postneonatal mortality by race, birth weight, and gestational age. Public Health Rep 1987;102:182-92. 33. Lambert DA. Unlinked death certificates from the National Infant Mortality Surveillance project (Master of Science thesis). Amherst, Massachusetts: University of Massachusetts, May 1986. 34. Lambert DA, Strauss LT. Analysis of unlinked infant death certificates from the NIMS project. Public Health Rep 1987;102:200-4. 35. Wilson AL, Fenton LJ, Munson DP. States reporting of live births of newborns weighing less than 500 grams: impact on neonatal mortality rates. Pediatrics 1986;78:850-4. 36. Friede AM, Baldwin W, Rhodes PH. Young maternal age and infant mortality: the role of low birth weight. Public Health Rep 1987;102:192-9. 37. Allen DA, Buehler JW, Hogue CJR, Strauss LT, Smith JC. Regional differences in birth weight-specific infant mortality, 1980. Public Health Rep 1987;102:138-45. 38. Buehler JW, Strauss LT, Hogue CJR, Smith JC. Birth weight-specific causes of infant mortality, United States, 1980. Public Health Rep 1987;102:162-71. 39. Berry RJ, Buehler JW, Strauss LT, Hogue CJR, Smith JC. Birth weight-specific infant mortality due to congenital anomalies, 1960 and 1980. Public Health Rep 1987;102:171-81. 40. Buehler JW, Kleinman JC, Hogue CJR, Strauss LT, Smith JC. Birth weight-specific infant mortality, United States, 1960 and 1980. Public Health Rep 1987;102:151-61. 41. Marks JS, Buehler JW, Strauss LT, Hogue CJR, Smith JC. Variation in state-specific infant mortality risks. Public Health Rep 1987;102:146-51. 42. McCarthy BJ, Terry J, Rochat RW, Quave S, Tyler CW Jr. The underregistration of neonatal deaths: Georgia 1974-77. Am J Public Health 1980;70:977-82. 43. Frost F, Kirkwood KS. Racial differences between linked birth and infant death records in Washington state. Am J Public Health 1980;70:974-6. 44. CDC. Birthweight-specific neonatal mortality rates--Kentucky. MMWR 1985;34:487-8. 45. Zahniser C, Halpin G, Hollinshead W, Kessel S, Koontz A. Using linked birth and infant death files for program planning and evaluation: NIMS workshop lessons. Public Health Rep 1987;102:211-6.
Appendix. NIMS TABLES INDEX FOR NIMS TABLES Set 1 (See unabridged report.) 1D. Number of Infant Deaths, by Birthweight, Plurality, Age at Death, Race, and Place of Residence, 1980 Birth Cohort, United States Set 2 (See unabridged report.) 1R. Infant Mortality Risks, by Birthweight, Plurality, Age at Death, Race, and Place of Residence, 1980 Birth Cohort, United States Set A/3 2. Infant Mortality, by Birthweight, Age at Death, Race, and Maternal Age: Singleton Infants, 1980 Birth Cohort, United States 3. Infant Mortality, by Birthweight, Age at Death, Race, and Number of Previous Live Births: Singleton Infants, 1980 Birth Cohort, United States 4. Infant Mortality, by Birthweight, Age at Death, Race, and Number of Other Terminations (greater than or equal to 20 weeks' gestation): Singleton Infants, 1980 Birth Cohort, United States (See unabridged report.) 5. Infant Mortality, by Birthweight, Age at Death, Race, and Maternal Education: Singleton Infants, 1980 Birth Cohort, United States (See unabridged report.) 6. Infant Mortality, by Birthweight, Age at Death, Race, and Sex of Infant: Singleton Infants, 1980 Birth Cohort, United States (See unabridged report.) 7. Infant Mortality, by Birthweight, Age at Death, Race, and Gestational Age at Birth: Singleton Infants, 1980 Birth Cohort, United States 8. Infant Mortality, by Birthweight, Age at Death, Race, and Type of Delivery: Singleton Infants, 1980 Birth Cohort, United States (See unabridged report.) 9. Infant Mortality, by Birthweight, Age at Death, Race, and Month Prenatal Care Began: Singleton Infants, 1980 Birth Cohort, United States 10. Infant Mortality, by Birthweight, Age at Death, Race, and Number of Prenatal Care Visits: Singleton Infants, 1980 Birth Cohort, United States (See unabridged report.) 11. Infant Mortality, by Birthweight, Age at Death, Race, and Grouped Cause of Death: Singleton Infants, 1980 Birth Cohort, United States Set B/4 12. Infant Mortality, by Birthweight and Race: Multiple Birth Infants, 1980 Birth Cohort, United States 13. Infant Mortality, by Birthweight and Plurality: Multiple Birth Infants, 1980 Birth Cohort, United States 14. Infant Mortality, by Birthweight and Gestational Age at Birth: Multiple Birth Infants, 1980 Birth Cohort, United States 15. Infant Mortality, by Birthweight and Type of Delivery: Multiple Birth Infants, 1980 Birth Cohort, United States (See unabridged report.) Appendices (See unabridged report.)
calculations of mortality risk
D indicates "death," i.e., number of infant deaths (Table 1 only).
R indicates "risk," i.e., infant mortality risk (Table 1 only). Tables are repeated for neonatal/postneonatal/infant, plurality, and racial group as follows:
1 indicates neonatal deaths, singleton infants. 2 indicates postneonatal deaths, singleton infants. 3 indicates infant deaths, singleton infants. 4 indicates infant deaths, multiple birth infants. A indicates all races. W indicates whites. B indicates blacks. Examples: 1D.1A indicates a table showing the number of neonatal deaths for singleton infants of all races. 1R.2W indicates a table showing the mortality risk for postneonatal deaths for singleton infants of white race. NOTES FOR TABLES
The following notes apply to the data tables for the National Infant Mortality Surveillance (NIMS) Report.
The symbols "#" and "=" may appear on any of the postneonatal mortality risk tables: # Zero deaths reported for cell, negative value for number of neonatal survivors (see Appendix IV, Section C, unabridged report) = Deaths reported for cell, negative value for number of neonatal survivors (see Appendix IV, Section C, unabridged report)
The situations represented by "*," "#," and "=" may occasionally arise because different data sources were used for the numerator and denominator in mortality risk calculations. 4. Mortality risk tables: United States definition
For mortality risk tables for the United States, the United States is defined for the various tables as follows (see Appendix III, Section B, unabridged report): AREA OF RESIDENCE (Tables 1R.)--U.S. mortality risks are based on data from 52 vital statistics reporting areas. (See unabridged report.) MATERNAL AGE (Tables 2.)--U.S. mortality risks are based on data from 49 vital statistics reporting areas. NUMBER OF MOTHER'S PREVIOUS LIVE BIRTHS (Tables 3.)--U.S. mortality risks are based on data from 48 vital statistics reporting areas. NUMBER OF MOTHER'S OTHER TERMINATIONS OF PREGNANCY greater than or equal to 20 WEEKS' GESTATION (Tables 4.)--U.S. mortality risks are based on data from 39 vital statistics reporting areas. (See unabridged report.) MATERNAL EDUCATION (Tables 5.)--U.S. mortality risks are based on data from 41 vital statistics reporting areas. (See unabridged report.) SEX OF INFANT (Tables 6.)--U.S. mortality risks are based on data from 50 vital statistics reporting areas. (See unabridged report.) GESTATIONAL AGE AT BIRTH, SINGLETON INFANTS (Tables 7.)--U.S. mortality risks are based on data from 45 vital statistics reporting areas. MONTH PRENATAL CARE BEGAN (Tables 9.)--U.S. mortality risks are based on data from 44 vital statistics reporting areas. NUMBER OF PRENATAL CARE VISITS (Tables 10.)--U.S. mortality risks are based on data from 43 vital statistics reporting areas. (See unabridged report.) GROUPED CAUSE OF DEATH (Tables 11.)--U.S. mortality risks are based on data from 52 vital statistics reporting areas. RACE, MULTIPLE BIRTH INFANTS (Table 12.)--U.S. mortality risks are based on data from 52 vital statistics reporting areas. PLURALITY, MULTIPLE BIRTH INFANTS (Table 13.)--U.S. mortality risks are based on data from 51 vital statistics reporting areas. GESTATIONAL AGE AT BIRTH, MULTIPLE BIRTH INFANTS (Table 14.)--U.S. mortality risks are based on data from 34 vital statistics reporting areas.
Disclaimer All MMWR HTML documents published before January 1993 are electronic conversions from ASCII text into HTML. This conversion may have resulted in character translation or format errors in the HTML version. Users should not rely on this HTML document, but are referred to the original MMWR paper copy for the official text, figures, and tables. An original paper copy of this issue can be obtained from the Superintendent of Documents, U.S. Government Printing Office (GPO), Washington, DC 20402-9371; telephone: (202) 512-1800. Contact GPO for current prices.**Questions or messages regarding errors in formatting should be addressed to email@example.com.
Page converted: 08/05/98
This page last reviewed 5/2/01