Alameda county birth certificate

ABBREVIATIONS. LBW, low birth weight; VLBW, very low birth weight; PROM, premature rupture of the membranes; RDS, respiratory distress syndrome; IUGR, intrauterine growth retardation; OR, odds ratio; CI, confidence interval; HMO, health maintenance organization.

Despite recent advances in neonatal technology and the improved survival of high-risk infants, the United States ranks 22nd in infant mortality among developed nations.[1] In large part this is attributable to the high incidence of low birth weight (LBW) among segments of the population.[2-4] LBW infants (weighing 2500 g or less) are 40 times more likely to die then are normal birth weight infants.[5] Very low birth weight (VLBW) infants (weighing 1500 g or less) are 200 times more likely to die than are infants of normal birth weight.[6,7] Because birth weight is the single most powerful predictor of infant mortality,[8-10] many assume that risk factors for LBW always predict infant mortality. However, only 10% of LBW infants die.[11] Which risk factors for LBW also contribute directly to infant mortality? If we are to refine our understanding of the causes of infant mortality, it is essential to proceed beyond birth weight and more rigorously study the direct relationship between maternal risk factors and infant survival.

The ease with which maternal risk factors can be linked to birth weight has contributed to a wide body of literature on maternal risk factors for LBW. In contrast, substantial studies directly linking maternal risk facto;:s to infant mortality are scant. Only two maternal (prenatal) risk factors have been clearly shown to exert a birth weight-independent effect on infant mortality: infant sex and race.[9-12] Twin status,[9-11] socioeconomic status,[13] history of fetal loss,[7] advanced maternal age,[7,11] and intrauterine growth retardation[9,14,15] are controversial. Interestingly, female sex and nonwhite race are associated with lower birth weight but higher survival rates. Wilcox[16] suggests that mean birth weight differs between races, and that studies on birth weight-specific mortality should consider birth weight with respect to deviation from the race-specific mean. This approach applies to female sex as well, as the mean birth weight differs from that of males.

Preliminary studies suggest that several maternal risk factors for LBW may increase the survival of LBW infants. These factors include heroin, premature rupture of the membranes (PROM), and cocaine use during pregnancy.[17-25] Glass and associates[17] found that premature infants born to heroin-using mothers were less likely to develop respiratory distress syndrome (RDS), the leading cause of death among VLBW infants. Studies on PROM have shown both a twofold decrease in the incidence of RDS[18] and a decrease in mortality[19] from RDS among infants with PROM [greater than or equal to]24 hours before delivery. Another study demonstrated lower mortality (of VLBW infants with PROM) due to all causes, but no decrease in the incidence of RDS.[20]

Although maternal cocaine use has been linked to a number of adverse pregnancy outcomes, such as preterm delivery, PROM, intrauterine growth retardation (IUGR), congenital anomalies, and LBW,[21,22] it has not been consistently linked to increased mortality. Phibbs and associates[23] compared cocaine-exposed infants to nonexposed controls, and found similar mortality rates despite a lower mean birth weight and more frequent pregnancy/delivery complications among cocaine-exposed infants. Several studies have shown that compared with nonexposed controls, cocaine-exposed babies have a lower rate of RDS. Maynard[24] found that among 16 cocaine-exposed infants weighing <2000 grams and of [less than or equal to]34 weeks' gestation, RDS was significantly less common (P = .014) than among 32 nonexposed controls of similar birth weight and gestation.[24] Zuckerman[25] found that 13 nonexposed infants of [less than or equal to]34 weeks' gestation had an odds ratio (OR) of 8.9 (95% confidence interval [CI]: 0.9, 83.5) for RDS compared with 8 cocaine-exposed infants of similar gestation. The adjusted OR remained essentially unchanged after controlling for PROM, black race, infant gender, and gestational age.[25]

To ascertain the independent effect of maternal risk factors on infant mortality, we conducted a retrospective population-based study, examining the relative incidence of maternal cocaine use, PROM, and other select risk factors among infants who died before their first birthday (cases) versus those who survived (controls). We restricted our infant population to VLBW infants, the subgroup of infants whose survival appears to differ with respect to several of the risk factors under study. To control for the confounding effects of birth weight, sex, and race, we matched cases and controls for these variables. The rest of the maternal risk factors under study were entered into univariate analysis for matched pairs. Factors deemed marginally significant by univariate analysis were entered into multivariate analysis to assess the independent contribution to infant mortality among VLBW infants.

METHODS

Subjects

The study population from which cases and controls were drawn included all deliveries to Oakland residents occurring in Alameda County between January 1,1988 and December 31, 1990. Only live-born singleton infants weighing 500 to 15(10 g at birth were included. We identified 316 infants who met the study population requirements. Cases were defined as infants in the study population who died before their first birthday. All causes of death were included. We identified 78 cases by linking county resident infant death certificates with county resident birth certificates. All cases except five were delivered at one of four county hospitals, including one public hospital, two private hospitals, and one health maintenance organization (HMO). Of the remainder, one was delivered at a naval hospital, and four were delivered out of county. These five infants were excluded from the study, as they were thought not to reflect the general population of Oakland residents. Of the remaining 73 infants in the case population, the hospitals were able to locate inpatient charts for the mothers of 41 infants (56%). One of the 41 charts revealed the infant case to be a therapeutic abortion and was excluded. Thus, the final sample of cases, for whom both birth certificate and hospital chart information were available, numbered 40. Of these, 100% were delivered prematurely (before 37 weeks gestation) and 25% had concurrent IUGR (birth weight less than the tenth percentile for a given gestational age).

Of the 32 cases for whom hospital charts were unavailable, demographic data from infant birth certificates was available for 18. Compared with the 40 infants for whom we obtained hospital charts, these infants were more likely to be in the 751 to 1000-g birth weight range, of female sex, of other race, and to have died during the postneonatal period. Differences were marginally significant for sex ([chi square] = 5.02, P = .025) and age at death ([chi square] = 4.70, P = .05). Thus, both females and postneonatal deaths were under-represented in the infant population analyzed. Because we ultimately matched controls to cases by sex, and stratified maternal risk factors according to neonatal versus postneonatal mortality, these differences are not likely to have influenced the associations seen between maternal risk factors and infant mortality.

For each case, a control infant who did not die before its first birthday was selected from the study population. Controls were matched for birth weight, sex, and race to control for potential confounding effects. Because the case population (n = 40) differed from the overall study population (n = 316) with respect to these factors, we were unable to unanimously match for all three characteristics. We prioritized matching by birth weight based on prior data from Alameda County which showed birth weight to be more closely associated than sex or race with infant mortality.[26] We ultimately selected 40 matched controls for 40 cases. Demographic characteristics are shown in Table 1. Case/control differences in birth weight, sex, and race were not significant, although there were slightly more controls of female sex and of black race.

[TABULAR DATA OMITTED]

Distribution by hospital of birth was not significantly different for cases versus controls. Fifteen percent of cases and 10% of controls were born at the county general hospital; 67.5% of both cases and controls were born at private hospital A; 5% of both cases and controls were born at private hospital B; and 12.5% of cases and 17.5% of controls were born at the HMO hospital. Confidentiality requirements as stipulated by the University of California at Berkeley Committee for the Protection of Human Subjects were followed throughout the study.

Procedure