The specific aims of this study are to 1) Compare 4 low socioeconomic status groups -- blacks, whites, Mexican- Americans in San Antonio and blacks in Detroit -- with respect to a) several adverse pregnancy outcomes (including low birthweight) and b) medical/obstetrical, environmental, demographic, social, cultural, psychological, and behavioral characteristics that may be associated with these outcomes. 2) Determine which factors or configurations of factors are independently associated with each outcome for the sample as a whole and for each group considered separately, with special emphasis given to patterns of interactions among independent variables. The objective here is to a) explain outcome differences among the three racial/ethnic groups in San Antonio, focusing on why Mexican-Americans have as favorable or more favorable outcomes than whites despite having social characteristics that would appear to place them at higher risk and much more favorable outcomes than blacks with whom they share many risk factors and b) determine the effects of relatively favorable (San Antonio) vs. unfavorable (Detroit) regional environments on outcomes among blacks. 3) Suggest culturally meaningfully interventions to modify risk factors. In order to achieve these aims, an explanatory model based on life events, qualify of social support, and cultural, valuation of pregnancy and childbearing will be tested. The sample consists of 3000 black, white and Mexican-American pregnant women (1000 per group) in San Antonio and 1000 pregnant blacks in Detroit utilizing health facilities for the indigent. Blacks (in San Antonio) and whites will be oversampled using weighted, stratified sampling procedures. Patients will be recruited at their first prenatal visit. Those who lack prenatal care will be recruited after delivery. Subjects will be interviewed at 3 time points: the first prenatal visit, 28-30 weeks gestation, and after delivery. Charts and birth/death certificates will be used to determine neonatal mortality. Data analysis will be based upon univariate statistical analyses (analysis of variance and chi- square or other nonparametric statistics) and multivariate analyses (logistic regression, ordinary least-squares regression). Maximum-likelihood methods for the analysis of covariance structure, specifically the SIMPLIS or LISREL computer software, will be used to estimate a measurement model of the dependent variable. New methods for the estimation of covariance structure model with data violate the maximum-likelihood assumption will be considered.
