This event history analysis of the Perinatal Substance Exposure Study investigates individual and community level correlates of the timing of first prenatal care among pregnant women in California. Data were collected anonymously at the time of delivery and include demographic information from hospital records and urine samples which were tested for a battery of substances. Zip-code level data from the 1990 census were appended to each record to assess absolute community effects. A discrete-time hazard rate was estimated for each trimester in six nested models using logistic regression. Results suggest that: poor, non-White, younger, native born, Spanish-speaking, substance-using women in poorer neighborhoods were least likely to receive prenatal care. While a positive urine test for alcohol was not associated with prenatal care initiation; overall drug positives and tobacco positives (self-report) were. The optimal model, with interactions, estimates that women in poorer communities were less likely to receive prenatal care in the first trimester, but more likely to receive care in the third trimester. This pattern is similar for most of the time-varying covariates in that non-Whites, English speakers and younger women were less likely to receive first trimester care, but more likely to receive second and third trimester care.