Most analyses of prenatal substance use focus on individual level correlate
s and ignore community level variables and the effect of the dependency of
respondents within communities. This analysis uses multilevel logistic regr
ession models to more accurately assess the correlates of perinatal substan
ce use in California. Statistical results indicate that a significant porti
on of substance use can be attributed to neighborhood heterogeneity, and th
at traditional models of substance use may inaccurately attribute this vari
ation to individual level regression coefficients. Substantive results indi
cate that levels of neighborhood public assistance had an independent, sign
ificant effect on the prevalence of all substances tested for except alcoho
l. Black women had higher predicted prevalence risks for alcohol and cocain
e while White women had higher predicted risks for tobacco, marijuana and a
mphetamines. Racial contrasts were nonsignificant for the overall illicit d
rug category and opiates, after controlling for neighborhood public assista
nce. Finally, individual level variables, with the exception of age, were n
ot moderated by levels of neighborhood public assistance. (C) 2001 Elsevier
Science Ltd. All rights reserved.