Crack cocaine risers are nt high risk for HIV with higher frequency crack u
sers engaging in higher rates of HIV-related sexual risk behaviours. This s
tudy will assess the variables impacting changes in crack use frequency. Ou
t-of-treatment crack users were street recruited in East Harlem, NY. Subjec
ts (n = 727) were 33% female, 92% minority, and 28% reported recent drug in
jecting. Baseline and 6-month follow-up interviews were administered. There
was a significant reduction it crack use over time (p < .001). Subjects we
re categorized according to the groups, based on their change in level of c
rack use between the two interviews, to predict those who stopped maintaine
d, or changed their level of use. Discriminant analyses identified six vari
ables as the best predictors of the five groups, including having been in d
rug treatment since baseline and having been a drug injector (both related
to reduced levels of crack use). The overall reduction in crack use for the
sample masked the fact that important subgroups remained at high use level
s or increased their use. The identification of subgroups who may be most r
esistant to reducing use drug can be helpful in developing more effective i
nterventions.