Many risk factors for HIV incidence among drug users have been reporte
d in the literature. However most of these studies have been with foca
l samples, and typically have had limited sample sizes. The National I
nstitute on Drug Abuse (NIDA) funded a Cooperative Agreement for AIDS
Community-Based Outreach/Intervention Research (CA) since 1991 that pr
oduced a national database. Associations found in the literature were
tested to determine whether they were replicated on this national data
base. The CA national database had complete data, including blood draw
s six months apart, on 6,970 drug users who were seronegative at basel
ine. Sites that had no seroincident cases were excluded from the analy
sis. Twenty-nine risk factors identified in the literature were tested
on the national database on a bivariate basis. There were 56 seroinci
dent cases (those who were HIV-seronegative at baseline and HIV-seropo
sitive at follow-up) out of 3752 person years at risk, for a seroconve
rsion rate of 1.49 (CI 1.05, 1.94) per 100 person years at risk. Data
were analyzed both as a whole dataset and with the sites stratified by
high and low prevalence. In the overall analysis, risk factors associ
ated with seroconversion were times injected any drug, ever injected a
ny drug, days injected cocaine, times injected cocaine, and times inje
cted speedball (these variables had a 30-day time referent). Also sign
ificant was how recently formerly-used injection equipment was used. H
igh-prevalence sites revealed associations for times injected any drug
, days injected cocaine, and times injected speedball. Low-prevalence
sites revealed associations for injected any drug and days injected co
caine. These findings highlight the importance of injection drug use,
especially cocaine injection, as a major risk factor in recent serocon
version.