Background:  In 2003, 2% of all state prison inmates in the US were known to be HIV+, a rate that is more than 5 times that in the community.  The NYS prison system has the largest numbers of male and female HIV positive inmates in the nation.  Knowledge of HIV prevalence trends in this high-risk population is important to plan for needed health care and HIV prevention.

Methods: Samples of consecutive male and female inmates entering 4 intake facilities were selected in 7 cross-sectional anonymous surveys from 1988-2003.  HIV antibody status was determined on remnant sera from routine intake.   Demographic, medical and risk data were abstracted from medical history and DOCS records.  Prevalence trends were estimated based on survey data without prison population adjustment.  Logistic regression modeling was performed for each study cohort to examine and compare the relative impact of selected predictors of HIV infection.

Results:  22,504 study records were compiled through 7 study cohorts, with 999 records for 1988, 1062 for 1990, 3404 for 1992, 3880 for 1994, 5323 for 1996, 3900 for 2000, and 3936 for 2003.  HIV prevalence declined steadily, yet the decline from 1988 to 2003 was about 75% (17.6% to 4.5%) for males and <40% (18.8% to 11.4%) for females.  Hispanic inmates showed the greatest HIV prevalence decline (male 25.2% to 6.6%, females 29.5% to 11.5%).  Black female inmates experienced virtually no change with a HIV+ rate of 14.5% in 1988 and 14.7% in 2003.  Self-reported injection drug use (IDU) was lower for male and female black inmates than for other groups.  Logistic regression analyses indicate changing patterns of predictors of HIV infection over time.  IDU was by far the single most important predictor of HIV in the 1988 model with an odds ratio (OR) of 9.22 (95% CI: 6.10, 13.94).  By 2003, IDU (OR:3.03, CI:2.06, 4.45) was similar to race (Black OR:2.71, CI:1.62,4.53; Hispanic OR:2.44, CI:1.41, 4.24), age >30 (OR: 5.12, CI:3.14, 8.34), and MSM (OR: 3.38, CI:1.59, 7.19) as a predictor of HIV infection.

Conclusions: Among incoming inmates of the NYS prison system, HIV prevalence declined for all large race and gender groups except for the black females.  Although IDU remains a major predictor of HIV infection, its predictive power has dramatically decreased, providing solid evidence of success in prevention.  Further research is needed to seek effective prevention efforts for the non-IDU inmates as they remain at high risk for HIV infection.