BACKGROUND In Latin America, insufficient data are available to improve local admission policies for human immunodeficiency virus (HIV) patients in the intensive care units (ICU). We undertook this study to evaluate the outcome and survival determinants of HIV patients in a Mexican ICU during three time periods. METHODS From December 1985 through January 2006, a clinical chart-based, retrospective study of all HIV patients admitted to the ICU was conducted. Demographic, clinical and laboratory data; disease severity score (APACHE II) and mortality were evaluated. A comprehensive database was created and data were analyzed using survival and regression models. RESULTS Ninety HIV patients were admitted to the ICU during the study: 16 (18%) in 1985-1992 (non-antiretroviral [ARV]-period), 21 (23%) in 1993-1996 (ARV-period), and 53 (58%) in 1996-2006 (highly active antiretroviral treatment [HAART] period). Leading reasons for admission were the need for mechanical ventilatory support (MVS, 85.5%), septic shock (23%), and non-HIV/AIDS complications (15.5%). Survival in the ICU increased from 12.5% (non-ARV period) to 57% (HAART period). Mortality during ICU stay was associated with MVS (HR: 3.2; 95% CI 1.0-10.2) and APACHE II > or =13 points (HR: 2.2; 95% CI 1.3-4.0). Use of steroids (HR: 0.4; 95% CI 0.2-0.8) and HAART (HR: 0.25; 95% CI 0.1-0.5) were associated with a lower risk of death. In multivariate analysis, septic shock was the main predictor of death in the ICU (HR: 2.4; 95% CI 1.1-5.2) and after discharge. HAART remained as a significant protective factor. CONCLUSIONS Overall survival in Mexican HIV patients admitted to an ICU has substantially increased in recent years. These data should encourage policies that consider HIV patients as good candidates for receiving intensive care.