ObjectivePrevious studies have shown a decrease in hospitalization rates associated with the introduction of highly active antiretroviral therapy (HAART). To evaluate hospitalization rates and patterns in discharge diagnoses that changed between 1995 and 1998 and to examine risk factors for hospitalization in HIV-positive patients, we conducted a cohort study.Patients and MethodsAll inpatient hospitalizations of 2,151 HIV-positive patients enrolled in our university-based HIV clinic between January 1, 1994 and December 31, 1998 with a CD4 count within a 6-month calendar semester were examined to evaluate hospitalization rates, discharge diagnoses, and intensive care department use. Negative binomial regression was used to assess the effect of various risk factors on hospitalization.ResultsHospitalization rates decreased between 1995 and 1996 but increased between 1997 and 1998. In multivariate regression, female gender (incidence rate ratio [IRR], 1.45;p< .001), injection drug use (IRR, 1.36;p< .001), and having received no antiretroviral therapy were strong predictors of total hospitalization. White race, low CD4 count, and no antiretroviral treatment were strong predictors of hospitalization for an opportunistic infection. Female gender (IRR, 1.45;p< .001), African-American ethnicity (IRR, 1.22,p= .05), no antiretroviral treatment, and low CD4 counts were predictive of higher hospitalization rates for nonopportunistic infection–related diagnoses. Intensive care department–use was associated with white ethnicity (IRR, 1.86;p= .028), heterosexual transmission of HIV (IRR, 1.90;p= .009), no antiretroviral treatment, and low CD4 count at enrollment.ConclusionsOur data indicate that hospitalization rates decreased between 1995 and 1997 after introduction of HAART, but that they then increased between 1997 and 1998, particularly for diagnosed nonopportunistic infections. If these trends continue, it indicates that patients may be developing previously unseen comorbidities and that HAART may have reached or exceeded a threshold in its effectiveness in reducing the clinical morbidity that results in hospital admission.