Abstract
The objective of this study was to obtain accurate estimates of the prevalence of psychiatric disorder in the population represented by the HIV Costs and Services Utilization Study cohort. We constructed logistic regression models to predict DSM-IV diagnoses of depression, generalized anxiety disorder, panic, and dysthymia among a subsample of the HCSUS cohort who in separate interviews completed the CIDI-SF and the full CIDI diagnostic interview. Diagnoses were predicted using responses to the CIDI-SF as well as other variables contained in the baseline and first follow-up interviews. Resulting regression equations were applied to the entire baseline and first follow-up samples to obtain new estimates of the prevalence of disorder. Compared to estimates based on the CIDI-SF alone, estimates obtained from this procedure provide a more accurate representation of the prevalence of the presence of any one of these four psychiatric disorders in this population, yielding more correct classifications and a lower false-positive rate. Prevalence rates reported in this study are as much as 16% lower than rates estimated using the CIDI-SF alone, but are still considerably higher than estimates for the general community population.