A Competing Risks Analysis of Presenting AIDS Diagnoses Trends

Robert E. Fusaro
Peter Bacchetti, Department of Epidemiology & Biostatistics, University of California, San Francisco
Nicholas P. Jewell, Division of Biostatistics, School of Public Health, University of California, Berkeley

Abstract

The proportions of gay men presenting with various AIDS diagnoses display temporal trends. For example, the proportion of initial diagnoses reported as Kaposi's sarcoma has declined over time. Epidemiologists have hypothesized that subjects with different presenting diagnoses may have (a) different infection time distributions or (b) different incubation time distributions. We examine these explanations for the observed trends in a competing risks framework. We nonparametrically estimate the relevant cause specific hazard functions from the doubly-censored data of the San Francisco City Clinic Cohort by maximizing a roughness penalized likelihood using an EM algorithm. These estimates provide insight into the natural history of HIV disease and allow us to compare the observed diagnoses trends to those expected under various hypotheses.