Modeling the HIV epidemic: The application of a discrete Markov model to HIV/AIDS surveillance data for estimating HIV incidence and detection rates in Louisiana
Description
Population based estimates of recent incidence and prevalence of HIV are needed to effectively plan and evaluate programs for prevention, treatment, and services. Historically, backcalculation techniques were used to reconstruct the HIV epidemic by applying estimated incubation period distributions to AIDS incidence data, yielding imprecise estimates. Since then, these methodologies have been complicated by additional theoretical considerations. In addition, population-based detection rates among infected persons in Louisiana have not been examined. The purpose of this dissertation was to develop a valid and reliable model to estimate HIV incidence and detection rates among infected persons in Louisiana A time-inhomogeneous discrete Markov model of disease progression, detection, and treatment was developed and applied to quarterly HIV/AIDS surveillance data reported to the Louisiana health department, including HIV (non-AIDS) data, from 1981 to 1996. The model estimated incidence and stage-specific detection rates based on joint distributions of HIV detection and AIDS diagnosis dates for selected demographic and risk subgroups. The development of the model addressed the 1993 change in the AIDS case definition, pre-AIDS mortality, reporting delays, flexible treatment uptake and response, changes in detection over calendar time and stage of disease progression, and inclusion of casts with missing risk Evenly distributed across the three major risk groups, the total estimated number of new infections was 1,300 cases (plausible range: 1,200--1,500) in 1996. Of the estimated 20,500 (19,850--21,250) cases infected by 1996, 14,800 were alive and two-thirds had been detected and reported. Incident cases among women and African-American men were double those of white men. Incidence trends indicated a steady increase in high-risk heterosexual transmission and a substantial decrease among men who have sex with men. Detection rates increased steadily from 1985 until 1993, after which they remained stable across all subgroups. Due to the changes in detection, trends in HIV surveillance data were not good indicators of incidence. In addition to elucidating issues regarding the interpretation of HIV surveillance data, the methodology provided a viable and flexible approach to addressing the current data needs while accounting for some of the major challenges in monitoring the HIV/AIDS epidemic in Louisiana