Dissertation Title:  "The Application of Decision Analytic Methods to Diverse Public Health Problems in Underserved Populations"

The unifying theme of this dissertation is the application of decision analytic methods to diverse public health problems. The research addresses the complexities of modeling screening and treatment of chronic and infectious diseases and the policy challenges given population heterogeneities, diverse treatment strategies, and the introduction of new technologies.

Chapter I present results our analysis of syphilis screening in pregnancy and the potential impact of new screening technologies in resource poor settings. In developing countries, syphilis infection in pregnancy poses significant deleterious consequences to maternal and infant health. While treatment with penicillin is highly effective and relatively inexpensive, identification and timely treatment of infected women often falls short due to logistical and technical obstacles. New rapid testing technologies have become available with the potential to improve diagnosis and treatment of infection during pregnancy. We developed a Markov model of the natural history of syphilis and pregnancy in women in the sub-Saharan Africa setting in order to assess the benefits, cost-effectiveness and policy implications of universal on-site prenatal syphilis screening and treatment using rapid diagnostic tests versus standard testing methods. Our analysis showed that syphilis screening with new rapid tests that allow for same day testing and treatment was highly cost-effective and had the potential to provide substantial benefits by preventing syphilis transmission and adverse pregnancy outcomes.

Chapter II describes our calibration of a 1st order Monte Carlo simulation model of the natural history and treatment of HIV infection, which incorporates components of both infectious and chronic disease modeling. Previously, model natural history parameter estimates were primarily derived from data from the Multicenter AIDS Cohort Study (MACS)—a longitudinal study of HIV/AIDS in gay and bisexual men. Recently, the model was adapted to address questions relevant to U.S. HIV-infected women using data from the Women’s Interagency HIV Study (WIHS). Motivated by parameter changes and assumptions related to use of the WIHS dataset, we assessed the internal consistency of the newly parameterized natural history model by exploring the ability to achieve good visual fits to the empiric survival data for untreated HIV-infected women. We evaluated model performance and characterized uncertainties associated with model assumptions using calibration techniques in addition to exploring model external validity in estimating treatment survival using a unique subset of WIHS empiric survival data for women who received ART.

Chapter III examines the cost-effectiveness and trade-offs between methods for treating panic disorder. Panic disorder is often debilitating, resulting in numerous lost workdays and reduced overall productivity. We developed a 1st order Monte Carlo simulation model of the natural history of panic disorder and treatment to assess the cost effectiveness and the impact of different treatment strategies on burden of disease; this analysis represents a significant extension of modeling research in the area of mental health. Our analysis supports conclusions that treatment of panic disorder is cost-effective even at low cost-effectiveness thresholds and can substantially reduce time spent with disease. However, intervention characteristics, patient heterogeneities and disease course have important implications on treatment choice.