An investigation of the age-sex pattern in Spectrum's estimates of HIV/AIDS mortality for generalized epidemics
Ortblad, Katrina Frances
MetadataShow full item record
University of Washington Abstract An investigation of the age-sex pattern in Spectrum's estimates of HIV/AIDS mortality for generalized epidemics Katrina Frances Ortblad Chair of Supervisory Committee: Christopher JL Murray, MD, DPhil Department of Global Health, Institute for Health Metrics and Evaluation Objective: Currently there are significant variations between the age-sex patterns of the Global Burden of Disease 2010 (GBD 2010) Study's estimates of all cause morality and UNAIDS' 2012 estimates of HIV/AIDS mortality for a number of sub-Saharan African (SSA) countries. While both estimates have their own unique limitations, the goal of this analysis was to explore assumptions used in UNAIDS' Spectrum model that influence their age-sex specific HIV/AIDS mortality estimates. Methods: UNAIDS' Spectrum/Estimation and Projection Package (EPP) was used as the vehicle for exploration. Spectrum currently applies the same assumptions and IRRs for all countries in SSA with generalized epidemic. Assumptions about the progression of HIV infection in the absence of antiretroviral treatment and age-sex specific incidence rate ratio (IRR) were identified as two areas in Spectrum that could influence the model's age-sex patterns of HIV/AIDS mortality. Within these, four scenarios were explored; adjusting the proportion of new infections in CD4+ count categories, adjusting years lived in various CD4+ count categories, adjusting the non-AIDS mortality rate, and adjusting the IRRs. Assumption modifications were made based on alternative cohort studies, estimates from the GBD 2010 study, or country-specific data. Age-sex-specific HIV/AIDS mortality estimates were generated using these modified assumptions and these were compared to those generated using Spectrum defaults. Results: There is wide variation in the Spectrum assumptions by cohorts and country-level DHS data. When these alternative assumptions for disease progression or IRRs are entered in Spectrum, this can shift the age group with the highest estimates of HIV/AIDS mortality and can also alter the level of mortality. The degree of estimate variation depends on the degree of assumption adjustment. When disease progression parameters are calibrated to the ALPHA survivorship data results are more similar to those generated using Spectrum's default values. The cohorts that drive Spectrum's assumptions have small sample sizes and are not necessarily representative of all SSA countries. Conclusions: Spectrum's age-sex patterns of HIV/AIDS mortality are sensitive to disease progression parameters and IRR adjustments. When compared to other studies or more detailed data sources, Spectrum's current assumptions for these inputs do not capture the wide variability seen at the individual and subpopulation level. A systematic review of data from all potential cohort studies should be conducted and Spectrum should better propagate uncertainty in parameter values. More country-level data should also be incorporated into Spectrum, when available, to better reflect each country's unique epidemic.
- Global health