Purpose: Data informing the natural history component of a previously published model of HIV disease (CEPAC model) were largely from the Multicenter AIDS Cohort Study (MACS). The model was recently adapted to address questions relevant to U.S. HIV-infected women. We assess the internal consistency of a newly parameterized model using data from the Women's Interagency HIV Study (WIHS), and explore the ability to achieve good visual fits to empiric survival data in women. Methods: Using the CEPAC 1st order Monte Carlo model, simulations were performed to project survival in HIV-infected women stratified by mean starting CD4 cell count. Cohort characteristics (age, CD4 and viral load distribution) were from the WIHS. Simulations were conducted first using probabilities of opportunistic infections (OI), attributable mortality, CD4 cell decline, and HIV-related mortality from the MACS and other published literature, and then from the WIHS. Kaplan-Meier survival curves based on data from WIHS were visually compared with model-estimated survival. After a series of one-way sensitivity analyses, we ranked uncertain assumptions a priori and systematically varied each individually to ascertain influence on visual fit to 3-year survival stratified by CD4 count. Results: Using WIHS data, model-projected survival over 36 months in women with CD4 <50 and CD4 >350 closely approximated empiric survival data in women. Model-projected survival in women with CD4 50-199 and 200-349 closely approximated the empiric data through 12 months of follow-up but substantially underestimated survival between 12 and 36 months. A series of one-way sensitivity analyses showed little influence of initial viral load distribution and mortality attributable directly to acute opportunistic infections when varied over plausible ranges; however, assumptions concerning both the incidence of opportunistic infections and rates of chronic AIDS death in patients with a prior opportunistic infection were influential. The best visual fit to the WIHS empiric survival data required a 50% reduction in either one of these in all CD4 strata except CD4 <50. Good visual fits to the empiric survival data in women were not achieved using data from the all-male MACS cohort. Conclusions: Iteratively assessing internal consistency, face validity, and performance of disease simulation models can identify assumptions that are inconsistent with observations based on empiric data, highlight the need for parameterization, and provide insight into areas of uncertainty likely to be influential.