Data-augmented multistate modeling for chronic disease processes using cross-sectional studies: application to HPV and cervical precancer.
Understanding the population-level natural history of human papillomavirus (HPV) infection and cervical precancer is essential for informing effective prevention strategies. Multistate models provide a principled framework for characterizing disease progression by modeling transitions among discrete health states. However, intensity-based multistate analyses typically require longitudinal data, which are often costly and logistically difficult to obtain, whereas cross-sectional data are more widely available, particularly in resource-limited settings. We develop a semi-Markov multistate model that captures HPV infection dynamics and cervical precancer progression, including recurrent transitions between HPV-infected and uninfected states. To enable estimation from cross-sectional data, we propose a two-step approach that combines local cross-sectional observations with external information on transferable transition intensities. First, biologically driven transition intensities are estimated from external data under a transferability assumption. Second, these estimates are incorporated into an approximate likelihood to infer population-specific transition intensities shaped by local epidemiological and behavioral factors. Simulation studies and applications to 2 populations demonstrate the performance of the proposed method and its practical utility.