Emerging Methods for Prevalence Estimation With Imperfect and Multisource Data

This session focuses on modern statistical methods for estimating prevalence and population-level burden from real-world data subject to imperfect and biased observation. Speakers present recent methodological advances that address heterogeneous ascertainment, dependence across data sources, linkage error, and misclassification—challenges that routinely undermine classical approaches such as capture–recapture and small-area estimation. The talks span complementary paradigms, including models for observation processes, Bayesian integrative population frameworks, hierarchical and small-area methods, and latent or pseudo–multisource constructions for recovering unobserved populations. Through applications ranging from ecology to substance use disorder surveillance, the session highlights how these approaches collectively improve uncertainty quantification and population inference in complex data environments.