BackgroundMorbidity estimates between different GP registration networks show large, unexplained variations. This research explores the potential of modeling differences between networks in distinguishing new (incident) cases from existing (prevalent) cases in obtaining more reliable estimates.MethodsData from five Dutch GP registration networks and data on four chronic diseases (chronic obstructive pulmonary disease [COPD], diabetes, heart failure, and osteoarthritis of the knee) were used. A joint model (DisMod model) was fitted using all information on morbidity (incidence and prevalence) and mortality in each network, including a factor for misclassification of prevalent cases as incident cases.ResultsThe observed estimates vary considerably between networks. Using disease modeling including a misclassification term improved the consistency between prevalence and incidence rates, but did not systematically decrease the variation between networks. Osteoarthritis of the knee showed large modeled misclassifications, especially in episode of care-based registries.ConclusionRegistries that code episodes of care rather than disease generally provide lower estimates of the prevalence of chronic diseases requiring low levels of health care such as osteoarthritis. For other diseases, modeling misclassification rates does not systematically decrease the variation between registration networks. Using disease modeling provides insight in the reliability of estimates.