Variation in Phenotype Definitions in Observational Clinical Research: A Review of Three Conditions

Our study examined the heterogeneity of phenotype algorithms (PA) in literature on Alzheimer's disease (AD), major depressive disorder (MDD), and pulmonary arterial hypertension (PAI), focusing on the impact of PA differences on patient overlap and incidence rate variability across conditions in six observational databases. We reviewed 49 replicated PAs (13 for AD, 23 for MDD, and 13 for PAI) and found significant heterogeneity. These varied PAs identified distinct patient cohorts, resulting in significant incidence rate heterogeneity. Despite some papers reporting primary condition codes and inclusion. comprehensive documentation ensuring reproducibility was often lacking, underscoring the need for more transparent and robust research practices.