Developing a Robust Computable Phenotype Definition Workflow to Describe Health and Disease in Observational Health Research

Health informatics can inform decisions that practitioners, patients, policymakers, and researchers need to make about health and disease. Health informatics is built upon patient health data leading to the need to codify patient health information. Such standardization is required to compute population statistics (such as prevalence, incidence, etc.) that are common metrics used in fields such as epidemiology. Reliable decision-making about health and disease rests on our ability to organize, analyze, and assess data repositories that contain patient health data. While standards exist to structure and analyze patient data across patient data sources such as health information exchanges, clinical data repositories, and health data marketplaces, analogous best practices for rigorously defining patient populations in health informatics contexts do not exist. Codifying best practices for developing disease definitions could support the effective development of clinical guidelines, inform algorithms used in clinical decision support systems, and additional patient guidelines. In this paper, we present a workflow for the development of phenotype definitions. This workflow presents a series of recommendations for defining health and disease. Various examples within this paper are presented to demonstrate this workflow in health informatics contexts.