Branching Out: Use of Decision Trees in Epidemiology

Purpose of Review Decision trees are a well-established tool for statistical modeling and machine learning, but they are not widely used in the epidemiological literature. In this review, we introduce the reader to the basic concept of the decision tree and describe three distinct ways that they can be used: for explanatory modeling, outcome prediction, and subgroup identification. Recent Findings We discuss varieties and generalizations of decision trees that are best-suited for analyzing epidemiological data and introduce some visualizations which can help researchers interpret decision tree outputs. Throughout, we provide diverse examples from recent literature of how decision trees have been applied to analyze epidemiological data. Summary The overall aim is to encourage epidemiologists to incorporate decision trees into their analytic toolkit.