Analyzing predictors of high opioid use in the U.S. Navy

Abstract

This study analyzes data from a select group of active duty (AD) service members enrolled to the Puget Sound area Navy military treatment facilities (MTF) in order to develop a model that identifies the risk that opioid users will become high opioid users, as defined by Navy Bureau of Medicine and Surgery (BUMED). The analysis examines the relationship between the response variable—high opioid user—as a function of a number of explanatory variables, including patient age, deployment history, sources of prescription and medical diagnoses. Logistic regression and machine learning models are used for data analysis. The study concludes that a simple, executable model that consolidates the variables to two explanatory factors performs as well, if not better than, the more complicated machine learning models. The two highly influential factors are the number of prescription sources for opioid medications and the total number of diagnoses. This logistic regression model has the potential to benefit Navy Medicine to make important decisions for their opioid-prescribed patients. With the ability to identify the risk that an opioid user becomes a high user, healthcare leaders can better manage resources to focus on the prevention and treatment of higher-risk patients. This concentrated coordination can result in improved patient care for this sub-population, reduced long-term cost for the military healthcare system and, overall, a more medically ready military force.