Predictor Profiles of Opioid Use Disorders and Overdose Among Post-9/11 Veterans

点击翻译按钮获取中文摘要

The overall aim of this proposed study is to use machine learning prediction models to evaluate the multifaceted, additive and multiplicative interactions of known and novel risk factors for opioid use disorder (OUD) and overdose in Post-9/11 Veterans. The proposed study will also investigate the short- and long-term impact of the coronavirus disease 2019 (COVID-19) pandemic on the risk of OUD and overdose. TRAINING PLAN: The CDA-2 training plan will facilitate the applicant’s primary career goal of becoming a fully funded, independent epidemiologic researcher at the Department of Veterans Affairs (VA), with a focus on addiction and suicidal behavior. The CDA-2 will provide additional training necessary to lead an independent program of research investigating the multifaceted sociodemographic, physical, psychological, and behavioral factors mediating and moderating the risk of addiction and suicidal behavior. The first step of achieving this goal is to complete the following training aims: 1) gaining expertise in the biological and behavioral basis of addiction; 2) gaining expertise in the assessment of the problems of TBI and blast exposure, psychiatric disorders, and suicidal behavior, which is pervasive in this generation of Veterans; 3) gaining expertise in advanced analytic techniques employed in health data science, including machine learning algorithms; and 4) professional development to achieve career independence as a VA funded epidemiologic researcher. RESEARCH DESIGN & METHODS: The proposed study will use Veterans Health Administration (VHA) electronic medical records to develop models predicting OUD and overdose risk. The sample will include Post- 9/11 Veterans who are aged 18-65, receive care in the VHA, and will have completed the VA primary TBI screen between October 2007 and February 2020 (n~1,267,000). We will assess the risk of incident and recurrent OUD and overdose events, as separate outcomes, using machine learning algorithmic models. We will examine whether overdose was 1) fatal and non-fatal and 2) intentional and unintentional. For Aims 1 and 2, we will examine the risk of OUD and overdose events between October 1, 2007 and February 29, 2020. For Exploratory Aim 3, we will examine the risk of OUD and overdose events between March 1, 2020 and September 30, 2025. We will use several machine learning classification-tree modeling approaches, including classification and regression trees, random forest, and gradient boosting, to develop predictor profiles of OUD and overdose incorporating important risk factors and interactions. The validity (sensitivity and specificity) and prediction accuracy (area under the curve) will be assessed for all prediction profile models. OBJECTIVES: Aim 1: Develop and evaluate the performance of predictor profiles incorporating known and novel risk factors and interactions for OUD and overdose over proximal (30, 60, and 90 days) and distal (180, 365, 730, 1095 and >1460 days) prediction intervals using machine learning classification algorithms. Hypothesis 1a: The machine learning algorithms will have high validity and prediction accuracy (e.g., sensitivity and specificity and area under the curve) >0.8. Hypothesis 1b: Accuracy and predictive ability will be higher in the proximal vs. distal prediction intervals. Aim 2: Examine gender, race/ethnicity, deployment-related trauma (e.g., TBI and prevalent psychiatric and substance disorders), and close-blast exposure as moderators of the risk of OUD and overdose. Hypothesis 2: There will be novel risk factors and differential variable importance impacting the risk of OUD and overdose within the subgroup-specific predictor profiles. Exploratory Aim 3: Investigate the short- and long-term impact of the COVID-19 pandemic on the risk of OUD and overdose using machine learning classification algorithms to develop predictor profiles of known and novel risk factors and interactions. Hypothesis 3: The COVID-19 pandemic will have both a direct effect on the risk for OUD and overdose and an indirect effect through the onset or exacerbation of mental health symptoms and psychiatric conditions.