Personalized Treatment for Opioid Use Disorder
40 Pages Posted: 12 Jun 2019
Date Written: May 16, 2019
In order to be cost effective, an opioid use disorder (OUD) treatment must collect and utilize information on how a patient responds to different treatment regimens. Traditional methods of evaluating patient response – urine tests and self-reports – have not been effective: the number of days that a urine test can detect drug usage is relatively small, and self-reports are subject to response bias. In contrast, wearable devices can potentially help detect patient craving episodes and health status in real-time. A variety of wearable devices with different features and costs are available; whether such devices are practical in OUD treatments, and if so how they should be used, are critical questions. We build a sequence of partially observable Markov decision processes (POMDPs) and a Markov decision process (MDP) with budget constraints to address these questions. We provide a fast solution method for the POMDP models: a novel heuristic algorithm with an analytic error bound. Using our models, we perform a numerical study to investigate the value of incorporating different wearables in OUD treatments under various scenarios of budget, wearable precision, and patient treatment adherence (TA). We find that wearables can be valuable at moderate budgets for patients with low or moderate TA. This benefit increases as the wearable accuracy increases and as we use wearables to learn patients’ personalized treatment dynamics (PTD).
Keywords: MDP, POMDP, Budget Constraint, Personalized Treatment, Opioid Addiction
Suggested Citation: Suggested Citation