Reducing Non-Medical Opioid Use: An Automatically Adaptive mHealth Intervention

Overview

In recent years in the U.S., problems associated with opioid prescriptions, including non-medical use and overdose, increased to historically unprecedented levels and represent a public health crisis. Emergency departments (EDs) play an important role in opioid prescribing, particularly to individuals at high risk for adverse opioid-related outcomes. The purpose of this study is to determine whether a new mobile health (mhealth) intervention can assist people in the safe use of opioid analgesic (OA) medications after leaving the emergency department (ED). The specific aims of this project are to: (1) adapt and enhance an existing motivational intervention to decrease non-medical opioid use after an ED visit by optimizing intervention intensity and duration through reinforcement learning (RL); (2) examine the impact of an RL-supported intervention on non-medical opioid use level during the six months post-ED visit; and (3) examine the impact of the RL intervention on the opioid-related behaviors and adverse outcomes of driving after opioid use, overdose risk behaviors, and subsequent opioid-related ED visits. The secondary aims of this project are to: (SA1) examine whether baseline level of non-medical opioid use moderates the effects of the intervention; and (SA2) understand barriers and facilitators of implementation of the intervention based on qualitative interviews with ED patients.

The proposed study will test the efficacy of an interactive voice response (IVR) and reinforcement learning (RL) supported motivational intervention delivered after an emergency department (ED) visit to participants with recent non-medical OA use who receive an OA in the ED or who are prescribed an OA at ED discharge, compared to enhanced usual care (EUC). In the intervention condition, IVR calls will ask participants to report information about their health and medications using their touch-tone phone, and based on their responses they may receive brief or extended motivational messages during the IVR call, or they may be assigned to receive a 20 minute motivational enhancement session with a study therapist over the phone. Because the most helpful intensity of intervention is unknown and likely to vary between patients, the project will use an artificial intelligence strategy called reinforcement learning (RL). The RL system will continuously "learn" from the success of prior actions in similar situations with similar patients in order to select the action most likely to reduce non-medical opioid use for each participant during each call. The proposed study will screen \~ 5,600 ED patients to enroll 600 ED participants in the randomized controlled trial (RCT). Participants will be randomized to the intervention condition (n=300) or to EUC (n=300). All participants will be re-assessed at 1, 3 and 6 months post-ED visit for level of non-medical OA use and related outcomes. The RCT will be complemented by qualitative interviews to inform later implementation.