Encouraging Abstinence Behavior in a Drug Epidemic: Optimizing Dynamic Incentives

Over the past decade, the annual number of drug-related deaths more than doubled in the United States (Swensen, 2015). In particular, over the 2001-2013 period, overdose deaths involving prescription pain relievers tripled while those involving heroin increased fivefold (NIDA, 2015). Further, the COVID-19 pandemic is thought to have significantly increased drug use, especially opioids, cocaine, and methamphetamines. This upward-sloping trend has steepened in the past few years. Drug overdoses are now the principal cause of death among Americans aged less than 50. A primary cause of this escalating public health crisis is the abuse of opioids (e.g., prescription pain relievers and heroin), which is estimated to concern more than two million Americans (New York Times, 2017).

Many studies in the medical literature have tested whether providing incentives to encourage abstinence from drugs can further reduce drug abuse in a drug-treatment setting. The results are very promising: Incentives to reduce opioid abuse increase the average duration of abstinence by 25 - 60% relative to medication and counseling alone (Petry et al., 2005; Schottenfeld et al., 2005; Petry et al., 2010; Ling et al., 2013). Similar effects have been demonstrated repeatedly across a wealth of populations, substance-abuse disorders, and payment methodologies (Lussier et al., 2006; Davis et al., 2016; Higgins, 2016). A meta-analysis of psychosocial treatments concluded that providing incentives for abstinence behavior was the intervention with the greatest effect size in treating substance use disorders (Dutra et al., 2008). Despite their costs, incentive programs have been estimated to be cost-effective, with the estimated benefits - including benefits to participants and to taxpayers from lower health care costs and higher earnings - estimated to be on the order of 20 times as large as normal program costs (WSIPP, 2017). Although such estimates are somewhat speculative, the case for scaling up incentive programs is strong.

And yet, despite evidence that incentives are effective and the ever-more-dire need for effective approaches to combat the addiction crisis, incentive programs have not been scaled up widely to date. A key barrier is that while the benefits are largely borne by patients and taxpayers, there are large logistical costs that must be borne by clinics: existing incentive programs involve manual, in-person measurement of behaviors, and prize or voucher purchase and delivery by clinic staff. The significant clinic-level legwork necessary to set up these programs, including setting up behavioral and payment tracking systems, training staff, etc., have prevented the programs from scaling widely (Benishek et al., 2014).

We propose to conduct the first randomized evaluation of an innovative, scalable incentives program for drug addiction delivered through a mobile application. The application, which was developed by our implementing partner, DynamiCare Health (henceforth "DynamiCare"), provides a "turnkey" solution that health clinics can easily prescribe. The app enables remote monitoring of behavior; for example, drug tests can be administered in patients' homes, as patients submit "selfie-videos" showing them taking saliva drug tests, which are then verified by trained remote staff. Treatment adherence can similarly be checked through GPS tracking for on-site methadone pharmacotherapy. The efficacy of this approach has not been tested rigorously before.

This study will address two key knowledge gaps in the logistics of existing incentive program design for drug addiction. First, we will test the first technology that we know of for remote monitoring of abstinence behavior for drug use. Remote monitoring of abstinence from cigarettes and alcohol has been integral in reducing the costs and extending the potential reach of incentive programs for people with nicotine/tobacco and alcohol use disorders (e.g. to vulnerable or rural populations), and our study promises to do the same for illicit drug addiction (see for a review of remote monitoring technologies for incentive delivery). Our second gap is in remote delivery of incentives. After a behavior is verified, the app will deliver incentives to patients as cash available on a linked debit card. The delay between monitoring of the target behavior and the delivery of financial incentives has been shown to be a significant moderator of treatment effect size (Lussier, Heil, Mongeon, Badger, \& Higgins, 2006). Our technology allows patients to receive incentives almost immediately following the undertaking of the incentivized behavior: a first in incentives for drug addiction.

The second question is how to optimize the size of incentives over time to maximize incentive effectiveness. We propose to do this by randomly varying the size and timing of incentives offered to participants across groups. We will then use the variation in incentive amounts across participants and time to fit a structural model of abstinence behaviors over time. We will then use the model to describe the optimal shape of incentives over time.

The results of this intervention will be directly relevant for potential users of this or similar mobile applications for incentive provision among people with substance use disorders, including insurers, treatment facilities, and governments.

• Percent of Psychotherapy Visits Attended [Срок оценки: month 3]
• Percent of Treatment Discharge Against Medical Advice [Срок оценки: month 3]
• Percent of Relapse [Срок оценки: month 3]
• Percent of transition to step-down care [Срок оценки: Day 14]