Prediction of Relapse Using Digital Technology in People in Recovery From Substance Use Disorders: Early Economic Evaluation With a Case Study of the Subreal App.

Background: Many people relapse after achieving abstinence in substance use disorders. Health care providers may scan the horizon for new technologies to predict response that allow interventions to be targeted rather than routine. Currently, no such predictive technologies are available in the United Kingdom. The Subreal app is available for use in research contexts, but no clinical data specific to the app are yet available. Early health economic modeling can use data from the literature to explore characteristics essential for the new technology to be cost-effective. This information can guide developers in setting performance targets and pricing and estimating potential cost savings and/or cost-effectiveness for health care providers. Objective: This study was supported by a UK industry funding body to explore the potential of digital technologies such as the Subreal app to offer cost savings or cost-effectiveness for health care providers. We explored the threshold price and clinical effectiveness required to deliver cost savings and cost-effectiveness in 2 subpopulations with substance use disorders in a UK setting. Methods: Deterministic models were used to estimate costs per relapse and quality-adjusted life years over 1-, 5-, and 20-year time horizons for people who have achieved abstinence after treatment for alcohol or opioid misuse. The intervention was a digital technology predicting relapse, provided-in addition to standard care-for 1 year post achievement of abstinence. In Subreal, biomarker data are collected daily through the app, and artificial intelligence-enhanced risk assessment flags patients who require additional support. The comparator was event-driven, reactive response to relapse. Costs and quality-of-life estimates were calculated using Markov models with data from existing published sources. The base-case estimate of 15% reduction in first-year relapse rates was based on a previous study on a similar but simpler digital technology. Results: Digital technologies such as the Subreal app have the potential to be cost-saving from a UK health and social care perspective, especially when used over a longer time horizon. Assuming a reduction of 15% in first-year relapse rates, digital technologies have the potential to be cost-saving, provided that they do not cost more than £300 (US $400.09) and £460 (US $613.47) per patient per annum for alcohol and opioid use disorders, respectively. No cost was included for postalert care, as it was assumed that this could be met within existing resources. Cost savings would be achieved predominantly through a reduction in treatment requirements as fewer people relapse. Price thresholds would reduce correspondingly if a <15% reduction in relapse rates were achieved. Conclusions: Developers of digital technologies that aim to reduce relapse need to focus on the generation of evidence of clinical effectiveness and develop a commercially sustainable pricing model that allows health care providers to benefit from cost savings.