OBJECTIVES: To assess the relationship between patient engagement with reSET-O, an FDA-cleared prescription digital therapeutic (PDT) for the treatment of opioid use disorder in adults, and change in inpatient (IP) and emergency department (ED) utilization using a machine learning approach.

METHODS: An unsupervised machine learning approach was applied to classify PDT patient subgroups based on two distinct engagement categories: count of distinct lessons completed, and count of weeks >= 1 lesson was completed. IP and ED resource utilization, along with cost ($12,476 and $536, respectively, in 2020 USD), was evaluated for the six months pre- and post-treatment initiation with reSET-O (index date). PDT subgroups were assessed relative to IP/ED utilization and cost.

RESULTS: Four distinct PDT subgroups were identified via k-means clustering: high (n=130), medium (n=88), low (n=104), and non (n=26) engagement. Twenty-one percent of patients in the high engagement subgroup had >=1 IP/ED event pre-index, versus 31%, 39%, and 19% in the medium, low, and non-engagement subgroups, respectively. All engaged subgroups experienced decreases in IP/ED utilization following treatment initiation (high: 39% pre- vs 16% post-index; medium: 31% vs 13%; low: 21% vs 9% for pre- relative to post-index, respectively). By contrast, the non-engagement subgroup experienced an increase in IP/ED utilization (19% pre- vs 23% post-index). The low engagement subgroup had the highest within-patient reduction in IP/ED utilization at 34% (95% CI 25.3-43.2%), compared to the medium (26%, 18.1-36.2%) and high (18%, 12.1-25.2%) engagement subgroups. Within-patient reduction in IP/ED utilization was lowest for the non-engagement subgroup (12%, 4-29%). Cost reductions were 1.8 - 2.7 times greater with any engagement ($2080, $2654 and $1749 for high, medium, and low PDT groups, respectively) relative to non-engagement ($979).

CONCLUSIONS: Compared to non-engagers, patients that engaged with reSET-O experienced reductions in IP/ED utilization and costs, supporting broad therapeutic value of the PDT.