Effect and Mechanisms of a Voice-based Coach using AI on Psychological Distress: A Phase 2 Randomized Trial

Abstract

Importance: Artificial Intelligence (AI) voice applications have the potential to address the unmet treatment needs among patients with depression and anxiety, but their therapeutic utility is largely unknown. Objective: To investigate the efficacy and mechanisms of an AI-driven voice-based coach, Lumen, delivering problem-solving treatment (PST) for patients with untreated, moderate depression and/or anxiety. Design: Phase 2, 3-arm randomized clinical trial. Setting: A public university and affiliated medical center in Chicago, IL. Participants: Adults with a Patient Health Questionnaire-9 score of 10-19 and/or a Generalized Anxiety Disorder Scale score of 10-14 were randomized to Lumen-coached PST (n=100), human-coached PST (n=50), or waitlist control group (n=50) groups. Interventions: PST was delivered by Lumen on a custom-developed application on Amazon's Alexa platform or a human coach via videoconferencing in 4-weekly and then 4-biweekly sessions. Main Outcome(s) and Measure(s): Hospital Anxiety and Depression Scale was used to assess depression and anxiety symptoms, which summed to an overall psychological distress score (primary outcome), at baseline and at 18 weeks. Measures of mechanisms included problem-solving abilities (using Social Problem-solving Index-Revised Short Form) and a priori neural targets. Results: Among 200 participants (Mean (SD) age, 36.6 (11.9) years), 77% were women, 25% Black American, and 28% Latino. At 18 weeks, Lumen-coached participants had significantly greater reductions in psychological distress (between-group mean difference=-3.56 [95%CI: -5.69, -1.43], Cohen's d=0.49), depression (-1.15 [95%CI: -2.25, -0.04]; d=0.19), and anxiety symptoms (-2.43 [95%CI: -3.72, -1.14]; d=0.61), compared with waitlist control participants. Mean differences in change from baseline to 18 weeks between Lumen- and human-coached participants were not statistically significant for all 3 symptoms. Both Lumen- and human-coached PST resulted in significant improvements in overall problem-solving ability and negative problem orientation, which were significantly correlated with improved psychological distress and anxiety symptoms. Neural targets did not change significantly with either treatment vs. waitlist control. One serious adverse event involving hospitalization, unrelated to this study, was detected. Conclusions and Relevance: An AI-driven voice-based coach delivering PST may be a viable treatment option for untreated, moderate depression and/or anxiety. However, clinical efficacy needs to be confirmed in a phase 3 trial. Trial Registration: clinicaltrials.gov,NCT05603923

Links & Resources

Authors

Cite This Paper