376 dCBT-I with Chatbot and Artificial Intelligence: a feasibility study in Brazil. (3rd May 2021)

• Record Type: Journal Article
• Authors: Castro, Laura; Baraças, Lucas; Hashioka, Guilherme; Carvalho, Adriana
• Is Part Of: Sleep. Volume 44(2021)Supplement 2
• Page Start: A149
• Page End: A150
• Publication Date: 2021-05-03
• Subjects: Sleep -- Physiological aspects -- Periodicals; Sleep disorders -- Periodicals; Sommeil -- Aspect physiologique -- Périodiques; Sommeil, Troubles du -- Périodiques; Sleep disorders; Sleep -- Physiological aspects; Sleep -- physiological aspects; Sleep Wake Disorders; Psychophysiology; Electronic journals; Periodicals; 616.8498
• DOI: 10.1093/sleep/zsab072.375
• Languages: English
• ISSNs: 0161-8105

Abstract

Abstract: Introduction: Digital cognitive-behavioral therapies for insomnia (dCBT-I) provide low-cost, evidence-based technology, effective in improving mental health and reducing healthcare costs. However, dropout rates still challenge dCBT-I scalability. Moreover, few solutions are available in middle-and-low-income economies where they are most needed. Our goal was to investigate feasibility, describing real-world data and preliminary findings of a novel, fully automated program, developed by Vigilantes do Sono (Sleep Watchers) using Chatbot and Artificial Intelligence (AI). Methods: A digital coach interacts with users daily for 5–10 minutes, asking them to complete tailored diaries and delivering CBT-I knowledge pills in ~50 sessions, during ~7 weeks. The Insomnia Severity Index (ISI) is used before and after sleep restriction cycles, weekly revised by an algorithm. Participants (18+ years) were recruited (Jan-Oct/2020) through advertisements on social media, organic search, or were referred by health-care professionals, without face-to-face evaluation. All electronically signed an informed consent. We estimated engagement dividing number of complete diaries by number of days in the program. Generalized Estimating Equations (GEE) evaluated changes in sleep parameters, adjusting for baseline characteristics. Results: Of 3, 887 individuals who completed initial assessment, 3, 139 (81%) had insomnia (ISI ≥11) and 1, 489 (42±11 years, 91% women) fulfilled 7+ diaries, commenced sleep restriction, and were included in analysis. Of them, 604 (41%) completed a second ISI and 326 (22%) finished the program. GEE analyzing 42, 802 diaries showed sleep duration increased 16.8 (11.9–21.6) minutes from first to second week and 67.3 (52.8–81.8) after week seven; parallel to a relative increase of 34% in sleep efficiency among women and 26% among men. Of 296 participants who reached therapeutic response (ISI reduction ≥8), 66% completed all sessions and 34% crossed half-way. Insomnia remission (ISI≤7) was seen for 55% and 33% of those with subthreshold (n=171) or clinical (n=419) baseline insomnia, respectively. Median (interquartile) engagement was 86% (65–98) and 90% of users recommend the program. Conclusion: Chatbot and AI provide a framework to customize dCBT-I and personalize insomnia therapy, potentially favoring engagement and effectiveness. Our findings demonstrate feasibility of the program and support moving forward to continued development and testing the effects in clinical trials. Support (if any):