Chatbot-delivered alcohol screening and brief intervention for working-age adults: A proof-of-concept randomised waitlist-controlled trial
Related Experiment Video
Updated: Oct 30, 2026
The Motivation for Alcohol Reward: Predictors of Progressive-Ratio Intravenous Alcohol Self-Administration in Humans
Published on: April 28, 2022
Siu Long Chau1, Janet Yuen Ha Wong2, Daniel Yee Tak Fong3
1School of Public Health, The University of Hong Kong, Hong Kong Special Administrative Region of China.
Background:
This study adapted a web-based chatbot to deliver screening and brief intervention (SBI) and assessed its usability, engagement, and preliminary between-group estimates for alcohol-related outcomes.
Methods:
This randomised waitlist-controlled trial recruited 304 community-dwelling working-age adults with an Alcohol Use Disorders Identification Test-C (AUDIT-C) score ≥ 3 in Hong Kong from 15 October 2024-12 May 2025 and randomised 1:1 to the intervention group (IG; n = 152) or waitlist control group (WL; n = 152). Both groups received standardised alcohol brief intervention (ABI) at baseline. The IG received web-based chatbot-delivered SBI immediately, while WL received it after the 4-week follow-up. The primary outcome was weekly alcohol consumption (grams/week) at 4-week follow-up. Secondary outcomes included AUDIT score, chatbot engagement (defined as having read the chatbot messages and interacted with the chatbot more than once during the intervention period, verified by system log files), and 15-item Bot Usability Scale [BUS]. To evaluate cost-effectiveness and scalability, an economic evaluation of intervention-related costs and quality-adjusted life years (QALYs) estimated from EQ-5D-5L utility scores were calculated.
Results:
By intention-to-treat analysis, the adjusted between-group difference in weekly alcohol consumption was -10.2g/week (95% CI: -33.8-13.4; P = 0.40) and AUDIT score was b = -2.6 (95% CI: -4.6-1.4; P = 0.46) at 4 weeks. Chatbot process indicators showed satisfactory use, with an engagement rate of 56.2% (171/304), a mean BUS score of 56.8/75 (SD 7.8), and overall satisfaction of 7.8/10. The mean intervention-related cost was HKD 489.5 (USD 62.9) per participant, and EQ-5D-5L utility scores increased from 0.870 to 0.909 in the IG and from 0.860 to 0.876 in the WL.
Conclusions:
The chatbot-delivered SBI showed satisfactory usability and moderate engagement, supporting its feasibility and acceptability. This trial did not show an incremental effect on alcohol consumption over the standardised ABI received by both groups. Future trial with longer follow-up and a control condition that better isolates the chatbot component is warranted.
