Supporting data for "Optimizing the mHealth intervention effect on addictive behaviors"
The rapid rise of mHealth has transformed healthcare delivery, and been found effective in addictive behaviors. However, challenges on participant disengagement and insufficient descriptions of the intervention content hinder further research. This thesis explored the mechanisms behind effective mHealth interventions for smoking cessation and alcohol reduction by examining both human-level factors (engagement) and intervention-specific content (Behavior change techniques, BCTs).
A community-based, two-arm, parallel group, pragmatic, cluster-randomized controlled trial was conducted in 2017 to test the effectiveness of a 3-month mHealth chat-based smoking cessation intervention. It integrated regular cessation-specific messages and real-time support by a live counselor. Biochemically validated smoking abstinence was measured at 3 and 6 months. Similarly, a two-arm, parallel group, pragmatic and individual-randomized controlled trial for risky alcohol drinkers (AUDIT 8+) visiting AEDs was started in 2022. This intervention consisted of brief alcohol advice in-person at baseline, followed by 3 months of mHealth chat-based support with regular alcohol-related messages and real-time support. Study outcomes include weekly alcohol consumption, AUDIT score, binge drinking, heavy drinking, planned drinking, alcohol-related harm and AED re-attendance. Secondary analyses included participants from the intervention groups of both trials. Weekly responses to the regular messages were characterized as engagement, and analyzed using group-based trajectory modelling. Poisson regression with robust variance or linear regression were used to examine the associations between engagement trajectories and study outcomes. Only participants who interacted with counselors during real-time support were included for characterizing BCTs. Predefined messages were excluded for analyzing, with only the counselor’s messages during real-time support being coded. Machine learning approaches were employed to identify predictive BCTs for smoking abstinence, and Poisson regression with robust variance was used to examine associations between these BCTs and abstinence outcomes. The associations between coded BCTs and alcohol-related outcomes were tested using Poisson or linear regression in alcohol reduction trial.
Different patterns of engagement were observed in smoking cessation and alcohol reduction intervention. Results from both studies indicated that higher engagement levels were associated with better outcomes. Significant BCTs associated with smoking cessation and alcohol reduction were identified. Future studies should focus on enhancing participant engagement, and incorporating effective BCTs to optimize the impact of mHealth interventions.
