Currently accepted at: Journal of Medical Internet Research
Date Submitted: Jun 25, 2026
Date Accepted: Nov 20, 2026
This paper has been accepted and is currently in production.
It will appear shortly on 10.2196/105525
The final accepted version (not copyedited yet) is in this tab.
Comparative Effectiveness of AI-based and Other Digital Health Interventions for Smoking Cessation by Decision and Adaptation Mechanism: A Systematic Review and Network Meta-Analysis
ABSTRACT
Background:
Rapid advances in artificial intelligence (AI), particularly natural language processing and machine learning, have expanded the potential for automated and adaptive smoking cessation support. Digital smoking cessation interventions vary substantially in how participant-specific or contextual information is used to determine subsequent intervention actions. However, their comparative efficacy across decision and adaptation mechanisms remains uncertain.
Objective:
This study aimed to compare and rank digital smoking cessation intervention packages categorized by their decision and adaptation mechanisms and to examine factors associated with intervention effectiveness. Exploratory analyses compared delivery platforms and mechanism-platform combinations.
Methods:
We searched 15 bibliographic databases, trial registries, and gray literature sources from inception to May 1, 2026. Randomized controlled trials evaluating digital interventions for smoking cessation were eligible. In the primary network meta-analysis, digital smoking cessation intervention packages were classified by decision and adaptation mechanisms as static pre-programmed, human-directed adaptive, rule-based adaptive, AI-based adaptive, or hybrid interventions. Exploratory analyses classified interventions by delivery platform and mechanism-platform combination. Random-effects frequentist network meta-analyses were conducted for 7-day point-prevalence abstinence (PPA), continuous abstinence rate, and quit attempts. Relative effects were expressed as risk ratios (RRs) with 95% confidence intervals (CIs), and rankings were summarized using P scores. Certainty of evidence was assessed using the Confidence in Network Meta-Analysis framework (CINeMA).
Results:
A total of 118 randomized controlled trials from 23 countries were included to evaluate digital interventions for smoking abstinence and quit attempts. In the primary networks, AI-based adaptive interventions were associated with higher 7-day PPA (RR 1.73, 95% CI 1.20-2.51; P score=0.87) and continuous abstinence rate (RR 6.74, 95% CI 1.44-31.65; P score=0.96) than the blank control. Rule-based adaptive interventions were also associated with higher 7-day PPA (RR 1.45, 95% CI 1.21-1.74; P score=0.58) and continuous abstinence rate (RR 1.96, 95% CI 1.33-2.88; P score=0.71). Interventions in the other decision and adaptation mechanisms showed benefits for one or more abstinence outcomes, while no intervention significantly increased quit attempts. In the delivery platform networks, social media showed the largest effect estimates for both 7-day PPA (RR 1.68, 95% CI 1.33-2.12; P score=0.96) and continuous abstinence rate (RR 2.22, 95% CI 1.48-3.33; P score=0.93) compared with the blank control. Meta-regression suggested intervention effects attenuated with longer follow-up and were larger in populations with higher educational attainment and in low- and middle-income countries.
Conclusions:
Several digital interventions were associated with higher smoking abstinence than blank control, with differences observed across decision and adaptation mechanisms. These findings support the potential value of adaptive digital cessation strategies but should be interpreted alongside the certainty and amount of available evidence. Larger, rigorously designed long-term trials are needed to confirm the effectiveness of AI-enabled cessation strategies. Clinical Trial: PROSPERO registration number: 420251089935
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.