Abstract
BACKGROUND: This systematic review aims to summarize the effectiveness, acceptability, and potential mechanisms of chatbot-based interventions in improving diet, physical activity, and tobacco use behaviors. METHODS: This protocol follows the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P). We will include individual- or cluster-randomized, parallel-group controlled trials that compare chatbot-based interventions with active-control, waiting-list, or usual-control comparators among children, adults, and the elderly irrespective of their behavioral patterns at baseline. We will also include the non-randomized or single-group trials to expand the evidence base. The primary outcomes will be the change in diet, physical activity, and tobacco use behaviors assessed by validated questionnaires or objective measures. The secondary outcomes will include the change in obesity-related outcomes, stage of behavioral change, change of motivation, emotion, knowledge, or other constructs that might mediate the intervention effect, chatbot use behaviors during the process of intervention implementation, the facilitators and barriers to chatbot use, and safety issues. We will search both the studies published in PubMed, EMBASE, ACM Digital Library, Web of Science, PsycINFO, CINAHL, and IEEE and the unpublished studies in the WHO's International Clinical Trials Registry Platform, ClinicalTrials.gov, conference proceedings, GitHub, and arXiv. We will group the included studies based on their consistency in the Population, Intervention, Comparator, Outcome and Study design (PICOS) elements for data synthesis. The random-effects meta-analysis will be used to quantitatively synthesize the results across studies if data permits; otherwise, we will synthesize the study results based on the guideline of Synthesis Without Meta-analysis (SWiM). We will use the correlation-based meta-analytical structural equation modeling approach to examine the presence of mediators of chatbot-based interventions. We will assess the risk of bias for each included study using the revised version of the Cochrane risk-of-bias tool for randomized trials (RoB 2) or the Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I), and appraise the certainty of the evidence for each synthesized result using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach. DISCUSSION: This systematic review will not only answer whether the state-of-the-art chatbot-based interventions are acceptable and effective in changing a person's diet, physical activity, and tobacco use behaviors but also explore the potential mechanisms underlying the effects of the chatbot-based interventions. The findings of this study will pave the way for optimizing future chatbot-based interventions in the field of health-related behaviors. SYSTEMATIC REVIEW REGISTRATION: PROSPERO CRD42023492013.