JMIR Publications

ABSTRACT

Background:

Empowered by the rapid development of artificial intelligence technology, chatbots have been increasingly used for the promotion of health-related behaviors to address the high demand for human resources in conventional behavioral interventions. However, the existing reviews have not focused on the rigorously designed randomized trials of the state-of-the-art chatbots (interacting with users through unconstrained natural language), thus calling for an updated review.

Objective:

We aimed to explore the effects of chatbot-based interventions on improving diet, physical activity, and tobacco use behaviors in the general population and to evaluate the chatbot use behaviors during the implementation process.

Methods:

We comprehensively searched twelve databases/registers for eligible studies published from January 1, 2010 until July 16, 2024. We included randomized controlled trial (RCT) studies that used unconstrained chatbots to promote diet, physical activity, or tobacco use behaviors among adults or children. Due to considerable heterogeneity across the included studies, we adopted the Synthesis Without Meta-analysis guidelines and used the new evidence-mapping method (bubble plot) to summarize the effectiveness of chatbot-based interventions. To evaluate the implementation process of intervention, we summarized users’ interaction (percentage, relevant rate, etc) with chatbots as well as their feelings (satisfaction, appreciation, etc) about chatbot use. In addition, we assessed the risk of bias of studies using the Risk of Bias 2.0 tool.

Results:

We finally included 7 RCT studies. For dietary and physical activity behaviors, the effectiveness of chatbot-based interventions was inconsistent among adults while no evidence of effect was observed among children. Concerning tobacco use behaviors, the included studies showed consistent evidence of improving this behavior among adults. A total of 4 studies evaluated the implementation process of chatbot intervention. For interaction with chatbots, the included studies showed an overall high percentage of general interaction between users and chatbots, but interactions specific to health behaviors were not highly satisfactory. Concerning feelings about chatbot use, users showed a positive impression of chatbot use, feeling it useful, credible, and financially feasible. Regarding the risk of bias of included studies, three were classified as overall low-risk and the remaining four were classified as overall high-risk.

Conclusions:

Chatbot-based interventions were beneficial for adults’ tobacco use behaviors, but no such evidence was found on diet or physical activity behaviors among adults and children. More RCTs with larger samples and lower risk of bias were urgently need to enhance our findings in the future. Clinical Trial: The protocol was registered in the International prospective register of systematic reviews (PROSPERO; https://www.crd.york.ac.uk/PROSPERO/) on December 21, 2023 (CRD42023492013)