Feasibility and effectiveness of artificial intelligence-driven conversational agents in healthcare interventions: A systematic review of randomized controlled trials

Background: A virtual conversational agent is a program that typically utilizes artificial intelligence technology to mimic human interactions. Many robust and high-quality clinical trials have been conducted to test the effectiveness of conversational agent-based interventions. However, there is a lack of systematic reviews of randomized controlled trials that evaluate the effects of artificial intelligence-driven conversational agents in healthcare interventions. Objective: To examine the feasibility and effectiveness of conversational agent-based interventions evaluated by randomized controlled trials in the healthcare context, as well as to evaluate the information quality of artificial intelligence-driven conversational agents. Design: A systematic review. Data source: A systematic search of relevant literature published in English in Scopus, Pubmed, Embase, PsycINFO, Cochrane Library, Information Science & Technology, and Web of Science, was performed. Only randomized controlled trials from the inception of the databases until May 2022 were included. Review methods: Two reviewers independently selected the articles according to the inclusion and exclusion criteria. Study findings were narratively synthesized and summarized. The studies' risk of bias was evaluated using the Risk of Bias 2.0 tool. The Silberg Scale was used to evaluate the quality of the conversational agent system utilized in each reviewed study. Results: Twenty-one studies were included in the data synthesis. The recruitment rates ranged from 34% to 100% (mean = 84%), and completion rates ranged from 40% to 100% (mean = 83%). A moderate to high level of intervention acceptability was reported. The intervention approaches included health counseling and education (n = 8), cognitive-behavioral interventions (n = 7), storytelling (n = 1), acceptance and commitment therapy (n = 1), and coping skills training (n = 1). Findings indicated inconsistent effects on improving participants' physical activity and function, healthy lifestyle modifications, knowledge of the diseases, and mental health and psychosocial outcomes. The overall risk of bias varied from low risk (n = 6) to high risk (n = 7) across the studies. The mean Silberg score of included studies was 5.4/9, with a standard deviation of 1.6. Conclusion: Our review findings indicated that conversational agent-based interventions were feasible, acceptable, and had positive effects on physical functioning, healthy lifestyle, mental health and psychosocial outcomes. Conversational agents can provide low-threshold access to healthcare services. They can serve as remote medical assistants to support patients' recovery or health promotion needs before or after medical treatments. The conversational agent-based interventions can also play adjunctive roles and be integrated into current healthcare systems, which could improve the comprehensiveness of services and make more efficient use of physicians' and nurses' time.