Artificial intelligence (AI) chatbots are able to mimic human interactions with the help of oral, written, or verbal communication with the user. AI chatbots can provide important health-related information and services, which ultimately lead to promising interventions facilitated by technology.
Study: Chatbots based on artificial intelligence (AI) in promoting behavioral changes in health: a systematic review. Image credit: TippaPatt / Shutterstock.com
AI chatbots in healthcare
Current therapeutic and digital telehealth interventions are associated with several challenges, such as unsustainability, low adherence, and inflexibility. AI chatbots are able to overcome these challenges and offer personalized on-demand support, greater interactivity, and greater sustainability.
AI chatbots use data input from multiple sources, followed by data analysis that is completed using natural language processing (NLP) and machine learning (ML). Data output helps users achieve their health behavior goals.
Thus, AI chatbots are able to promote various health behaviors by performing interventions effectively. In addition, this technology can provide additional benefits to changes in health behavior through integration into built-in features.
Most previous studies conducted on AI chatbots aimed to improve mental health outcomes. In comparison, recent studies are increasingly focusing on the use of AI chatbots to incite changes in health behavior.
However, a systematic review of the impact of AI chatbots on lifestyle modification was associated with several limitations. These include the inability of authors to differentiate AI chatbots from other chatbots. In addition, this study addressed only a limited set of behaviors and did not discuss all potential platforms that AI chatbots could use.
A new systematic review published on the medRxiv * prepress server analyzes the results of previous studies on the features, functionality, and intervention components of the AI chatbot, as well as its impact on a wide range of health behaviors.
About the study
The current study was conducted in June 2022 and followed PRISMA guidelines. Here, three authors searched seven bibliographic databases, including IEEE Xplore, PubMed, JMIR Publications, EMBASE, ACM Digital Library, Web of Science, and PsychINFO.
The search involved a combination of keywords that belonged to three categories. The first category included keywords related to AI-based chatbots, the second included keywords related to health behaviors, and the third focused on interventions.
The inclusion criteria for the search were studies involving intervention research focused on health behaviors, those that were developed on existing AI platforms or AI algorithms, empirical studies using chatbots, articles in English which were published between 1980 and 2022, as well as studies that reported quantitative or qualitative intervention results. All data were extracted from these studies and subjected to a quality assessment according to the National Institutes of Health (NIH) quality assessment tool.
Study results
A total of 15 studies matched the inclusion criteria, most of which were distributed among developed countries. The average sample size was 116 participants, while the average was 7,200 participants.
Most studies included adult participants, while only two included participants under 18 years of age. All study participants had pre-existing conditions and included individuals with less exercise, obese, smokers, substance abusers, breast cancer patients, and Medicare recipients.
Targeted health behaviors included quitting smoking, promoting a healthy lifestyle, reducing substance abuse, and adhering to medication or treatment. In addition, only four studies reported using randomized controlled trials (RCTs), while others used a quasi-experimental design.
The risk of reporting the results and the bias in the randomization process was low, the risk of bias of the planned interventions was low to moderate, the risk of bias in the measurement of the results was moderate and the risk of results of unwanted sources was high. All factors for the description of the AI components were sufficient, except the handling of the unavailable input data and the characteristics of the input data.
Of 15 studies, six reported feasibility in terms of the average number of messages exchanged with the chatbot per month and security. In addition, 11 studies reported usability in terms of content usability, ease of use of the chatbot, user-initiated conversation, unjudged safe space, and out-of-office support. Acceptability and commitment were reported in 12 studies in terms of satisfaction, retention rate, technical issues, and duration of commitment.
An increase in physical activity was reported in six studies, along with an improvement in diet in three studies using chatbots-based interventions. In all four studies evaluated, smoking cessation was reported, while one study reported a reduction in substance use and two studies reported an increase in adherence to treatment or medication through use of chatbots.
Various behavior change theories were integrated into chatbots, including the transtheoretical model (TTM), cognitive-behavioral therapy (CBT), social cognitive theory (SCT), the habit-forming model, the motivational interview, Mohr’s supportive responsibility model and emotionally focused therapy to provide motivational support and track participants ’behavior. Most studies focused on setting behavioral goals, used behavioral tracking, and provided behavior-related information, while four studies also provided emotional support.
Most studies used different AI techniques such as ML, NLP, hybrid health referral systems (HHRS), hybrid techniques (ML and NLP), and facial tracking technology to offer personalized interventions. Chatbots primarily used text-based communication and were integrated into pre-existing platforms or delivered as stand-alone platforms. In addition, most chatbots required data on users ’background information, their goals, and behavioral performance feedback to ensure the provision of personalized services.
Conclusions
Taken together, AI chatbots can efficiently promote a healthy lifestyle, quit smoking, and adherence to treatment or medications. In addition, the current study found that AI chatbots demonstrated significant usability, viability, and acceptability.
Taken together, AI chatbots are able to provide personalized interventions and can be scalable to diverse and large populations. However, additional studies are needed to obtain an accurate description of AI-related processes, as AI chatbots interventions are still in an incipient stage.
Limitations
The current study did not include a meta-analysis and focused on only three behavioral outcomes. In addition, articles from unselected databases, articles in other languages, gray literature, and unpublished articles were not included in the study.
An additional limitation was that the interventions could not provide a clear description of the excluded AI chatbots. The study also lacked generalization and information on patient safety was limited.
* Important news
medRxiv publishes preliminary scientific reports that are not peer-reviewed and therefore should not be considered conclusive, guided by clinical practice or health-related behavior, or treated as established information.