ÆßÃõ¼º½

Chatbots may help improve physical activity and diet across various populations

Singh B, Olds T, Brinsley, J, et al. Systematic review and meta-analysis of the effectiveness of chatbots on lifestyle behaviours Digital Medicine. 2023;6:118.

Review question

Do chatbots improve physical activity, diet, and sleep across various populations?

Background

Adopting unhealthy lifestyle behaviours, such as consuming a poor diet, not getting enough exercise or sleep, and sitting too much, can increase the risk of developing chronic diseases. Tailored support from health professionals is effective for improving lifestyle behaviours, but there are many barriers to accessing this type of care. Technology, such as chatbots, can help increase access to personalized and less costly care. Chatbots (aka conversational agents or virtual assistants) are tools used to reproduce or imitate human communication via text, voice/speech, and visuals. Early research shows they have the potential to positively impact certain health behaviours, but meta-analyses are lacking and needed.      

How the review was done

This is a systematic review and meta-analysis of 19 studies, the majority of which were randomized controlled trials. The studies were published between 2007 and 2022 and included 3567 participants.

Key features of the studies were:

• Participants, on average, ranged in age from 9 - 71 years old and included healthy adults, older adults, adult cancer survivors, overweight or obese adults, adults with insomnia, physically inactive adults, adolescents, children with sleep problems, and families or parent-child pairs. 
• Participants engaged in chatbot only strategies or multicomponent strategies that involved chatbots and another component (e.g., wearable trackers, diaries/logs, mailouts, etc.) for 6 weeks or less (shorter-term) or more than 6 weeks (longer-term).   
• Chatbots could either be text-based, voiced-based, or a combination of text, voice/speech, and visuals (e.g., avatars, images, graphs), and some used artificial intelligence (e.g., technology with the ability to learn and make predictions and identify and evaluate language). 
• Chatbot strategies were delivered in various ways, such as through smartphones, apps, telephone-linked computer systems, text-messages, smart speakers, and websites.
• Researchers measured changes in total physical activity, moderate-to-vigorous physical activity, daily steps, fruit and vegetable consumption, sleep quality, and sleep duration.
• Results were compared with people in diverse control groups (e.g., receiving usual care, no treatment, waiting to engage in a strategy, etc.).

What the researchers found

Overall, chatbot-based strategies may have the potential to increase total physical activity, moderate-to-vigorous physical activity, daily step counts, fruit and vegetable consumption, sleep duration, and sleep quality by small to moderate amounts. However, when we take a closer look at the outcomes and the factors that may affect them, we see some interesting results. First, some of the studies combined the chatbot with other strategies, and if we separate these from the studies of chatbots on their own, the benefits seen for sleep duration and quality don’t hold up.  However, chatbots combined with other strategies may still be the better of the two for sleep duration and quality. 

Second, total physical activity, daily steps, and fruit and vegetable consumption may benefit from both chatbot-only strategies and multi-component chatbot strategies, with no significant difference between the two. Additionally, there doesn’t seem to be much of a difference in benefits for outcomes when comparing studies where chatbots were used for a shorter or longer duration. However, it should be noted that text-based chatbots and chatbots using artificial intelligence tend to be better at improving fruit and vegetable consumption than speech/voice chatbots and chatbots that don’t use artificial intelligence. Given the evidence base is still new and growing, future high-quality research is needed to support these results.   

Conclusion

Preliminary research shows that chatbot only and multicomponent chatbot strategies may be effective for improving lifestyle behaviours such as physical activity (e.g., total activity, moderate-to-vigorous activity, and daily steps) and diet (e.g., fruit and vegetable consumption) across various populations.