Chatbots: Transforming Health Behaviors for a Fitter Life

By Dr. Sampath LonkaReviewed by Susha Cheriyedath, M.Sc.Jul 5 2023

In a paper published in the journal npj digital medicine, researchers presented a systematic review and meta-analysis of chatbot interventions for physical activity, diet, and sleep. Nineteen trials involving participants of various ages were included. Most interventions focused on physical activity, with low-quality ratings.

The analysis showed positive effects on physical activity, steps, fruit and vegetable consumption, sleep duration, and quality. Text-based and artificial intelligent (AI) chatbots were effective for diet, while multicomponent interventions improved sleep. Overall, chatbot interventions were effective across populations, age groups, and intervention types.

Background

Insufficient physical activity, sedentary behavior, poor diet, and inadequate sleep contribute to global health issues and chronic diseases. Previous reviews support interventions for improving these behaviors, but traditional healthcare services are costly and limited to specific populations. Technology, including chatbots utilizing AI and natural language processing (NLP), offers accessible and personalized behavior change interventions. Chatbots can provide health-related information and services, enhancing accessibility and efficacy. They have effectively addressed depression, anxiety, medication adherence, smoking cessation, and substance abuse.

The researchers conducted a systematic review and meta-analysis to assess the effectiveness of chatbot interventions in enhancing physical activity, diet, and sleep. Findings indicate significant effects on these outcomes, with text-based and AI chatbots effectively promoting fruit and vegetable consumption. Sleep improvement is more effectively achieved through the implementation of multicomponent interventions. Chatbot interventions demonstrate effectiveness across diverse populations and age groups, exhibiting intervention durations and components variability.

Analysis and results

The literature search yielded 2514 records, leading to the inclusion of 19 studies in this work. The studies consisted of 11 randomized controlled trials (RCTs), Two non-RCTs, Five single-group, pre-post studies, and one-panel design trial. The majority of studies received a rating indicating a low risk of bias. Meta-analyses were performed to analyze diverse outcomes such as physical activity, daily steps, fruit and vegetable consumption, moderate-to-vigorous physical activity (MVPA), sleep quality, and sleep duration. The analysis showed that chatbot interventions significantly increased total physical activity, MVPA, daily steps, and fruit and vegetable consumption. The interventions varied in duration, participant characteristics, and components, with some involving AI or NLP. Multicomponent interventions were more effective in improving sleep duration and quality.

Overall, chatbot interventions received a Grade A recommendation for increasing physical activity and fruit and vegetable consumption and a Grade B recommendation for improving sleep duration and quality. Subgroup analyses showed that multi-component interventions and text-based chatbots were more effective in some instances.

Chatbot interventions for health behavior

The researchers reviewed 19 trials with 3,567 participants to assess the effectiveness of chatbot interventions in promoting physical activity, diet, and sleep. The findings indicate that chatbot interventions effectively improve these health behaviors, significantly increasing physical activity, fruit and vegetable consumption, sleep duration, and sleep quality. Short-term and long-term interventions, as well as chatbot-only and multicomponent interventions, demonstrated effectiveness.

Text-based chatbots were found to be more effective in promoting fruit and vegetable consumption. At the same time, multicomponent interventions were more beneficial for enhancing sleep duration and quality compared to chatbot-only interventions. The observed effect sizes for each health behavior were small to moderate and aligned with previous research.

Chatbot interventions offer advantages such as delivering behavior change interventions in existing clinical settings without burdening healthcare providers. They can be implemented through various platforms, making them highly accessible. However, limitations include functional constraints and the potential exacerbation of health disparities due to unequal access to digital health tools.

Future research should explore different aspects of the chatbot user experience, including text-based and voice-based interactions. Longer-term follow-up is needed to assess sustained effects. Chatbots' ability to replicate human interaction and provide a safe environment for discussing sensitive issues is a strength. However, privacy and data security must be carefully addressed to prevent data breaches and protect user privacy.

Methods

The protocol for this systematic review was registered on PROSPERO (ID: CRD42022353633) and followed PRISMA guidelines. The study searched 14 electronic databases for peer-reviewed journal articles published in English until September 1st, 2022. The eligibility criteria were based on the PICOS framework, including any population, interventions targeting physical activity, sedentary behavior, diet, or sleep using chatbots. Chatbots were defined as conversational agents enabling natural language conversations. Various platforms and modalities were included.

Data extraction and screening were performed independently by 10 authors, with disagreements resolved through discussion. Study quality was assessed using the EPHPP tool, and outcomes of interest included physical activity, fruit and vegetable consumption, and sleep quality and duration.

Meta-analyses were conducted using RevMan software, and publication bias was evaluated using funnel plots. Subgroup analyses were performed based on intervention components, output type, AI or NLP use, and duration. Researchers used standardized classifications for effect size and Oxford Centre for Evidence-Based Medicine levels of evidence. Sedentary behavior was excluded due to insufficient data, and additional subgroup analyses were added for output type and AI or NLP use.

Conclusion

In conclusion, chatbot interventions effectively improve physical activity, diet, and sleep across populations, offering personalized behavior change interventions. Addressing privacy and access challenges is crucial for maximizing their potential in addressing global health issues. Further research is needed to explore user experience and long-term effects.