Chatbot interventions effective in improving physical activity, diet, and sleep

In a recent study published in npj Digital Medicine, researchers evaluate the efficacy of chatbot interventions in improving individuals’ activity, diet, and sleep patterns across various age- and ethnic groups. These virtual assistants were found to improve a person’s food choices and lifestyle, with text-based chatbots found to be more effective than their more vocal counterparts.

Study: Systematic review and meta-analysis of the effectiveness of chatbots on lifestyle behaviors. Image Credit: TZIDO SUN / Shutterstock.com

The benefits of healthcare chatbots

Poor lifestyle choices, specifically those pertaining to physical activity, diet, and quality and quantity of sleep, are significant contributors to mental health issues and chronic diseases. Since 2010, the economic costs of chronic diseases alone will likely reach $45 trillion USD globally by 2025. Fortunately, these choices can be significantly improved with interventions from trained healthcare practitioners and, more recently, virtual assistants.

Chatbots, a type of software that utilizes various forms of communication to emulate human interaction, can improve physical and mental health through positive prompts, timely reminders, and information delivery aimed at lifestyle, diet, and sleep interventions. Unlike their living healthcare counterparts, chatbots present numerous advantages, including significant time-, travel-, and money savings. These benefits, combined with the ubiquity of the internet and smartphones, make chatbots ideal for tackling the global pandemics of poor diet, sleep, and fitness.

Previous research has analyzed the efficacy of chatbots in improving lifestyle choices. However, these studies have focused on specific improvements in diet or physical activity in isolation or have been population or region specific, thereby inherently lacking a holistic, global outlook.

About the study

In the present study, researchers conducted a review of scientific studies to elucidate the effectiveness of various chatbots aimed at improving sleep, diet, and physical activity habits. Two thousand five hundred fourteen papers were reviewed through four rounds of title and abstract screened, with only 19 studies included in the final review. The final dataset comprised 3,567 participants between nine and 71 years of age, with sample sizes in individual studies between 25 and 958.

Research target groups included spanned the health spectrum, including those with insomnia to sufficient sleep and cancer survivors. Moreover, fifteen of the 19 papers analyzed physical activity, seven focused on diet patterns, and five involved sleep.

Chatbot-only interventions comprised 58% of the raw data. The remaining studies involved multi-component interventions that utilized both virtual assistants and additional wearables or trackers that helped provide the chatbot with additional individualized information.

The aim of this review was to elucidate changes in total physical activity, which was evaluated by the number of steps and activity minutes per week, the number of fruits and vegetables consumed as a proxy for diet, and the duration and quality of sleep of the study participants.

Different types of chatbots, including those utilizing artificial intelligence [AI], text- and voice-based, as well as non-AI chatbots, were also compared. This information allowed the researchers to determine which chatbots performed best at improving overall participant lifestyle.

Study findings

Every study reported a positive effect of chatbot intervention on total physical activity and the number of weekly steps. On average, chatbot interventions contributed to the increase of 735 steps each day across the studies, irrespective of age, population, or region.

Similar results were noted for sleep, with an average increase of 45 minutes per night, and diet, as demonstrated by at least one additional fruit or vegetable serving each day, in chatbot-intervention groups. These results remained positive independent of the duration or intervention type, thereby establishing the benefits of employing chatbot interventions in healthcare.

Multi-component interventions may be more beneficial than chatbot-only interventions in regard to sleep. To this end, wearable trackers may provide chatbots with information on individual sleep quality that the participants themselves could not provide to improve the accuracy of chatbot recommendations ultimately.

Although no difference was noted between text- and voice-based chatbots in modifying sleep or physical activity, text-based chatbots were superior to voice-based chatbots in improving diet. This may be due to the implementation of text-based chatbots in smartphone apps that are always readily available, while voice-based chatbots are restricted to home use, such as smart speakers.

This observation contradicts previous research highlighting how the ease of use of voice-based chatbots over text-based chatbots improves engagement and retention.

Conclusions

The study findings highlight how chatbots can provide a cost-effective, as well as a time- and travel-efficient alternative to traditional ‘in-person’ visits to healthcare professionals to combat the growing pandemic of lifestyle-related chronic health issues. Both AI and non-AI-based chatbots were found to be effective at improving individuals’ physical activity, diet, and sleep patterns. Thus, chatbots are promising well-being management tools, irrespective of age, population, or geographical region of deployment.

Users may not fully understand the implications of sharing personal information with chatbots, which may collect data beyond their expectations and control.”

Importantly, data breaches are becoming increasingly common. Thus, the sensitivity of data collected by chatbots, including age, sex, mental and physical state, and even personal preferences, can no longer be circumvented by individually masking data points through grouping/categorizing.

This field of research is still in its infancy and listing out recommendations for future work, including larger sample sizes, improved a priori study design, and long-term follow-up, which, if done correctly, might see us in a world where our smart devices ubiquitously aid healthcare practitioners (and apples) in keeping us fit and healthy.