A conversational-based dietary assessment tool for cardiovascular patients

Abstract

Abstract Funding Acknowledgements Type of funding sources: None. Background Cardiovascular diseases (CVDs) are currently the leading cause of death worldwide. Accumulated evidence indicates that a healthy diet contributes significantly to health promotion and an increase of quality of life for those living with CVDs. While dietary intake management is an integral part of cardiac rehabilitation (CR), tracking dietary intake is burdensome, resulting in measurement errors caused by low adherence, incomplete data, or erroneous information. Objective Dietary assessment tools, such as 24-hour Dietary Recall (24HR) or Food Frequency Questionnaire (FFQ), are developed to help patients to reflect on their dietary intake but are often burdensome. To lower tracking burden and reduce measurement errors, conversational agents, i.e., chatbots, promise to improve self-reported dietary assessment by increasing flexibility and reliability. Therefore, this qualitative study aimed to evaluate the utility of a conversational-based dietary assessment tool among experts. Methods We developed a nutrition chatbot to track dietary intake and provide feedback to support people in understanding their dietary choices. The chatbot provides web-based dietary assessment either in 3h recalls, following standard procedure, or tailored recall periods. Participants are reminded to register their food intake through push notifications. Users can start reporting by choosing food categories as derived from the Dutch Food Composition Database (FCDB). When selecting a food category, follow-up questions are asked regarding the food items and the corresponding portion size in household measures, i.e., spoons, cups, glasses, handfuls, or weight in grams. We applied Fuzzy Terminology Recognition, which enabled users to enter the food item, reducing overall selection time. Meanwhile, the traffic light system, based on the Dutch Nutrition Guidelines, was implemented to support self-reflection. We conducted a study with 9 experts specialized in the field of CR or dietary behavior management, who were asked to use the chatbot to self-report their dietary intake for one week. Results The results show that the chatbot is easy to interact with and experts praised the simplicity and flexibility of using the chatbot. According to the experts, the chatbot could bridge the cognitive gap between how patients perceive their dietary habits and their actual habitual intake. They pointed out that self-reporting and feedback of the dietary choices could potentially raise awareness for dietary intake. Further, the experts mentioned that to maintain simplicity, the kind of information presented should be carefully considered. Conclusion The results from the expert reviews are promising and suggest the benefits of applying the chatbot for self-reporting dietary intake. Next, the chatbot will be included in a clinical trial with cardiovascular patients to further evaluate adherence, user experience, and the effects on measurement errors in dietary intake assessment.