The Use of Conversational Agents in Self-Management: A Retrospective Analysis

Background Understanding user engagement with conversational agents (CAs) in mobile health apps is crucial for improving sustained usage. We analyzed CA interactions in a mobile health app to identify usage patterns and potential barriers. Materials and Methods Retrospective data from 100,571 active users of the Albert Health app in 2022 were analyzed. Interactions with CA were categorized by demographics (gender and age), interaction type (health information, medication-related, clinical parameters, and non-clinical), and engagement method (text, voice). Descriptive statistics were used to identify trends and patterns in app usage. Results Out of the active users, 19,051 (18.9%) engaged with the CA. The majority were female (61%), with 43% aged 30-45 years and 23% older than 45 years. The analysis showed that 94.5% engaged in general health management, while 5.3% used disease-specific programs. Average usage per user was highest in cardiovascular and respiratory diseases. Interaction types varied, with health information and medication-related interactions. The varied messaging behavior suggests different user engagement levels, with some users seeking quick information and others engaging more deeply for health management. Engagement was high initially but decreased over time. Discussion This study provides insights into user interactions with a healthcare CA, highlighting a preference for general health management and diverse usage patterns. The significant number of single-session users indicates potential barriers to sustained engagement, highlighting the need for strategies to enhance user experience and retention. Future research should investigate the CA’s performance, effectiveness and extend observations to broader healthcare contexts by using large language models.