Healthcare | Free Full-Text | Engagement with mHealth Alcohol Interventions: User Perspectives on an App or Chatbot-Delivered Program to Reduce Drinking
4.1. Lessons Learned
This study focused on determining barriers and facilitators of engagement with an app or chatbot-based intervention for alcohol use reduction. Results support the previous literature that providing feedback, notifications, and personalization features are motivators for engaging in behavior change apps, while technical difficulties create barriers to engagement [26]. Participants in both intervention groups reported that the features most important to their continued engagement were those that supported their ability to track their drinking and monitor their progress, prompts and reinforcing messages from the app or bot, and self-assessments and feedback that increased their awareness of their drinking patterns and helped them to be more mindful about their drinking. Users reported that the notifications to check in with the app or bot encouraged their engagement, and when there were technological errors with these notifications, they engaged less. Overall, participants reported appreciating the ability to moderate their drinking instead of abstaining and to work on reducing their alcohol consumption without going to in-person, traditional forms of treatment.
This study adds new information related to engagement with a chatbot version of the app. The Step Away chatbot was created to facilitate engagement with the intervention, and there were lessons learned from this pilot study on how the chatbot provided its own unique facilitators and barriers to utilization. A barrier that was unique to bot users was the lack of self-direction the bot script provided the opportunity for. Participants indicated a desire to utilize the parts of the bot they wanted to use, such as tracking their drinks, without responding to the full script first. This supports previous research suggesting that continued guidance after setup with the app may be a nuisance to users [32] and that reducing the actions users need to take to implement the target behavior is a facilitator of engagement [27]. Additionally, the repeated bot conversation during the daily interview feature may have hindered utilization: many bot participants indicated that the conversation script felt rote or repetitive. Bot participants were divided on whether they appreciated the affirmations given by the bot, with some stating these affirmations were motivating and others reporting they felt the conversation was too affirmative. As tracking drinks and receiving visual feedback emerged as such a strong facilitator, it is recommended to other developers that, if utilizing a chatbot script, allow users to access these primary features without having to engage with the bot.
Consistency in providing outreach and prompting users, a hypothesized advantage of the chatbot, fell short due to platform constraints. Numerous participants mentioned using the bot less because they were no longer being prompted daily, while the app users were prompted regularly regardless of their response to the notifications.
One user suggestion was an intervention that combines the features of the app and chatbot into one. Combining preferred features of the two interventions would allow for more self-direction from the user to engage in tracking, monitor their own progress, receive feedback, and engage in positive dialogue. Combining the app and bot would also remove some barriers, such as the Facebook platform, repetitive conversation, and ill-timed prompts. Participants could have the option of speaking to the bot but would not be required to converse with it to access various features of the intervention. Having the chatbot included could also provide some conversational assistance with setup and introduction to the key features, as well as built-in opportunities for positive conversation, reinforcement of progress, and connection. It is possible that a combined app and bot intervention would lead to fewer technological barriers and more engagement with preferred features that are sustained over time.
Finally, the COVID-19 pandemic had varying effects on participant utilization of either intervention and varying effects on participant alcohol consumption. Some reported they used the app or bot less because they were drinking less: these participants referenced social drinking as why they were drinking less, stating they had fewer opportunities for social drinking during stay-at-home orders or establishments closing. Another reason participants stated for using the intervention less during the pandemic was the stress the pandemic caused, making reducing their drinking less of a priority. Others reported using the intervention more, referencing increased drinking or increased free time as their reasons.
4.2. Recommendations for mHealth Apps
The feedback from users of the bot and app highlights essential recommendations for designers and researchers in the field of mHealth behavior change interventions. Users consistently emphasized the importance of three key aspects: (1) notifications to encourage continued use; (2) ability to easily track the target behavior; and (3) feedback with progress reporting on the tracked behaviors. Incorporating visual aids such as graphs that display progress toward goals over time is strongly recommended. Additional feedback mechanisms that enhance user insights may include weekly reports on the target behavior and feedback on behavior patterns that illustrate progress over several months.
Notifications also emerged as a critical factor driving user engagement. It is advisable to develop personalized notifications that align with the user’s preferred notification schedule. As users in this study appreciated the ability to self-direct their usage of the app or bot, an easily accessible menu and settings feature is encouraged.
Another noteworthy recommendation is the regular rotation and updating of content to prevent it from becoming repetitive. Incorporating large language models (LLMs) like GPT-4 for more sophisticated and context-aware interactions could be considered, providing a more dynamic and responsive experience to users. Additionally, allowing the intervention to connect users with online educational resources or local support networks, such as group meetings or counselors, can enhance the overall user experience and effectiveness of the intervention.
While this study focused on the development of an alcohol intervention app, the insights into facilitators and barriers to use are likely applicable to other mHealth apps aiming to foster behavior change. User engagement remains a critical factor for the efficacy of mHealth apps, and these findings can inform the design and promotion of engagement in various mHealth interventions.
