Predicting Dropouts in Digital Weight Loss Programs with AI Algorithm
Obesity and lack of exercise are among the leading causes of preventable health issues plaguing our society. As the world progresses into a digital age, digital behavioral intervention programs provide a convenient and effective way to support health intervention and lifestyle changes of individuals. However, engagement in these digital programs is crucial to achieving successful behavior change and improved health outcomes. In this regard, a new study from CSIRO has shown the promise of AI in predicting when users are likely to drop out of an online weight loss program, thereby opening up opportunities for timely intervention. In this article, we will delve into the details of the study and discuss how AI algorithms could revolutionize digital health programs.
The study, published in The Journal of Medical Internet Research, analyzed over 59,000 participants from CSIRO’s Total Wellbeing Diet program and made use of a CSIRO-developed AI algorithm. The algorithm analyzed user data such as BMI, frequency of website logins, and food diary entries, among others, to predict the likelihood of dropouts. The algorithm identified eight behavioral patterns that indicated users may quit the program, such as skipping meals, not logging food regularly, or failing to reach certain weight loss milestones.
The AI algorithm was successful in predicting participant attrition with an accuracy rate of 64 percent, with predictions becoming even more accurate as the user spent more time on the program. The study concluded that identifying participants who are likely to drop out could help health professionals intervene and provide targeted support to such users.
The potential benefits of the AI algorithm don’t stop at predicting dropouts. It could also help personalize interventions and support for individual users. The study showed that two people with the same BMI, age, gender, and starting weight may have different behavioral patterns and, therefore, require different levels of support and intervention. With AI, health professionals can identify these differences and tailor programs and interventions accordingly.
Digital health programs have traditionally struggled with high dropout rates, with some reports suggesting that up to 90 percent of users drop out within a year of starting. Predicting such occurrences can serve as a valuable tool for program developers by enabling them to modify programs to improve engagement. Predictive algorithms such as this will also encourage design efforts for new digital health programs so that the user journey can be carefully considered to maximize persistence, engagement and better outcomes.
The potential of AI in improving digital health programs cannot be overstated. As the world becomes increasingly digital, personalized interventions have become even more important to improve the health outcomes of individuals. With this study, we have seen that AI algorithms could help predict user attrition, enabling health professionals to intervene and provide targeted support at critical points in the user’s journey. Additionally, intervention strategies can be tailored better according to behavioral data, and digital health programs can be modified to improve engagement. The future looks promising for AI in healthcare, and we anticipate more groundbreaking studies in this field in the years to come.