Integrating principles of recommender systems into mHealth for smoking cessation in patients

Abstract

Introduction: Health recommender systems (HRS) are can predict on the basis of previously acquired knowledge which items (for instance, messages) each user prefers to receive in the future. Using this system will lead to a better matching of user needs and thus to more satisfaction and less drop-out. Methods:   Smoking patients are randomly assigned to the usual care or experimental condition. Smoking patients from the experimental condition received the HRS – using the trans-theoretical behavioural change model and previous studies of smoking cessation messages. Messages were sent via a mobile app, as part of a pharmacologic and behavioural change treatment in a hospital smoking cessation unit. We measure the quality of the recommender system, and the user engagement. Results: After an initial 3-month period, a total of 283 messages have been sent to the 30 recruited participants. They have rated 278 messages as positive (matching their interests), 4 as negative, and 1 as neutral. The results of an in-depth analysis comparing those who were satisfied versus those who were less satisfied will be discussed in detail as well as comparisons between the experimental and control condition. Discussion: The early results showed positive findings. Further work is needed to analyse the complete data set when the intervention finishes, to see how we can improve the algorithm Conclusion: HRS grounded in psychology theories to support behavioural change and increase engagement in smoking cessation is still in its infancy. Yet, early results from this study are promising for future applications.