A wide variety of algorithms has been developed to provide efective support in eHealth applications for behavior change. However, an overview of the types of algorithms is missing. We aim to provide such an overview by conducting a scoping review of papers published in the Scopus database. We are currently screening the 44 remaining papers based on their full texts and collecting information on the characteristics of the algorithms themselves, what the algorithms optimize in an intervention, and the domain in which the algorithms are employed. We also keep track of how the algorithms have been evaluated. Our review will provide insights into what types of algorithms are currently used and how they can be improved in the future.
CEUR ceur-ws.org
Given that 18.5% of the disease burden in the Netherlands is caused by unhealthy behavior [
Various algorithms have been designed to address this discrepancy by adapting what these applications ofer (e.g., diferent physical activity suggestions [ 9]), how (e.g., using diferent persuasive strategies such as commitment and authority [10]), when (e.g., optimizing the timing of physical activity notifications [ 11]), and with whom (e.g., deciding when to add human support [12]). The decisions these algorithms make can be based on theories such as the Transtheoretical Model (e.g., [13]), expert knowledge (e.g., [14]), as well as ofline and online data (e.g., [10, 14]). Moreover, the algorithms can be forward- (e.g., [10]) or backward-directed (e.g., [15]), include a positive feedback loop (e.g., [14]) or a negative one (e.g., [16]), consider (W. Brinkman) users’ future states (e.g., [10]) or the efects of repetitions (e.g., [ 17]), and balance exploration and exploitation (e.g., [18]).
In light of this variety of algorithms, we seek to provide a review of algorithms for adaptive health behavior change support. The focus thereby lies on the characteristics of these algorithms as well as how their efectiveness has been evaluated (e.g., controlled experiments [ 10], simulations [11]). To this end, we are conducting a scoping review using journal and conference articles published in the Scopus database. The general goal of a scoping review is to ”identify and map the available evidence” [19]. For example, scoping reviews can be used to examine how research is conducted in a field or to identify important characteristics related to a concept [19]. We expect that our scoping review will give us insights into the types of algorithms that are currently developed to support health behavior change and how they can be improved.
We formulated a search query consisting of four components. Specifically, we wanted to obtain papers about 1) digital interventions, 2) algorithms, 3) behavior change, and 4) health. The resulting query (Table 1) led to 993 results in Scopus in March 2024.
digital health intervention recommender system* mHealth algorithm eHealth machine learning digital intervention deep learning mobile health reinfocement learning
artificial intelligence
Subsequently, we removed papers using the first three exclusion criteria presented in Table 2, leading to 678 remaining papers. Next, papers were excluded based on their titles and abstracts if they were review papers or did not mention a behavior change algorithm. The remaining 235 papers were screened based on their full texts. 29 of these papers were beavio* change physical activity intervention obesity health self management smoking health promotion sleep non-communicable disease mental health cessation health excluded because we did not have access to the full texts, 76 because they did not describe a behavior change algorithm, and 86 because they did not provide enough information about a behavior change algorithm. Currently, we are examining the 44 remaining papers in more detail. The primary goal is to characterize the algorithms based on their characteristics (e.g., based on online data, expert-devised rules). Moreover, we will investigate what the algorithms are used for (e.g., reminder timing, intervention selection), the domain they are employed in (e.g., mental health, smoking cessation), and how they have been evaluated.
This work is part of the multidisciplinary research project Perfect Fit, which is supported by several funders organized by the Netherlands Organization for Scientific Research (NWO), program Commit2Data - Big Data & Health (project number 628.011.211). Besides NWO, the funders include the Netherlands Organisation for Health Research and Development (ZonMw), Hartstichting, the Ministry of Health, Welfare and Sport (VWS), Health Holland, and the Netherlands eScience Center. Austria, April 5-7, 2016, Proceedings, volume 9638 of Lecture Notes in Computer Science, Springer, 2016, pp. 276–287. doi:10.1007/978-3-319-31510-2\_24. [8] T. Greenhalgh, J. Wherton, C. Papoutsi, J. Lynch, G. Hughes, S. Hinder, N. Fahy, R. Procter, S. Shaw, Beyond adoption: a new framework for theorizing and evaluating nonadoption, abandonment, and challenges to the scale-up, spread, and sustainability of health and care technologies, Journal of Medical Internet Research 19 (2017) e367. doi:10.2196/jmir.8775. [9] M. Rabbi, A. Pfammatter, M. Zhang, B. Spring, T. Choudhury, et al., Automated personalized feedback for physical activity and dietary behavior change with mobile phones: a randomized controlled trial on adults, JMIR mHealth and uHealth 3 (2015) e4160. doi:10.2196/mhealth.4160. [10] N. Albers, M. A. Neerincx, W.-P. Brinkman, Addressing people’s current and future states in a reinforcement learning algorithm for persuading to quit smoking and to be physically active, PloS One 17 (2022) e0277295. doi:10.1371/journal.pone.0277295. [11] S. Wang, C. Zhang, B. Kröse, H. van Hoof, Optimizing adaptive notifications in mobile health interventions systems: reinforcement learning from a data-driven behavioral simulator, Journal of Medical Systems 45 (2021) 102. doi:10.1007/s10916-021-01773-0. [12] J. D. Piette, S. Newman, S. L. Krein, N. Marinec, J. Chen, D. A. Williams, S. N. Edmond, M. Driscoll, K. M. LaChappelle, R. D. Kerns, M. Maly, H. M. Kim, K. B. Farris, D. M. Higgins, E. Buta, A. A. Heapy, Patient-centered pain care using artificial intelligence and mobile health tools: a randomized comparative efectiveness trial, JAMA Internal Medicine 182 (2022) 975–983. doi:10.1001/jamainternmed.2022.3178. [13] R. A. J. de Vries, Theory-Based and Tailor-Made: Motivational Messages for Behavior
Change Technology, Ph.D. thesis, University of Twente, Netherlands, 2018. [14] S. Hors-Fraile, S. Malwade, F. Luna-Perejon, C. Amaya, A. Civit, F. Schneider, P. Bamidis, S. Syed-Abdul, Y.-C. Li, H. De Vries, Opening the black box: Explaining the process of basing a health recommender system on the i-change behavioral change model, IEEE Access 7 (2019) 176525–176540. doi:10.1109/ACCESS.2019.2957696. [15] O. A. Blanson Henkemans, P. J. Van Der Boog, J. Lindenberg, C. A. Van Der Mast, M. A.
Neerincx, B. J. Zwetsloot-Schonk, An online lifestyle diary with a persuasive computer assistant providing feedback on self-management, Technology and Health Care 17 (2009) 253–267. doi:10.3233/THC-2009-0545. [16] D. Hartanto, Computer-Based Social Anxiety Regulation in Virtual Reality Exposure
Therapy, Ph.D. thesis, Delft University of Technology, Netherlands, 2019. [17] Y. Mintz, A. Aswani, P. Kaminsky, E. Flowers, Y. Fukuoka, Nonstationary bandits with habituation and recovery dynamics, Operations Research 68 (2020) 1493–1516. URL: https://doi.org/10.1287/opre.2019.1918. [18] H. Ritschel, A. Seiderer, K. Janowski, S. Wagner, E. André, Adaptive linguistic style for an assistive robotic health companion based on explicit human feedback, in: F. Makedon (Ed.), Proceedings of the 12th ACM International Conference on PErvasive Technologies Related to Assistive Environments, PETRA 2019, Island of Rhodes, Greece, June 5-7, 2019, ACM, 2019, pp. 247–255. doi:10.1145/3316782.3316791. [19] Z. Munn, M. D. Peters, C. Stern, C. Tufanaru, A. McArthur, E. Aromataris, Systematic review or scoping review? guidance for authors when choosing between a systematic or scoping review approach, BMC Medical Research Methodology 18 (2018) 1–7. doi:10.