OBJECTIVES: With the emerging use of machine learning (ML) techniques, there has been particular interest in utilizing data from wearable devices for health economics and outcomes research (HEOR). We aimed to understand the emerging patterns of how ML has been applied to data collected from wearables in HEOR.

METHODS: We identified studies published in PubMed between January 2016 through March 2021. Studies that included at least one HEOR-related MeSH term, applied an ML technique and used wearable data were eligible for inclusion. Two reviewers abstracted information on ML application types, data on which ML was applied, and specific ML methods used. Extracted data were analyzed using descriptive analyses.

RESULTS: A total of 148 studies were identified from PubMed, among which 33 studies met the inclusion criteria. Between 2016-2021, there has been an increase over time in the number of ML studies using wearable data. ML has been more frequently used for monitoring events in real time (79%) than to predict future events (21%). There has been a wide range of outcomes examined, ranging from general physical or mental health (45%) to more disease-specific outcomes such as disease incidence (18%) and disease progression (15%). ML has also been used for treatment-related outcomes such as treatment adherence (9%), treatment outcomes (9%), and healthcare resource utilization (3%). Data for ML models were more often derived from wearable devices with specific medical purposes (58%) than those without (42%). Studies have used tree-based methods (e.g., random forests and boosting) most frequently (25%).

CONCLUSIONS: There has been a wide range of applications of ML to wearable data, from predicting general health outcomes to disease- or treatment-specific health outcomes. Both medical and non-medical wearable devices have been used as a data source, showing the potential for providing rich data for ML studies in HEOR.