Examining the Use of Consumer Wearable Devices and Digital Tools for Stress Measurement in College Students: Scoping Review of Methods

Abstract Background College-aged students face persistent academic and social stress that adversely affects their mental and physical health. Digital phenotyping with wearable devices enables real-time stress monitoring from continuous physiological signals, supporting just-in-time therapeutic interventions to improve student well-being. Despite rapid advances in wearables and analytical methods, it remains unclear which devices, physiological signals, and machine learning or deep learning approaches are most commonly used for stress detection in this population. Objective This study aimed to systematically review the literature to identify best practices and emerging trends in stress measurement using wearable technology and digital tools among college-aged students. We sought to evaluate commonalities in sensor types, datasets, and machine learning approaches used for stress detection. Methods A systematic search was conducted across medical and computer science databases, including Embase, PubMed, IEEE Xplore, and ACM Digital Library, for studies published between January 2020 and December 2025. Studies were included if they examined psychological stress detection using wearable or digital tools among college-aged students and were excluded if they focused on nonpsychological stress, were reviews or prototypes without a defined study population, or lacked clear population information. Two reviewers independently screened studies and extracted data on the wearable sensors, physiological signals, datasets, and modeling approaches to summarize trends in stress prediction. Results A total of 134 studies met the inclusion criteria and were included in the review from the original 792 papers. Electrodermal activity was the most frequently used physiological signal, appearing in 57.5% (n=77) of studies, and wrist-worn wearable devices were the predominant sensing modality. Among studies that compared algorithms, support vector machines were identified as the most commonly applied and best-performing model in 33.3% (n=45) of cases. Overall, 62.8% (n=84) of included studies relied on preexisting datasets, and approximately 80% (n=67) of those used the Wearable Stress and Affect Detection dataset, which contains only 15 participants. Demographic reporting was inconsistent, as 27.6% (n=37) of studies did not report sex distribution, and only 4 studies justified the sample size. The use of temporal modeling algorithms was limited, despite their importance for capturing the dynamic, time-varying nature of stress. This review highlights persistent gaps and underscores the need for more diverse datasets and advanced modeling approaches to improve stress detection accuracy. Conclusions Our review innovatively synthesizes wearable-based stress detection research focused on college-aged students. Unlike prior reviews that aggregate heterogeneous populations or focus primarily on algorithmic performance, this review focused on wearable sensors, physiological signals, modeling approaches, and methodological quality to identify persistent gaps limiting real-world deployment. These findings inform the development of more generalizable monitoring systems to support early mental health intervention in students.