Patient-generated health data from wearables and smart home devices, particularly smart plugs monitoring television use patterns, can classify mental health risk in older cancer survivors, with the best-performing model combining smart plug and smart scale features achieving a mean F1-score of 0.77 and mean AUC of 0.85.
Key Findings
Results
The best-performing machine learning configuration, combining smart plug and smart scale features, achieved good classification performance for mental health risk in older cancer survivors.
Combined smart plug and smart scale features achieved a mean F1-score of 0.77 (SD 0.15) and a mean AUC of 0.85 (SD 0.10)
Performance was measured across 3 repeated train-test splits
Tree-based gradient boosting models showed the best overall performance
Mental health risk was defined based on PHQ-4 questionnaire scores
Results
Standalone smart plug models based solely on passive television use patterns outperformed models relying only on activity tracker data.
Smart plug models achieved a mean F1-score of 0.66 (SD 0.04) and a mean AUC of 0.71 (SD 0.06)
Activity tracker-only models achieved a mean F1-score of 0.59 (SD 0.2)
Smart plug data captured television viewing behavior passively without user burden
Television use was used as a proxy for sedentary television viewing behavior
Results
Multimodal combinations of sensor data tended to improve average performance but did not consistently yield large gains over the strongest single-modality configurations.
Lack of consistent large gains was likely attributed to adherence-related data loss for wearables and scales
Multiple sensor modalities were tested both independently and in combination
Modalities included activity tracker, smart scale, and smart plug
Activity tracker measured physical activity, sleep, and physiological metrics
Methods
The study recruited 41 older cancer survivors monitored over a 12-week period using multiple smart home and wearable devices.
Participants had a mean age of 72.3 years (SD 6.81 years)
Participants were recruited from the LifeChamps project
Monitoring devices included an activity tracker, smart scale, and smart plug
Mental health status was self-reported via the PHQ-4 questionnaire through a mobile app
Data collection occurred 'in the wild,' meaning in participants' natural home environments
Results
Passive monitoring of television use patterns emerged as a promising behavioral proxy measure of mental health states in older cancer survivors.
Smart plugs captured behavioral patterns without requiring active user input or burden
Standalone smart plug performance reached a mean AUC of 0.71 (SD 0.06)
This represents the first reported use of passively collected smart plug data for mental health monitoring in older cancer survivors
The authors describe this as positioning smart plugs as 'a promising low-burden modality'
Background
Older survivors of cancer face heightened risk of depression and anxiety related to cancer experiences, fear of recurrence, and aging-related difficulties.
Conventional mental health monitoring approaches including clinical assessments and electronic patient-reported outcomes are limited by recall bias, patient burden, and infrequent data collection
Emerging patient-generated health data from wearables and smart home devices offer passive, low-burden, continuous monitoring as an alternative
The ability of such technologies to capture mental health risks specifically in older cancer survivors was previously unclear
What This Means
This research suggests that everyday smart home devices, particularly smart plugs that track when a television is in use, can help detect signs of anxiety and depression in older people who have survived cancer. Researchers monitored 41 cancer survivors (average age 72) for 12 weeks using a fitness tracker, a smart scale, and a smart plug connected to their television. Participants also regularly reported their mental health through a short questionnaire on a mobile app. Computer models were then trained to predict whether someone showed signs of poor mental health based on patterns in the device data.
The most effective approach combined data from the smart plug and smart scale, correctly identifying mental health risk with an accuracy score (AUC) of 0.85 out of 1.0. Notably, the smart plug alone — simply tracking television viewing patterns — performed better than the fitness tracker alone, suggesting that how and when people watch TV may reflect their emotional state. This is significant because smart plugs require no effort from the user; the device passively collects data without anyone needing to wear anything or actively record information.
This research suggests that low-cost, unobtrusive smart home technology could offer a practical way to continuously monitor the mental health of vulnerable older adults without adding to their daily burden. Traditional mental health monitoring relies on clinic visits or self-reported questionnaires, which can miss day-to-day changes and may be difficult for older or less tech-savvy individuals. The authors caution that the study involved a small number of participants and that findings need to be confirmed in larger groups before this approach could be used in routine care.
Petridis G, Billis A, Meditskos G, Tsanousa A, Lagakis P, Naranjo J, et al.. (2026). Identification of Early Signs of Mental Health Disorders in Older Survivors of Cancer Using Patient-Generated Health Data: Observational Study.. JMIR cancer. https://doi.org/10.2196/75050