JMIR Publications

ABSTRACT

Background:

Older cancer survivors encounter distinct challenges, including a heightened risk of mental health conditions like depression and anxiety due to psychological distress from cancer experiences, fear of recurrence, and aging-related difficulties. These factors negatively affect quality of life (QoL), highlighting the need for non-intrusive monitoring approaches to enable timely interventions. Conventional mental health monitoring approaches, such as clinical assessments and even electronic patient-reported outcomes, face limitations like memory recall bias, patient burden, and infrequent data collection. Emerging technologies like wearables and smart home devices offer passive, continuous monitoring, potentially enabling early detection with minimal patient effort. However, their effectiveness in accurately identifying mental health risks in vulnerable groups, such as older cancer survivors, requires further exploration.

Objective:

This study aims to explore the added value that patient-generated health data collected in-the-wild, either passively or actively, may contribute in terms of older cancer survivors’ health-related QoL monitoring. The main objective is to explore the value that data from completely passive monitoring modalities, such as smart plugs, offer towards that end.

Methods:

This study recruited 41 older cancer survivors (mean age: 72.3, SD = 6.81) from the LifeChamps project . Over a 12-week period, participants were monitored using an activity tracker to measure physical activity, sleep, and physiological metrics; a smart scale to capture weight and body composition; and a smart plug to track television usage as a proxy for daily living activities. Participants also self-reported their mental health status using the PHQ-4 questionnaire via a mobile application. Machine learning models were trained to classify mental health risk levels based on features derived from each sensor modality, both independently and in combination.

Results:

Random Forest models demonstrated superior performance in classifying mental health risk. Notably, smart plug data alone achieved the highest predictive accuracy, with an average F1 score of 73% and peaks of 100% in specific instances—outperforming models using Fitbit and Withings data. While combining sensor data enhanced cross-validation results, it failed to consistently surpass the standalone smart plug model in test scenarios. The reduced efficacy of wearables and smart scales may reflect declining user adherence over time. Crucially, passive monitoring of TV usage patterns emerged as a robust behavioral indicator of mental health states.

Conclusions:

This study pioneers the use of passively collected data (e.g., smart plugs) for mental health monitoring in older cancer survivors, demonstrating their potential. Smart plugs capture behavioral patterns without user burden, with strong standalone performance (average F1: 73%) positioning them as promising tools. Future work should validate findings in larger cohorts and integrate non-intrusive sensors to enhance robustness. Such technologies could transform monitoring for vulnerable populations, enabling scalable, inclusive care while reducing healthcare burdens.