JMIR Publications

ABSTRACT

Background:

Describing changes in health and behavior that precede and follow a sentinel health event, like a cancer diagnosis, is challenging because of a lack of longitudinal, objective measurements that are collected frequently enough to capture varying trajectories of change leading up to and following the event. A continuous passive assessment system that continuously monitors older adults’ physical activity, weight, medication-taking behavior, pain, health events, and mood could enable the identification of more specific health and behavior patterns leading up to a cancer diagnosis and whether and how patterns change thereafter.

Objective:

To conduct a proof-of-concept retrospective analysis in which we identified new cancer diagnoses in older adults and compared trajectories of change in health and behaviors before and after cancer diagnosis.

Methods:

Participants were 10 older adults (age = 71.8±4.9 years, 30% women) with various self-reported cancer types from a larger prospective cohort study of older adults. A technology-agnostic assessment platform using multiple devices provided continuous data on daily physical activity via wearable sensors (actigraphy), weight via a WiFi-enabled digital scale, daily medication behavior using electronic Bluetooth-enabled pillboxes, and weekly pain, health events, and mood with online, self-report surveys.

Results:

Longitudinal linear mixed-effects models revealed significant differences in the pre- and post-cancer trajectories of step counts (P<.001), step count variability (P=.004), weight (P<.001), pain severity (P<.001), hospital/ER visits (P=.03), days away from home overnight (P=.01), and the number of pillbox door openings (P<.001). Over the year preceding a cancer diagnosis, there were gradual reductions in step counts and weight, gradual increases in pain severity, step count variability, hospital/ER visits, and days away from home overnight, compared to one year after a cancer diagnosis. Across the year after a cancer diagnosis, there was a gradual increase in the number of pillbox door openings as compared to one year before a cancer diagnosis. There was no significant trajectory change from pre- to post-cancer in terms of low mood (P=.60) and loneliness (P=.22).

Conclusions:

A home-based, technology-agnostic, multi-domain assessment platform could provide a unique approach to monitoring in parallel different types of behavior and health markers before and after a life-changing health event. Continuous passive monitoring that is ecologically valid, less prone to bias, and limits participant burden could greatly enhance research that aims to improve early detection efforts, clinical care, and outcomes for people with cancer.