Consumer wearables and medical care utilization in gastrointestinal cancers: A Fitbit data analysis in NIH All of Us Research Program.

Abstract

11100 Background: Patients with gastrointestinal (GI) cancers may experience delayed or forgone care due to cost, travel, or health system barriers. Wearable devices are widely used and capture physical activity that may reflect health status, mobility, and social factors. Whether wearable activity data can help identify patients at risk for delayed care is not well understood. We examined associations between activity levels and delayed or forgone care among adults with GI cancers in the NIH All of Us Research Program. Methods: Adults with GI cancers who contributed wearable activity data and healthcare utilization surveys were included; ≥30 valid Fitbit days were required. Mean daily step counts were calculated and categorized into cohort-specific quartiles: Q1 (≤3,621 steps/day), Q2 (3,621–5,591), Q3 (5,591–8,272), and Q4 ( > 8,272). Delayed or forgone care was determined from survey items asking whether participants postponed or avoided medical care due to financial or access barriers. Analyses were performed in the overall GI cohort (n = 407) and stratified by cancer site (colorectal, pancreatic, upper GI, hepatobiliary, and other). Logistic regression evaluated associations between activity and delayed care using categorical quartiles, continuous log-transformed steps, and age-adjusted models. Results: The analytic cohort included 407 participants with GI cancers. After age adjustment, higher daily step counts—analyzed continuously and across ordered quartiles—were associated with lower odds of delayed or forgone care. Age itself was independently associated with fewer delays (OR per 10 years 0.62, p = 0.001–0.002). In categorical models using Q3 as the reference, participants in Q1 had higher odds of delayed care (OR 3.57, 95% CI 1.06–12.10, p = 0.041), while Q2 and Q4 did not differ significantly from Q3, supporting a non-linear association. Site-stratified analyses showed similar non-linear patterns; for example, pancreatic cancer (n = 60) had delayed care prevalences of Q1 11.1%, Q2 0%, Q3 5.0%, Q4 9.1%, and upper GI cancers (n = 59) had Q1 9.1%, Q2 0%, Q3 6.3%, Q4 13.3%. In colorectal cancer (n = 210), Q3 had the lowest delayed care vs Q1 (OR 0.25, 95% CI 0.05–1.32), although other quartile estimates were less precise due to small event counts. Conclusions: Among patients with GI cancers who contributed wearable and survey data, delayed or forgone care showed a non-linear association with activity, with the lowest odds in the moderate–high activity range. Participants in Q1 had higher odds of delayed care compared with Q3 after age adjustment. Wearable activity data may help identify patients who face barriers to timely care, although this cross-sectional study cannot show cause and effect. Longitudinal studies are needed to determine whether passive wearable data can support care navigation and healthcare utilization in oncology.