Frailty in motion: How Fitbit data reflects patients with cancers' functional status in the All-of-Us database.

1561 Background: Frailty is a vital determinant of outcomes in adults with cancer, leading to heightened toxicity, morbidity, and reduced survival. Despite its significance, comprehensive frailty assessment remains constrained by time, resources, and knowledge. Deficit-based frailty indices (FIs) require extensive surveys, limiting real-world use. Wearable devices (e.g., Fitbit) that track daily steps offer a straightforward measure of physical function, yet their associations with FIs are uncertain in oncology. Leveraging the All of Us Research Program (Controlled Tier Dataset Version 7), one of the most diverse NIH resources, we examined the association between steps and the validated All of Us Frailty Index (AoU-FI) [Wong et al. J Gerontol A Biol Sci Med Sci. 2023]. We hypothesized that higher step counts would correspond to lower (fitter) AoU-FI scores. Methods: We identified adults ≥50 years old with cancer and 3–7 days of Fitbit data between 2017–2025. The AoU-FI comprises 33 deficits encompassing lifestyle, comorbidities, overall health, and healthcare access. Frailty was categorized as Fit ( < 0.15), Pre-Frail (0.15–0.25), or Frail ( > 0.25). Scores were matched with wearable data ± 30 days; outliers ( < 300 or > 20,000 steps/day) and records missing ≥20% of FI items were excluded. Demographics (age, sex, race/ethnicity) and cancer diagnoses were documented. Mean step counts were compared via t-tests. A linear regression assessed the step–frailty link, and a multivariable logistic model for discerning Fit vs Pre-Frail or Frail (age≥65, male, > one cancer) employed normalized step counts (per 1,000). Results: Among 361 participants (mean age 65.7±7.8; 66.8% female; 92.5% White; 97.5% non-Hispanic/Latino), 86.2% had one cancer (skin 55.7%, breast 24.8%, prostate 9.9%). Frailty categories were 44.0% Fit, 40.4% Pre-Frail, and 15.5% Frail. Mean steps declined with increasing frailty (Fit vs. Pre-Frail 8,136±3,251 vs. 7,067±3,598; p = 0.007; vs. Frail 5,486±3,623; p < 0.0001). A linear regression confirmed an inverse link between frailty and steps. (slope = –7.38×10⁻⁶; p < 0.0001). Adjusted analysis showed steps remained highly significant (OR = 1.14, 95% CI: 1.07–1.21, p < 0.0001), whereas gender, > one cancer, and age≥65 were not. Conclusions: All-of-Us is designed to represent many populations, and our sample showed older adults adopting wearable technologies, highlighting their potential for clinical use. However, the sample was primarily White individuals with skin cancer, reflecting coverage gaps in wearable data across the cancer population. Step counts strongly correlated with fitness, supporting wearable-based functional assessment in oncology. Chronological age did not correlate with frailty, aligning with guidelines prioritizing holistic assessments. Future efforts should integrate broader populations, minute-level metrics, and cancer and treatment information to refine frailty classification as recruitment grows.