Assessment of Wrist-Worn Activity Tracker Accuracy for Measuring Sedentary Behavior in Outpatients With Cardiovascular Disease: Validation Study.
Wearable monitoring of sedentary behavior is increasingly used in digital cardiovascular care, but the validity of consumer-grade wrist-worn trackers in older patients with cardiac issues is unclear.
This study aimed to evaluate the agreement between sedentary time measured by a consumer-grade wrist-worn tracker (Fitbit Inspire 3; Fitbit LLC) and a research-grade waist-worn accelerometer (ActiveStyle Pro HJA-750C; Omron Healthcare Co, Ltd) used as the comparator, in Japanese outpatients with cardiac issues under free-living conditions.
Fifty-one outpatients with heart failure (28/51, 55%) or ischemic heart disease (23/51, 45%; mean age 78.4, SD 9.0 years; 25/51, 49% male) simultaneously wore both devices for 7 consecutive days. Sedentary time was assessed within a standardized 12-hour daytime window (9 AM to 9 PM) at 1-minute epochs. Agreement was examined using Pearson correlation, Bland-Altman analysis (after confirming normality of the paired differences), Lin concordance correlation coefficient (CCC), and Cohen kappa for distribution-based tertile classification. A constant bias correction was applied, with leave-one-out cross-validation (LOOCV) used as an internal-consistency check on the stability of the sample mean bias. Moderators of the bias were explored using multivariable regression.
Mean sedentary time within the 12-hour window was 497.9 (SD 86.5) minutes for the comparator and 559.9 (SD 84.3) minutes for the Fitbit. The devices were strongly correlated (r=0.905, 95% CI 0.838-0.945; P<.001). The Fitbit showed a significant fixed bias of 62.0 minutes (95% CI 51.5-72.5; P<.001) with no proportional bias (slope -0.03; P=.67); the 95% limits of agreement were wide (-11.3 to 135.2 minutes). Uncorrected concordance was moderate (CCC=0.71, 95% CI 0.57-0.80) and improved to high after subtracting the 62-minute bias (CCC=0.904, 95% CI 0.84-0.94); LOOCV indicated that this in-sample improvement was internally stable (out-of-sample CCC=0.90). No candidate factor (sex, age, diagnosis, and musculoskeletal disorder) significantly moderated the bias (model P=.12). Tertile classification showed moderate agreement (κ=0.588, 95% CI 0.405-0.772; 37/51, 72.5% agreement), with all misclassifications between adjacent categories.
In older Japanese outpatients with cardiac issues, a consumer-grade wrist-worn tracker overestimated sedentary time by approximately 62 minutes within a standardized 12-hour daytime window compared with a waist-worn comparator, with fixed but no proportional bias. A simple constant correction markedly improved group-level concordance (cross-validated CCC≈0.90), supporting calibrated use of consumer wearables for scalable monitoring in ambulatory cardiac care. However, wide limits of agreement constrain individual-level precision, and the population-specific correction factor requires independent validation before transfer to other settings.
This study aimed to evaluate the agreement between sedentary time measured by a consumer-grade wrist-worn tracker (Fitbit Inspire 3; Fitbit LLC) and a research-grade waist-worn accelerometer (ActiveStyle Pro HJA-750C; Omron Healthcare Co, Ltd) used as the comparator, in Japanese outpatients with cardiac issues under free-living conditions.
Fifty-one outpatients with heart failure (28/51, 55%) or ischemic heart disease (23/51, 45%; mean age 78.4, SD 9.0 years; 25/51, 49% male) simultaneously wore both devices for 7 consecutive days. Sedentary time was assessed within a standardized 12-hour daytime window (9 AM to 9 PM) at 1-minute epochs. Agreement was examined using Pearson correlation, Bland-Altman analysis (after confirming normality of the paired differences), Lin concordance correlation coefficient (CCC), and Cohen kappa for distribution-based tertile classification. A constant bias correction was applied, with leave-one-out cross-validation (LOOCV) used as an internal-consistency check on the stability of the sample mean bias. Moderators of the bias were explored using multivariable regression.
Mean sedentary time within the 12-hour window was 497.9 (SD 86.5) minutes for the comparator and 559.9 (SD 84.3) minutes for the Fitbit. The devices were strongly correlated (r=0.905, 95% CI 0.838-0.945; P<.001). The Fitbit showed a significant fixed bias of 62.0 minutes (95% CI 51.5-72.5; P<.001) with no proportional bias (slope -0.03; P=.67); the 95% limits of agreement were wide (-11.3 to 135.2 minutes). Uncorrected concordance was moderate (CCC=0.71, 95% CI 0.57-0.80) and improved to high after subtracting the 62-minute bias (CCC=0.904, 95% CI 0.84-0.94); LOOCV indicated that this in-sample improvement was internally stable (out-of-sample CCC=0.90). No candidate factor (sex, age, diagnosis, and musculoskeletal disorder) significantly moderated the bias (model P=.12). Tertile classification showed moderate agreement (κ=0.588, 95% CI 0.405-0.772; 37/51, 72.5% agreement), with all misclassifications between adjacent categories.
In older Japanese outpatients with cardiac issues, a consumer-grade wrist-worn tracker overestimated sedentary time by approximately 62 minutes within a standardized 12-hour daytime window compared with a waist-worn comparator, with fixed but no proportional bias. A simple constant correction markedly improved group-level concordance (cross-validated CCC≈0.90), supporting calibrated use of consumer wearables for scalable monitoring in ambulatory cardiac care. However, wide limits of agreement constrain individual-level precision, and the population-specific correction factor requires independent validation before transfer to other settings.
Authors
Naito Naito, Izawa Izawa, Maeda Maeda, Kasai Kasai, Tani Tani, Kondo Kondo, Iwama Iwama
View on Pubmed