Predicting hospitalization from real-world measures in patients with chronic kidney disease: A proof-of-principle study.

Objective: To investigate if in-clinic measures of physical function and real-world measures of physical behavior and mobility effort are associated with one another and to determine if they predict future hospitalization in participants with chronic kidney disease (CKD).

Methods: In this secondary analysis, novel real-world measures of physical behavior and mobility effort, including the best 6-minute step count (B6SC), were derived from passively collected data from a thigh worn actigraphy sensor and compared to traditional in-clinic measures of physical function (e.g. 6-minute walk test (6MWT). Hospitalization status during 2 years of follow-up was determined from electronic health records. Correlation analyses were used to compare measures and Cox Regression analysis was used to compare measures with hospitalization.

Results: One hundred and six participants were studied (69  ±  13 years, 43% women). Mean  ±  SD baseline measures for 6MWT was 386  ±  66 m and B6SC was 524  ±  125 steps. Forty-four hospitalization events over 224 years of total follow-up occurred. Good separation was achieved for tertiles of 6MWT, B6SC and steps/day for hospitalization events. This pattern persisted in models adjusted for demographics (6MWT: HR  =  0.63 95% CI 0.43-0.93, B6SC: HR  =  0.75, 95% CI 0.56-1.02 and steps/day: HR  =  0.75, 95% CI 0.50-1.13) and further adjusted for morbidities (6MWT: HR  =  0.54, 95% CI 0.35-0.84, B6SC: HR  =  0.70, 95% CI 0.49-1.00 and steps/day: HR  =  0.69, 95% CI 0.43-1.09).

Conclusion: Digital health technologies can be deployed remotely, passively, and continuously to collect real-world measures of physical behavior and mobility effort that differentiate risk of hospitalization in patients with CKD.