Accelerometers' Validity in Counting Number of Steps in the Elderly Subjects Before Discharge From Rehabilitation Units, Having Reached Their Maximal Walking Ability After Physiotherapy. Comparison of the Accuracy of 3 Accelerometer Positions: Wrist, Ankle, Hip

Accelerometers enhance physical activity. Only one study assessed accelerometers' validly in an elderly robust population, over 100m walking distance. The accelerometers' validly was demonstrated especially for the ankle position (Floegel et al., 2017). The primary objective was to assess accelerometers validity for counting steps in an elderly frail population ready for discharge from a rehabilitation unit. The comparison gold standard was steps count by 2 physicians blind to accelerometers data, over a filmed 10-meter walk (minimal capacity to walk inside their own living place). The second objective was to evaluate the best position of the accelerometer: wrist, ankle, hip.

Investigators included prospectively subjects aged ≥ 70 years, hospitalized in the Paul Brousse geriatric rehabilitation ward, able to walk at least 10 meters (maximum functional recovery according to the physiotherapist opinion) and with a Mini mental state examination ≥ 20/30 (able to understand). All subjects gave written informed consent and the study was approved by the local ethics committee. The secondary objective was to evaluate the position of the accelerometer that give the most accurate step count: wrist, ankle, hip. The study took place in the Paul Brousse hospital in Villejuif in rehabilitations units. Falls during the protocol were considered as the only risk. To prevent this risk a physiotherapist walked behind the participants during the 10-meter walk. Demographic, clinical, physiological data were recorded and anonymized. In this monocentric non-randomized study, the number of participants to include was 120. Twenty participants were planned to be included each month during 6 months. Statistical analysis will be made by a T test to measure the difference between the gold standard and the count of the accelerometers. To eliminate measurement bias, Bland Altman analysis will be performed. Interclass correlation will be performed to measure the differences between accelerometers' positions. Logistic regressions will be done with measurement's variability determinants. In all analyses, the 2-sided α-level of 0.05 was used for significance testing.. All analysis will be performed using R statistical software.