Accuracy of wrist-worn heart rate monitors during physical therapy sessions among hemiparetic inpatients with stroke

[Purpose] We aimed to examine the accuracy of heart rate monitors worn on the wrist by patients with stroke. The wrist worn heart rate monitor could improve the quality of rehabilitation by monitoring exercise intensity during physical therapy. [Participants and Methods] Thirty inpatients with subacute hemiparetic stroke wore heart rate monitors on both (non-paretic and paretic) wrists, as well as a chest heart rate monitor. We recorded the heart rate values measured at the wrist and chest every minute during physical therapy sessions. The wrist monitors were an optical heart rate measurement device based on photoplethysmography, and the chest monitor was a traditional chest device based on electrocardiography. The relative and absolute reliabilities between the heart rate measurements from the wrist and chest monitors were calculated. [Results] The intraclass correlation coefficients for model 2.1 ranged from 0.75 to 0.79. Bland-Altman analysis revealed a very slight fixed bias; however, no significant proportional bias was observed. For the non-paretic and paretic sides, the lower and upper limits of agreement ranged from -21.8 to 23.8 beats/min and from -20.8 to 21.6 beats/min, and the mean absolute percentage errors were 6.7% and 5.9%, respectively. The Cohen's d value was small. [Conclusion] The relative reliability of the wrist heart rate monitors was substantial. The absolute reliability as bias in wrist heart rate and chest heart rate was small, but heart rates estimated from wrist monitors were not particularly accurate.