Agreement of a wearable activity monitor with reference devices in post-operative inpatients: A method-comparison study

Objective To evaluate the agreement of wrist-based step count and walking distance metrics from an investigational wearable device with consumer- and research-grade reference devices in post-operative inpatients, and to explore performance at alternative anatomical placements. Methods In this single-centre observational study, 99 patients (median age 58.0 years [IQR 48.0-66.5]) were monitored during the early post-operative period following cardiac or thoracic surgery. The investigational device was worn simultaneously on the wrist, upper arm, and chest. Wrist analyses were prespecified as primary, whereas upper-arm and chest analyses were exploratory. Reference data were collected using an Apple Watch Series 8 (wrist) and an Axivity AX6 (ankle). Axivity-derived reference step count outputs were obtained from raw accelerometry data processed offline using the open-source NiMBaLWear analytics pipeline. Agreement was assessed using Lin’s concordance correlation coefficient (CCC), intraclass correlation coefficient (ICC 2,1), and repeated-measures Bland-Altman analysis. Results In the prespecified primary wrist analysis, agreement with the Axivity AX6 reference was strong (CCC = 0.850, ICC = 0.852, MAPE = 40.7%). Apple Watch Series 8 agreement with Axivity was lower (CCC = 0.621, ICC = 0.622, MAPE = 67.8%). In exploratory analyses, orientation-aware axis alignment substantially improved agreement at the upper arm and chest (CCC=0.917 and 0.916, respectively). Conclusions The wrist-based implementation demonstrated agreement with the evaluated reference devices for step count and walking distance in early postoperative inpatients. These findings reflect device-to-device agreement rather than absolute accuracy against a criterion standard. Orientation-aware alignment improved agreement at alternative wear locations, but these exploratory findings require independent prospective validation.