Logo image
Back
Calibration Of Steps From Two Generations Of ActiGraph Accelerometers: 7164 And WGT3x-BT: 488
Abstract   Peer reviewed

Calibration Of Steps From Two Generations Of ActiGraph Accelerometers: 7164 And WGT3x-BT: 488

Erin E. Dooley, Bjoern Hornikel, Amanda E. Paluch, Kara M. Whitaker, Sylvia E. Badon, David R. Jacobs, Barbara Sternfeld, Mehedi Hasan, Zhe Wang and Kelley Pettee Gabriel
Medicine and science in sports and exercise, Vol.58(10S), 488
10/2026
DOI: 10.1249/01.mss.0001260132.20138.0c

View Online

Abstract

Longstanding cohorts, like Coronary Artery Risk Development in Young Adults (CARDIA), are subject to advances in technology across data collection. CARDIA has accelerometry measures across midlife (ActiGraph 7164 [2005-06] and wGT3X-BT [2015-16 & 2020-22]). Differences in steps estimated by the two devices may impact life course analyses. PURPOSE: To examine the agreement in step measurement of two generations of devices (ActiGraph 7164 and wGT3X-BT) and, if differences are observed, to develop and test a minute-level calibration formula to enhance comparability of steps between the two devices. METHODS: We analyzed a subsample (n=100) of adults who simultaneously wore both devices for 7 days. Minutes were screened for wear and non-wear was removed. Days with <10 hrs of wear were excluded. Daily steps between the devices were compared using paired t-tests. Agreement was assessed using mean difference (wGT3X-BT - 7164), mean percentage error (MPE), mean absolute percentage error (MAPE), and intraclass correlation coefficients (ICC). A linear regression model was fit to estimate a calibration factor. Bland-Altman plots were used to visually assess bias and limits of agreement pre and post calibration. RESULTS: There were 95 with adherent wear (mean age = 55.4 y; 46% Black). Mean daily steps were significantly higher for wGT3X-BT compared with 7164 (16,366 ± 6,215 vs. 10,652 ± 4,844; mean difference = 5,714, p < 0.001) with MPE = 88.3%, MAPE = 88.3%, and moderate agreement (ICC = 0.56) between devices. Bland-Altman plots show overestimation by the wGT3X-BT across the range of steps (Figure 1A). A linear calibration factor of 0.714 was applied to the wGT3X-BT and improved agreement (mean difference = 1033) with reduced bias and narrower limits of agreement (Figure 1B). CONCLUSIONS: Our calibration improved agreement and will harmonize data across CARDIA. This approach can be used in other studies that change devices/manufacturers to retain their longitudinal structure.

Details

Metrics

6 Record Views
Logo image