Discussion on “INTACT: a method for integration of longitudinal physical activity data from multiple sources” by Jingru Zhang, Erjia Cui, Hongzhe Li, and Haochang Shou
Biometrics · 09.04.2026 · 82 (2)
DOI: 10.1093/biomtc/ujag114 ↗Kurz gefasst
Fachartikel in der Zeitschrift Biometrics, online erschienen am 17.06.2026. Verfasst von Daniel Fuller und 1 weiteren Autor:innen. Beteiligte Einrichtung: Department of Community Health and Epidemiology, University of Saskatchewan, Saskatoon, S7N 5E5, Canada. Hauptthemen laut MEDLINE: Exercise, Accelerometry. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 23 Literaturangaben.
Abstract (Originalsprache)
Abstract The INTACT (INtegration of Time varying data from weArable sensors for physiCal acTivity) method proposed by Zhang et al. represents a timely and important contribution to the harmonization of accelerometer data across studies. The authors provide a mathematically rigorous presentation and an open source R package. INTACT has clear utility for large scale data integration initiatives. INTACT draws inspiration from techniques used in domains where latent biological mechanisms are poorly defined. We argue that physical activity research is distinct in that the primary latent constructs, time and intensity, are well understood. Accelerometer summary metrics represent proxies for these constructs, raising questions about whether abstraction through an eigenmodel improves interpretation of the physiological meaning. The post harmonization “activity intensity” metric produced by INTACT lacks clear units, scale consistency, and it is unclear how the harmonized metric would be applied to activity classification methods. The method’s reliance on eigenfunctions is both a strength and a limitation. There are challenges for comparability across studies, and it will be important for applied users to conduct sensitivity analyses to evaluate eigenfunction selection. We also highlight the critical assumption of a shared eigenspace across datasets and populations, which may be violated in contexts with different physical activity patterns. Additional methodological development and clear interpretability will play an important role in the adoption of this method by the research community.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Community Health and Epidemiology, University of Saskatchewan, Saskatoon, S7N 5E5, Canada.
- Department of Computer Science, University of Victoria, Victoria, V8W 2Y2, Canada.
ORCID-Profile: Daniel Fuller
Zitieren
APA
Fuller, D., & Stanley, K. (2026). Discussion on “INTACT: a method for integration of longitudinal physical activity data from multiple sources” by Jingru Zhang, Erjia Cui, Hongzhe Li, and Haochang Shou. Biometrics, 82(2). https://doi.org/10.1093/biomtc/ujag114
Vancouver
Fuller D, Stanley K. Discussion on “INTACT: a method for integration of longitudinal physical activity data from multiple sources” by Jingru Zhang, Erjia Cui, Hongzhe Li, and Haochang Shou. Biometrics. 2026;82(2). doi:10.1093/biomtc/ujag114
BibTeX
@article{Fuller2026,
author = {Fuller, Daniel and Stanley, Kevin},
title = {Discussion on “INTACT: a method for integration of longitudinal physical activity data from multiple sources” by Jingru Zhang, Erjia Cui, Hongzhe Li, and Haochang Shou},
journal = {Biometrics},
year = {2026},
volume = {82},
number = {2},
doi = {10.1093/biomtc/ujag114}
}
EUVIMED