1. |
- Coenen, Pieter, et al.
(författare)
-
Calibration of self-reported physical behaviours among office workers: A compositional data analysis
- 2019
-
Ingår i: ICAMPAM 2019. - Maastricht : ICAMPAM.
-
Konferensbidrag (refereegranskat)abstract
- The aim of this study was to develop and evaluate calibration models to predict objectively measured time spent sitting, standing and walking during office work from self-reported time-use compositions using a compositional data analysis (CoDA) approach. Ninety-nine office workers (49 women) at the Swedish Transport Administration participated in an intervention study on relocation to activity-based offices. At baseline and at a 3-months follow-up, physical behaviours (sitting, standing and walking) at work were assessed for five days using a thigh-mounted accelerometer (Actigraph) and by self-report (IPAQ). The time-use composition of the three behaviours was expressed in terms of isometric log-ratios (ILR). Calibration models predicting accelerometry-based time-use from self-reported compositions were constructed using linear regression on baseline data, and then validated using follow-up data. The accelerometer data showed that, on average, workers spent 69.9% of their day sitting, 23.7% standing, and 6.4% walking. The corresponding percentages for self-reports were 71.7%, 21.6%, and 7.4%, respectively. Non-calibrated self-reports were biased: the RMS errors obtained from the ILRs expressing sitting, standing and walking were 0.73, 1.09 and 1.05, respectively. Calibration models reduced these errors by 45% (sitting), 56% (standing), and 76% (walking). Validation of the calibration models using follow-up data from the same workers showed calibration remained equally effective; RMS errors were reduced by 55% (sitting), 58% (standing), and 75% (walking). In conclusion, calibration models for compositional time-use data were effective in reducing bias in self-reported physical behaviours at work, and the models remained effective when used on new data from the same workers. Calibrated self-reports may represent a cost-effective method for obtaining physical behaviour data with a satisfying accuracy in large-scale cohort and intervention studies.
|
|