Correction of bias in self-reported sitting time among office workers - a study based on compositional data analysis

Area of Science:

• Occupational Health
• Biostatistics
• Physical Activity Epidemiology

Background:

• Excessive sitting is linked to adverse health outcomes.
• Research on sitting time often relies on self-reports, which are prone to bias.
• Objective measurement of sitting time is needed to overcome self-report limitations.

Purpose of the Study:

• To develop and validate a calibration model for estimating true occupational sitting time from self-reported data.
• To correct for known biases in self-reported sitting time using objective measures.

Main Methods:

• Used self-reports (International Physical Activity Questionnaire) and thigh-worn accelerometers to measure sitting time in 99 Swedish office workers.
• Applied compositional data analysis to transform sitting estimates into isometric log-ratios (ILR).
• Developed linear regression calibration models (simple and full) to predict objectively measured sitting ILR from self-reported sitting ILR, validated with follow-up data.

Main Results:

• Uncalibrated self-reported sitting time had a root-mean-square (RMS) error of 0.767.
• Simple and full calibration models reduced RMS error by 55% and 52%, respectively.
• Model performance decreased slightly during validation, but calibration remained effective.

Conclusions:

• Calibration models substantially improve the accuracy of sitting time estimates compared to uncalibrated self-reports.
• Simple calibration models incorporating objective measures are effective and perform similarly to more complex models.
• Calibration is a valuable tool for more accurate assessment of sedentary behavior in research.