Who Are We Excluding From Physical Activity Research? Examining the Potential for Exclusion Bias in Pedometer Data...

Area of Science:

• Physical activity measurement
• Biostatistics
• Public health research

Background:

• Pedometer-derived physical activity (PA) data are widely used but may contain systematic bias due to varying data cleaning criteria.
• The impact of different inclusion criteria on sample size and sociodemographic representation in pedometer studies remains unclear.

Purpose of the Study:

• To explore existing pedometer data cleaning criteria.
• To examine how different inclusion criteria affect sample size retention and participant exclusion based on sociodemographic factors.

Main Methods:

• Data from a community survey in Aotearoa/New Zealand were analyzed.
• Participants wore Yamax CW300 pedometers for 7 days, with sociodemographic data collected via surveys.
• Analyses included outlier removal, minimum steps per day determination, day 1 data removal assessment, and bias risk evaluation.

Main Results:

• Pedometer data were obtained from 895 participants, with 100 steps/day set as a valid day threshold.
• Increasing the stringency of inclusion criteria led to reduced participant retention and increased bias.
• Models with lower exclusion bias demonstrated convergent and concurrent validity.

Conclusions:

• More stringent pedometer data inclusion criteria can lead to substantial, biased sample size reduction.
• Transparent reporting of data cleaning methods and their potential biases is crucial for pedometer-based physical activity research.