The prevalence of gas exchange data processing methods: a semi-automated scoping review.

Abstract

Cardiopulmonary exercise testing involves collecting variable breath-by-breath data and sometimes requiring data processing of outlier removal, interpolation, and averaging before later analysis. These data processing choices, such as averaging duration, affect calculated values such as ˙VO₂max. However, assessing the implications of data processing without knowing popular methods worth comparing is difficult. In addition, such details aid study reproduction. We conducted a semi-automated scoping review of articles with exercise testing that collected data breath-by-breath from three databases. Of the 8,344 articles, 376 (mean: 4.5% and 95% confidence interval: 4.1-5.0%) and 581 (mean: 7.0% and 95% confidence interval: 6.4-7.5%) described outlier removal and interpolation, respectively. A random subset of 1,078 articles revealed (mean: 60.9% and 95% confidence interval: 57.9-63.7%) the reported averaging methods. The commonly documented outlier cutoffs were±3 or 4 SD (39.1 and 51.6%, respectively). The dominating interpolation duration and procedure were 1 s (93.9%) and linear interpolation (92.5%). Averaging methods commonly described were 30 (30.9%), 60 (12.4%), 15 (11.6%), 10 (11.0%), and 20 (8.1%) second bin averages. This shows that studies collecting breath-by-breath data often lack detailed descriptions of data processing methods, particularly for outlier removal and interpolation. While averaging methods are more commonly reported, improved documentation across all processing steps will enhance reproducibility and facilitate future research comparing data processing choices.