Response to commentary by Skinner et al. on Regression to the Mean (RTM) in Burke et al.

Rachel M Burke,Christi Kay,Julie Gazmararian

doi:10.1186/s12966-015-0220-6

Rachel M Burke, Christi Kay + Show 1 more

Open Access

https://doi.org/10.1186/s12966-015-0220-6

Copy DOI

Abstract

This letter is a response to commentary by Skinner et al. on an evaluation by Burke et al. of the HealthMPowers program, an elementary-school-based program developed to improve child health and wellness. In their commentary, Skinner et al. make the criticism that our results for changes in BMI-for-Age Z score were simply reflective of Regression to the Mean (RTM). In this response, we show that while some of our results are consistent with RTM, with adjustment we do still observe some small effects in BMI-for-Age Z score over the course of the school year. We conclude that while our evaluation was not definitive, we still believe that programs of similar design to HealthMPowers merit further rigorous study.

Highlights

This letter is a response to commentary by Skinner et al on an evaluation by Burke et al of the HealthMPowers program, an elementary-school-based program developed to improve child health and wellness
As mentioned in the original paper, this analysis had the severe limitation of lacking a control group. This is an effect of the fact that the HealthMPowers program was not designed as a research study, and the paper itself was conceived as an evaluation of the program based on existing data from its implementation
As Skinner et al explain, one-group designs are vulnerable to the potential effects of Regression to the Mean (RTM)

Summary

Introduction

This letter is a response to commentary by Skinner et al on an evaluation by Burke et al of the HealthMPowers program, an elementary-school-based program developed to improve child health and wellness. This is an effect of the fact that the HealthMPowers program was not designed as a research study, and the paper itself was conceived as an evaluation of the program based on existing data from its implementation.

Results

Conclusion