The following study illustrates some of the statistical and methodological challenges associated with examining short-term changes in depressive symptoms among individuals participating in substance abuse treatment.
Participants (N=146) were male military veterans that were participating in primary residential or day hospital-based substance abuse treatment. All participants completed a self-report measure of depressive symptoms upon admission and just before program completion.
Initial statistical analyses showed significant improvements, however, additional analyses indicated that some of the observed change was likely because of regression to the mean.
The confounding effects of regression to the mean are discussed, along with an examination of analytical, ethical, and methodological issues that make it difficult to control for regression to the mean artifacts. Alternative methodological and analytical approaches are explored and directions for future research are discussed.
*Department of Health Disparities Research, University of Texas M.D. Anderson Cancer Center, Houston, TX
†Department of Anesthesiology, Wake Forest University School of Medicine, Winston-Salem, NC
‡G.V. (Sonny) Montgomery VA Medical Center, Mental Health Service
§Department of Psychiatry, University of Mississippi School of Medicine, Jackson, MS
The authors declare no conflict of interest.
Reprints: Claire E. Adams, PhD, Department of Health Disparities Research—Unit 1440, The University of Texas M.D. Anderson Cancer Center, P.O. Box 301402, Houston, TX 77230-1402 (e-mail: firstname.lastname@example.org; email@example.com).