Details

Autor(en) / Beteiligte

• Jewett, Lisa R
• Newton, Evan G
• Smith, Sascha
• Thombs, Brett D
• Kaplan, Robert M

Titel

Time to Abandon Small-Sample Cohort Research in Health Psychology?

Ist Teil von

• Health psychology, 2010-07, Vol.29 (4), p.339-340

Ort / Verlag

United States: American Psychological Association

Erscheinungsjahr

2010

Quelle

Applied Social Sciences Index & Abstracts (ASSIA)

Beschreibungen/Notizen

• Comments on the original article, "Avoidant coping as a predictor of mortality in veterans with end-stage renal disease" by E. J. Wolf and D. L. Mori (see record 2009-06704-009), in which the authors report that among 61 patients with endstage renal disease (ESRD), avoidant coping more than doubled the odds of mortality in multivariable analysis. The authors concluded that evaluating coping style in patients with ESRD would be clinically useful and that interventions targeting avoidant coping should be investigated. In this commentary, the current authors note that Wolf and Mori's (2009) finding was based on a sample where only 23 patients died over 9 years, and results were evaluated using a Cox regression model with six covariates, a ratio of only 3.8 outcome events per variable--well below recommended thresholds (e.g., Babyak, 2004; Peduzzi et al., 1995). Beyond this, the model did not assess key variables typically included in studies of ESRD outcomes (e.g., transplant status, cardiovascular problems). Wolf and Mori chose to focus on the relationship between avoidant coping and mortality. However, bivariable correlations and confidence intervals reported or generated from data in the article showed that the association between avoidant coping and mortality (r=.28, 95% CI [.03 to .50]) was not significantly stronger than two other correlations that were in the opposite direction of what would have been expected: hypertension with mortality (r=-.39, 95% CI [-.15 to -.59]) and medication compliance with mortality (r=.16, 95% CI [-.10 to .40]). Oddities of this sort are often encountered in very small datasets and, as a result, stronger evidence must be accrued from larger, more robust samples.