How To Detect Effects: Statistical Power and Evidence-Based Practice in Occupational Therapy Research

Abstract

Abstract The findings from 30 research investigations examining the effectiveness of occupational therapy interventions were reviewed and analyzed. The statistical conclusion validity was determined by computing post hoc power coefficients for the statistical hypothesis tests included in the examined studies. Data analysis revealed the median power values to detect small, medium, and large effect sizes were .09, .33, and .66, respectively. These results suggest a high probability of Type II errors in the sample of occupational therapy intervention research examined. In practical terms, this means the intervention produced a potentially useful treatment effect, but the effect was not detected as significant. Examples are provided that illustrate how low statistical power contributes to increases in Type II errors and inhibits the development of consensus through replication in the research literature. The presence of low-power studies with high rates of false negative findings prevents the establishment of guidelines for evidence-based practice and impedes the scientific progress of rehabilitation professions such as occupational therapy.