Use of the odds ratio to translate neuropsychological test scores into real-world outcomes: From statistical significance to clinical significance

Abstract

Standard parametric tests generate p values and effect sizes, but often these are difficult to translate into real-world outcomes. In this study, the odds ratio was applied to neuropsychological testing and was compared to parametric approaches. Participants were 26 community-dwelling adults with possible or probable Alzheimer's disease and 25 matched healthy community-dwelling volunteers. Odds ratios were computed to estimate the probability of concurrent diagnosis given neuropsychological performance level. Odds ratios discriminated the groups at magnitudes that could not be discerned from t-test significance tables. These values were compared to sensitivity, specificity, and overall accuracy. Clinical and research applications and implications were addressed.