Interpretation of a Self-Esteem Subscale for Erectile Dysfunction by Cumulative Logit Model

Abstract

A content-based interpretation of a health status measure uses an item, along with its response choices, internal to the measure to understand the meaning of its scores. We applied a cumulative logit model to content-based interpretation of a validated self-esteem subscale (four items) for men with erectile dysfunction (score range: 0 to 100, where higher scores indicate a more favorable response). Data were obtained from a nontreatment cross-sectional study with 98 men with erectile dysfunction and 94 controls. The ordinal response item “I had good self-esteem” over the past 4 weeks (1 = Almost never/never, 2 = A few times, 3 = Sometimes, 4 = Most times, 5 = Almost always/Always) was regressed on the self-esteem subscale score to which the item belongs. The proportional odds assumption was not refuted (P = .085) and supported graphically. At relatively high self-esteem scores (65+), the estimated probability of reporting good self-esteem (ie, good self-esteem most of the time or almost always/always) was 68% or more. For a 10-point increase (eg, 65 to 75), the odds of reporting good self-esteem increased slightly more than threefold (3.2). A content-based interpretation using a cumulative logit model can increase sensitivity, enhance meaning, and provide succinct and simple interpretation to scores of a health status measure.

A content-based interpretation of health status measures using the cumulative logit model can increase sensitivity and enhance meaning of scores on a health status measure, as was illustrated in this article with a self-esteem subscale for men with erectile dysfunction. The cumulative logit model capitalizes on the ordered levels of a response variable and provides a parsimonious model-based approach to elucidate and understand scores from a health status measure. In particular, when the proportional odds assumption is met and the model supports the data, the cumulative logit model is an attractive ordinal logistic regression model that allows for succinct and simple interpretation in designing studies, evaluating interventions, informing consumers and health policy makers, and providing information for formulary and reimbursement decisions.