Item response theory analysis and properties of decisional conflict scales: findings from two multi-site trials of men with localized prostate cancer

Abstract Background Decisional conflict is associated with decision quality and may affect decision outcomes. In the health sciences literature, the Decisional Conflict Scale is widely used to measure decisional conflict, yet limited research has described the psychometric properties of the Decisional Conflict Scale subscales and of the low literacy version of the scale. The purpose of this secondary data analysis was therefore to examine properties of the original (DCS-12) and low literacy (LL DCS-10) Decisional Conflict Scales using Classical Measurement Theory and Item Response Theory. Methods Data from two multi-site trials of men with prostate cancer were used to analyze the DCS-12, LL DCS-10, and an aggregated DCS-12 dataset in which five response options were aggregated into three. Internal consistency was estimated with Cronbach’s alphas. Subscale correlations were evaluated with Pearson’s correlation coefficient. Item difficulty, item discrimination, and test information were evaluated using Graded Response Modeling (GRM). The likelihood ratio test guided model selection. Results Cronbach’s alphas for the total scales and three of four subscales were ≥ 0.85. Alphas ranged from 0.34–0.57 for the support subscales. Subscale correlations ranged from 0.42–0.71 (P < 0.001). Items on the DCS-12 exhibited the widest range of difficulty. Two items on the support subscale had low to moderate discrimination and contributed little information. Only the DCS-12 was informative across the full range of decisional conflict values. Conclusions Lack of precision in the support subscale raises concerns about subscale validity. The DCS-12 is most capable of discriminating between respondents with high and low decisional conflict. Evaluation of interventions to reduce decisional conflict must consider the above findings.

Abstract licence: CC BY-NC-ND

33 more references, whose wording the paper’s licence keeps private.

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.