Are the Items of the Starkstein Apathy Scale Fit for the Purpose of Measuring Apathy Post-stroke?

Importance: Given the importance of apathy for stroke, we felt it was time to scrutinize the psychometric properties of the commonly used Starkstein Apathy Scale (SAS) for this purpose. Objectives: The objectives were to: (i) estimate the extent to which the SAS items fit a hierarchical continuum of the Rasch Model; and (ii) estimate the strength of the relationships between the Rasch analyzed SAS and converging constructs related to stroke outcomes. Methods: Data was from a clinical trial of a community-based intervention targeting participation. A total of 857 SAS questionnaires were completed by 238 people with stroke from up to 5 time points. SAS has 14 items, rated on a 4-point scale with higher values indicating more apathy. Psychometric properties were tested using Rasch partial-credit model, correlation, and regression. Items were rescored so higher scores are interpreted as lower apathy levels. Results: Rasch analysis indicated that the response options were disordered for 8/14 items, pointing to unreliability in the interpretation of the response options; they were consequently reduced from 4 to 3. Only 9/14 items fit the Rasch model and therefore suitable for creating a total score. The new rSAS was deemed unidimensional (residual correlations: < 0.3), reasonably reliable (person separation index: 0.74), with item-locations uniform across time, age, sex, and education. However, 30% of scores were > 2 SD above the standardized mean but only 2/9 items covered this range (construct mistargeting). Apathy (rSAS/SAS) was correlated weakly with anxiety/depression and uncorrelated with physical capacity. Regression showed that the effect of apathy on participation and health perception was similar for rSAS/SAS versions: R(2) participation measures ranged from 0.11 to 0.29; R(2) for health perception was ∼0.25. When placed on the same scale (0–42), rSAS value was 6.5 units lower than SAS value with minimal floor/ceiling effects. Estimated change over time was identical (0.12 units/month) which was not substantial (1.44 units/year) but greater than expected assuming no change (t: 3.6 and 2.4). Conclusion: The retained items of the rSAS targeted domains of behaviors more than beliefs and results support the rSAS as a robust measure of apathy in people with chronic stroke.