Responder analysis confirms results of a stroke transitional care trial but provides more interpretable results.

OBJECTIVES:Interpreting between-group differences in patient-reported outcome measures can be challenging. Responder analyses, which compare the proportions of patients who achieve a meaningful clinical change, represent a more interpretable approach. We conducted a secondary responder analysis of the Michigan Stroke Transitions Trial (MISTT). STUDY DESIGN AND SETTING:The MISTT randomized 265 patients with stroke to three treatment groups: usual care [UC], social work case management [SWCM], or social work case management plus access to a patient-oriented website [SWCM + website]. Two Patient-Reported Outcomes Measurement and Information System (PROMIS) Global-10 subscales (representing physical and mental health) and 5 additional patient-reported outcomes were collected at baseline and 90-days. Responder analyses were conducted using modified Poisson and linear regression using published minimal important differences. Multiple imputation was used to address missing data. RESULTS:For the PROMIS-10 global physical health subscale, responders were 80% more common in the SWCM + website group compared to the UC group (relative risk = 1.8, 95% confidence interval [CI]: 1.0, 3.1), with a number needed to treat of 7 (95% CI: 3, 112). No significant treatment effects were observed for the PROMIS-10 global mental health subscale. CONCLUSION:Results of this responder analysis were largely consistent with the original trial analysis but have the advantage of presenting treatment effects using more clinically interpretable number needed to treat metrics.