Retrospective collection of 90-day modified Rankin Scale is accurate.

Background: The 90-day modified Rankin Scale is a widely used outcome after stroke but is sometimes hard to ascertain due to loss to follow-up. Missing outcomes can result in biased and/or inefficient estimates in clinical trials. The aim of this study is to assess the validity of acquiring the 90-day modified Rankin Scale at a later point of time when the patient has been lost at 90 days to impute the missing value.

Methods: Participants who had prospectively completed a 90-day modified Rankin Scale questionnaire on their own in the Benefits of Stroke Treatment Using a Mobile Stroke Unit study were randomly interviewed to recall the 90-day modified Rankin Scale at 6, 9, or 12 months after hospital discharge over the phone. Concordance between the two scores was assessed using kappa and weighted kappa statistics. Logistic regression was used to identify factors associated with inconsistent reporting of the 90-day modified Rankin Scale.

Results: Substantial agreement was observed between in-the-moment and retrospective 90-day modified Rankin Scale recalled at 6, 9, or 12 months (weighted kappa = 0.93, 95% confidence interval: 0.89-0.98; weighted kappa = 0.93, 95% confidence interval: 0.85-1.00 and weighted kappa = 0.89, 95% confidence interval: 0.82-0.95, respectively).

Conclusion: Retrospective recall of 90-day modified Rankin Scale at a later time point is a valid means to impute missing data in stroke clinical trials.