Improving the Assessment of Outcomes in Stroke

Sixty-three patients with stroke 6 to 24 months previously were interviewed and graded independently on the modified Rankin Scale by 2 observers. These observers then underwent training in use of a structured interview for the scale that covered 5 areas of everyday function. Eight weeks after the first assessment, the same observers reassessed 58 of these patients using the structured interview.

Interrater reliability was measured with the kappa statistic (weighted with quadratic weights). For the scale applied conventionally, overall agreement between the 2 raters was 57% (kappa(w)=0.78); 1 rater assigned significantly lower grades than the other (P=0.048). On the structured interview, the overall agreement between raters was 78% (kappa(w)=0.93), and there was no overall difference between raters in grades assigned (P=0.17). Rankin grades from the conventional assessment and the structured interview were highly correlated, but there was significantly less disagreement between raters when the structured interview was used (P=0.004).