Accuracy of physical therapists' early predictions of upper-limb function in hospital stroke units: the EPOS Study

Abstract

Background: Early prediction of outcome after stroke is becoming increasingly important, as most patients are discharged from hospital stroke units within several days after stroke.

Objectives: The primary purposes of this study were: (1) to determine the accuracy of physical therapists' predictions at hospital stroke units regarding upper-limb (UL) function, (2) to develop a computational prediction model (CPM), and (3) to compare the accuracy of physical therapists' and the CPM's predictions. Secondary objectives were to explore the impact of timing on the accuracy of the physical therapists' and CPM's predictions and to investigate the direction of the difference between predicted and observed outcomes. Finally, this study investigated whether the accuracy of physical therapists' predictions was affected by their experience in stroke rehabilitation.

Design: A prospective cohort study was conducted.

Methods: Physical therapists made predictions at 2 time points-within 72 hours after stroke onset (T(72h)) and at discharge from the hospital stroke unit (Tdischarge)-about UL function after 6 months in 3 categories, derived from the action research arm test. At the same time, clinical variables were measured to derive a CPM. The accuracy of the physical therapists' and CPM's predictions was evaluated by calculating Spearman rank correlation coefficients (r(s)) between predicted and observed outcomes.

Results: One hundred thirty-one patients and 20 physical therapists participated in the study. For the T(72h) assessment, the rs value between predicted and observed outcomes was .63 for the physical therapists' predictions and .75 for the CPM's predictions. For the Tdischarge assessment, the rs value for the physical therapists' predictions improved to .75, and the rs value for the CPM's predictions improved slightly to .76.

Limitations: Physical therapists administered a test battery every 3 days, which may have enhanced the accuracy of prediction.

Conclusions: The accuracy of the physical therapists' predictions at T(72h) was lower than that of the CPM's predictions. At Tdischarge, the physical therapists' and CPM's predictions are about equally accurate.