# The Accuracy of the PREP2 Prediction Tool for Upper Limb Outcomes After Stroke as Part of Routine Clinical Care

## TL;DR

This study evaluates the accuracy of the PREP2 tool for predicting upper limb recovery after stroke in clinical practice, finding it highly accurate for excellent and poor outcomes but less so for good outcomes.

## Contribution

The study validates the PREP2 prediction tool in routine clinical care and identifies limitations in predicting 'Good' outcome categories.

## Key findings

- Overall accuracy of PREP2 in clinical practice was 66%, with highest accuracy for Excellent (80%) and Poor (100%) categories.
- Prediction accuracy for Good outcomes was 67% without TMS and 27% with TMS, with finger extension differentiating outcomes.

## Abstract

The Predict REcovery Potential-2 (PREP2) prediction tool uses clinical assessments and transcranial magnetic stimulation (TMS) within 1 week post-stroke to predict individuals’ upper limb functional outcome at 3 months (3M) post-stroke. PREP2 was successfully implemented in clinical care at Auckland City Hospital, New Zealand in 2017.

The primary aim was to evaluate the accuracy of PREP2 predictions made by clinicians during routine clinical care, with a threshold of 70% accuracy for validation. A secondary aim was to identify new baseline predictors that could increase PREP2 accuracy.

Eighty-three patients who received PREP2 predictions were recruited within 1 week of stroke and had their upper limb outcome assessed at 3M post-stroke with the Action Research Arm Test. Cognition and sensation were evaluated within 1 week of stroke.

Overall accuracy of the PREP2 prediction tool in clinical practice was 66% (95% confidence interval, 55%-76%). Accuracy was highest for the Excellent (80%) and Poor (100%) categories and lowest for the Good category (36%). Prediction accuracy for Good outcomes was 67% for patients who did not require TMS and 27% for patients who did. Finger extension differentiated participants predicted to have a Good outcome using TMS who did and did not have a favorable upper limb outcome.

Excellent and Poor predictions are highly accurate when used in clinical practice, however the full PREP2 tool is not yet validated in clinical practice. Future studies with larger samples could investigate additional measures to enhance accuracy of the Good prediction category.

Clinical trial registration number ACTRN12619000225112, https://anzctr.org.au/Trial/Registration/TrialReview.aspx?ACTRN=12619000225112.