Accuracy of an Algorithm in Predicting Upper Limb Functional Capacity in a United States Population

Area of Science:

• Neurology
• Rehabilitation Medicine
• Clinical Prediction Modeling

Background:

• Stroke often results in upper limb (UL) paresis, impacting functional recovery.
• Accurate prediction of UL capacity is crucial for tailoring rehabilitation strategies.
• Current prediction methods may lack precision or rely on extensive assessments.

Purpose of the Study:

• To evaluate the accuracy of a clinical measures-only algorithm for predicting UL capacity in first-ever stroke patients in the US.
• To test the hypothesis that algorithm accuracy would range between 70%-80%.

Main Methods:

• Secondary analysis of a prospective, observational, longitudinal cohort study.
• Included 49 adults with first-ever stroke and UL paresis within 48 hours.
• Two assessments: 1 week and 12 weeks post-stroke, comparing predicted vs. observed UL capacity.

Main Results:

• Overall algorithm accuracy was 61% (weighted κ=62%), significantly better than chance (25%).
• High sensitivity for Excellent (95%) and Poor (81%) categories; high specificity for Good (82%), Limited (98%), and Poor (95%) categories.
• Positive and negative predictive values were high across categories, indicating clinical utility.

Conclusions:

• The algorithm demonstrates moderate to high accuracy in predicting UL capacity 3 months post-stroke.
• Clinical measures-only algorithm shows utility in US healthcare settings for stroke recovery.
• Further validation may enhance its role in guiding rehabilitation decisions.