Predict the Medicare Functional Classification Level (K-level) using the Amputee Mobility Predictor in people with unilateral transfemoral and transtibial amputation: A pilot study.

Background: While Amputee Mobility Predictor scores differ between Medicare Functional Classification Levels (K-level), this does not demonstrate that the Amputee Mobility Predictor can accurately predict K-level.

Objectives: To determine how accurately K-level could be predicted using the Amputee Mobility Predictor in combination with patient characteristics for persons with transtibial and transfemoral amputation.

Study Design: Prediction.

Method: A cumulative odds ordinal logistic regression was built to determine the effect that the Amputee Mobility Predictor, in combination with patient characteristics, had on the odds of being assigned to a particular K-level in 198 people with transtibial or transfemoral amputation.

Results: For people assigned to the K2 or K3 level by their clinician, the Amputee Mobility Predictor predicted the clinician-assigned K-level more than 80% of the time. For people assigned to the K1 or K4 level by their clinician, the prediction of clinician-assigned K-level was less accurate. The odds of being in a higher K-level improved with younger age and transfemoral amputation.

Conclusion: Ordinal logistic regression can be used to predict the odds of being assigned to a particular K-level using the Amputee Mobility Predictor and patient characteristics. This pilot study highlighted critical method design issues, such as potential predictor variables and sample size requirements for future prospective research. Clinical relevance This pilot study demonstrated that the odds of being assigned a particular K-level could be predicted using the Amputee Mobility Predictor score and patient characteristics. While the model seemed sufficiently accurate to predict clinician assignment to the K2 or K3 level, further work is needed in larger and more representative samples, particularly for people with low (K1) and high (K4) levels of mobility, to be confident in the model's predictive value prior to use in clinical practice.