Category: Limb Restoration Rehabilitation; Clinical Practice (assessment, diagnosis, treatment, knowledge translation/EBP, implementation science, program development)
Objective : Eligibility for many prosthetic components is based on an individual's assigned functional ambulation level. Recent efforts to collect big data on prosthetic outcomes has provided the ability to develop predictive models regarding functional ambulation potential. This effort describes the development of a model to assist in determining whether an individual will function at either a limited community/household ambulator (LA) or an unlimited community ambulator (UA)
Design : Classification and Regression Tree (CART) Analysis
Setting : Multicenter Free-Standing Prosthetic Care Facilities
Participants (or Animals, Specimens, Cadavers) : Convenience sample of records taken from 2,770 lower limb prostheses users
Interventions : Records consisted of general patient demographics including age, weight and amputation etiology, assigned Medicare K-level, functional mobility reported as the T-score on the Prosthetic Limb Users Survey of Mobility (PLUS-M) and comorbid health reported using the Functional Comorbidity Index.
Twenty percent of the original sample (n=554, age 57.0±14.7 years) were selected to build the predictive model. The remaining 2216 patients (age 57.2±14.5 years) were used to test the model.
CART analysis was implemented with Matlab® to develop a classification tree with functional potential as the predicted dependent variable.
Main Outcome Measure(s) : Classification tree for determining UA versus LA
The resultant classification tree had classification accuracy of 87.4% (SEM 1.4%) for the training sample and 81.6% (SEM 0.82%) for the testing sample. Twelve of the nodes were found to have class assignment significantly different from a random assignment (p
The collection of large volumes of data within rehabilitation affords the ability to implement various predictive analytics models. The current study model used PLUS-M T-scores, age, cause of amputation, and body weight to provide probability of functional potential.
SHANE WURDEMAN– Director of Clinical Research, Hanger Clinic- Dept of Clinical and Scientific Affairs, Sugar Land, Texas
Phillip Stevens– Director, Department of Clinical and Scientific Affairs, Hanger Clinic, Salt Lake City, Utah
James H Campbell– Chief Clinical Oficer, Hanger Inc, Austin, Texas