Developing artificial neural network models to predict functioning one year after traumatic spinal cord injury

Model System:

Reference Type:

Accession No.:

Journal:

Year, Volume, Issue, Page(s):

Publication Website:

Abstract:

Study developed mathematical models for predicting level of independence with specific functional outcomes 1 year after discharge from inpatient rehabilitation for spinal cord injury (SCI). Specifically, the study compared the accuracy and performance of statistical analyses using artificial neural networks (ANNs) and logistic regression methods to estimate longer-term ambulation and non-ambulation outcomes. Participants were 3,142 subjects with traumatic SCI who contributed data to the National SCI Model Systems Database longitudinal outcomes studies. Outcome measures included self-reported ambulation ability and Functional Independence Measure-derived indices of level of assistance required for self-care activities (bed-chair transfers, bladder and bowel management, eating, and toileting). Results indicated that the models for predicting ambulation status were highly accurate (>85 percent case classification accuracy; areas under the receiver operating characteristic curve were between .86 and .90). Models for predicting non-ambulation outcomes were moderately accurate (76 to 86 percent case classification accuracy; areas under the receiver operating characteristic curve between .70 and .82). The performance of models generated by ANNs closely paralleled the performance of models analyzed using logistic regression constrained by the same independent variables. With further prospective validation, such predictive models may allow clinicians to use data available at the time of admission to inpatient SCI rehabilitation to accurately predict longer-term ambulation status, and whether individual patients are likely to perform various self-care activities with or without assistance from another person.

Author(s):

Participating Centers: