Developing Artificial Neural Network Models to Predict Functioning One Year After Traumatic Spinal Cord Injury

• Published in: Archives of physical medicine and rehabilitation Vol. 97; no. 10; pp. 1663 - 1668.e3
• Main Authors: Belliveau, Timothy, Jette, Alan M., Seetharama, Subramani, Axt, Jeffrey, Rosenblum, David, Larose, Daniel, Houlihan, Bethlyn, Slavin, Mary, Larose, Chantal
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.10.2016
• Subjects: Activities of Daily Living; Decision support techniques; Female; Humans; Logistic Models; Male; Medical informatics; Models, Theoretical; Neural Networks (Computer); Physical Medicine and Rehabilitation; Physical Therapy Modalities; Recovery of Function; Rehabilitation; Reproducibility of Results; Retrospective Studies; Self Care; Spinal cord injuries; Spinal Cord Injuries - rehabilitation; Walking - physiology; ANN; Activities of daily living; AIS; SCIMS; AUC; Decision support techniques; ASIA; NLR; PLR; SCI; Spinal cord injuries; Rehabilitation; Medical informatics; positive likelihood ratio; spinal cord injury; artificial neural network; area under the curve; American Spinal Injury Association Impairment Scale; negative likelihood ratio; American Spinal Injury Association
• ISSN: 0003-9993; 1532-821X; 1532-821X
• DOI: 10.1016/j.apmr.2016.04.014

To develop mathematical models for predicting level of independence with specific functional outcomes 1 year after discharge from inpatient rehabilitation for spinal cord injury. Statistical analyses using artificial neural networks and logistic regression. Retrospective analysis of data from the national, multicenter Spinal Cord Injury Model Systems (SCIMS) Database. Subjects (N=3142; mean age, 41.5y) with traumatic spinal cord injury who contributed data for the National SCIMS Database longitudinal outcomes studies. Not applicable. Self-reported ambulation ability and FIM-derived indices of level of assistance required for self-care activities (ie, bed-chair transfers, bladder and bowel management, eating, toileting). Models for predicting ambulation status were highly accurate (>85% case classification accuracy; areas under the receiver operating characteristic curve between .86 and .90). Models for predicting nonambulation outcomes were moderately accurate (76%–86% case classification accuracy; areas under the receiver operating characteristic curve between .70 and .82). The performance of models generated by artificial neural networks closely paralleled the performance of models analyzed using logistic regression constrained by the same independent variables. After further prospective validation, such predictive models may allow clinicians to use data available at the time of admission to inpatient spinal cord injury 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.