Comparison of Multiple Prediction Models for Ambulation Following Spinal Cord Injury

The study compares three different models to predict whether someone will be able to walk after a spinal cord injury. The models used are logistic regression, neural networks, and rough sets. The models are tested using data from a large database of spinal cord injury patients. The goal is to see which model is best at predicting ambulation at the time of discharge from rehabilitation. The study looks at various factors available at the time of admission to the hospital, such as the type of injury, age, and other medical information, to see how well they predict walking ability.

• 1
  All models performed well in predicting ambulation, with areas under the ROC curve in the 0.88-0.91 range.
• 2
  The neural network and logistic regression models performed statistically better than the rough sets model.
• 3
  The logistic regression model was considered the method of choice due to its good performance, calibration, and potential insight into variable relevance.