Validation of the Dutch clinical prediction rule for ambulation outcomes in an inpatient setting following traumatic spinal cord injury

Abstract

STUDY DESIGN: Retrospective study. OBJECTIVES: To determine the accuracy of a previously described Dutch clinical prediction rule for ambulation outcome in routine clinical practice. SETTING: Adult (⩾18 years) patients who were admitted to the Austin Hospital with a traumatic spinal cord injury between January 2006 and August 2014. METHODS: Data from medical records were extracted to determine the score of the Dutch clinical ambulation prediction rule proposed by van Middendorp et al. in 2011. A receiver-operating characteristics (ROC) curve was generated to investigate the performance of the prediction rule. Univariate analyses were performed to investigate which factors significantly influence ambulation after a traumatic spinal cord injury. RESULTS: The area under the ROC curve (AUC) obtained during the current study (0.939, 95% confidence interval (CI) (0.892, 0.986)) was not significantly different from the AUC from the original Dutch clinical prediction model (0.956, 95% CI (0.936, 0.976)). Factors that were found to have a significant influence on ambulation outcome were time spent in the ICU, number of days hospitalised and injury severity. Age at injury initially showed a significant influence on ambulation however, this effect was not apparent after inclusion of the 24 patients who died due to the trauma (and therefore did not walk after their injuries). CONCLUSION: The Dutch ambulation prediction rule performed similarly in routine clinical practice as in the original, controlled study environment in which it was developed. The potential effect of survival bias in the original model requires further investigation.