A simplified clinical prediction rule for prognosticating independent walking after spinal cord injury: a prospective study from a Canadian multicenter spinal cord injury registry

Abstract

A simplified predictive model with similar accuracy to a more complex model for predicting independent walking was created, which improves utility in a clinical setting. Such models will allow clinicians to better predict the prognosis of ambulation in individuals who have sustained a traumatic SCI.

Highlighting discrepancies in walking prediction accuracy for patients with traumatic spinal cord injury: an evaluation of validated prediction models using a Canadian Multicenter Spinal Cord Injury Registry

“…When comparing individuals with AIS A+D with those with AIS B+C, there was a considerably higher degree of prognostication present with the AIS A+D cohort, in keeping with our hypothesis. These results were similarly reflected utilizing the simplified, three-variable model from Hicks et al [7], collectively suggesting that current prediction models have achieved erroneously high predictive accuracies by combining all tSCI patients with all severities of injury into a single cohort.…”

“…These prognostic variables were age (dichotomized at 65 years old), motor scores of the quadriceps femoris (L3) and gastrocsoleus (S1) muscles, and light touch sensation of the corresponding L3 and S1 dermatomes. Using the cohort from Hicks et al [7] (see their paper for full cohort details), we tested the van Middendorp and Hicks models in the individual AIS categories of A, B, C, and D patients, as well as in the subgroups AIS A+D and AIS B+C patient categories. For each model, classification accuracy, sensitivity, specificity, and the area under the receiver-operating curve (AUC, a measure of diagnostic performance via confidence intervals) were calculated.…”

“…This prediction model was initially tested on European data, and it was externally validated in the United States [5] and Australia [6]. Most recently, Hicks et al [7] were able to validate this five-variable predictive rule using data from the Rick Hansen Spinal Cord Injury Registry (RHSCIR), a Canadian multicenter database [8]. Furthermore, they were able to generate an abridged three-variable model to predict walking ability with a similar level of accuracy based on an individual's Functional Independence Measure (FIM) score [9,10] at 1-year postinjury.…”

“…Despite the reported success of recent prognostic models, it is important to note that they were developed for all severities of tSCI as a collective whole, from the most severe injuries (AIS [American Spinal Injury Association] Impairment Scale A) to the least severe injuries (AIS D). Hicks et al [7] observed that the sample populations in many of these studies consisted predominantly of AIS A+D patients, likely biasing their models' predictive abilities. Our objective was to examine the effect of such a bias on the prognostic accuracy of the model in those whose outcomes are least clinically predictable, patients with AIS B +C injuries, for whom outcome prediction is most useful.…”

“…Our hypothesis is that the prognostication of AIS B+C patients is less accurate than that of AIS A+D, and that former models achieve high predictive accuracy by means of sampling bias. As such, we aim to examine and compare the predictive accuracy of the van Middendorp et al [4] five-variable and Hicks et al [7] three-variable prediction models (hereafter referred to as "van Middendorp model" and "Hicks model") in the individual AIS classifications, that is, AIS A, B, C, and D using the Hicks cohort. Furthermore, we will compare and contrast the predictive ability of these two models in the subgroups of AIS A+D patients as well as AIS B+C patients.…”