International standards for neurological classification of spinal cord injury: classification skills of clinicians versus computational algorithms.

Abstract

This is a retrospective analysis. The objective of this study was to describe and quantify the discrepancy in the classification of the International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI) by clinicians versus a validated computational algorithm. European Multicenter Study on Human Spinal Cord Injury (EMSCI). Fully documented ISNCSCI data sets from EMSCI's first years (2003-2005) classified by clinicians (mostly spinal cord medicine residents, who received in-house ISNCSCI training by senior SCI physicians) were computationally reclassified. Any differences in the scoring of sensory and motor levels, American Spinal Injury Association Impairment Scale (AIS) or the zone of partial preservation (ZPP) were quantified. Four hundred and twenty ISNCSCI data sets were evaluated. The lowest agreement was found in motor levels (right: 62.1%, P=0.002; left: 61.8%, P=0.003), followed by motor ZPP (right: 81.6%, P=0.74; left 80.0%, P=0.27) and then AIS (83.4%, P=0.001). Sensory levels and sensory ZPP showed the best concordance (right sensory level: 90.8%, P=0.66; left sensory level: 90.0%, P=0.30; right sensory ZPP: 91.0%, P=0.18; left sensory ZPP: 92.2%, P=0.03). AIS B was most often misinterpreted as AIS C and vice versa (AIS B as C: 29.4% and AIS C as B: 38.6%). Most difficult classification tasks were the correct determination of motor levels and the differentiation between AIS B and AIS C/D. These issues should be addressed in upcoming ISNCSCI revisions. Training is strongly recommended to improve classification skills for clinical practice, as well as for clinical investigators conducting spinal cord studies. This study is partially funded by the International Foundation for Research in Paraplegia, Zurich, Switzerland.