The validity of administrative data to classify patients with spinal column and cord injuries

International Classification of Diseases (ICD) codes are used to document patient morbidity in administrative databases. Although administrative data are used for research purposes, the validity of the data to accurately describe clinical diagnostic information is uncertain. We compared the clinical diagnoses for spinal cord and column injuries from a longitudinal patient registry, the Rick Hansen Spinal Cord Injury Registry (RHSCIR), to the ICD-10 spinal injury codes from the Discharge Abstract Database (DAD) at one institution. There were 603 RHSCIR participants with data describing the spinal cord injury, and 341 had data on the spinal column injury. The validity of DAD data to describe spinal injuries was evaluated using the sensitivity and positive predictive values of specific ICD-10 codes; 5.3% of the spinal column injuries and 10.9% of the spinal cord injuries documented in RHSCIR were missed in data from the DAD using ICD-10 codes. The most problematic spinal column ICD-10 code was the dislocation of the cervical vertebra (S13.1); only 14.0% of the dislocations of the cervical vertebrae in RHSCIR were correctly coded in the DAD. The most problematic spinal cord injury ICD-10 code was the incomplete lesion of the lumbar spinal cord (S34.1X); 66.7% of incomplete lesions of the lumbar spinal cord in RHSCIR were correctly coded in the DAD. The validity of DAD data to code spinal injuries is variable, and cannot be reliably used to classify all types of spinal injuries. Patient registries, such as RHSCIR, should be used if accurate detailed diagnostic data are required.