Statement of purpose International Classification of Diseases, Tenth Revision (ICD-10), Clinical Modification codes are used to identify Veterans with traumatic brain injury (TBI) for surveillance and research. We examined the validity of using ICD-10 codes in Veterans Health Administration (VA) administrative data to identify Veterans with TBI.

Methods/Approach Clinician-confirmed TBI cases and non-cases after October 1, 2016 were extracted from the VA Comprehensive TBI Evaluation (CTBIE) database, a standardized TBI evaluation tool, and used as the criterion standard. Accessing VA administrative healthcare data, we compared three series of TBI-related ICD-10 codes recommended by the Centers for Disease Control and Prevention (CDC), the Defense and Veterans Brain Injury Center (DVBIC), and the VA TBI Program Office (VA) to clinician- confirmed TBI diagnoses. Sensitivity, specificity, and concordance were calculated. The validity of distinguishing TBI severity (mild versus moderate/severe), using DVBIC-recommended codes, was also examined.

Results Among 29,767 Veterans who received a CTBIE between 2016 and 2019, 20,670 (69.4%) received a clinician-confirmed TBI diagnosis and 20,760 (69.7%) had at least one TBI-related ICD-10 code assigned from any series. The CDC series had low sensitivity (40.8%) but high specificity (80.7%; 53.0% concordance). The DVBIC series had moderate sensitivity and specificity (71.7% and 63.9%, respectively; 69.3% concordance). The VA series had high sensitivity (81.8%), but moderate specificity (58.8%; 74.7% concordance). The sensitivity of DVBIC-recommended codes to distinguish TBI severity was high (83.4%), but specificity was low (21.5%; 76.4% concordance).

Conclusion We found that all three series of ICD-10 codes for TBI resulted in considerable misclassification of clinician-confirmed TBI cases and non-cases. The potential misclassification in studies using ICD-10 codes to examine TBI prevalence and outcomes among Veterans should inform surveillance and research efforts.

Significance Examining the validity of methods used for enumerating TBI cases allows researchers to interpret and translate their findings in the context of potential error.