Case ascertainment in pediatric traumatic brain injury: challenges in using the NEISS

Abstract

Primary objective: To evaluate mild traumatic brain injury (TBI) case ascertainment in the US National Electronic Injury Surveillance System (NEISS).

Research design: A sample of NEISS injury cases was analyzed. Two trained researchers independently reviewed diagnostic codes based on the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). The sensitivity and specificity of two NEISS TBI definitions were evaluated, and Kappa coefficients were calculated to evaluate the agreement between NEISS TBI definitions and the standard TBI definition based on the ICD-9-CM.

Results: Our sampling procedures provided a total of 1,018 NEISS cases, of which 880 were injury cases and 138 were non-injury cases. Out of the injury cases, 54 cases were defined as mild TBI cases based on the ICD-9-CM diagnosis codes. Compared with non-TBI injury cases, a significantly higher percentage of TBI cases (33.3% versus 15.4%) were infants less than 1 year old and TBI cases were more likely to be admitted for hospitalization. If TBI was identified using NEISS definition 1 (diagnosis code 52 for 'concussion'), the sensitivity was only 38.9% (95% confidence intervals (CI) = 25.9-51.9%). However, if TBI was identified using NEISS definition 2 (diagnosis code 52 for 'concussion' or 62 for 'internal organ injury' and 'head' was the affected body part), the sensitivity increased dramatically to 79.6% (95% CI = 68.9-90.4%). The Kappa coefficient for overall agreement between the NEISS and ICD-9-CM system was 0.528 (95% CI = 0.393-0.664) for NEISS definition 1 and 0.807 (95% CI = 0.723-0.892) for NEISS definition 2.

Conclusions: The current approach of pediatric TBI case ascertainment in the NEISS faces some challenges in identifying mild TBI. Future research efforts are needed to refine TBI case ascertainment in the NEISS.