Predictive value of seizure ICD-9 codes for vaccine safety research

Post-licensure vaccine safety studies often monitor for seizures using automated screening of ICD-9 codes. This study assessed the positive predictive value (PPV) of ICD-9 codes used to identify seizure visits in children aged 6 weeks to 23 months who were enrolled in seven managed care organizations during January 2000 to December 2005. ICD-9 codes were used to identify visits for seizures in the 0–30-day period following receipt of a pneumococcal vaccine. Visits were stratified by setting of diagnosis (emergency department (ED), outpatient, and inpatient). Review of medical records confirmed whether the visit represented a true acute seizure event. 3233 visits for seizures were identified; 1024 were randomly selected for medical record review and 859 (84%) had records available. The PPV of ICD-9 codes was highest in the ED setting (97%), followed by the inpatient setting (64%). In the outpatient setting, computerized codes for seizures had very low PPV: 16% on days 1–30 following vaccination and 2% for visits on the same day of vaccination. An estimated 77% of true seizures identified were from the ED or inpatient settings. In conclusion, when using ICD-9 codes to identify seizure outcomes, restricting to the ED and inpatient settings of diagnosis may result in less biased preliminary analyses and more efficient vaccine safety studies.