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

Area of Science:

• Vaccinology
• Epidemiology
• Health Informatics

Background:

• Post-licensure vaccine safety surveillance frequently uses automated systems to detect adverse events like seizures.
• International Classification of Diseases, Ninth Revision (ICD-9) codes are commonly employed for this automated screening.
• The accuracy of these codes in identifying true seizure events, particularly following vaccination, requires careful evaluation.

Purpose of the Study:

• To assess the positive predictive value (PPV) of ICD-9 codes for identifying seizure visits in young children.
• To compare the PPV of ICD-9 seizure codes across different healthcare settings (emergency department, outpatient, inpatient).
• To evaluate the accuracy of ICD-9 codes for seizure detection within 30 days after pneumococcal vaccination.

Main Methods:

• A retrospective study analyzed data from seven managed care organizations (2000-2005).
• Identified seizure visits using ICD-9 codes in children aged 6 weeks to 23 months post-pneumococcal vaccination.
• Randomly selected visits for medical record review to confirm true seizure events and calculate PPV by diagnostic setting.

Main Results:

• The PPV of ICD-9 codes for seizures was highest in the emergency department (97%) and inpatient (64%) settings.
• In outpatient settings, the PPV was low: 16% within 30 days post-vaccination and 2% on the same day.
• An estimated 77% of confirmed seizure events occurred in ED or inpatient settings.

Conclusions:

• ICD-9 codes for seizures have high PPV in emergency and inpatient settings, making them reliable for vaccine safety studies.
• The low PPV in outpatient settings suggests this data source may introduce bias in vaccine safety analyses.
• Restricting analyses to ED and inpatient data can improve the efficiency and reduce bias in vaccine safety studies monitoring for seizures.