INTRODUCTION: Post-licensure surveillance of adverse events following vaccination or prescription drug use often relies on electronic healthcare data to efficiently detect and evaluate safety signals. The accuracy of seizure-related diagnosis codes in identifying true incident seizure events in vaccine safety studies is influenced by factors such as clinical setting of diagnosis and age. To date, most studies of post-vaccination seizure have focused on pediatric populations. More information is needed on how well seizure can be identified in adults and children using algorithms that rely on electronic healthcare data. METHODS: This validation study was part of a larger safety study of influenza vaccination during the 2009-2010 and 2010-2011 influenza seasons. Children and adults receiving influenza vaccination were drawn from an administrative claims database of a large United States healthcare insurer. Potential seizure events were identified using an algorithm of ICD-9 diagnosis codes associated with an emergency department (ED) visit or hospitalization within pre-specified risk windows following influenza vaccination. Seizure events were confirmed through medical record review. The positive predictive value (PPV) of the algorithm was calculated within each diagnostic setting and stratified by age group, ICD-9 code group, and sex. RESULTS: Review confirmed 113 out of 176 potential seizure events. The PPVs were higher in the ED setting (93.9%) than in the inpatient setting (38.3%). The PPVs by age varied within the ED setting (98.2% in <7 years, 76.9% in 7-24 years, 92.3% in ≥25 years) and within the inpatient setting (64.7% in <7 years, 33.3% in 7-24 years, 32.3% in ≥25 years). CONCLUSIONS: Our algorithm for identification of seizure events using claims data had a high level of accuracy in the emergency department setting in young children and older adults and a lower, but acceptable, level of accuracy in older children and young adults.
INTRODUCTION: Post-licensure surveillance of adverse events following vaccination or prescription drug use often relies on electronic healthcare data to efficiently detect and evaluate safety signals. The accuracy of seizure-related diagnosis codes in identifying true incident seizure events in vaccine safety studies is influenced by factors such as clinical setting of diagnosis and age. To date, most studies of post-vaccination seizure have focused on pediatric populations. More information is needed on how well seizure can be identified in adults and children using algorithms that rely on electronic healthcare data. METHODS: This validation study was part of a larger safety study of influenza vaccination during the 2009-2010 and 2010-2011 influenza seasons. Children and adults receiving influenza vaccination were drawn from an administrative claims database of a large United States healthcare insurer. Potential seizure events were identified using an algorithm of ICD-9 diagnosis codes associated with an emergency department (ED) visit or hospitalization within pre-specified risk windows following influenza vaccination. Seizure events were confirmed through medical record review. The positive predictive value (PPV) of the algorithm was calculated within each diagnostic setting and stratified by age group, ICD-9 code group, and sex. RESULTS: Review confirmed 113 out of 176 potential seizure events. The PPVs were higher in the ED setting (93.9%) than in the inpatient setting (38.3%). The PPVs by age varied within the ED setting (98.2% in <7 years, 76.9% in 7-24 years, 92.3% in ≥25 years) and within the inpatient setting (64.7% in <7 years, 33.3% in 7-24 years, 32.3% in ≥25 years). CONCLUSIONS: Our algorithm for identification of seizure events using claims data had a high level of accuracy in the emergency department setting in young children and older adults and a lower, but acceptable, level of accuracy in older children and young adults.
Authors: Hajime Uno; Debra P Ritzwoller; Angel M Cronin; Nikki M Carroll; Mark C Hornbrook; Michael J Hassett Journal: JCO Clin Cancer Inform Date: 2018-12
Authors: Richard L Morrow; Colin R Dormuth; Michael Paterson; Muhammad M Mamdani; Tara Gomes; David N Juurlink Journal: BMJ Open Date: 2019-03-13 Impact factor: 2.692