Temporal data mining for adverse events following immunization in nationwide Danish healthcare databases

Abstract

Background: A rarely used opportunity in pharmacovigilance is data mining for adverse drug reactions (ADRs) in population-based healthcare databases.

Objective: To evaluate the potential of data mining for ADRs in the nationwide Danish healthcare databases. We specifically considered hospital contacts following measles, mumps and rubella (MMR) immunization.

Methods: We constructed a cohort consisting of all children born in Denmark from 1995 to 2007 (n = 918,831) with individual-level linked data on childhood vaccinations and hospital contacts from the nationwide Danish healthcare databases. We applied a cohort-based data mining methodology to compare the observed versus the expected incidence of adverse event in different time periods relative to immunization. With this approach we evaluated temporal associations between MMR immunization and 5915 different diagnoses occurring in the cohort. In order to evaluate the ability of our approach to detect signals, we singled out a set of four adverse events previously recognized as being associated with the MMR vaccine.

Results: We were able to link a total of 3,162,251 hospital contacts and 5915 different diagnoses to the children in the cohort. Previously recognized temporal associations between adverse events (febrile convulsions, idiopathic thrombocytopenic purpura, lymphadenopathy and rash) and MMR immunization were identified in the Danish databases by our method.

Conclusions: Data mining in the Danish population-based healthcare databases provides adequate ability to detect adverse events. Pharmacovigilance using electronic healthcare databases holds potential as an important supplement to traditional pharmacovigilance.