Assessing misclassification of vaccination status: Implications for studies of the safety of the childhood immunization schedule

Background To address public concern about the safety of the childhood immunization schedule, the Institute of Medicine recommended observational studies comparing adverse health outcomes of fully vaccinated children to children under-vaccinated due to parental choice. Misclassification of vaccination status could bias such studies. Objective To assess risk of misclassification of vaccination status within the Vaccine Safety Datalink (VSD). Design/methods A retrospective cohort study was conducted in three phases. In phase 1, electronic health record (EHR) data were used to identify patterns of under-vaccination during the first 24 months of life potentially due to parental choice. In phase 2, a random sample of records of under-vaccinated children was manually reviewed. In phase 3, a separate sample of parents were surveyed to assess whether EHR data accurately reflected their child's vaccination status. Phases 1 and 2 were conducted at 6 VSD sites, phase 3 at 1 site. Results The study cohort included 361,901 children born 2004 through 2012. By 24 months of age, 198,249 (54.8%) were fully vaccinated with no delays, 84,698 (23.4%) experienced delays but were fully vaccinated by 24 months of age, 4865 (1.3%) received no vaccines, 3789 (1.0%) delayed starting vaccination until ≥4 months of age, 4781 (1.3%) had consistent vaccine-limiting (≤2 vaccines per visit), and the remaining 65,519 (18.1%) were missing vaccine series or doses. When a diagnosis code for vaccine refusal was present in EHR data, encounter notes confirmed vaccine refusal as the reason for under-vaccination for nearly 100% of sampled records. Parent surveys confirmed these findings. Parents of under-vaccinated children were more likely to report visiting an alternative medical provider than parents of fully vaccinated children. Conclusions Specific groups of children, under-vaccinated due to parental choice, can be identified with relatively low likelihood of misclassification of vaccination status using EHR-based vaccine data and diagnosis codes.