Accuracy of Influenza Vaccination Status In the Electronic Medical Record

Misclassification of vaccination status may lead to biased assessments of safety and effectiveness in studies based solely on electronic medical records. Influenza immunization in nontraditional settings (e.g., workplace) may not be captured in patient medical records. To inform studies relying on electronic records, we conducted a study to describe the magnitude of this misclassification of influenza vaccination status.

In a large managed care organization where immunizations are tracked in an electronic registry, a random sample was selected consisting of approximately 500 subjects ages 50-79 with and without a record of influenza vaccination after September 2007. A telephone survey was conducted from February to April 2008 asking these subjects whether they had received a flu shot this season. To evaluate the accuracy of vaccination status in the medical record as compared to the “gold standard” survey responses, we calculated positive and negative predictive values (PPV and NPV), sensitivity, and specificity.

Influenza vaccination status in the medical record had a high PPV and specificity (both >99%). The NPV was 79% and sensitivity was 77%. Of those who reported receiving influenza vaccine, 89% received it within their health care system and 11% in nontraditional settings. Of those whose record indicated they were unvaccinated, 10% reported receiving influenza vaccine within their health care system and 11% in nontraditional settings.

These data suggest that influenza immunization studies based solely on registry information may rely on exposure status that is misclassified, leading to biased estimates of effect.