Validation of self‐reported influenza vaccination in the current and prior season

Self‐reported influenza vaccination is generally accurate for the current season, but the accuracy of self‐report for vaccination in prior seasons is largely unknown. This study evaluated the accuracy of self‐report for current and prior season influenza vaccination among patients with medically attended acute respiratory illness enrolled in a study of influenza vaccine effectiveness during the 2014‐15 influenza season. It demonstrates there is a greater potential for exposure misclassification when prior season vaccinations are ascertained by self‐report. Percent agreement between self‐report and final status was high for both current and prior season vaccination: 97.7% and 93.2%, respectively.


| BACKG ROU N D
Observational studies of influenza vaccine effectiveness (VE) are conducted annually in North America, Europe, and Australia.
Accurate information on influenza vaccination status is important to minimize exposure misclassification and bias. These studies often rely on a combination of self-report and information obtained from medical records to determine vaccination status. In the United States, influenza vaccines are readily available outside the healthcare system, and self-report may be the only practical method to ascertain vaccination status in some research settings. This is especially true for ascertainment of vaccines received in prior seasons.
Vaccine effectiveness may be reduced in persons who are repeatedly vaccinated, and recent VE studies have assessed this by collecting information on prior season vaccination. Self-report of influenza vaccination is generally accurate for the current season, but the accuracy of self-report for vaccination in prior seasons is largely unknown. 1 Self-report may be particularly susceptible to recall bias for more distant seasons and for individuals with sporadic vaccination patterns. We evaluated the accuracy of self-reported influenza vaccination in the current season and the prior season among patients with medically attended acute respiratory illness who were enrolled in a study of influenza vaccine effectiveness during the 2014-15 season.

| ME THODS
This study was conducted within an influenza vaccine effectiveness study in the 2014-15 northern hemisphere influenza season.
The methods of the influenza vaccine effectiveness study have been previously reported. 2 In brief, patients with acute respiratory illness (including cough) with symptom duration ≤7 days were recruited during an outpatient visit. Enrollment was restricted to a predefined cohort of individuals living near Marshfield, Wisconsin, who receive care from Marshfield Clinic. This analysis is restricted to the individual's first enrollment in the season. After consent, adult participants and parents of children were interviewed to assess symptoms, onset date, and vaccination status. Participants were asked whether they had received a seasonal influenza vaccine since July 1, 2014.
A separate question asked whether the previous season's influenza vaccine was received.
Vaccination status was ascertained from a regional immunization registry. The Registry for Effectively Communicating Self-reported influenza vaccination is generally accurate for the current season, but the accuracy of self-report for vaccination in prior seasons is largely unknown. This study evaluated the accuracy of self-report for current and prior season influenza vaccination among patients with medically attended acute respiratory illness enrolled in a study of influenza vaccine effectiveness during the 2014-15 influenza season. It demonstrates there is a greater potential for exposure misclassification when prior season vaccinations are ascertained by self-report. Percent agreement between self-report and final status was high for both current and prior season vaccination: 97.7% and 93.2%, respectively.

K E Y W O R D S
human, influenza, influenza vaccines, self-report Immunization Needs (RECIN) is a web-based, population-based immunization registry to capture vaccines for both adults and

| D ISCUSS I ON
This study confirms previous, published reports indicating that selfreport of influenza vaccination in the current season can provide a valid measure of vaccine exposure when medical records or registry data are not available. 1,3,4 To our knowledge, this is the first study to assess the accuracy of self-reported influenza vaccination in the prior season. We observed a higher level of misclassification for self-reported vaccination in the prior season, although the percent agreement was above 90% for all age groups. The sensitivity of selfreported vaccination in the prior season was also greater than 90% in all age groups. The specificity of self-report (ie, 100 × self-reported unvaccinated/all unvaccinated) was substantially lower for the prior season (88.2%) compared to the current season (97.0%). These findings demonstrate a greater potential for bias in analyses that rely on self-report of vaccination in the prior season.
The lowest percent agreement for report of prior season vac- This population is also likely to have sporadic vaccination patterns making recall of prior season vaccination more subject to error.
Adults aged 50 years and over had the highest agreement in both the current and prior season. These findings are concordant with studies showing high reliability of and high sensitivities for self-report of influenza vaccinations in older adults. 4,7 Older adults were among the first groups recommended for annual vaccination and often have more frequent contact with the healthcare system so may have more consistent annual vaccination which facilitates recall.
In both seasons, the number of vaccinated individuals denying vaccination was less than those reporting vaccination without documentation. There was no observable pattern of late-season enrollments coupled with early season vaccination which might suggest recall bias within a season, but this misclassification did occur more for prior season vaccination than current season.
A limitation of this study is the lack of racial/ethnic diversity.
Results may not be generalizable to more diverse urban populations. We were also unable to obtain responses from all vaccination providers, particularly companies hired by employers to implement workplace vaccination campaigns, which led to exclusions from the analyses.
In conclusion, this study demonstrates a greater potential for exposure misclassification when prior season vaccinations are ascertained by self-report. The impact of this exposure misclassification on vaccine effectiveness estimates requires further assessment.