Background Published influenza vaccination coverage in health care workers (HCW) are calculated using two sources: self-report and vaccination records. The objective of this study was to determine whether self-report is a good proxy for recorded vaccination in HCW, as the degree of the relationship is not known, and whether vaccine behaviour influences self-reporting. Methods A cross-sectional study was conducted using a self-administered survey during September 2010. Considering the vaccination record as the gold standard of vaccination, the properties of self-report as a proxy of the record (sensitivity, specificity, positive predictive value, negative predictive value) were calculated. Concordance between the vaccination campaigns studied (2007-2010) was made using the Kappa index, and discordance was analyzed using McNemar's test. Results 248 HCW responded. The 95% confidence intervals of coverage according to the vaccination record and to self-report overlapped, except for 2007, and the Kappa index showed a substantial concordance, except for 2007. McNemar's test suggested that differences between discordant cases were not due to chance and it was found that the proportion of unvaccinated discordant cases was higher than that of vaccinated discordant cases. Conclusions In our study population, self-reported influenza vaccination coverage in HCW in the previous two years is a good proxy of the vaccination record. However, vaccination behaviour influences the self-report and explains a trend to overestimate coverage in self-reporting compared to the vaccination record. The sources of coverage should be taken into account whenever comparisons are made.