Abstract
Administrative and vital statistics databases are frequently used for public health surveillance of influenza incidence and outcomes. We used population based, probabilistic record linkage of laboratory diagnosed influenza, emergency department, hospital admission and death registration databases to determine how frequently laboratory diagnosed influenza is coded as influenza or recorded as a cause of death. Influenza was substantially under-recorded as a cause of emergency presentation, hospitalization and death. Influenza type A infection was more likely than type B to lead to coding of influenza. Despite under-coding, time series of coded influenza from databases do reflect trends in influenza incidence.