Under-ascertainment of severe outcomes of influenza infections in administrative databases has long been recognised. After reviewing registered deaths following an influenza epidemic in 1847, William Farr, of the Registrar-General's Office, London, England, commented: ''the epidemic carried off more than 5,000 souls over and above the mortality of the season, the deaths referred to that cause [influenza] are only 1,157"[1]. Even today, studies of the population epidemiology, burden and cost of influenza frequently assume that influenza's impact on severe health outcomes reaches far beyond the number of influenza cases counted in routine clinical and administrative databases. There is little current evidence to justify the assumption that influenza is poorly identified in health databases. Using population based record linkage, we evaluated whether the assumption remains justified with modern improvements in diagnostic medicine and information systems.
Objective
To estimate the degree to which illness due to influenza is under-ascertained in administrative databases, to determine factors associated with influenza being coded or certified as a cause of death, and to estimate the proportion of coded influenza or certified influenza deaths that is laboratory confirmed.