Improving seasonal and pandemic influenza surveillance, epidemiology and disease burden estimation in populations using administrative and vital statistics databases and population surveys

The impact of influenza virus on populations is often under-estimated, a thorough understanding of its epidemiology remains elusive and common methods of measuring influenza lack validation. To determine limitations of data sources used for epidemiological study of influenza, record linkage was used to combine person-level information from emergency department, hospital admission, death and laboratory-notified influenza databases, from New South Wales (NSW), Australia. In each of the hospital and death databases, a minority of patients with a virologically confirmed influenza infection were recorded as having influenza as a cause of their illness. Among database records with influenza recorded as a cause of the illness, a minority were based on laboratory diagnosis. The emergence of pandemic influenza in 2009 caused by the influenza A(H1N1)pdm09 virus provided an opportunity to attempt more complete estimates of the outcomes of influenza infection. Questions on influenza-like illness (ILI) were introduced into the continuous NSW Health Survey to provide population representative estimates of influenza-like illness. During July to September 2009, 23% of the NSW population reported an ILI, of which 55% sought general practice care and 5% presented to a hospital. Smoking and obesity were risk factors for reporting ILI, and the effectiveness of the pre-pandemic seasonal vaccine against ILI was ~20%.Time series analysis was used to estimate mortality attributable to the pandemic virus and seasonal influenza epidemics in preceding years, for NSW and Australia. Using age-specific weekly counts of all-cause death registrations revealed that overall mortality in NSW was lower than seasonally expected during the cooler months of 2009 in which the pandemic virus circulated. Consistent with earlier pandemics, reduced mortality appeared largely driven by the relative immunity of older persons to the pandemic virus. The national study used cause-specific death and laboratory reported influenza virus time series and a novel statistical method. All-cause, all-age mortality reached levels as high as 12.0 per 100,000 when influenza A(H3N2) predominated, while 2009 pandemic mortality was 2.5 per 100,000 (not statistically significant), and influenza B mortality was negligible. Ongoing application of these methods can provide improved and consistent comparisons of the impact of influenza on populations over time