Abstract
Background
In temperate zones, all‐cause mortality exhibits a marked seasonality, and influenza represents a major cause of winter excess mortality. We present a statistical model, FluMOMO, which estimate influenza‐associated mortality from all‐cause mortality data and apply it to Danish data from 2010/11 to 2016/17.
Methods
We applied a multivariable time series model with all‐cause mortality as outcome, influenza activity and extreme temperatures as explanatory variables while adjusting for time trend and seasonality. Three indicators of weekly influenza activity (IA) were explored: percentage of consultations for influenza‐like illness (ILI) at primary health care, national percentage of influenza‐positive samples, and the product of ILI percentage and percentage of influenza‐positive specimens in a given week, that is, the Goldstein index.
Results
Independent of the choice of parameter to represent influenza activity, the estimated influenza‐associated mortality showed similar patterns with the Goldstein index being the most conservative. Over the 7 winter seasons, the median influenza‐associated mortality per 100 000 population was 17.6 (range: 0.0‐36.8), 14.1 (0.3‐31.6) and 8.3 (0.0‐25.0) for the 3 indicators, respectively, for all ages.
Conclusion
The FluMOMO model fitted the Danish data well and has the potential to estimate all‐cause influenza‐associated mortality in near real time and could be used as a standardised method in other countries. We recommend using the Goldstein index as the influenza activity indicator in the FluMOMO model. Further work is needed to improve the interpretation of the estimated effects.
In temperate zones, all‐cause mortality exhibits a marked seasonality, and influenza represents a major cause of winter excess mortality. We present a statistical model, FluMOMO, which estimate influenza‐associated mortality from all‐cause mortality data and apply it to Danish data from 2010/11 to 2016/17.
Methods
We applied a multivariable time series model with all‐cause mortality as outcome, influenza activity and extreme temperatures as explanatory variables while adjusting for time trend and seasonality. Three indicators of weekly influenza activity (IA) were explored: percentage of consultations for influenza‐like illness (ILI) at primary health care, national percentage of influenza‐positive samples, and the product of ILI percentage and percentage of influenza‐positive specimens in a given week, that is, the Goldstein index.
Results
Independent of the choice of parameter to represent influenza activity, the estimated influenza‐associated mortality showed similar patterns with the Goldstein index being the most conservative. Over the 7 winter seasons, the median influenza‐associated mortality per 100 000 population was 17.6 (range: 0.0‐36.8), 14.1 (0.3‐31.6) and 8.3 (0.0‐25.0) for the 3 indicators, respectively, for all ages.
Conclusion
The FluMOMO model fitted the Danish data well and has the potential to estimate all‐cause influenza‐associated mortality in near real time and could be used as a standardised method in other countries. We recommend using the Goldstein index as the influenza activity indicator in the FluMOMO model. Further work is needed to improve the interpretation of the estimated effects.
Originalsprog | Engelsk |
---|---|
Tidsskrift | Influenza and Other Respiratory Viruses |
Vol/bind | 12 |
Udgave nummer | 5 |
Sider (fra-til) | 591-604 |
ISSN | 1750-2640 |
DOI | |
Status | Udgivet - 2018 |