tetano
Editor, Senior Moderator
Vaccine
Volume 29, Supplement 2, 22 July 2011, Pages B49-B55
Historical Influenza Pandemics
doi:10.1016/j.vaccine.2011.03.061 | How to Cite or Link Using DOI
Permissions & Reprints
The perils of using annual all-cause mortality data to estimate pandemic influenza burden
Viggo Andreasena, b, Corresponding Author Contact Information, E-mail The Corresponding Author and Lone Simonsenb, c
a Department of Science, Roskilde University, Denmark
b Fogarty International Center, NIH, Bethesda, MA, USA
c Department of Global Health, School of Public Health and Health Services, George Washington University, USA
Received 21 January 2011;
revised 18 February 2011;
accepted 17 March 2011.
Available online 12 July 2011.
Abstract
Measuring the burden of historic pandemics is not straightforward and often must be based on suboptimal mortality data. For example, the critical 1918 pandemic global burden estimate was based on excess in annual all-cause mortality – calculated as the difference between deaths during 1918–1920 and the surrounding 3-year periods. One intriguing result was a not, vert, similar40-fold between-country variation in pandemic mortality burden: not, vert, similar0.2% of Danes died, compared to not, vert, similar8% of populations in some Indian provinces (Murray et al., 2006 [16]).
Using the same methodology and data source we explore the robustness of this methodology for different age-groups. For infants the country estimates varied 100-fold, from 15 to 1500 excess deaths/10,000 population, while for adults ≥45 years estimates ranged from −70 to 170/10,000 population. In contrast, estimates for children, 1–14 years, and adults aged 15–44 years, were far more stable.
We next used detailed mortality data from Copenhagen to compare such estimates to the more precise estimates obtained from monthly mortality time series data and respiratory deaths. We found that the all-cause annual method substantially underestimated due to an unexplained depression in all-cause mortality in Denmark in 1918 and deaths caused by other epidemic diseases during the baseline periods.
We conclude that country estimates for infants and older adults were highly variable by the Murray method due to substantial variability in annual all-cause mortality. A more precise 1918 pandemic burden estimate would be gotten from either focusing analysis on persons age 1–44 who suffered 95% of all pandemic deaths and had a substantial rise over their baseline mortality level, or if possible focus analysis on annual respiratory deaths. For less severe pandemics, including the ongoing 2009 H1N1 pandemic, the use of all-cause mortality data requires careful consideration of excess deaths in defined pandemic periods and a focus on age groups known to be at risk.
http://www.sciencedirect.com/science/article/pii/S0264410X11004439
Volume 29, Supplement 2, 22 July 2011, Pages B49-B55
Historical Influenza Pandemics
doi:10.1016/j.vaccine.2011.03.061 | How to Cite or Link Using DOI
Permissions & Reprints
The perils of using annual all-cause mortality data to estimate pandemic influenza burden
Viggo Andreasena, b, Corresponding Author Contact Information, E-mail The Corresponding Author and Lone Simonsenb, c
a Department of Science, Roskilde University, Denmark
b Fogarty International Center, NIH, Bethesda, MA, USA
c Department of Global Health, School of Public Health and Health Services, George Washington University, USA
Received 21 January 2011;
revised 18 February 2011;
accepted 17 March 2011.
Available online 12 July 2011.
Abstract
Measuring the burden of historic pandemics is not straightforward and often must be based on suboptimal mortality data. For example, the critical 1918 pandemic global burden estimate was based on excess in annual all-cause mortality – calculated as the difference between deaths during 1918–1920 and the surrounding 3-year periods. One intriguing result was a not, vert, similar40-fold between-country variation in pandemic mortality burden: not, vert, similar0.2% of Danes died, compared to not, vert, similar8% of populations in some Indian provinces (Murray et al., 2006 [16]).
Using the same methodology and data source we explore the robustness of this methodology for different age-groups. For infants the country estimates varied 100-fold, from 15 to 1500 excess deaths/10,000 population, while for adults ≥45 years estimates ranged from −70 to 170/10,000 population. In contrast, estimates for children, 1–14 years, and adults aged 15–44 years, were far more stable.
We next used detailed mortality data from Copenhagen to compare such estimates to the more precise estimates obtained from monthly mortality time series data and respiratory deaths. We found that the all-cause annual method substantially underestimated due to an unexplained depression in all-cause mortality in Denmark in 1918 and deaths caused by other epidemic diseases during the baseline periods.
We conclude that country estimates for infants and older adults were highly variable by the Murray method due to substantial variability in annual all-cause mortality. A more precise 1918 pandemic burden estimate would be gotten from either focusing analysis on persons age 1–44 who suffered 95% of all pandemic deaths and had a substantial rise over their baseline mortality level, or if possible focus analysis on annual respiratory deaths. For less severe pandemics, including the ongoing 2009 H1N1 pandemic, the use of all-cause mortality data requires careful consideration of excess deaths in defined pandemic periods and a focus on age groups known to be at risk.
http://www.sciencedirect.com/science/article/pii/S0264410X11004439
