Annual influenza epidemics result in substantial mortality, especially among adults aged 65 years and older. Previous estimates attributed to WHO1 indicated that 250 000–500 000 influenza-associated deaths occur annually, corresponding to estimates of 3·8–7·7 deaths per 100 000 individuals calculated using 2005 UN Department of Economic and Social Affairs World Population Prospects.2 The methods used to calculate this WHO estimate have not been published and might not have accounted for annual variability in the incidence of influenza virus infection, the age and health status of populations, or risk of influenza death across countries. A 2013 study,3 which used data from 2005 to 2009, suggested that 148 000–249 000 annual influenza respiratory deaths might occur each year.3 Current, reliable global and country-specific influenza-associated mortality estimates are needed to inform decisions about the value of influenza prevention and control and to inform global public health priorities.
Estimating the burden of annual influenza epidemics is challenging for many countries because of the requirement for high-quality systematic vital records and local viral surveillance data. As a result, most influenza-associated mortality estimates have been obtained from high-income countries with a temperate climate.4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26 Many developing or recently industrialised countries have leveraged improvements in influenza surveillance data27 to develop country-specific influenza-associated mortality estimates, which are generally higher (4·3–31·6 per 100 000 individuals) than WHO-attributed estimates,28, 29, 30, 31 implying that the current global influenza death estimates might underestimate the global mortality burden of influenza. The wide variation in influenza-associated excess mortality estimates between countries and between seasons and years highlights the effect of specific circulating influenza viruses on mortality and the need to include more country-specific estimates in global mortality estimation to better quantify global disease burden. Furthermore, global and country-specific estimates should be updated periodically to account for changes in population demographics, improvements in health care, and viral evolution.
Research in context
Evidence before this study
Previous estimates commonly attributed to WHO indicate that 250 000–500 000 deaths occur annually worldwide due to seasonal influenza viruses. However, no information has been published about the methods or data sources used to calculate these global estimates. These estimates began to be cited in publications and on WHO's website around 2004, suggesting that the estimates might have been generated using data from the 1990s. Since the 1990s, improvements in influenza virus surveillance and vital records systems in many countries have led to more estimates of influenza-associated mortality for countries across the world, including some estimates from middle-income and low-income countries. Global estimates of pandemic influenza deaths are available; however, these estimates might not meet the requirements for global seasonal influenza estimates because deaths during pandemic periods are likely to differ from seasonal epidemics. We searched PubMed for articles that estimated influenza-associated excess deaths or excess mortality published between Jan 1, 1960 and Dec 31, 2012, using the search terms “influenza”, “death”, “mortality”, “excess death”, “excess mortality”, “Serfling”, “negative binomial”, “time series”, “respiratory”, “circulatory”, “pneumonia”, and “influenza” with no language restrictions. Previously published data were not suitable for our extrapolation model to estimate global influenza deaths because of heterogeneity between the age groups studied and death outcomes (eg, pneumonia and influenza, respiratory, all-cause, or circulatory) investigated. Thus, we initiated an effort to directly collect data from partners around the world using common age groups and death outcomes to update and improve global seasonal influenza-associated mortality estimates.
Added value of this study
Previous global influenza mortality estimates were calculated more than 10 years ago and detailed information about the methods used are not available. Since these estimates were made available by WHO the number of countries with capacity to calculate national estimates for seasonal influenza-associated excess mortality has increased. We initiated a project to use this additional information to update and improve global estimates of influenza-associated respiratory mortality. We worked with collaborators from 47 countries to develop an innovative statistical model to calculate global estimates of influenza-associated mortality using vital records and viral surveillance data, including estimates for 1999–2015 from countries with data, which were extrapolated to countries without such data. Our study presents a comprehensive analysis of influenza-associated mortality, in which we provide extensive details about methods used and account for differences between countries. In this study, we calculated country-specific estimates for influenza-associated respiratory deaths in three age groups (<65 years, 65–74 years, and ≥75 years) and did a subanalysis for children younger than 5 years, which might help country-level policy makers to understand the impact of influenza virus infection on their populations. Additionally, our analysis complies with the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) recommendations.
Implications of all the available evidence
Our estimates of influenza-associated respiratory mortality are higher than previous estimates. The results of this study help to improve understanding about the burden of influenza viral infections and emphasise the need for continued support to detect, prevent, and control influenza viruses. Furthermore, this study provides age-specific and country-specific influenza-associated respiratory death estimates that could be used by countries to inform prevention and control measures for influenza virus infection in their population. This study improves on earlier estimates by including primary data from 47 countries to estimate global influenza-associated deaths and by validating our extrapolation models.
Improved methods to estimate baseline and influenza-associated mortality have been developed.32, 33, 34, 35, 36, 37 However, influenza virus infections are rarely confirmed systematically by laboratory diagnosis, and thus influenza deaths might be attributed to other comorbid conditions or secondary infections. Ecological models are commonly used to estimate influenza-associated mortality. Vital records death data, which have varying levels of quality, completeness, and population coverage, are systematically coded for cause of death using the International Classification of Diseases (ICD),38, 39 and categories of death commonly associated with influenza, including respiratory or circulatory causes,40 are modelled with virological data to identify periods of influenza virus circulation and to estimate influenza-associated excess deaths. The inclusion of virological data is important for the quantification of influenza-associated deaths because the circulating strains of the virus subtypes vary from year to year, which can affect annual mortality.40, 41 The application of these methods is challenging in many low-income and middle-income countries because they often have year-round or multiple peaks in influenza virus circulation, vital records data that are not of sufficient quality, or they have too few years of reliable surveillance data for robust estimates despite recent efforts to improve and expand surveillance.27, 35, 42, 43
We aimed to estimate the number of country-specific influenza-associated respiratory deaths and to update global and regional influenza-associated respiratory death estimates among individuals younger than 65 years, 65–74 years, and 75 years and older, using excess death estimates from contributing countries extrapolated to countries without such data.