Advances in measuring influenza burden of disease

Influenza is a global public health threat, with seasonal and pandemic influenza resulting in substantial impact on health, the economy and society. The World Health Organization (WHO) has recently estimated that every year, 290 000 to 650 000 deaths are associated with respiratory diseases from seasonal influenza.1 This estimate takes into account findings from recent influenza respiratory mortality studies, including a study conducted by Iuliano et al.2 Many highincome countries (HICs) that have invested in measuring the impact of influenza epidemics and the costeffectiveness of interventions against influenza have also spent substantial resources in preventing spread and mitigating health outcomes through vaccination, clinical management of severe cases and other public health measures. At the same time, many lowand middleincome countries (LMICs), especially those in the tropics, are grappling with understanding the impact of influenza in their local setting, and to determine whether such interventions are costeffective visàvis interventions for other diseases.3 Furthermore, LMICs are likely to have the highest burden of influenza in children, but these are also the countries with the least data available.4

Influenza is a global public health threat, with seasonal and pandemic influenza resulting in substantial impact on health, the economy and society. The World Health Organization (WHO) has recently estimated that every year, 290 000 to 650 000 deaths are associated with respiratory diseases from seasonal influenza. 1 This estimate takes into account findings from recent influenza respiratory mortality studies, including a study conducted by Iuliano et al. 2 Many high-income countries (HICs) that have invested in measuring the impact of influenza epidemics and the cost-effectiveness of interventions against influenza have also spent substantial resources in preventing spread and mitigating health outcomes through vaccination, clinical management of severe cases and other public health measures. At the same time, many low-and middle-income countries (LMICs), especially those in the tropics, are grappling with understanding the impact of influenza in their local setting, and to determine whether such interventions are cost-effective vis-à-vis interventions for other diseases. 3 Furthermore, LMICs are likely to have the highest burden of influenza in children, but these are also the countries with the least data available. 4

| EFFORTS TO EXPAND BURDEN OF DISEASE STUDIES
It is in this context that burden of disease studies are important to document the potential impact of influenza on various aspects of health.
This includes the number of cases among higher-risk populations, hospitalizations and deaths. Estimates of national disease burden during influenza seasons will raise awareness on the impact of seasonal influenza in the local setting. It will allow for cost-effectiveness analyses to aid decision-making on investments such as vaccination programmes, and can lead to an expansion of efforts to tackle the spread of influenza in regions where impact is highest. In countries that have conducted burden of disease and cost-effectiveness studies, influenza vaccination programmes have likewise taken root. [5][6][7][8] Undertaking burden of disease studies for seasonal influenza also helps countries to develop surveillance and analytical capabilities for use during pandemics.
Previous influenza burden studies were performed mostly in HICs, whose results are not necessarily generalizable to LMICs because of differences in underlying determinants such as age structure, nutrition, prevalence of high-risk conditions, uptake of preventive strategies and access to medical care. Influenza-related research has been expanding in LMICs over the last few years, showing that influenza causes substantial clinical cases, hospitalizations and deaths across various geographical and social settings. [9][10][11][12][13] Research has also shown that the outcomes from influenza infection may be more severe in LMICs compared to HICs, 14,15 and that LMICs contribute a disproportionate burden towards global influenza disease. 4 However, influenza disease burden information is still sparse in regions such as Africa, Asia and South America where there are differences in way that healthcare services are organized, accessed and financed across countries, even within the same region. There is also a relative lack of data from the tropics, where influenza exhibits different seasonal patterns compared to temperate countries. 16 In addition to national considerations, determining more accurate regional and global burden of disease estimates for influenza is necessary to provide perspective on the impact caused by influenza compared to other diseases, and to determine investments to develop better pharmaceuticals such as vaccines with broad immunogenicity and improved vaccine production methods. As the WHO global estimates are extrapolations from available national data, building accurate global estimates requires better data from representative areas from all regions of the world. The heterogeneity of study designs is apparent and expected.
Eight studies used the WHO manual in developing their methodology, 18,19,22,24,26,28,29,34 although this ranged from hospital admission surveys to adapting suggested analytic methods. Most studies brought together information from various sources, including primary data from sentinel sites, provincial or national databases, and virological surveillance networks. Statistical models used range from regression analyses to semi-parametric generalized additive models. The diversity of methods shows that there are different ways in which countries can develop burden estimates, depending on the resources available.
Two studies in this Special Edition, from India and South Africa, evaluated various data sources and methods and determined the best approaches for developing disease burden estimates in their country.
The study from India found that while the Sample Registration System provided the most appropriate national mortality data set, other mortality data sources could be used for subregional estimates. 36 The South African study showed that weekly proportion and influenza subtype-specific proxies provided the best model fit with nonsignificant differences in the estimates. 25 The outcomes also varied across studies. The most common outcomes were deaths attributable to influenza, used in nine studies, 18,20,21,25,29,32,34,36 and severe acute respiratory infections (SARI) or its equivalent, used in eight studies. 19,22,24,26,28,29,34 Another seven studies measured excess or absolute hospitalizations due to influenza, [27][28][29]31,32,34,35 and five studies measured community or outpatient visits. [28][29][30]32,33 Some studies were less representative for the entire country than others-there were 10 studies that collected data from across the country, although most national estimates were calculated with the help of extrapolations and appropriate adjustments for demographic differences. 18,24,27,29,31,33 Among the wealth of information in this supplement, five countries measure their influenza burden for the first time-Cambodia, Chile, Romania, Rwanda and Zambia. 19,24,26,28,34 Fresh perspectives on influenza burden are also offered by the other studies. For example, studies from the United States and the United Kingdom shed light on the burden of disease in the community, including its effect on qualityadjusted life days and years. 30,32,33 This Special Edition also includes a review of the risk factors for severe outcomes associated with influenza illness in HICs versus LMICs, showing differences in determinants for severe outcomes between the two settings. Pregnancy, living with HIV or AIDS and young age were found to be additional risk factors in LMICs but not in HICs.
Furthermore, children with neurological conditions in LMIC also had a higher risk of severe outcomes than those from HICs. 37 Lastly, to assist countries in translating their burden of disease data into economic and policy decisions on the use of vaccines, WHO has also developed an economic evaluation tool. This is described in a paper in this Special Edition, which summarizes the key components of the tool and its implications for public health. 38

| THE WAY FORWARD
The diversity in methods used across various burden of disease studies is reflective of the multifaceted and complex nature of these estimations. Certain methods may be easier to implement in some contexts  Knowing the national and global burden of disease of seasonal influenza is an important first step in providing clarity on the magnitude of the problem at hand. However, its true utility is when such information is used to guide public health actions. There is a need for guidance to support countries to translate these data into policies and practices that would help to reduce this burden. The WHO economic evaluation tool described above is one such example. 38