Annual estimates of the burden of seasonal influenza in the United States: A tool for strengthening influenza surveillance and preparedness

1Estimates of the burden of seasonal influenza are broadly useful for public health, helping national and local authorities monitor epidemiologic trends, plan and allocate resources, demonstrate the impact of vaccine programs as well as other public health and clinical interventions, and inform the public, clinicians, and policymakers about the importance of influenza and influenza prevention.

2Estimates of the burden of seasonal influenza in the United States have evolved over time. First, estimates focused on the number of deaths due to influenza in the 1960s.1 With better access to hospital records, estimates were expanded to include influenza‐related hospitalizations.2, 3 During the 2009 H1N1 pandemic, there was a need to describe the burden of less severe outcomes, which further expanded the burden estimation to include outpatient medical visits and illness in the community.4

3Changes were also made to the methods used to generate estimate of influenza burden. Statistical models were initially used to estimate excess deaths and hospitalizations, those that occur above what is predicted based on historical trends.1, 5, 6 During the 2009 H1N1 pandemic, there was a move toward using a multiplier that could extrapolate rates of hospitalization to rates of less severe disease.

4Historically, new estimates of influenza‐related mortality or hospitalization over multiple influenza seasons were published periodically, as new data became available.5, 6, 7 However, a hallmark of influenza is its variability from one season to the next and periodic assessments of the burden of influenza fail to capture the full extent of seasonal variation. For this reason, the Centers for Disease Control and Prevention (CDC) has transitioned from providing periodic estimates to reporting annual estimates of influenza burden in the United States. Annual estimates of disease burden, in combination with annual assessments of influenza vaccine coverage and vaccine effectiveness in preventing disease, allow for timely evaluation of influenza prevention and control efforts.

5Data from the 2010‐2011 through 2015‐2016 influenza seasons (October through April) have been used on an annual basis to estimate the burden of seasonal influenza and the disease burden averted by influenza vaccination in the United States.8, 9, 10, 11, 12, 13 The methods have previously been described in detail (Fig. S1).7, 9, 11, 12, 13, 14

6Briefly, rates of hospitalization with laboratory‐confirmed influenza were obtained from the Influenza Hospitalization Surveillance Network (FluSurv‐NET), a population‐based surveillance conducted in 14 geographically distributed states.15 Hospitalization rates were generated by age group (0‐4, 5‐17, 18‐49, 50‐64, and ≥65 years). Rates were adjusted for influenza testing practices and test sensitivity and then applied to the US population to obtain estimates of the number of influenza‐associated hospitalizations that occurred each season.

7Estimates of excess deaths related to influenza were based on a statistical model of the weekly number of deaths obtained from the National Center for Health Statistics.16 The model accounts for seasonal trends in mortality and weekly circulation of influenza and respiratory syncytial virus, obtained from national virologic surveillance.16, 17 The model was fitted using Markov chain Monte Carlo methods, yielding “point estimates” (mean or median of the empirical posterior distribution) and “confidence intervals” (95% credible intervals) for the number of deaths attributable to influenza. Data on deaths with pneumonia or influenza listed as a cause of death were used in the statistical model because they are available in near real time. However, most influenza‐related deaths are likely not due directly to influenza virus infection but may be due to secondary bacterial infection or worsening of underlying chronic health conditions, such as chronic heart or lung disease. Even when influenza likely contributed to the events leading to a death, it may not be recognized and is rarely listed on the death certificate. From prior analyses, the number of deaths associated with influenza may be two to four times higher than the number of deaths related to influenza that have pneumonia or influenza listed on the death certificate.7, 18 Deaths with any respiratory or circulatory causes listed on the death certificate are likely more inclusive of deaths related to influenza than deaths with pneumonia or influenza causes; therefore, additional statistical models were created using death from respiratory or circulatory causes. Data on respiratory and circulatory deaths were available with a 3 year lag; therefore, in 2016, data were available and summarized for the 2010‐2011 season through the 2013‐2014 season only.

8Estimates of the case‐to‐hospitalization ratio, obtained from studies during the 2009 pandemic in the United States,4 were used to calculate the number of illness episodes that occurred in the community from the number of hospitalizations with laboratory‐confirmed influenza. Estimates of the proportion of ill persons who sought medical care, obtained from a nationwide behavior survey conducted in the United States during the 2009 pandemic,19 were used to estimate the number of outpatient medical visits relative to the estimated number of influenza illnesses in the community.

9Estimates of the numbers of influenza‐associated illnesses, outpatient medical visits, hospitalizations, and deaths prevented by influenza vaccination were derived from burden estimates, influenza vaccination coverage, and vaccine effectiveness, as previously described.9 Briefly, estimates of monthly influenza vaccine coverage by age group (6 months‐4 years, 5‐17, 18‐49, 50‐64, and ≥65 years) and annual vaccine effectiveness estimates were used to estimate the number of influenza outcomes that would have occurred in the absence of vaccination, assuming equal vaccine effectiveness against each outcome.20, 21 From the estimates of burden estimates, vaccination coverage, and vaccine effectiveness, we calculated age group‐specific hypothetical numbers of illness, medical visits, hospitalizations, and respiratory or circulatory deaths related to influenza per month that would have occurred in the absence of vaccination (assuming only a direct vaccination effect). The outcomes prevented by vaccination were the difference between these hypothetical numbers and the burden estimates for the actual population. The fraction of hospitalizations prevented by vaccination was the total number of hospitalizations averted divided by the number of hospitalizations that would have occurred in the absence of vaccination.

10The model of averted outcomes was also used to estimate the incremental benefits expected from increasing vaccine coverage or vaccine effectiveness. The number needed to vaccinate to prevent one influenza‐associated illness was estimated by dividing the size of the vaccinated population in the United States by the number of illnesses prevented by vaccination.

11Estimates of the number of illnesses, outpatient medical visits, hospitalizations, and respiratory and circulatory deaths related to seasonal influenza in the United States and the number of illnesses, medical visits, hospitalizations, and deaths related to influenza that were prevented by influenza vaccination are updated annually and published online.13

12Over the past six influenza seasons in the United States (2010‐2011 through 2015‐2016), we estimate that influenza‐associated illnesses have ranged from a low of 9.2 million to a high of 35.6 million illnesses, with variation by age (Table 1). Outpatient medical visits related to influenza have ranged from 4.3 million to 16.7 million, while influenza‐associated hospitalizations have ranged from 139 000 to 708 000. From 2010‐2011 through 2013‐2014 influenza seasons, influenza‐associated respiratory and circulatory deaths have ranged from a low of 12 000 to a high of 56 000, and associated pneumonia and influenza deaths have ranged from 4000 to 12 000 over the six seasons from 2010‐2011 through 2015‐2016. Figure 1 is a visual representation of the burden of influenza, which depicts the relative magnitude of each season's burden in the United States.

13Given the estimates of seasonal incidence of influenza, the associated burden of severe disease, and estimates of influenza vaccine effectiveness and coverage in the United States, we estimate that influenza vaccination has prevented between 1.6 million and 6.7 million illnesses, 790 000‐3.1 million outpatient medical visits, 39 000‐87 000 hospitalizations, and 3000‐10 000 respiratory and circulatory deaths related to influenza each season (Table 2).

14Surveys in the US population suggest that overall vaccination coverage was 42%‐47% over the past six influenza seasons, although coverage varies considerably by age.22 Increases in vaccination coverage can translate to large reductions in influenza‐associated disease burden when influenza viruses in the vaccine are similar to circulating viruses. For example, during the 2015‐2016 influenza season, overall vaccination coverage was 46% and we estimate that vaccination prevented more than 5 million illnesses and 71 000 hospitalizations. If vaccination coverage had been increased by five percentage points across all age groups, more than 500 000 additional illnesses and 6000 additional hospitalizations would have been prevented.

15The burden models can also serve to assess the impact of changes in vaccine effectiveness on an annual basis. For example, during the 2014‐2015 influenza season, the influenza A/H3N2 virus that circulated widely drifted, both genetically and antigenically, after vaccine virus recommendations were made, leading to reduced overall vaccine effectiveness (19%) against influenza A viruses.23 Given the reduced vaccine effectiveness and high burden of illness, we estimated that, on average, 92 people needed to be vaccinated to prevent a single case of influenza and 3115 people needed to be vaccinated to prevent a single hospitalization. Changes in the vaccine composition and improved vaccine effectiveness (47%) against the circulating influenza virus types during the 2015‐2016 season, along with fewer hospitalized patients, meant that only 29 people needed to be vaccinated, on average, to prevent a single case of influenza and 2033 needed to be vaccinated to prevent one influenza‐associated hospitalization.21, 24

16Periodic estimates of the burden of seasonal influenza in the United States have been made for more than 50 years, focused mainly on mortality and hospitalization related to influenza. In response to the 2009 pandemic and the need for timely data on burden and severity, CDC now generates annual burden estimates and has also expanded burden estimates to include less severe outcomes. These estimates have allowed for timely communication about the importance of vaccination to prevent influenza and helped frame discussions about influenza program goals with policymakers. Annual estimates have reinforced the message that influenza is ever‐changing and will likely differ from one season to the next and we use a range to describe burden in order to more accurately reflect the annual variability of influenza.

17In addition to being reported annually, CDC estimates of influenza disease burden were expanded to include outpatient medical visits and symptomatic community illness. Estimates focused on influenza‐related mortality and hospitalizations reinforce the potentially serious nature of influenza, but are a small fraction of the total burden of influenza and can be biased, as they are highly influenced by patterns and policies for hospital admission, influenza testing, and reporting.11 On the other hand, estimates of the number of symptomatic community illnesses, for which medical care is not sought but may still result in missed school or work, and outpatient medical visits due to influenza underscore the frequency of influenza illness and its widespread societal impact. We estimate that for every influenza‐related hospitalization, between 11 and 365 more non‐hospitalized cases occur in the community, depending on the age group.4, 8

18Generating annual estimates of influenza burden helped CDC recognize gaps in influenza surveillance activities. For example, there were no means to directly estimate medically‐attended and community illness during the 2009 pandemic. Instead, these portions of influenza burden were indirectly estimated using the rates of influenza‐associated hospitalization and field‐validated multipliers of healthcare utilization and case‐to‐hospitalization ratios from the 2009 H1N1 pandemic. In an effort to fill these gaps, there are now several ongoing efforts and collaborations to gather data that can directly estimate the burden of medically‐attended illness related to influenza as well as symptomatic community illness on a routine basis.25, 26, 27 Not only are these improvements to surveillance helpful during seasonal epidemics, but the creation and optimization of surveillance activities that are routine, robust, and near real time will be helpful when a pandemic occurs.16, 28

19The methods and estimates of seasonal influenza burden are not without limitations. First, influenza vaccination coverage estimates and the multipliers used for estimating outpatient medical visits and symptomatic community illness were derived from survey respondents and are subject to recall bias and non‐response bias. Second, the model of disease prevented through vaccination only calculated outcomes averted through direct protection of persons who were vaccinated, and not indirect or herd protection. Third, this same model assumed a single estimate of vaccine effectiveness against all outcomes and constant effectiveness over the course of the season within each age group, which may be oversimplifications. Fourth, estimates of influenza‐associated mortality were based on an ecologic analysis at the national level and may wrongly attribute mortality above an imputed “baseline” to influenza. In addition, it may not reflect the burden on a different level, such as the state or local level.

20Despite their limitations, the models we use for estimating the burden of seasonal influenza are simple and provide timely information that is valuable for public health activities. Burden estimates are invaluable for estimating the economic and societal costs of influenza and making decisions about procurement of vaccines and influenza antivirals before the influenza season begins. At CDC, we have also found that providing estimates of burden on an annual basis has served to strengthen existing influenza surveillance activities, allowed for timely communication of the value of vaccination, improved our understanding of the epidemiology of seasonal influenza in the United States, and enhanced preparedness for future influenza pandemics.