Providing reliable indicators of the severity of emerging pathogens is key to public health decision making. In particular, real-time estimates of the case fatality risk, defined as the risk of death among patients classified as cases, can help determine the type and strength of interventions required to mitigate an emerging outbreak.
Estimating case fatality risk has proved to be particularly challenging in recent emerging infectious disease outbreaks, including the 2003 SARS epidemic,1 the 2009 A/H1N1influenza pandemic,1 and more recently the influenza A/H7N92 and MERS-coV outbreaks.3 In this issue of EPIDEMIOLOGY, Wong and colleagues4 present a systematic review of case fatality risk estimates of the 2009 influenza A/H1N1 pandemic based on 50 studies representing 33 countries or regions. This is a most welcome study assessing the lessons learned from a well-documented influenza pandemic, associated with an estimated 100,000–300,000 deaths globally—a substantial number, although modest in comparison with earlier influenza pandemics (L. Simonsen, unpublished data, 2012).5
The most salient finding of the review by Wong et al is the high level of heterogeneity in published case fatality risk estimates of 2009 pandemic influenza. Such heterogeneity greatly undermines the use of these data for policy purposes. It is well known that case fatality risk values are highly sensitive to the choice of denominator (persons classified as cases).1,4 Three categories of cases are typically considered. One is laboratory-confirmed cases. These usually provide a gross underestimate of the total number of cases due to limitations in testing capacity and high propensity to diagnose severe cases. Nonetheless, lab-confirmed cases are often the only available denominator in the early stages of an outbreak.1,4 A second category is symptomatic cases. These can be limited to medically attended clinical infections or can include cases who do not seek care. The third is serologically confirmed infections. This category may produce more reliable case fatality estimates, but it requires expensive and time-consuming biological measurements. In contrast, the choice of numerator is more standard in case fatality risk studies, and typically relies on laboratory-confirmed deaths (69 of 77 pandemic influenza estimates reviewed by Wong and colleagues), although some studies have used alternative metrics such as deaths from pneumonia and influenza, influenza-like illness deaths, or model-derived excess deaths.4
Unsurprisingly, the review by Wong and colleagues4 found dramatic variation in case fatality risk estimates depending on the choice of denominator. These estimates ranged from 100 to 5000 deaths per 100,000 laboratory-confirmed influenza A/H1N1 pandemic cases, 0 to 1200 deaths per 100,000 symptomatic cases, and 1 to 10 deaths per 100,000 serologic infections. What is perhaps more surprising is the high level of residual heterogeneity even after controlling for the choice of denominator. For instance, there was up to 10-fold difference in published symptomatic case fatality risk estimates based on data from the same country and influenza surveillance system.6,7
Early estimates of case fatality can shape policy decisions in pandemic crises, but they can also evolve rapidly as more epidemiological information becomes available and treatment guidelines change. It is informative to consider the well-characterized Mexican 2009 pandemic experience.6–10 The earliest documented case fatality risk estimate of 2009 pandemic influenza was an astounding 4%, based on the first 1100 laboratory-confirmed cases reported by 5 May 2009 in Mexico.8 This estimate is higher than that of the catastrophic 1918 pandemic. Given the apparent severity of this new pandemic virus, the Mexican government closed the nation’s schools for 18 days (24 April to 11 May 2009).7 By June, case fatality risk estimates had been quickly downgraded to 0.4% (0.03–1.8%), as more robust data on laboratory-confirmed cases and deaths accumulated.9
The review by Wong et al suggests that estimates of case fatality risk based on laboratory-confirmed cases are nearly meaningless in absolute terms owing to dramatic under-reporting.4 However, such estimates can arguably be useful for monitoring time trends in disease severity and the impact of treatment strategies. As an illustration, laboratory-confirmed case fatality risk estimates declined throughout the spring 2009 pandemic in Mexico, as the lag between disease onset and hospital admission shortened.10 It is well-accepted that delayed hospital admission complicates influenza case management, limiting the effectiveness of antiviral treatment and, in turn, exacerbating disease severity. Then there followed a 3-fold rise in case fatality risk between the summer and fall 2009 in Mexico, coinciding with a drop in antiviral use from 50% to 9%.10 The large impact of admission delay and antiviral use is further underscored by a large hospital study of more than 10,000 laboratory-confirmed pandemic influenza patients characterized by very short median admission delay and very high antiviral treatment rates—and reporting no pandemic-related death.11
The systematic review by Wong and colleagues4 includes information on case fatality risk from 33 countries. These data highlight interesting regional differences worthy of further investigation. Case fatality risk estimates were particularly high for Argentina, Mexico, and Colombia,4,12 consistent with a high excess mortality burden of the pandemic in this region as assessed by vital statistics data (L. Simonsen, unpublished data, 2013). This suggests that between-country differences in case fatality risk estimates are not purely methodological artifacts and may reflect underlying geographic differences in influenza case management, access to antibiotics and antivirals,10 uptake of seasonal influenza vaccines,6,13 prevalence of underlying comorbid conditions,14 coinfections with other respiratory agents, baseline health, and population demographics.4 Overall, a better understanding of the determinants of influenza severity globally would be important for improving pandemic preparedness plans.
A clear data gap identified by Wong et al’s review4 is the lack of case fatality risk estimates for seasonal influenza epidemics. Such estimates would have been highly informative for comparison purposes at the outset of the 2009 pandemic. We greatly need a population-level epidemiologic tool to compare influenza severity across pandemic and epidemic seasons.
A further issue is the paucity of robust influenza information from low-income regions, which undermines the validity of global projections of pandemic influenza mortality burden.5 The relationship among influenza severity, income, and geography is not straightforward, as shown by the moderate case fatality risk estimates of 2009 pandemic influenza in Mauritius, India, and Nepal4 and the relatively mild excess mortality burden of the pandemic in Bangladesh and South Africa (L. Simonsen, unpublished data, 2013), relative to estimates in high-income countries. Clearly much more work needs to be done to understand these differences.
In conclusion, we wholeheartedly concur with Wong and colleagues4 that rigorous guidelines are needed to monitor influenza severity during seasonal and pandemic outbreaks and across geographic settings. The continued threat posed by novel influenza viruses (H5N1, 2009 H1N1, and most recently H7N9) underscores the urgency of establishing long-term surveillance systems that can assess severity in near real-time and identify changes that might signal the emergence of atypical influenza viruses.15–17 Ideally, these surveillance systems would integrate measures informing the severity pyramid,18 including the risk of hospitalization among medically attended infections and the risk of death among laboratory-confirmed hospitalized patients (see Yu et al2). A globally representative sample of such surveillance systems could shed light on the spectrum of factors that drive the risk of influenza-related death in interpandemic seasons and provide an essential evidence base for decision making when the next health crisis comes.
ABOUT THE AUTHORS
GERARDO CHOWELL is an associate professor in the School of Human Evolution and Social Change at Arizona State University and a research fellow at the Fogarty International Center, US National Institutes of Health. His research interests include mathematical and statistical modeling of infectious disease transmission and control interventions.
CÉCILE VIBOUD is a senior research scientist at the Fogarty International Center, focusing on the transmission dynamics and health burden of acute viral infection at the interface of mathematical modeling, epidemiology, evolutionary genetics, and public health.
1. Garske T, Legrand J, Donnelly CA, et al. Assessing the severity of the novel influenza A/H1N1 pandemic. BMJ. 2009;339:b2840
2. Yu H, Cowling BJ, Feng L, et al. Human infection with avian influenza A H7N9 virus: an assessment of clinical severity. Lancet. 2013;382:138–145
4. Wong JY, Kelly H, Ip DKM, Wu JT, Leung GM, Cowling BJ. Case fatality risk of influenza A(H1N1pdm09): a systematic review. Epidemiology. 2013;24:830–841
5. Dawood FS, Iuliano AD, Reed C, et al. Estimated global mortality associated with the first 12 months of 2009 pandemic influenza A H1N1 virus circulation: a modelling study. Lancet Infect Dis. 2012;12:687–695
6. Echevarría-Zuno S, Mejía-Aranguré JM, Mar-Obeso AJ, et al. Infection and death from influenza A H1N1 virus in Mexico: a retrospective analysis. Lancet. 2009;374:2072–2079
7. Chowell G, Echevarría-Zuno S, Viboud C, et al. Characterizing the epidemiology of the 2009 influenza A/H1N1 pandemic in Mexico. PLoS Med. 2011;8:e1000436
8. Lipsitch M, Finelli L, Heffernan RT, Leung GM, Redd SC2009 H1n1 Surveillance Group. . Improving the evidence base for decision making during a pandemic: the example of 2009 influenza a/H1n1. Biosecur Bioterror. 2011;9:89–115
9. Fraser C, Donnelly CA, Cauchemez S, et al.WHO Rapid Pandemic Assessment Collaboration. Pandemic potential of a strain of influenza A (H1N1): early findings. Science. 2009;324:1557–1561
10. Chowell G, Viboud C, Simonsen L, et al. Impact of antiviral treatment and hospital admission delay on the risk of death of 2009 A/H1N1 pandemic influenza in Mexico, April-December 2009. BMC Infect Dis. 2011;12:97
11. Torres JP, O’Ryan M, Herve B, et al. Impact of the novel influenza A (H1N1) during the 2009 autumn-winter season in a large hospital setting in Santiago, Chile. Clin Infect Dis. 2010;50:860–868
12. Nishiura H. The relationship between the cumulative numbers of cases and deaths reveals the confirmed case fatality ratio of a novel influenza A (H1N1) virus. Jpn J Infect Dis. 2010;63:154–156
13. Skowronski DM, De Serres G, Crowcroft NS, et al. Association between the 2008–09 seasonal influenza vaccine and pandemic H1N1 illness during Spring-Summer 2009: four observational studies from Canada. PLoS Med. 2010;7:e1000258
14. Chowell G, Echevarria-Zuno S, Viboud C, et al. Epidemiological characteristics and underlying risk factors for mortality during the autumn 2009 pandemic wave in Mexico. PLoS One. 2012;7:e41069
15. Chowell G, Bertozzi SM, Colchero MA, et al. Severe respiratory disease concurrent with the circulation of H1N1 influenza. N Engl J Med. 2009;361:674–679
16. Louie JK, Jean C, Acosta M, Samuel MC, Matyas BT, Schechter R. A review of adult mortality due to 2009 pandemic (H1N1) influenza A in California. PLoS One. 2011;6:e18221
17. Wilking H, Buda S, von der Lippe E, et al. Mortality of 2009 pandemic influenza A(H1N1) in Germany. Euro Surveill. 2010;15:19741
18. Presanis AM, De Angelis D, Hagy A, et al.New York City Swine Flu Investigation Team. The severity of pandemic H1N1 influenza in the United States, from April to July 2009: a Bayesian analysis. PLoS Med. 2009;6:e1000207