Conceived and designed the experiments: CvdW LvA M. Koopmans WvP NN MvdS WvdH M. Kretzschmar. Analyzed the data: CvdW. Contributed reagents/materials/analysis tools: CW AvL CH AM GD FD. Wrote the paper: CvdW LvA M. Koopmans WvP NN CW AvL WvdH AM GD FD CH MvdS M. Kretzschmar.
The authors have declared that no competing interests exist.
We assessed the severity of the 2009 influenza pandemic by comparing pandemic mortality to seasonal influenza mortality. However, reported pandemic deaths were laboratory-confirmed – and thus an underestimation – whereas seasonal influenza mortality is often more inclusively estimated. For a valid comparison, our study used the same statistical methodology and data types to estimate pandemic and seasonal influenza mortality.
We used data on all-cause mortality (1999–2010, 100% coverage, 16.5 million Dutch population) and influenza-like-illness (ILI) incidence (0.8% coverage). Data was aggregated by week and age category. Using generalized estimating equation regression models, we attributed mortality to influenza by associating mortality with ILI-incidence, while adjusting for annual shifts in association. We also adjusted for respiratory syncytial virus, hot/cold weather, other seasonal factors and autocorrelation. For the 2009 pandemic season, we estimated 612 (range 266–958) influenza-attributed deaths; for seasonal influenza 1,956 (range 0–3,990). 15,845 years-of-life-lost were estimated for the pandemic; for an average seasonal epidemic 17,908. For 0–4 yrs of age the number of influenza-attributed deaths during the pandemic were higher than in any seasonal epidemic; 77 deaths (range 61–93) compared to 16 deaths (range 0–45). The ≥75 yrs of age showed a far below average number of deaths. Using pneumonia/influenza and respiratory/cardiovascular instead of all-cause deaths consistently resulted in relatively low total pandemic mortality, combined with high impact in the youngest age category.
The pandemic had an overall moderate impact on mortality compared to 10 preceding seasonal epidemics, with higher mortality in young children and low mortality in the elderly. This resulted in a total number of pandemic deaths far below the average for seasonal influenza, and a total number of years-of-life-lost somewhat below average. Comparing pandemic and seasonal influenza mortality as in our study will help assessing the worldwide impact of the 2009 pandemic.
The 2009 influenza A/H1N1 pandemic has led to major additional surveillance and control efforts by public health authorities worldwide. Assessing the severity of the pandemic has been complicated. A concrete measure of severity would be the number of deaths (mortality) and years-of-life-lost due to the pandemic compared to the numbers caused by seasonal epidemics. However, the numbers of pandemic deaths reported to WHO
Another methodological issue is that young children seem to have been disproportionately affected by the pandemic, whereas the elderly have been relatively spared
Two studies have estimated age-specific pandemic and seasonal influenza mortality for the US
Nevertheless, associating all-cause mortality to ILI-incidence can be complicated by annual heterogeneity in the association between mortality and ILI. In an earlier study we already reported that such annual heterogeneity in association is likely to be due to differences in severity per ILI case driven by viral factors such as dominant (sub)types (influenza A(H1N1), A(H3N2) or B) and/or antigenic characteristics
In this study we compared the 2009 pandemic mortality to the 10 preceding seasonal epidemics in the Netherlands (population size 16.5 million) by attributing age-specific all-cause mortality to influenza in 1999–2010, using the modeled association of weekly all-cause mortality with ILI-incidence. We also calculated years-of-life-lost from the age-specific influenza mortality estimates.
We obtained weekly mortality data from Statistics Netherlands (100% coverage) for the period 1999–2010, aggregated by week and age category (0–4, 5–24, 25–44, 45–64, 65–74 and ≥75 yrs of age). In addition to all-cause mortality, physician-reported primary cause of death data were available for analysis. Weekly ILI-incidence data for 1999–2010 were obtained from a sentinel network of general practitioners (GP-ILI data, 0.8% coverage)
To adjust for possible mortality caused by pathogens other than influenza, we also investigated total weekly counts of RSV, as reported by a routine laboratory surveillance system (the Weekly Diagnostic Laboratory Reports of the Dutch Working Group on Clinical Virology, which covers 38–73% of viral diagnostics in the Netherlands; number of laboratory submissions (denominator) and patient age were not available
Another contributor to (excess) mortality can be weather conditions, such as hot and/or cold temperatures
Similar to
We then used the model coefficients to estimate the weekly number of ILI-attributed deaths by age category. Based upon these weekly estimates of influenza-attributed deaths, we calculated the annual influenza-attributed numbers of deaths, scaled to the 2009 population. Finally, we compared the pandemic mortality to the influenza mortality in earlier years.
We also performed a sensitivity analysis to evaluate the impact of our model choices on the influenza-attributed mortality; we evaluated the impact of in- or excluding RSV or the hot/cold weather variables in the model, using ILI-incidence vs influenza detections as a measure of influenza activity, and using pneumonia/influenza (ICD-10 codes J09-J18) or respiratory/cardiovascular (J00-J99 and I00-I99) deaths as explained variables instead of all-cause deaths.
We calculated the years-of-life-lost (YLL)
Age group | Average age at death in 2009 (yrs) | Life expectancy in 2009 (yrs) |
0–4 | 0.62 | 80.26 |
5–24 | 17.93 | 63.23 |
25–44 | 38.13 | 43.44 |
45–64 | 57.63 | 25.33 |
65–74 | 70.43 | 14.93 |
≥75 | 85.65 | 5.52 |
Fitting the regression models with annual ILI-coefficients to our data showed heterogeneity in the influenza-attributed mortality incidence both among years, and among age categories.
For these estimates we used regression models with annual ILI-coefficients, corrected for RSV activity, hot/cold weather, seasonal factors and a time trend. The green colored areas reflect the estimated ILI-attributed mortality, and the black dotted lines present the actually observed weekly mortality; the orange colored areas reflect the estimated basis seasonal mortality, corrected for a time trend; the brown and red colored areas reflect mortality attributed to RSV and hot/cold weather conditions respectively. The grey-shaded bars on the background of the graphs indicate influenza epidemic episodes (defined as −3/+3 weeks around successive weeks with an overall ILI-incidence ≥5.1/10,000 population); only ILI-incidence in these episodes was included in the models. The strong downward timetrend in 0–4 yrs of age seems mainly due to a decrease of 39% in neonatal deaths in the last decade as reported by the national perinatal audit
In the models, for all age categories the (lagged) ILI-incidence was included as explanatory variable for all-cause mortality. The annual ILI regression coefficients showed significant heterogeneity between years for all age categories (F-test, P≤0.0001). Some of the annual ILI regression coefficients estimates were below zero – reflecting the mild influenza impact in those years, as also observed in
Number of influenza-attributed deaths | YLL | ||||||||
By age category | Total | ||||||||
0–4 | 5–24 | 25–44 | 45–64 | 65–74 | ≥75 | ||||
|
|
4 (0–11) | 0 (0–2) | 13 (0–36) | 60 (14–106) | 291 (182–400) | 2,310 (1,891–2,730) | 2,679 (2,234–3,108) | 19,512 |
|
37 (29–45) | 38 (28–48) | 12 (0–37) | 0 (0–0) | 0 (0–13) | 0 (0–279) | 87 (0–183) | 5,906 | |
|
19 (10–29) | 22 (7–37) | 30 (0–70) | 142 (76–209) | 222 (73–372) | 2,257 (1,673–2,840) | 2,693 (2,085–3,301) | 23,650 | |
|
0 (0–0) | 21 (12–31) | 10 (0–41) | 176 (128–224) | 328 (230–426) | 1,716 (1,361–2,070) | 2,251 (1,859–2,604) | 20,614 | |
|
24 (18–30) | 3 (0–13) | 62 (36–89) | 233 (193–273) | 393 (314–472) | 2,232 (1,920–2,543) | 2,947 (2,622–3,271) | 28,933 | |
|
29 (24–34) | 20 (7–33) | 41 (14–68) | 170 (117–222) | 205 (123–286) | 3,170 (2,829–3,511) | 3,634 (3,278–3,990) | 30,243 | |
|
15 (8–22) | 15 (0–31) | 0 (0–21) | 8 (0–73) | 0 (0–0) | 1,096 (661–1,532) | 1,134 (435–1,341) | 8,388 | |
|
0 (0–0) | 0 (0–13) | 15 (0–40) | 0 (0–0) | 0 (0–19) | 415 (105–725) | 430 (0–619) | 2,942 | |
|
26 (18–33) | 13 (0–28) | 62 (36–88) | 111 (38–185) | 0 (0–12) | 396 (69–723) | 608 (182–879) | 10,589 | |
|
10 (0–21) | 12 (0–25) | 60 (29–91) | 187 (94–279) | 397 (303–491) | 2,434 (1,975–2,893) | 3,100 (2,622–3,579) | 28,298 | |
|
16 | 15 | 31 | 109 | 184 | 1,603 | 1,956 | 17,908 | |
|
|
77 (61–93) | 30 (13–48) | 29 (0–58) | 46 (0–121) | 313 (232–394) | 116 (0–442) | 612 (266–958) | 15,845 |
The 95% confidence intervals for the influenza-attributed deaths by age category are given in brackets behind the point estimates. In the last two columns on the right-side of the table for each year in the study period the total number of influenza-attributed deaths is given as well as the total years-of-life-lost attributed to influenza.
The horizontal black dotted lines indicate the average total number of deaths and years-of-life-lost for the 10 seasonal epidemics in 1999–2009.
Sensitivity analysis showed that RSV and hot/cold weather only minimally influenced the influenza-attributed mortality. Replacing ILI-incidence by influenza detections in the models did not change the estimates much either. When replacing all-cause mortality by respiratory/cardiovascular or pneumonia/influenza deaths, the estimates for pandemic and seasonal influenza deaths were much lower; but pandemic mortality was consistently estimated lower than during an average seasonal influenza epidemic, with a relatively low impact in the elderly combined with a relatively high impact in the youngest age category. See
Because of the increased number of influenza-attributed deaths in the 0–4 yrs of age during the pandemic year, we further explored the distribution of age and reported cause of death in this age category. After excluding neonatal deaths, deaths at <1 yrs of age occurred relatively more often during the pandemic than in earlier seasonal influenza years (right-sided Fisher's exact tests for 2×2 tables; P≤0.022), but to a lesser extent this was also observed for the seasonal epidemic in 2000/2001. Among the reported causes of death, cardiomyopathy (International Classification of Diseases, 10th revision, ICD-10 code I42) and spinal muscular atrophy (ICD-10 code G12.9) were overrepresented during the pandemic exclusively (P≤0.002 and P≤0.012 respectively). Intracerebral haemorrhage in neonates was also disproportionately high during the pandemic (ICD-10 P52.4; P≤0.004), but also in 2004/2005 although then only borderline significant. The two other causes of death that showed elevations (fetus/newborn with necrotizing enterocolitis and fetus/newborn affected by multiple pregnancy, ICD-10 P77 and P01.5) seemed unlikely to be related to the pandemic, since the first had been increasing since 2007 and the second showed elevations for two other years in the study. See
By comparing mortality estimates of the 2009 pandemic with those for seasonal influenza, we found that deaths in young children disproportionately contributed to the number of pandemic deaths compared to seasonal influenza. However, the overall pandemic impact was only moderate with the total years-of-life-lost somewhat below but close to the average of seasonal influenza, and the total number of deaths far below average. This limited impact – despite disproportionate mortality in young children – was mainly due to low mortality in the elderly (≥75 yrs of age) compared to most seasonal epidemics.
We were able to compare pandemic to seasonal influenza mortality by using the same statistical methodology and data types for both pandemic and seasonal influenza years. Our use of annual instead of constant ILI-coefficients made it possible to estimate the influenza-attributed mortality per year, adjusted for shifts in the association of ILI-incidence and mortality.
We found a large range of total numbers of deaths and years-of-life-lost for the 10 seasonal epidemics. In 2003/2004 the influenza A/Fujian/411/02(H3N2) drift variant emerged and Fujian-like H3N2 viruses remained the dominant subtype in the next 3 years
For the 0–4 yrs of age, we estimated 77 pandemic deaths, which is higher than for any seasonal influenza epidemic in the study. Remarkably, in the Netherlands only 5 laboratory-confirmed deaths in 0–4 yrs of age were reported during the pandemic, including one child <1 yrs of age with a medical condition
The higher proportion of deaths in <1 yrs of age (excluding neonates) that we observed, suggests that a disproportionate fraction of the 77 estimated influenza-attributed deaths may have occurred in infants <1 yrs of age who may be less likely to show sufficiently specific symptoms of influenza infection to prompt requests for influenza laboratory diagnostics. In England and Argentina, high mortality rates in infants have been reported among pediatric deaths due to laboratory-confirmed influenza A(H1N1) 2009 infection
Our observation of 8 deaths by cardiomyopathy during the pandemic in the 0–4 yrs of age, and 5 deaths by spinal muscular atrophy – even if confirmed to be related to pandemic influenza infection – cannot quantitatively explain our estimate of 77 pandemic deaths. Nevertheless, cardiomyopathy and spinal muscular atrophy are not unlikely to be related to influenza. Heart disease is a known complication of influenza and has also been specifically reported for pandemic influenza A(H1N1) 2009
Unlike two earlier Dutch studies on the impact of the pandemic
In the other Dutch study
Several methods can be used to estimate influenza deaths from mortality data, like annual regression methods
Nevertheless, our estimates for influenza mortality can be affected by the way we chose to model the association of mortality with influenza activity in the population; another study
Some consider mortality categories such as pneumonia/influenza and respiratory/cardiovascular deaths more specific for influenza than all-cause mortality
Finally, the residual life expectancy that we used to estimate the years-of-life-lost may be lower for people with underlying medical conditions (who are more likely to die from influenza; the so called “harvesting effect”), leading to a potential overestimation of years-of-life-lost. In an earlier study on laboratory-confirmed pandemic deaths, modifying the life expectancy based on reported underlying disease resulted in a reduction of years-of-life-lost by over a third
We found that the pandemic had an overall moderate impact on mortality compared to the 10 preceding seasonal epidemics; disproportionate mortality in young children was observed, as opposed to low mortality in ≥75 yrs of age. This resulted in a far below average total number of pandemic deaths compared to seasonal influenza, but a total number of years-of-life-lost only somewhat below average. To assess the impact of future pandemics, attributing age-specific all-cause mortality to influenza should be preferred above comparing laboratory-confirmed deaths to historical estimates for seasonal influenza. For other countries and continents, pandemic mortality should be estimated relative to seasonal influenza as well, to assess the worldwide impact of the 2009 pandemic.
Weekly all-cause mortality by age category and ILI-incidence in time.
(DOC)
Details on regression models.
(DOC)
Sensitivity analysis.
(DOC)
Deaths in 0–4 yrs of age during the pandemic compared to seasonal influenza; distribution of age and causes of death.
(DOC)
We thank the members of the Dutch Working Group on Clinical Virology, for collecting and providing weekly positive diagnostic results.