Introduction
With the emergence of the Omicron variant of SARS-CoV-2, it became clear that the relative risk for severe outcomes of COVID-19 in adults is quite lower for the Omicron variant compared to the Delta variant [Reference Nyberg, Ferguson, Nash, Webster, Flaxman, Andrews, Hinsley, Bernal, Kall, Bhatt, Blomquist, Zaidi, Volz, Aziz, Harman, Funk, Abbott, Hope, Charlett, Chand, Ghani, Seaman, Dabrera, de Angelis, Presanis, Thelwall, Nyberg, Ferguson, Nash, Webster, Flaxman, Andrews, Hinsley, Lopez Bernal, Kall, Bhatt, Blomquist, Zaidi, Volz, Abdul Aziz, Harman, Funk, Abbott, Hope, Charlett, Chand, Ghani, Seaman, Dabrera, de Angelis, Presanis and Thelwall1, Reference Auvigne, Vaux, Strat, Schaeffer, Fournier, Tamandjou, Montagnat, Coignard, Levy-Bruhl and Parent du Châtelet2]. At the same time, the appearance of the Omicron variant led to an increase in the volume of severe outcomes (including hospitalisations and ICU admissions) in SARS-CoV-2-positive patients that were for a cause other than COVID-19 – see, for example, the data from France on hospitalisations and ICU admissions for a cause other than COVID-19 but with a SARS-CoV-2 infection versus admissions for COVID-19 [3]. This is related to differences in disease manifestation for Omicron infections versus Delta infections for both emergency department admissions [Reference Bouzid, Visseaux, Kassasseya, Daoud, Fémy, Hermand, Truchot, Beaune, Javaud, Peyrony, Chauvin, Vaittinada Ayar, Bourg, Riou, Marot, Bloom, Cachanado, Simon, Freund, Salmona, le Goff, Veyer, Chocron, Augustin, Rozenberg, Meritet, Schnuriger, Fourati, Khellaf, Coulomb, Gindre, Aloui, Jacquier, Eyer, Choquet, Ing, Pavlovsky, Houhou-Fidouh, Ferré, Landraud, Goulet, Yordanov, Berard, Rousseau, Morel, Vieux, Cahen, Saib, Salmona, le Goff, Veyer, Chocron, Augustin, Rozenberg, Meritet, Schnuriger, Fourati, Khellaf, Coulomb, Gindre, Aloui, Jacquier, Eyer, Choquet, Ing, Pavlovsky, Houhou-Fidouh, Ferré, Landraud, Goulet, Yordanov, Berard, Rousseau, Morel, Vieux, Cahen and Saib4], hospitalisations [Reference Leiner, Pellissier, Hohenstein, König, Schuler, Möller, Nachtigall, Bonsignore, Hindricks, Kuhlen and Bollmann5], and admissions to critical care [Reference de Prost, Audureau, Heming, Gault, Pham, Chaghouri, de Montmollin, Voiriot, Morand-Joubert, Joseph, Chaix, Préau, Favory, Guigon, Luyt, Burrel, Mayaux, Marot, Roux, Descamps, Meireles, Pène, Rozenberg, Contou, Henry, Gaudry, Brichler, Timsit, Kimmoun, Hartard, Jandeaux, Fafi-Kremer, Gabarre, Emery, Garcia-Sanchez, Jochmans, Pitsch, Annane, Azoulay, Mekontso Dessap, Rodriguez, Pawlotsky and Fourati6]. In particular, Omicron epidemics make a substantial contribution to mortality for cardiac causes, cancer, Alzheimer’s disease/neurological disorders, and other causes [7]. The number of deaths for different underlying causes that are triggered by an initial Omicron infection is expected to be quite higher compared with the number of deaths for COVID-19, though further work is needed to better characterise severe disease episodes for different principal causes for which the initial Omicron infection was not detected or treated (e.g. Omicron-associated acute myocardial infarction mortality in persons aged 25–44 years [Reference Yeo, Wang, He, Lv, Zhang, Zu, Li, Jiao, Ebinger, Patel, Cheng and Ji8]).
In December 2022, high rates of all-cause mortality, not seen since April 2020, were recorded in France [9]. Those high rates of excess all-cause mortality were related to the Omicron and the influenza epidemics during that time. Those data provide an opportunity to estimate the volume of Omicron- and influenza-associated mortality and compare it with the number of deaths with COVID-19 or influenza listed on the death certificate. Studies suggest that only 23%–38% of influenza-associated deaths have an underlying respiratory cause of death [Reference Goldstein, Viboud, Charu and Lipsitch10–Reference Quandelacy, Viboud, Charu, Lipsitch and Goldstein12], for which only a further fraction has influenza listed on the death certificate, with an additional significant contribution of influenza infections to mortality for circulatory and other non-respiratory causes [Reference Goldstein, Viboud, Charu and Lipsitch10, Reference Quandelacy, Viboud, Charu, Lipsitch and Goldstein12, Reference Kwong, Schwartz, Campitelli, Chung, Crowcroft, Karnauchow, Katz, Ko, McGeer, McNally, Richardson, Rosella, Simor, Smieja, Zahariadis and Gubbay13]. Despite the high rates of influenza-associated mortality for different principal causes, frequency of testing for influenza and the related use of antiviral medications may be relatively low for various illness episodes (including pneumonia hospitalisations) during periods of active influenza circulation in the community. A U.S. study found that frequency of testing for influenza in cases of community-acquired pneumonia in the United States had increased significantly between 2010 and 2015, whereas oseltamivir use on the first day of hospitalisation in influenza-positive cases was associated with a relative risk of 0.75; 95% CI (0.59–0.96) for 14-day in-hospital mortality [Reference Deshpande, Klompas, Yu, Imrey, Pallotta, Higgins, Haessler, Zilberberg, Lindenauer and Rothberg14]. A French study of ICU patients with severe influenza illness showed a relatively low uptake of influenza antiviral medications prior to ICU admission [Reference Fartoukh, Voiriot, Guérin, Ricard, Combes, Faure, Benghanem, de Montmollin, Tandjaoui-Lambiotte, Vieillard-Baron, Maury, Diehl, Razazi, Lemiale, Trouiller, Planquette, Savale, Heming, Marey, Carrat, Lapidus, Djibré, Teboul, Messika, Demoule, Mira, Timsit, Cohen, Page, Mekontso Dessap, Azoulay, Sanchez, Humbert, Annane and Roche15]. Another study of French patients with laboratory diagnosed influenza showed a higher influenza antiviral uptake [Reference Cizeron, Saunier, Gagneux-Brunon, Pillet, Cantais and Botelho-Nevers16] supporting the need for wider laboratory detection of influenza infections in France.
In addition to wider detection/treatment of Omicron and influenza infections, particularly in older individuals/persons with underlying health conditions, boosting vaccination coverage for both SARS-CoV-2 and influenza should help mitigate the spread of those infections in the community and the associated burden of severe outcomes, including deaths. Vaccination coverage for the second booster for COVID-19 in France is limited and largely restricted to persons aged over 60 years [17]. At the same time, there is evidence that vaccination reduces both the risk of acquisition of Omicron infections [Reference Chin, Leidner, Lamson, Lucas, Studdert, Goldhaber-Fiebert, Andrews and Salomon18] and the risk of onward transmission of infection [Reference Tan, Kwan, Rodríguez-Barraquer, Singer, Park, Lewnard, Sears and Lo19]. A French study of ICU patients with severe influenza illness found that 84.4% of those patients were eligible for influenza vaccination according to the vaccination recommendations in France, but only 52 out of 245 patients (21.2%) in whom this information was reported had been vaccinated [Reference Fartoukh, Voiriot, Guérin, Ricard, Combes, Faure, Benghanem, de Montmollin, Tandjaoui-Lambiotte, Vieillard-Baron, Maury, Diehl, Razazi, Lemiale, Trouiller, Planquette, Savale, Heming, Marey, Carrat, Lapidus, Djibré, Teboul, Messika, Demoule, Mira, Timsit, Cohen, Page, Mekontso Dessap, Azoulay, Sanchez, Humbert, Annane and Roche15]. Influenza vaccination coverage levels in different population groups in France are moderate-to-low [Reference Robert, Detournay, Levant, Uhart, Gourmelen and Cohen20], and significantly lower than influenza vaccination coverage levels in the corresponding population groups in the United States [21]. This pertains to both influenza vaccination coverage in risk groups for influenza-associated complications, and to population groups that play an important role in spreading influenza infections in the community. In particular, studies have found that children experience higher than average influenza infection rates [Reference Monto, Koopman and Longini22] and play an important role in the spread of influenza infection in different countries [Reference Worby, Chaves, Wallinga, Lipsitch, Finelli and Goldstein23], including France [Reference Cauchemez, Valleron, Boëlle, Flahault and Ferguson24]. Finally, we note that during the winter of 2022–2023, the greatest relative increases in mortality in France took place in long-term care facilities for older individuals (EHPAD) [25]. Influenza vaccination rates for healthcare workers (HCWs) in EHPAD in France are quite low [26] despite evidence of a significant benefit of HCW vaccination against influenza for all-cause mortality in nursing home residents during influenza seasons [Reference Goldstein27].
In our earlier work [Reference Goldstein, Viboud, Charu and Lipsitch10, Reference Quandelacy, Viboud, Charu, Lipsitch and Goldstein12], we developed a method for combing data on syndromic surveillance with data on virologic surveillance to estimate rates of mortality associated with the major influenza subtypes in the United States. Subsequently, this method was applied to the estimation of influenza-associated mortality in other countries [Reference Pebody, Green, Warburton, Sinnathamby, Ellis, Mølbak, Nielsen, de Lusignan and Andrews28–Reference Nielsen, Vestergaard, Richter, Schmid, Bustos, Asikainen, Trebbien, Denissov, Innos, Virtanen, Fouillet, Lytras, Gkolfinopoulou, Heiden, Grabenhenrich, Uphoff, Paldy, Bobvos, Domegan, O’Donnell, Scortichini, de Martino, Mossong, England, Melillo, van Asten, de Lange, Tønnessen, White, da Silva, Rodrigues, Larrauri, Mazagatos, Farah, Carnahan, Junker, Sinnathamby, Pebody, Andrews, Reynolds, McMenamin, Brown, Adlhoch, Penttinen, Mølbak and Krause30], including the EU population [Reference Nielsen, Vestergaard, Richter, Schmid, Bustos, Asikainen, Trebbien, Denissov, Innos, Virtanen, Fouillet, Lytras, Gkolfinopoulou, Heiden, Grabenhenrich, Uphoff, Paldy, Bobvos, Domegan, O’Donnell, Scortichini, de Martino, Mossong, England, Melillo, van Asten, de Lange, Tønnessen, White, da Silva, Rodrigues, Larrauri, Mazagatos, Farah, Carnahan, Junker, Sinnathamby, Pebody, Andrews, Reynolds, McMenamin, Brown, Adlhoch, Penttinen, Mølbak and Krause30]. Here, we adopt this framework to evaluate the contribution of influenza infections to all-cause mortality for the 2014–2015 through the 2018–2019 influenza seasons, as well as the contribution of SARS-CoV-2 and influenza infections to all-cause mortality between week 33, 2022 and week 12, 2023. The reason for omitting the period between 2020 and August 2022 from the analyses is that patterns of non-COVID-19 mortality have changed during the pandemic due to lockdowns and other factors. For example, examination of mortality in England and Wales between March 2020 and June 2022 suggested that for deaths for which COVID-19 was not coded as the underlying cause of death, there was a reduction of 367,000 in the number of deaths for respiratory causes compared to averages during the previous 5 years [31]. Additionally, the heat wave in France resulted in significant levels of excess mortality in July 2022 [32]. However, starting mid-August 2023, baseline rates of mortality not associated with SARS-CoV-2 or influenza infections in France have returned to more regular patterns, which allowed for the inclusion of the period starting mid-August 2022 into the inference model in [Reference Goldstein, Viboud, Charu and Lipsitch10, Reference Quandelacy, Viboud, Charu, Lipsitch and Goldstein12]. The aim of this work is to help inform vaccination policies for SARS-CoV-2 and influenza, and policies for detecting/treating SARS-CoV-2 and influenza infections, particularly in persons with underlying health conditions (with Omicron-associated mortality for cardiac causes and for mental&behavioural disorders examined further in the Supplementary Material).
Methods
Data availability
This manuscript is based on aggregate, de-identified publicly available data. Data on weekly numbers of influenza-like illness (ILI) consultations in metropolitan France are available from the French sentinel surveillance [33]. Sentinel data on testing of respiratory specimens for the different influenza subtypes (A/H1N1, A/H3N2, B/Victoria, and B/Yamagata) are available from WHO FluNET [34]. Data on the daily number of deaths in France starting 2015 are available from [9]. Data on population in France are available from [35]. Electronic records for deaths with COVID-19 listed on the death certificate are available from [36]. We note that many SARS-CoV-2-associated deaths are not listed in [36], and weekly death rates in [36] are used as a covariate in the regression model rather than the estimate of the contribution of COVID-19, or SARS-CoV2 infections to mortality.
Incidence indicators for the major influenza subtypes and for SARS-CoV-2-associated deaths
Not all ILI consultations in the sentinel data [33] correspond to influenza infections, and those that do, correspond to infection with different influenza subtypes. For each influenza subtype (e.g. A/H3N2), we define an indicator for the incidence of that subtype on week
$ t $
(e.g.
$ A/H3N2(t) $
) as
$$ A/H3N2(t)={\displaystyle \begin{array}{l}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}\ \mathrm{o}\mathrm{f}\hskip2pt \mathrm{I}\mathrm{L}\mathrm{I}\hskip2pt \mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s}\mathrm{u}\mathrm{l}\mathrm{t}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{s}\hskip2pt \mathrm{p}\mathrm{e}\mathrm{r}\hskip2pt 100,000\hskip2pt \mathrm{p}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{o}\mathrm{n}\mathrm{s}\hskip2pt \mathrm{o}\mathrm{n}\hskip2pt \mathrm{w}\mathrm{e}\mathrm{e}\mathrm{k}\hskip2pt t\hskip2pt \mathrm{i}\mathrm{n}\hskip1.5pt [33]\\ {}\cdotp \mathrm{P}\mathrm{e}\mathrm{r}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{t}\ \mathrm{o}\mathrm{f}\ \mathrm{r}\mathrm{e}\mathrm{s}\mathrm{p}\mathrm{i}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{y}\ \mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{s}\hskip2pt \mathrm{o}\mathrm{n}\hskip2pt \mathrm{w}\mathrm{e}\mathrm{e}\mathrm{k}\hskip2pt t\hskip2pt \mathrm{i}\mathrm{n}\hskip1.5pt [34]\\ {}\hskip1.6em \mathrm{p}\mathrm{o}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{v}\mathrm{e}\ \mathrm{f}\mathrm{o}\mathrm{r}\ \mathrm{i}\mathrm{n}\mathrm{f}\mathrm{l}\mathrm{u}\mathrm{e}\mathrm{n}\mathrm{z}\mathrm{a}\hskip2pt \mathrm{A}/\mathrm{H}3\mathrm{N}2\end{array}} $$
To relate the incidence indicators for the major influenza subtypes to weekly levels of all-cause mortality per 100,000 persons in France [9], we note that age distribution of influenza cases changes with the appearance of antigenically novel influenza strains, and this changes the relation between rates of influenza-associated ILI (incidence indicators in equation (1)) and rates of influenza-associated mortality. The relevant antigenic changes for our study period were (a) the appearance of an antigenically/genetically novel A/H3N2 strain during the 2014–2015 influenza season [Reference Skowronski, Chambers, Sabaiduc, de Serres, Winter, Dickinson, Krajden, Gubbay, Drews, Martineau, Eshaghi, Kwindt, Bastien and Li37]; (b) the appearance of the novel influenza B/Yamagata strain during the 2017–2018 season [Reference Nielsen, Vestergaard, Richter, Schmid, Bustos, Asikainen, Trebbien, Denissov, Innos, Virtanen, Fouillet, Lytras, Gkolfinopoulou, Heiden, Grabenhenrich, Uphoff, Paldy, Bobvos, Domegan, O’Donnell, Scortichini, de Martino, Mossong, England, Melillo, van Asten, de Lange, Tønnessen, White, da Silva, Rodrigues, Larrauri, Mazagatos, Farah, Carnahan, Junker, Sinnathamby, Pebody, Andrews, Reynolds, McMenamin, Brown, Adlhoch, Penttinen, Mølbak and Krause30]; and (c) the 2018–2019 A/H3N2 epidemic strains belonging to several clades and exhibiting different immunity profiles for different birth cohorts [Reference Kissling, Pozo, Buda, Vilcu, Gherasim, Brytting, Domegan, Gómez, Meijer, Lazar, Vučina, Dürrwald, van der Werf, Larrauri, Enkirch, O’Donnell, Guiomar, Hooiveld, Petrović, Stoian, Penttinen and Valenciano38]. Correspondingly, for the model relating A/H3N2 circulation to associated mortality, we split the A/H3N2 incidence indicator into three:
$ A/H3N{2}_1 $
, equalling the A/H3N2 incidence indicator for the period from January 2015 to September 2015, and equalling to 0 for later weeks;
$ A/H3N{2}_2 $
, equalling the A/H3N2 incidence indicator for the period from October 2015 to September 2018, and equalling to 0 for other weeks; and
$ A/H3N{2}_3 $
, equalling the A/H3N2 incidence indicator for the period starting October 2018, and equalling to 0 for other weeks. Similarly, we split the B/Yamagata incidence indicator into two, corresponding to the periods before and starting the 2017–2018 B/Yamagata epidemic. Finally, we note that it takes 1–2 weeks between influenza illness and influenza-associated mortality [Reference Goldstein, Viboud, Charu and Lipsitch10]. Correspondingly, we use a regression framework (equation (2))) to relate the mortality rate
$ M(t) $
on week
$ t $
to the shifted incidence indicators, e.g.
$$ {A/H3N2}^s(t)=s\cdotp A/H3N2(t-2)+(1-s)\cdotp A/H3N2(t-1), $$
where the parameter
$ s $
(common for all incidence indicators) is chosen to minimise the R
2 for the model fit. For the estimation of the rates of SARS-CoV-2-associated deaths, we use the weekly rates
$ {Cov}_e(t) $
of deaths with COVID-19 listed on the electronic death certificate in [36] as an indicator (covariate in the regression model).
Regression model: The regression model that we use is
$$ M(t)={\displaystyle \begin{array}{l}{\beta}_1\cdotp {A/H1N1}^s(t)+{\beta}_2\cdotp {A/H3N2}_1^s(t)+{\beta}_3\cdotp {A/H3N2}_2^s(t)\\ {}+{\beta}_4\cdotp {A/H3N2}_3^s(t)+{\beta}_5\cdotp {B/Victoria}^s(t)+{\beta}_6\cdotp {B/Yamagata}_1^s(t)\\ {}+\hskip2px {\beta}_7\cdotp {B/Yamagata}_2^s(t)+{\beta}_8\cdotp {Cov}_e(t)+Baseline(t)\\ {}+\hskip2px Trend(t)+Noise.\ \end{array}} $$
The time period that we include in the analysis is week 3, 2015 through week 2, 2020, and week 33, 2022 through week 12, 2023. The baseline represents weekly rates of mortality not associated with influenza or SARS-CoV-2 circulation, and
$ Baseline(t) $
is modelled to have annual periodicity in week
$ t $
. We use periodic cubic splines to model the baseline mortality rates whose shape is unknown [Reference Goldstein, Viboud, Charu and Lipsitch10, Reference Quandelacy, Viboud, Charu, Lipsitch and Goldstein12]. The trend
$ Trend(t) $
is modelled as a quadratic polynomial in week
$ t $
. Finally, to account for the autocorrelation in the noise, we use a bootstrap procedure (resampling the noise on different weeks) to estimate the confidence bounds for various quantities evaluated in the model [Reference Goldstein, Viboud, Charu and Lipsitch10].
Results
Figure 1 plots the results of the regression model in equation (2). The model fits in Figure 1 are generally temporally consistent, with baseline rates of non-influenza mortality + trend + the contribution of influenza infections to all-cause mortality (difference between the red and the green curves) explaining the patterns of mortality prior to the COVID-19 pandemic, while baseline + trend + contribution of influenza infections + contribution of SARS-CoV-2 infections to mortality explaining the patterns of mortality between week 33, 2022 and week 12, 2023 (with SARS-CoV-2 infections being responsible for most of excess mortality during that period).
Figure 1. Weekly mortality rates for all causes between week 3, 2015 and week 2, 2020, and week 33, 2022 and week 12, 2023 (black curve), the baseline + trend for the rates of mortality not associated with influenza or SARS-CoV-2 infections in France (green curve), the model fit (baseline + trend + contribution of influenza infection +contribution of SARS-CoV-2 infection to mortality, red curve), and the model fit between week 33, 2022 and week 12, 2023 with the contribution of SARS-CoV-2 infection removed.
Table 1 gives the estimates of the contribution of influenza infections to all-cause mortality for the 2014–2015 through the 2018–2019 influenza seasons, and the contribution of SARS-CoV-2 and influenza infections to all-cause mortality between week 33, 2022 and week 12, 2023. For the 2014–2015 through the 2018–2019 seasons, influenza was associated with an annual average of 15,654 (95% CI (13,013, 18,340)) deaths, whereas between week 33, 2022 and week 12, 2023, we estimated 7,851 (5,213, 10,463) influenza-associated deaths and 32,607 (20,794, 44,496) SARS-CoV-2-associated deaths.
Table 1. The contribution of influenza infection to all-cause mortality for the 2014–2015 through the 2018–2019 influenza seasons, and the contribution of SARS-CoV-2 and influenza infections to all-cause mortality between week 33, 2022 and week 12, 2023
Data on mortality with COVID-19 listed on the death certificate in France are available through the end of 2022 [36]. For the period between weeks 33–52 in 2022, we estimated 23,983 (15,307, 32,620) SARS-CoV-2-associated deaths in France, compared with 12,811 deaths with COVID-19 listed on the death certificate [36], and 8,639 in-hospital deaths with COVID-19 during the same period [39].
Discussion
During the winter of 2022–2023, high levels of excess mortality (not seen since the first COVID-19 wave in April 2020) were recorded in France [9], with excess mortality being primarily associated with the Omicron epidemic [3], but also the influenza epidemic during that period [40]. The emergence of the Omicron variant saw a higher proportion of ICU admissions and deaths with a SARS-CoV-2 infection which were for causes other than COVID-19 compared to the earlier SARS-CoV-2 variants [3]. Additionally, most influenza-associated deaths do not have influenza listed on the death certificate [Reference Goldstein, Viboud, Charu and Lipsitch10, Reference Quandelacy, Viboud, Charu, Lipsitch and Goldstein12]. A better understanding is required of the contribution of SARS-CoV-2 and influenza infections to hospitalisations and mortality for different principal causes to inform vaccination policies for SARS-CoV-2 and influenza, and policies for detecting/treating SARS-CoV-2 and influenza infections, particularly in persons with underlying health conditions.
We have estimated (15,307, 32,620) SARS-CoV-2-associated deaths in France between weeks 32–52 in 2022, compared with 12,811 deaths with COVID-19 listed on the death certificate during the same period. In the Supplementary Material, we show a significant contribution of Omicron infections to mortality for cardiac disease and for mental&behavioural disorders without COVID-19 being listed on the death certificate. A significant contribution of Omicron infections to mortality for cardiac causes, cancer, Alzheimer’s disease/neurological disorders, and other causes was also found in [7]. Those findings suggest the need to widen the detection/treatment of Omicrons infections, particularly in individuals with underlying health conditions, as well as the need to increase COVID-19 booster vaccination coverage in the whole population to mitigate SARS-CoV-2 transmission in the community. We have also estimated that influenza was responsible for high levels of associated mortality prior to the pandemic, with an average of 15,654 (95% CI (13,013, 18,340)) annual deaths associated with influenza infections during the 2014–2015 through the 2018–2019 seasons, as well as 7,851 (5,213, 10,463) influenza-associated deaths between week 33, 2022 and week 12, 2023. Influenza vaccination coverage in non-elderly adults and in children in France is quite lower compared with the United States [Reference Robert, Detournay, Levant, Uhart, Gourmelen and Cohen20, 21], with children known to play a significant role in influenza transmission in the community [Reference Monto, Koopman and Longini22–Reference Cauchemez, Valleron, Boëlle, Flahault and Ferguson24]. Residents of establishments for dependent elderly persons (EHPAD) in France represent a sizeable share of all-cause mortality [Reference Botton, Drouin, Bertrand, Jabagi, Weill, Zureik and Dray-Spira41] and influenza-associated mortality in the French population. Rates of influenza vaccination for HCWs in EHPAD in France are quite low [26], while HCW vaccination against influenza has a significant effect on all-cause mortality in nursing home residents during influenza seasons [Reference Goldstein27]. Additionally, influenza vaccine effectiveness in older individuals can be quite low [42], and types of influenza vaccines administered to older individuals play a role in preventing adverse outcomes associated with influenza infections. In the United States, recent Advisory Committee on Immunization Practices recommendations stipulate that adults aged ≥65 years should preferentially receive any one of the following higher dose or adjuvanted influenza vaccines: quadrivalent high-dose inactivated influenza vaccine (HD-IIV4), quadrivalent recombinant influenza vaccine (RIV4), or quadrivalent adjuvanted inactivated influenza vaccine (aIIV4) [43]. Finally, antiviral medications for influenza are prescribed quite frequently in France when an influenza infection is laboratory-detected [Reference Cizeron, Saunier, Gagneux-Brunon, Pillet, Cantais and Botelho-Nevers16], while antiviral treatment of influenza infection in hospitalisations for respiratory causes, including pneumonia, was found to have benefits in terms of reducing mortality rates in hospitalised patients [Reference Deshpande, Klompas, Yu, Imrey, Pallotta, Higgins, Haessler, Zilberberg, Lindenauer and Rothberg14].
Our results have some limitations. Influenza surveillance data in France pertain to mainland France [33], whereas we have used data on mortality for the whole of France. Additionally, sentinel data on testing for viral specimens [34] have a moderate sample size and may not represent all of France. We note that influenza epidemics exhibit a great deal of temporal synchrony [Reference Schanzer, Langley, Dummer and Aziz44, Reference Morris, Freiesleben de Blasio, Viboud, Wesolowski, Bjørnstad and Grenfell45] which should help address the above limitations. Finally, despite the fact that we split some of the influenza subtype incidence indicators into several time periods, where might still be temporal variability in the relation between the incidence indicators used in this paper and rates of associated mortality? For example, while model fits are generally temporally consistent (Figure 1), the model fit for the mortality data for the 2017–2018 season is worse compared with other influenza seasons, which might be related to the fact that the influenza subtypes that circulated during that season (A/H1N1 and B/Yamagata) have different age distributions feeding into one ILI data stream.
Conclusions
Our findings about the high rates of influenza-associated mortality in France, the very high rates of influenza-associated mortality in individuals aged over 75 years during certain influenza seasons in France [46], and the low influenza vaccination rates in HCWs in France [26] suggest the need for (a) boosting influenza vaccination rates for HCWs and other population groups, possibly including children; and (b) wider testing for influenza infection in respiratory illness episodes with different principal diagnoses (including pneumonia) during periods of active influenza circulation in combination with the use of antiviral medications [Reference Deshpande, Klompas, Yu, Imrey, Pallotta, Higgins, Haessler, Zilberberg, Lindenauer and Rothberg14]. In the event of the recrudescence of Omicron circulation in France, efforts should be undertaken to administer booster vaccination for COVID-19 to HCWs, with coverage levels for booster vaccines containing the Omicron component in HCWs in France being relatively low [3]. Wider detection and treatment of Omicron infections, particularly in older individuals/persons with underlying health conditions such as cardiac disease and mental/behavioural disorders (Supplementary Material) should help mitigate the mortality burden of future Omicron epidemic waves.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0950268823001358.
Data availability statement
This manuscript is based on aggregate, de-identified publicly available data that can be accessed through the links in references [9, 33–36].
Author contribution
Conceptualization: E.G.; Formal analysis: E.G.; Investigation: E.G.; Methodology: E.G.; Software: E.G.; Writing – original draft: E.G.; Writing – review & editing: E.G.
Financial support
This work received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interest
The author declares no competing interests exist.
Open access
