Zylberman (2021) noted that “this pandemic began with the first case, but it will not end with the last case (…) one cannot date the end of a pandemic and the beginning of an endemic. Yet, in mid-April, the French president slipped in an interview (Garnier (2022)) that “society [is] on its way out of COVID” implying that the SARS-CoV-2 pandemic was over. At the same time, the virus was still killing more than 100 people a day, according to official statistics. While it is legitimate to question what exactly is a “COVID death“, it may seem surprising that 100 deaths per day (for more than three consecutive months) were met with such indifference, and that such a level was interpreted as the end of the pandemic.
(this is the English translation of a blog post published a few days ago)
The value of life, and the value of health care
Comparing – or prioritizing, as Karpf (2001) puts it – deaths is a difficult exercise. Keating (2013) asked the question “is it wrong to care more about 4 deaths in Boston than 80 in Syria?” (following the double bombing at the Boston Marathon) without really finding a morally satisfactory answer. And it is even more confusing when you look at the same territory and the same pandemic: in mid-March, the first containment was announced after 175 cumulative deaths (in a few weeks). Two years later, the pandemic was announced as over, while exceeding 175 deaths per day was still “the norm” (the day after Garnier (2022) postponed the declaration that “the company [is] out of COVID”, there were 189 deaths). This is a higher level than the “start of the second wave” in mid-October 2020. Clearly, the deaths at the beginning and the deaths at the end do not have the same meaning, not to mention the same value. If the two situations are not comparable (60% of the population is vaccinated at the end of spring 2022), the indifference around the current deaths may surprise and question. And as Charters (2021) stated, “declaring an end to the pandemic is about deciding how much illness, death and disruption is accepted and acceptable as a part of normal life“.
Any public policy that improves the well-being of all is naturally justified, this is the idea of the Pareto optimum. But what about a policy that satisfies some citizens and displeases others, a policy that destroys the economy to save lives? Every political decision is a trade-off, a choice between two (or sometimes more) alternatives. If doctors like to remind us that health has no price, it has a cost, and this cost is often ignored. In France, patients are often unaware of the magnitude of the cost of a treatment, the only visible part being the remaining “reste à charge“, which is often very low compared to the real cost (for the hospital). For instance, in France, a day in intensive care costs at least 3000 euros. As a comparison, assuming that 11% of the GDP in France is devoted to current health expenses, this means that a French person spends, on average, a little more than 3500 euros per year on health.
Attributing a monetary or pecuniary value to a treatment, or more generally to a life, does not aim to attribute a market value to humans, but simply to help public decision-makers make choices on the basis of less passionate and more “rational” criteria, if not “objective” ones, as explained by Charpentier & Cherrier (2019). This does not prevent debates on the quantification of our society, but giving a value to life makes decision-making more transparent, and sometimes also more coherent with the scenarios selected, especially when it is a question of making decisions between two objectives expressed in such different ways, between saving lives on the one hand, and maintaining economic activity on the other (as the question may have arisen when confinements were put in place, or lifted). In health (and it will be the same in public health), with a fixed budget, giving priority to a leukaemia patient means financing a treatment that will cost more than 100,000 euros, and that will therefore be at the expense of an investment in palliative care, or the recruitment of nurses. And if we do not wish to impose this budgetary constraint, arguing that the health budget must be increased, we must keep in mind that increasing the health budget means decreasing the budget for education or justice.
It is natural enough to ask how much we would be willing to pay, collectively, for an additional year of life (the “Value of a Life Year“, VOLY), or an additional year of life adjusted for health status (the “Quality Adjusted Life-Year“, QALY, or “Willingness-To-Pay for QALY“, WTP-QALY). While these quantities allow us to make choices about health, they also allow us to quantify the value of a death due to (or related to) the pandemic. This approach may bring to mind Foot’s (1978) streetcar thought experiment, repeated in the moralmachine.net experiment (related to ethics and autonomous vehicles) where one asks which option is less morally reprehensible, between running over a 10 year old child and running over 10 70 year old people. As an illustration, Figure 1 shows not the number of deaths, week after week, but an estimate of the number of years of life lost (as an example, the death of a 65 year old woman corresponds to 25 years of life lost, corresponding to the remaining life expectancy).
Fig 1: Number of years of life lost, each week1 (YLL), between 2020 and 2021 (author’s calculations, INED mortality data and based on residual life expectancy at each age).
In Great Britain, a year of life in good health (in other words, weighted by its quality, as defined by the QALY) has been estimated at around 35,000 euros, as mentioned by Wang et al. (2018), with a higher threshold of acceptability at the end of life or for more serious pathologies (one might be prepared to spend 100,000 euros, for example, for an extra year of life for a person with cancer). Cameron et al (2018) return to the difficulty of imposing explicit thresholds, and propose instead to calculate implicit life values based on choices made by public health authorities. “So far, Australia’s national government has committed some $300 billion to soften the economic consequences of state governments’ enforced social distancing. Even if mandatory shut-down really was all that avoided the initially-predicted 150,000 deaths, that still works out at about $2 million per life saved. If the average age of those who would have died is 80, even with roughly 10 years of expected life left, that’s still $200,000 per quality life year – or substantially beyond what governments are usually prepared to pay for life-saving drugs” (or 135,000 euros) in the words of Abbot (2021), the Australian Prime Minister. A quick reasoning in France shows that for a cost of 200 billion Euros (corresponding to 8% of the GDP, a loss estimated by many economists in 2020), if 700,000 lives were saved (as estimated by Flaxman et al. (2020)), one death cost a little less than 300,000 Euros, or 30,000 Euros if one life lost corresponds to a dozen QALYs. However, as we can see, these calculations are based on many estimates which are, for the most part, very uncertain (with orders of magnitude that can vary by a factor of 1 to 10). QALYs are likely to be an important tool when the issue of “long-COVID” (and all the other symptoms mentioned by Fernández-de-Las-Peñas et al. (2021)) is finally addressed seriously.
Although the long-term consequences of the SARS-CoV-2 pandemic are still difficult to quantify, it can be noted that the human cost (in terms of deaths or years of life lost) has been significant, greater than annual influenza and other endemic diseases. For comparison, the 2018-2019 influenza epidemic lasted 8 weeks, causing about 12,000 deaths of all ages (although the vast majority of those who died were over 85 years of age), or 215 deaths per day during the epidemic phase (nearly equivalent to 2 years of COVID-19 – considered continuously); the 2017-2018 influenza lasted longer (about 16 weeks) and killed about 13,000 people, or 115 deaths per day. These orders of magnitude seem clearly “acceptable” (or “tacitly acceptable“) to society, based on the lack of headlines in the newspapers. In contrast, during the same period, about 3,500 people lost their lives in road accidents, or just under 10 per day, and this figure is often seen as too high2. But perhaps this has something to do with the fact that the number of deaths in road accidents is still very low. But perhaps it has something to do with who dies?
Dying during the pandemic, from covid or with covid?
Counting the dead is not an easy exercise. In 1918-1919, the vast majority of deaths were not due to the influenza virus stricto sensu but to other diseases often caused by bacterial super-infections (pneumonia or tuberculosis) that attacked individuals largely weakened by influenza. Antibiotics and resuscitation services being absent a hundred years ago, it is clear that it will be complicated to compare pandemics. In particular, in 1918, adults between 15 and 45 years of age were most affected, whereas the SARS-CoV-2 pandemic affected mostly older people.
Some people died with the virus, but from another cause (e.g. a traffic accident). Others died of pneumonia caused by the virus. The WHO guide stated that if the deceased had certain chronic conditions, these should be added to the second part of the medical certificate of cause of death3. Specifically, when an infected person is admitted to the ICU with COVID-19 infection and then dies there, that person can be said to have died “because” of COVID-19, not because of their comorbidities. Nevertheless, a positive test was required, and many cases of failure to test in the first few months have been reported, especially for deaths in nursing homes.
Recall that “cause of death” is a clearly defined concept. For example, WHO (1967) indicated that “the causes of death to be mentioned on the medical certificate of cause of death are all the diseases, morbid conditions or injuries that led to or contributed to the death and the circumstances of the accident or violence that led to these injuries“. Thus, obesity or hypertension are not “causes of death” but risk factors. People do not die from obesity but from the diseases it encourages, such as cardiovascular disease. There is an almost exhaustive list of “causes of death”, with unfortunately many “garbage codes”, i.e. causes of death that are either useless or non-plausible. To say that a cause of death is a cardiorespiratory arrest is not very informative, for example, because everyone dies of a cardiac arrest. Much thought has gone into producing the Global Burden of Disease classifications, as described by Naghavi et al. (2010).
But one difficulty persists: there may be a sequence of causes of death that fall within a temporal sequence. For example, a person who dies from a pulmonary embolism (direct cause), from the aftermath of a pathological fracture (intermediate cause), which was encouraged by a secondary cancer of the femur (another intermediate cause), itself caused by a breast cancer (the initial cause). Which cause(s) should be retained? The prevailing logic is that breast cancer will be identified in the statistics as the main cause of death (because it is the initial cause). In a way, it is the cause (in the sense of disease or accident) that starts the sequence leading to death. Désesquelles et al (2010) return at length to the analysis of multiple causes. COVID-19 makes the analysis complex precisely because the virus can cause multiple problems – pneumonia, respiratory failure, blood clots, stroke, heart attack, etc. – any of which can lead to death. – each of which can lead to death. Most people who have died after contracting COVID-19 have had, overwhelmingly, one or more comorbidities. How do we know how many people died from COVID-19, and how many died with COVID-19? Again, WHO has attempted to set rules, “a death due to COVID-19 is defined as a death resulting from clinically compatible disease in a probable or confirmed case of COVID-19, unless there is another obvious cause of death unrelated to COVID-19 (e.g., trauma),” WHO (2022). But each country has adapted the definition. For example, in England, any death within 28 days of a positive PCR test for SARS-CoV-2 was recorded as a COVID-19-related death. As recounted by Ayoubkhani et al (2021), more than 40,000 people with COVID-19 discharged from hospital were followed for 140 days, and a readmission rate of 31% was found, and on the order of 10% died on readmission. While not all of these deaths were directly due to COVID-19 (they were not counted as such), it is conceivable that many of the deaths were accelerated by the virus.
Another interesting statistic was the analysis of excess mortality. This is the method used by Turquan (1891) to estimate the number of deaths following an 1890 influenza epidemic, as recounted by Vagneron (2020). The difficulty is to have a reference against which to compare oneself, often an average of past years, corrected for the aging of the population. In 2020, about one in five Europeans is over 65, whereas in 2000, only 20 years ago, it was one in eight. And the older a population gets, the more deaths we can expect to see. The advantage of the excess mortality analysis is that in addition to the deaths from (or with) COVID-19, we will find the people who died because of the disease, or rather the health management of the disease. As Arturo Casadevall, Professor of Immunology and Infectious Diseases at the Johns Hopkins School of Public Health, stated, “the number of deaths attributed to COVID-19 does not include other deaths associated with the pandemic, such as those caused by the absence of proper care for other conditions because the healthcare system was focused on COVID-19, and much routine care such as cancer screening was slowed or postponed” (quoted by Lang (2022)). For we are indeed in the aftermath of the pandemic.
Looking back at the beginning of the pandemic
At the very beginning of the pandemic, in March 2020, three strategies were on the table: do nothing and rely on collective immunity (as Sweden and Brazil did, believing in a form of collective mithridatization); implement social restriction measures that were reinforced and then reduced according to the data (Aghion et al. (2021) referred to a “stop and go” approach) used in much of Europe; and finally, a very localized and strong initial containment (“test, trace, and isolate” by Smith et al. (2021), sometimes also called “zero-covid“) implemented in much of Asia.
The idea of containment gained acceptance relatively quickly, although historically, such measures have rarely been taken. During the “Spanish flu” in 1918-1919, the authorities had refused to do so, by not cancelling shows or public events, mainly in order not to undermine the morale of the population, which had just come out of several years of war. Forty years later, the “Asian flu” vaccine of 1957-1959 travels around the world, without anyone doing anything. No debate in parliament, “the Asian flu remains a laboratory curiosity” as it was said in front of the French Academy of Science in October 1957. If a city councilor of Paris proposes to postpone the beginning of the school year (without being listened to), the only priority is the fear of an increase in absenteeism in the companies. The same thing happened with the “Hong Kong flu” in 1968-1970. As a reminder, these two influenzas caused 12,000 (officially, and according to the statistics of the time, such as Pressat (1960)) and 31,000 deaths in France. Closer to us, epidemiologists sounded the alarm following 6 deaths in 1997 due to an avian epizootic in Hong Kong, and two years later due to the “swine flu” in Mexico, as Zylberman (2009) reports. After these two false alarms, the other example is of course the “bird flu” of 2009. At the time, the French Minister of Health, Roselyne Bachelot, had built up a large stockpile of equipment to deal with a pandemic announced and named by the WHO “influenza A (H1N1)”. One billion surgical masks and 700 million FFP2 masks have been purchased. In France, 312 deaths have been recorded. At the beginning of 2020, the SARS-CoV-2 virus kills in China, and starts to arrive in Europe. The official reaction will be a “strict containment” of almost two months. One wonders if this reaction would not have been weaker if the leaders had not been obliged to follow the Chinese example.
A dedicated scientific council is constituted, seeking a subtle balance between scientific truth (based on knowledge that is constituted in real time) and a democratic, or political, will. The great difficulty in reading the beginning of the crisis is the context of “depoliticization” as defined by Burnham (2001), namely “to place at one remove the politically contested character of governing“, describing the attempt to distance the politically contestable character of governance, and to escape any responsibility. For Holdren (2022), this is a rule of denial: the government will make decisions without appearing to make them, treat the consequences as inevitable, and try to place authority elsewhere in order to avoid accountability.
The end of the pandemic? For whom?
In mid-April 2022, two years after the first containment, the head of state spoke of the end of the pandemic. But at the same time, epidemiologists such as Rochelle Walensky (CDC directive in the United States, quoted by Healy (2022)) announced that we were still far from the end. Indeed, unlike vaccines against diseases such as measles, polio and diphtheria, currently available vaccines against COVID-19 do not provide lifelong immunity. Nor does a past infection, even a very severe one. And since the virus is now present in animals (mink, hamsters and deer), the threat of a resurgence will always be present, possibly with a contagious and highly lethal variant (and one that existing vaccines will not be as effective against). If we place ourselves in an optimistic scenario of “end of pandemic” (as suggested by the decrees published in mid-May 2022, in France, which abolish the obligation to wear a mask in public transport, for example), can we believe that the pandemic is over for everyone? Because the pandemic has not affected everyone equally.
Based on the 2020 and 2021 statistics, people over 75 years of age were almost 150 times more likely to die than a 20 year old. Unvaccinated persons (including immunocompromised persons) were 50 times more likely to die than vaccinated persons. Using data from the United States, Pathak et al. (2021) found that blue-collar workers were 5 times more likely to die from COVID than college graduates. People of Hispanic or black origin died at much higher rates than whites.
In other words, for a white, vaccinated male between the ages of 30 and 60, the risk of dying from COVID-19 may be perceived as very low, and therefore socially acceptable. Taking cynicism a step further, for the antivaxers, the risk is zero, so why worry? As for the elderly or sick, they would have died anyway, right? We find ourselves in the situation described at the beginning, where we tend to be indifferent to distant death, the deaths in Syria, or in the Mediterranean Sea. Yet the hundred or so deaths a day occur among us, affecting members of our families, and not to be concerned about them shows a lack of empathy that can be frightening. To make these deaths invisible is also to hide the evidence of a political failure, and unfortunately this strategy prevents us from learning from the management of the crisis.
What consequences should be drawn?
Before pretending to manage crises of the magnitude of the SARS-COV-2 pandemic, the first thought is that it might be wiser to try to avoid them. The pandemic “turned out to be an exceptional disaster (…) we were forced to take imperfect measures (…) No one was prepared for the spread or the nightmarish scale of the pandemic“. These sentences sum up the assessment of the situation in March 2020, except that they are from Zylberman (2017) and are taken from Doctors Without Borders reports about Ebola in 2014 and 2015. At that time, the Haut Conseil de la Santé Publique estimated that caring for a single patient would require 2 to 3 caregivers, rotating every 6 hours, and that 20 beds would be identified and sanctuarized in a dozen hospitals. In retrospect, these capacities seem incredibly paltry. Beyond the value of life, it is the cost of prevention that should be calculated, because it is difficult to make a fair decision, which is neither in under-reaction, nor over-reaction. For the latter case, we can mention the H1N1 flu in 2009, but we can also think back to the “swine flu” in 1976 (described in Sencer & Millar (2006)), where, following a few cases of flu in New Jersey, and the fear of a new “Spanish flu” in the context of a presidential election, Gerald Ford vaccinated millions of people, before stopping due to the lack of a proven epidemic.
As Bénéplanc et al (2022) remind us, the difference between risk and threat marks the limits of uncertainty. Predicting a disaster does not mean issuing a simple warning, a prediction, as Bill Gates did in 2015, when he announced that a major health crisis would occur in the next 10 years. The problem with disasters is that they often require securing goods (or services) that are totally unnecessary “in normal times.” In the case of the pandemic, it was the supply of masks or respirators that was so lacking in early 2020. We find here the effect highlighted by Gillingham et al (2009) on energy markets: in (strong) regulation, the private market cannot cope with consumption peaks. Without wishing to play the bird of ill omen, it is worrying that, for a country that has invested so much in nuclear energy, France has so few fallout shelters. While Switzerland has a little more than 300,000 bunkers capable of accommodating nine million people (i.e. more than its population), apart from a few well-known examples (the bunker under the Elysée Palace, or in the underground passages of Radio France), the number of places for civilians in France would be close to 0. What can we say about the risks linked to climate change, where every year we seem to be surprised to break records? Is it so difficult to anticipate the risks, and to make preventive expenditures that will save thousands of lives tomorrow? More than two and a half years after the first cases of a pandemic that will go down in history, we are left with the bitter taste that nothing has been learned, nothing has been understood, and that the desire to turn the page by pretending that nothing happened has finally prevailed.
1 For the 2003 heat wave, as INSEE reminds us, there are no official statistics on the count of deaths “caused by the heat wave” but the heat wave affected older people. Also, the 15,000 deaths of the heat wave. Counting a loss of 5 years of life per death, we arrive at 75,000 years lost, compared to 1,400,000 for the pandemic! That is almost 20 times more.
2 To continue the comparisons, the September 11, 2001 attacks against the World Trade Center killed 2977 people. At the end of August 2005, Hurricane Katrina killed 1833 people. During the month of January 2021, there were more than 3000 COVID-19 deaths per day in the United States.
3 Many thanks to Bruno Masquelier of the Center for Demographic Research in Leuven for his insights.
Abbot, T. Australia and the Coronavirus Crisis. Policy Exchange (2021)
Aghion, P., Artus, P., Oliu-Barton, M., & Pradelski, B. Aiming for zero Covid-19 to ensure economic growth. VoxEU.org, 31 (2021)
Ayoubkhani, D., Khunti, K., Nafilyan, V., Maddox, T., Humberstone, B., Diamond, I., & Banerjee, A. Epidemiology of post-COVID syndrome following hospitalisation with coronavirus: a retrospective cohort study. MedRxiv. (2021)
Bénéplanc, G., Charpentier, A. & Thourot, P. Manuel d’assurance. Presses Universitaires de France (2022)
Burnham, P. New Labour and the politics of depoliticisation. The British Journal of Politics & International Relations, 3(2), 127-149. (2001).
Cameron, D., Ubels, J., & Norström, F. On what basis are medical cost-effectiveness thresholds set? Clashing opinions and an absence of data: a systematic review. Global health action, 11(1), 1447828 (2018)
Charpentier, A. & Cherrier, E. La valeur de la vie humaine. Risques, 107 (2019)
Charters, E, Heitman, K. How epidemics end. Centaurus, 63: 210– 224 (2021)
Désesquelles, A., et al. Revisiting the mortality of France and Italy with the multiple-cause-of-death approach, Demographic research, 23 (28), p.771-806 (2010)
Fernández-de-Las-Peñas, C., Palacios-Ceña, D., Gómez-Mayordomo, V., Cuadrado, M. L., & Florencio, L. L. (2021). Defining post-COVID symptoms (post-acute COVID, long COVID, persistent post-COVID): an integrative classification. International journal of environmental research and public health, 18(5), 2621
Flaxman, S. et al. Estimating the effects of non-pharmaceutical interventions on COVID-19 in Europe. Nature 584(7820), 257-261 (2020)
Foot, P. The Problem of Abortion and the Doctrine of the Double Effect, Virtues and Vices, Oxford, Basil Blackwell, (1978)
Ghabri, S., Poullié, A. I., Autin, E., & Josselin, J. M. Choix méthodologiques pour l’analyse de l’impact budgétaire à la HAS (No. halshs-01615097) (2016)
Gillingham, K., Newell, R. G., & Palmer, K. Energy efficiency economics and policy (No. w15031). National bureau of economic research. (2009)
Healy, M. How many COVID deaths are ‘acceptable’? Answer is key to moving to a post-pandemic world. Los Angeles Times, Avril, 1er (2022)
Holdren, N. Depoliticizing Social Murder in the COVID-19 Pandemic. Bill of Health Blog, Harvard Law School (2022).
Ingrand, C. Covid-19 : le “manque d’anticipation” de l’État pointé par la Cour des comptes. TF1 Info, 19 mars (2021)
Johansson LA, Pavilion G, Anderson R, Glenn D, Griffiths C, Hoyert D, Jackson G, Notzon FS, Rooney C, Rosenberg HM, et al.: Counting the dead and what they died of. Bulletin of the World Health Organization 84: 254, (2006)
Karpf, A. The hierarchy of deaths. The Guardian, 11 septembre (2001)
Leah Z. Rand, & Aaron S. Kesselheim. International Reference Pricing for Prescription Drugs in the United States: Administrative Limitations and Collateral Effects. Value in Health, 24:4 (2021)
Keating, J. Is it wrong to care more about 4 deaths in Boston than 80 in Syria? Foreign Policy. (2013)
Lang, K. How are COVID-19 deaths counted, and what does this mean? Medical News Today, 11 mars (2022)
MacKillop, Eleanor & Sally Sheard. Quantifying life: understanding the history of quality-adjusted life-years (QALYs). Social Science & Medicine 211 359-366, (2018)
Naghavi, M., Makela, S., Foreman, K. et al. Algorithms for enhancing public health utility of national causes-of-death data. Population Health Metrics 8, 9 (2010)
OECD/European Observatory on Health Systems and Policies. France : Profils de santé par pays 2021, OECD Publishing, Paris, (2021)
Oliver D. Mistruths and misunderstandings about covid-19 death numbers. British Medical Journal. 10;372-352. (2021)
OMS. Projet de troisième rapport de la commission du programme et du budget, A20/P&B/16 (1967)
OMS. Surveillance de la santé publique dans le contexte de la COVID-19, 14 février (2022)
Pathak, E.B., Menard, J., Garcia, R.B. & Salemi, J.L. Social Class, Race/Ethnicity, and COVID-19 Mortality Among Working Age Adults in the United States. MedRxiv, 11.23.21266759 (2021)
Pressat, R. La situation démographique. Population, 15:3. (1960)
Sencer, D. J., & Millar, J. D. Reflections on the 1976 swine flu vaccination program. Emerging Infectious Diseases, 12(1), 29 (2006)
Smith, L. E., Potts, H. W., Amlôt, R., Fear, N. T., Michie, S., & Rubin, G. J. Adherence to the test, trace, and isolate system in the UK: results from 37 nationally representative surveys. British Medical Journal, 372 (2021)
Turquan, V. Étude statistique de la dernière épidémie de grippe à Paris. Journal de la société française de statistique, 32, 62-67 (1891).
Vagneron, F. Déchiffrer la grippe russe. Quand une pandémie devient un événement statistique (1889-1893). Population, 75(2), 359-389. (2020).
Wang, S., Gum, D., & Merlin, T. Comparing the ICERs in medicine reimbursement submissions to NICE and PBAC—does the presence of an explicit threshold affect the ICER proposed?. Value in Health, 21(8), 938-943. (2018)
Zylberman, P. Les 90 premiers jours d’une pandémie. Pour la Science 383 (2009)
Zylberman, P. Beyond apocalyptical epidemics: Out of a paradox. Epidémies et sociétés, passé, présent et futur. B. Francine Eds. Edizioni ETS (2017)
Zylberman, P., Oublier Wuhan : Essais sur l’histoire contemporaine des crises sanitaires, La fabrique éditions, (2021)