Skip to Main Content

How best to represent the true toll of the Covid-19 pandemic on human lives is an urgent matter. Though loss of life represents the clearest indicator, limited testing, inconsistencies in assigning the cause of death, and even political influence are creating uncertainty over how deaths are being counted and attributed (or not) to Covid-19.

It’s simple, really: Limited testing gives a limited picture of confirmed Covid-19 cases and deaths. While many deaths show fairly clear evidence of Covid-19 infection even without testing (and so may be tallied as “suspected” Covid-19 deaths), even adding confirmed and suspected deaths together risks a significant undercount of the true magnitude of the epidemic’s toll.


We should be using a more encompassing metric for measuring mortality: the total number of lives lost in excess of historically expected levels. Here’s why.

A focus on just confirmed and suspected deaths misses out on those deaths from other causes resulting from more indirect effects of Covid-19. For example, deaths can occur when health systems are strained or overwhelmed and unable to provide sufficient or quality care — think of non-Covid-19 patients requiring ICU beds in units already over capacity due to the pandemic. Deaths can also arise from delays in going to the hospital among those needing care due to fear of getting infected during their stay. And lastly there are deaths stemming from Covid-19’s interactions with noncommunicable diseases such as diabetes, heart disease, cancer, kidney disease, and others.

The concept of “excess mortality” — the difference between the total number deaths from all causes during a specified period of time and the expected number of deaths for the same place and time of year, on an historical average — has been gaining attention as a useful metric. Countries from Belgium to South Africa, as well as major metropolitan areas like New York City, have adopted excess mortality counts to understand the true burden of Covid-19 outbreaks within their borders. Even news outlets are providing estimates of excess deaths in countries around the world.


As several colleagues and I recently wrote in the Bulletin of the World Health Organization, “Enumeration of all deaths, when compared to historically expected mortality, produces a picture of excess death, capturing both the direct burden of the epidemic and its indirect mortality burden …” Looking at all deaths, regardless of cause, brings unique insights to understanding the pandemic’s impact.

Countries worldwide are using excess mortality analyses to support their Covid-19 response, for a number of reasons.

Testing tells only part of the story

Testing — the only way to be certain if an illness or death is directly due to Covid-19 — is often limited or lacking in low-resource communities and in places where the need for testing greatly exceeds the number of tests available.

Beyond the global shortage of test kits, testing tends to occur almost exclusively in hospitals, leaving uncounted the deaths related to Covid-19 outside these facilities. This heavy reliance on hospital data excludes the tremendous number of deaths that occur at home in low- and middle-income countries. Even in the absence of a public health emergency, for example, some governments have estimated in unpublished reports that up to 80% of deaths happen outside of hospitals, with no doctor in attendance.

In a pandemic, this takes on new urgency, leaving governments without a clear understanding of the Covid-19 situation among their residents. When deaths are counted from both hospitals and communities, the resulting bird’s-eye view of excess mortality provides a more accurate window into the epidemic’s true human scope and scale.

Looking back to look forward

While cause of death is certainly of great importance, the process of assigning the correct cause of death during the Covid-19 pandemic — especially in countries without automated systems to do the job — can delay analysis and provide only an incomplete picture of the extent of the pandemic. Weekly tallies of all deaths, when compared to historically expected levels, can more rapidly produce a picture of excess mortality.

In Brazil, Colombia, and Peru — countries in the Bloomberg Philanthropies Data for Health Initiative — systems for registering all deaths within a day or so of occurrence are digitized with high coverage, permitting the sort of analysis shown here of deaths in Manaus, Brazil.

Patrick Skerrett / STAT Source: Civil Register COVID Portal/Brazil

A sharp increase in total mortality in 2020 is starkly evident, illustrating the enormous impact of the pandemic on the city’s death rate. Cumulatively, between March 16 and June 6, Manaus registered 3,549 excess deaths and a 153% increase in expected mortality. By contrast, only 1,462 confirmed Covid-19 deaths were reported during this period — a likely undercount due to the unavailability of testing.

In New York City, data published earlier this year show that between March and June, all-cause mortality increased by 700% above baseline at the peak of the pandemic, resulting in 25,100 excess deaths.

Patrick Skerrett / STAT Source: Daniel M. Weinberger, Yale School of Public Health

Health authorities in Bangladesh, Rwanda, and Colombia are leveraging existing community-based data collection systems to furnish weekly counts of mortality, either remotely via cell phone reporting or from data collectors in the field. This work has been underpinned by new guidance from several collaborating governmental, nongovernmental, and international organizations. In New Zealand, where it is possible to rapidly register a death online, death notifications and medical certifications are being used to track mortality daily. The information is disseminated within hours, allowing the Covid-19 response team to monitor death rates in near real-time.

Excess mortality surveillance is a powerful advocacy tool

The speed of the pandemic’s spread demands speed of response, and an informed response demands timely data. In the face of rapidly changing conditions, strong data and rational decision-making are our best defenses against Covid-19.

Weekly reports of excess mortality — disaggregated by age group, sex, location, and by race (where appropriate) — can provide part of the backdrop for policymakers and decision-makers as they assess how, when, where, and among who to implement public health measures like stay-at-home requirements.

The World Health Organization recently called for weekly reporting in its World Health Statistics Report. Yet more than half a year into the pandemic, the international community is only now beginning to reach consensus on the human toll of Covid-19 by employing excess mortality tools to understand the scale and scope of national outbreaks.

As the global Covid-19 situation continues to evolve and endure, it is of utmost importance that world leaders rely on epidemiological evidence and data to guide the return of open societies, economies, and a new normal for all.

Philip Setel is vice president and director of civil registration and vital statistics at Vital Strategies, a global health organization headquartered in New York City.

  • If the total # of China virus deaths is 165,000 since January 2020 and the population of the USA is 330.000,000 dividing 165,000 into 330,000,000 equals .0005% every death China virus or not is a tragedy. Like the flu testing positive for the China virus is not a death sentence. The media has created a fear of the China virus by announcing the death total every day.

  • How about an even less biased or influence prone statistic?

    Show deaths by all causes over time. Compare this during the pandemic to previous times.

    Step back from the data several pces and lets see if US deaths over time is increasing from now going back across years. I want expectations left out of the thing. Just count deaths and don’t expect or assign anything and look at it.

    • That’s what excess deaths is. When you look at yearly death rates on large populations they are largely predictable. You might see 0.3% to 1% more than expected, but those are usually attributable to something like a particularly dangerous flu strain. So when you see a significant spike, you can attribute it as being related to something like COVID19 that puts outsized pressure on the healthcare system.

      See the link below for the CDC numbers. In the 2 years prior to COVID, we were under the predicted numbers. COVID has us sharply higher.

  • I like this article partially because it shows some potential questions.
    1) With underlying health conditions would the person have died soon anyway even without Covid-19?
    2) Testing doesn’t create Covid-19 illnesses but it discovers them so the counts go up.
    3) The graph for New York shows the impact of the excesses but it would be nice to see equivalent graphs for other cities or regions.

    • 1) People who would have died anyway are part of the ‘expected’ curve and do not inflate the ‘excess mortality’ curve.
      2) Agree
      3) I’d be very curious about some other states

  • This seems to be suggesting we should take the pandemic even more seriously because of excess deaths, but how can we distinguish excess deaths from the reactions and lockdowns?

    My own community has seen empty hospitals, while many patients have delayed and refused to seek treatment, resulting in death. This has also been reported many places. These are the result of “the panic” of the pandemic and not related to the actual virus. How are we to distinguish these?

    Crime increase, suicide, supply disruption (especially in other parts of the world), all these are causing excess deaths and are not related to the virus itself, but the reaction.

    Yes, we need to understand excess deaths to better inform decisions, but we have to be willing to let those decisions reflect lessening measures when non-virus deaths become an issue. I don’t see many people advocating for this. Which means we’ve demonized one virus at the expense of neglecting others.

    • Well, in some countries, like Sweden and Russia, they did not demonize the virus, so there is no reaction effect, but the excess mortality is as high as anywhere else, if not higher.

  • If we also had statistics on excess morbidity, the true toll of the Covid-19 epidemic on human lives in the USA would be much worse. Surviving patients with strokes, heart attacks, lung fibrosis, retinal embolism, and kidney failure resulting from a SARS-CoV-2 infection are only included in the data for infection rates with no indication of their morbidities.

  • the only problem is that the Trump administration is doing everything it can to muddy the waters and hide those excess deaths. By analyzing CDC data, it is clear someone over there has been manipulating the data at least for the past month. Analyzing provisional covid tables for the period Feb1-Jun20, excess deaths as compared to same period for years 2015-2018 (adjusted for demographic increase) amounted to 149k, while the official covid deaths were listed at 112k. And if those official deaths were fitted to the correct timeline, the excess was of 230k. While, if you analyze the latest tables, those excess deaths have pretty much disappeared. Indeed, grotesquely, the numbers published suggest there have been less deaths this year than in 2017/18! And now, that data have to go through the WH, it will get even more impossible to get real data, as already evidenced buy the latest new daily infections data coming out of the administration.

    • Actually, it would be impossible for the WH to hide deaths, as those are recorded at every level, from county to state to federal.

      The WH is only summing numbers from other sources, and can’t be a source of data pollution any more than the CDC could, by presenting numbers in confusing manners.

Comments are closed.