New estimates by WHO point to a developing country-driven intensification and broadening of excess mortality across age cohorts

The WHO has just released new estimates of excess mortality over the course of the calendar years 2020 and 2021. These are not the first estimates of excess mortality, but what sets them apart is their level of granularity. The estimates differentiate by age cohort and sex and the insights we can develop based on them are astonishing. 

This post will discuss the new excess death estimates by age distribution and it will do so through the lens of the World Bank income classification. It will highlight the expansion of excess mortality between 2020 and 2021 and examine how much of that has been carried by an intensification and broadening of the mortality toll in the developing world. 

A treasure trove of data 

Excess mortality – the deviation of all-cause mortality from what one could have reasonably expected – is the best method available to us to capture the true and total impact of the pandemic.  It is able to capture misattributed COVID-19 deaths and account for other deaths indirectly caused or prevented during the pandemic. 

Unfortunately, many countries lack the data. Out of 194 countries, only 73 provide full national all-cause mortality data to the WHO. This is where the excess death model comes in and fills in the blanks on the basis of co-variates that are meaningful predictors of excess mortality. The technical note has the full details and the actual estimates can be downloaded here

A big picture of inequality and escalation

WHO’s mid-point estimate of excess mortality during the 24 months of 2020 and 2021 is 14.9 million people. The 95% confidence interval around this estimate ranges from 13.3 to 16.6 million. The estimate is mid-point estimate is conservative, but in the ballpark of other estimates. For example, The Economist estimated, using an entirely different method, a number of 17.8 million at the end of 2021 (95% CI: 12.3-21.0 million), which has since grown to 21.1 million as of May 7, 2022 (95% CI: 14.7-25 million). 

Let us first have a look at the big picture in the aggregate (i.e. without disaggregating the data by age cohorts). That big picture appears to be one of inequality between rich and poor countries and one of escalation between 2020 and 2021. Check out the media carrousel below, which has four charts offering alternating views of the excess death toll (absolute vs per capita; and monthly vs cumulative). Throughout we refer to the World Bank income classification, which divides the world into high-income (HIC), upper-middle-income (UMIC), lower-middle-income (LMIC) and low-income (LIC) countries. Note also that the developing world comprises UMICs, LMICs and LICs.

Here’s a few eye-catching results:

  • Developing countries represent 85% of the global excess death toll. That may sound unsurprising as the share of developing countries in the global population also happens to be 85%. But since developing countries are on average much younger, we would think, all else equal, that the share in excess mortality is smaller than their share in population.  
  • Lower-middle-income countries – not high-income countries as you might expect – have suffered the worst excess death rates. The high excess death rates in LMICs are mostly driven by India. HIC and UMIC rates are about 2/3 to 3/4 of LMIC rates, whereas LIC rates hover around 1/2 of LMIC rates. All rates are significantly higher than what the reported statistics suggest.
  • Unlike the high-income countries, excess mortality in the developing world expanded a lot between 2020 and 2021. LMICs saw excess mortality grow by 247%, LICs by 120% and UMICs by 85%. HICs, however, saw a much more modest rise of just 9%. This inequality across income groups reflects among other factors: (1) the time it took for the virus to spread globally in 2020, (2) the substantially negative excess mortality early on in 2020, (3) the arrival of hyper-contagious and more lethal Delta in population-dense developing countries and (4) the unequal roll-out of vaccines.

The unclustering of age-specific mortality rates

Let’s now examine the granular detail that distinguishes the WHO’s estimates from other sources: the excess death impact by age distribution. The two charts below look at excess mortality across the age distribution. The first one looks at age-specific excess mortality rates, whereas the second one looks at age-specific excess mortality in the absolute. Both split the data by World Bank income group and pandemic year. 

As to excess mortality rates (the first chart), it is immediately obvious that age-specific mortality rates rise exponentially as we progress along the age distribution. That conforms with the fact that COVID-19 is more risky for the elderly (of course, along with infection fatality risk, infection prevalence should be considered also when assessing how mortality among the elderly has been affected). Note the following:

  • Mortality curves in 2020 were clustered into two groups. LMICs and LICs had a virtually identical mortality curve and UMICs and HICs had another virtually identical curve. But the two clusters themselves were quite different from each other as the mortality curves for LMICs/LICs were about twice as high as those for UMICs/HICs. 
  • 2021 saw an escalation and decompression in mortality curves. Age-specific mortality escalated considerably for all income groups, except HICs. The mortality curves became unclustered: LMICs rose significantly above LICs; and UMICs rose significantly above HICs. 
  • The mortality rates of the elderly cohorts in LMICs stand out in particular. In 2021, a whopping 4% of the 80+ population died. Among the 70-79 and 60-69 cohorts, the mortality rate during that year was an alarming 2% and 1%, respectively.
 

When it comes to absolute excess mortality (the second chart), we see the following patterns:

  • The absolute excess death count 2020 was concentrated among the 60+ cohorts. LMICs and UMICs contributed large numbers across three elderly buckets (60-69, 70-79 and 80+). HICs contributed especially to the 80+ bucket.  Note also the considerable negative excess mortality in the 25-39 age bucket among LMICs. 
  • 2021 was entirely different, with HICs fading into the background in relative terms. The contributions from LMICs and UMICs became disproportionate. HICs did not increase by that much, in part because of the prioritized roll-out of vaccines to the elderly. 
 
The intensification and broadening of excess mortality
 
Let’s now look at the cumulative totals in some greater detail. The chart below shows again cumulative excess deaths by age bucket but this time it stacks the excess death data, which helps visualize developments at the global scale.  

The chart shows:

  • The global totals by age cohort suggest a considerable intensification and broadening of excess mortality. We see higher totals for each age bucket (see label on the outside of each bar). The elderly cohorts saw excess mortality doubling from an already high base, but interestingly the increase broadened as well to other age groups. Especially the 50-59 and 40-49 cohorts were affected, which rose by a factor of 3 and 10 respectively. Note also that negative excess mortality in the 25-39 cohort turned positive.  
  • Aggregating the three elderly buckets into one 60+ group, we see large differences across income groups. Absolute excess deaths among the 60+ rise by 170% in LMICs, 65% in UMICs and 60% in LIC. But they decline by 2% in HICs. The broadening of excess mortality into the non-elderly age cohorts occurred everywhere but was especially large again in LMICs and UMICs. The reversal of excess mortality from negative to positive among the 25-39 cohort was driven mostly by LMICs.
 

The non-elderly age cohorts

Finally, let us dig a little deeper into the observed broadening of excess mortality and focus on the non-elderly age cohorts (the 60- group). The two charts below show the distribution of excess deaths by age and year. The first chart looks at the global level and shows how much each age cohort contributed to the total excess mortality of that year. The second chart shows the results by income group, meaning that the shares represent how much each age cohort contributed to total excess mortality during a given year for that income group.

 

The following patterns emerge:

  • The share of non-elderly age groups in global excess mortality rises from 7% to 23% between 2020 and 2021. The shares are net shares reflecting the fact that the younger age groups show negative excess mortality especially in 2020. 
  • Within income groups, the broadening of excess mortality is happening especially in LMICs and LICs. As the second chart shows, the share of non-elderly age groups rise from 10% to 19% in HICs and 10% to 20% in UMICs and from 3% to 24% in LMICs and close to 0% to 27% in LICs. These are percentage shares within each income group. The absolute increases of course differ as mentioned earlier.

In conclusion

The excess death estimates released by the WHO shed light on the true and total impact of the pandemic. Some of the big picture patterns and trends are not new – as discussed herehere and here on the basis of alternative estimates.  But it is encouraging that the WHO’s excess death estimates confirm these results and thereby adds further credibility to them. 

The greatest value added of the dataset, however, lies in its granularity. Covering the distribution of excess deaths by age and income group, this post has highlighted the considerable expansion of excess mortality. This has happened due to a concurrent intensification of excess mortality within age buckets and a broadening across them. And it has emphasized that both trends have been carried mostly by the developing world.

These are important findings. All too often the public discourse dismisses the severity of the impact in developing countries on the basis of the mistaken belief that their populations are mostly young and therefore shielded from impact. Such a conclusion is entirely unwarranted. This post has mentioned two reasons why:

  • Developing country population are proportionately young but they count many, many elderly in the absolute. This is why their contributions to global excess deaths are so large, especially given the disproportionate increase in age-specific mortality risks. 
  • The excess mortality toll has very considerably broadened into the non-elderly age cohorts of developing countries. This is particularly visible in the 50-59 and 40-49 buckets but is also observed in the reversal of negative into positive excess mortality among younger buckets.

The above is all the more reason to fully take into account the cause of the developing world, especially the poorer countries among them, as we seek solutions to the pandemic in terms of diagnostics, , therapeutics and vaccines.