Introduction
Given the complexity of the question, and the different facets, we are breaking our commentary on this question into three bulletins:
 What is the likely mortality of the ‘survivor pool’ in the short term?
 What are appropriate ways to consider setting mortality assumptions at the ‘postpandemic’ stage?
 What are the longterm mortality implications of the pandemic?
In this bulletin we look at the mortality of the ‘survivor pool’ in the short term. The objective is to consider what the mortality of the population will be after the initial pandemic, not allowing for any ‘nth wave’ returns of the coronavirus.
What will affect the mortality of survivors?
The overall mortality of the survivor pool will be different from the equivalent prepandemic mortality to the extent that the COVID19 fatalities exhibited different mortality (prepandemic) from those who survived.
What do we know about the COVID19 fatalities that may have affected their ‘prepandemic’ mortality? Clearly they were primarily the old, and primarily male: however, demographers and actuaries (amongst others) will take account of age and sex in considering mortality, and hence these aspects are not relevant to the question here.
To estimate the mortality of the ‘survivor pool’, we can work through the following thought process:
 What aspects of COVID19 lead to disproportionate deaths for some types of people?
 What would be the likely mortality (prepandemic) of these people?
 Thus, what would be the likely mortality (prepandemic) of all those dying from COVID19?
 If we remove those people from the overall population, what is the remaining average mortality?
What aspects of COVID19 make deaths arise disproportionately?
So far, we know that the following factors are particularly relevant (where we provide also some indication of typical mortality impacts, expressed as ‘odds ratio’ multipliers – so e.g. the figure of 1.5 for obesity means an obese person would have 150% the mortality of a nonobese person, all other things assumed equal – that can be combined, so eg an obese diabetic would have mortality of approximately x 1.5 x 2.0 = x 3 normal mortality):
Factor 
Comment 
Obesity 
An important factor. The effect of obesity is of the order of 1.5. 
Diabetes 
Very important. The effect is of the order of 2.0. 
Other chronic conditions 
For common conditions (eg history of heart disease, cancer), of the order of 1.25 
SE Class 
Top quintile 0.8 of middle (ie average) quintile, lowest quintile 1.4 of middle 
Ethnicity 
BAME 1.52.0 
Note that the figures in the above table are approximate and indicative only. We have commented specifically on risk factors in [link to bulletin], and since then further papers have emerged (many of which we have noted in our Friday Reports).
What would be the likely mortality (prepandemic) of the people in those groups?
We know from existing research that people in the above groups have normal (ie allcause mortality absent the coronavirus) mortality somewhat different from the average. Using the same way of presenting this effect as we did in the above table, the factors as they would apply to people in the 6080 year old age group are, very approximately:
Factor 
Mortality effect (allcause) 
Obesity 
1.2 
Diabetes 
1.3 
Other chronic conditions 
2.0 
SE Class (by IMD) 
1.5 (ratio bottom to top IMD quintile) 
What would be the prepandemic mortality of those who die from COVID19?
We can use the known (disproportionate) nature of COVID19 to estimate how the deaths break down, for instance as follows (considering here just diabetes and obesity for simplicity). We can then calculate the expected prepandemic mortality of these subgroups, and hence the overall average ‘prepandemic’ mortality of the whole group of COVID19 fatalities.
Splitting just by diabetes and obesity, this table shows the proportion of COVID19 deaths we might expect (from the first table above on COVID19 mortality risk factors in conjunction with population prevalence).
Group 
Population 
Proportion of 
Allcause mortality multiplier 
Normal 
65% 
48% 
1 
Obese, nondiabetic 
20% 
22% 
1.3 
Obese, diabetic 
10% 
22% 
1.6 
Diabetic, normal BMI 
5% 
7% 
1.2 
Total 
100% 
100% 

Allcause mortality 
1.10 
1.20 
This table also shows the allcause mortality multiplier, weighted by population prevalence, and also weighted by proportions of COVID19 deaths.
We can therefore calculate that the expected ‘prepandemic’ mortality of the group who died from COVID19 is 1.2 / 1.1 = 109% of the same number of people randomly selected from the population.
With the COVID19 deaths no longer in the overall population, what is the survivor mortality?
Suppose 1% of the population die from COVID19. Although this is a high (hopefully unrealistically high) figure for the whole UK population, it is a reasonable figure (perhaps even an underestimate) for older age groups.
Having estimated the mortality of the group of all those who died from COVID19, we can estimate the relative mortality of the survivor pool:
Group 
Proportion 
Relative mortality 
COVID19 deaths 
1% 
109% of normal 
COVID19 survivors 
99% 
99.9% of normal 
Combined population 
100% 
100% of normal by definition 
The 99.9% above is ‘solved for’ by finding the value that, combined with the 109% of normal mortality for the COVID19 deaths group, takes us back to 100% for the combined group. Thus the mortality of the survivor pool is (in this example) around 0.1% below normal (prepandemic) mortality.
The effect is light in this example largely because we have assumed a small proportion of deaths from COVID19 in the ‘base’ group. For high age (and male) segments of the population, we expect higher proportions to die from COVID19 and hence a greater eventual differential between pre and postpandemic mortality.
The figure is also lighter than the ‘real’ figure will be because, in the above, we have looked at only two categories (obesity and diabetes). Introducing more categories (e.g., other common medical conditions, or socioeconomic splits) accentuates the effect (if, for each extra category, we have both an increased risk of death from COVID19 and increased ‘allcause’ mortality, which is generally the case).
Further work by the author testing the effect of this extra granularity increases the allcause prepandemic mortality ratio of COVID19 deaths to normal population from 109% to of the order of 130%.
Conclusion and further work
COVID19 mortality is associated with various risk factors (such as obesity and diabetes) that are themselves associated with higher ‘normal’ (allcause) mortality. If we consider this aspect only, ignoring for now other aspects (noted below), then the overall mortality of the postpandemic ‘survivor pool’ will be lighter than overall population mortality prepandemic (allowing for age and sex effects).
The example calculation above, which illustrates the underlying dynamics, shows that the effect is likely to be low other than in subgroups of the population with a high proportion of COVID19 deaths.
The other aspects to be borne in mind in estimating the likely mortality of postpandemic survivors are:
 How are future mortality improvements likely to differ because of the pandemic? (For instance, what effect might the associated economic shock have on healthcare expenditure or personal healthrelated expenditure that may affect mortality?)
 What is the effect on human physiology of surviving a ‘severe symptoms’ (hospitalisation necessary) infection of COVID19? (For instance, the Spanish flu was associated with longterm features that had a material morbidity / mortality effect.)
In the next two bulletins on this subject, we consider the points above, and we also consider how, once we are ‘postpandemic’, we could conduct an experience analysis in a way that allows appropriately for the largely ‘oneoff’ nature of the pandemic.
Matthew Edwards
27 May 2020