The paper addresses the measurement of ‘excess deaths’ related to the Covid-19 pandemic for the period up to 29 May 2020. In particular, it examines excess death rates by age, gender and region and over time, and their relation to the changes in the proportion of deaths occurring in care homes.
The analysis uses registered deaths based on doctors’ death certificates reported by the UK Office of National Statistics (ONS Statistical Bulletin, Deaths registered weekly in England and Wales, provisional: week ending 29 May 2020, ONS 9 June). It uses weekly data covering the first 22 weeks of 2020, with the last data being for the week ending 29 May, and ONS monthly mortality data from January 2006 through the end of April 2020. From 13 March 2020 the weekly figures also show the number of deaths involving coronavirus (COVID-19), based on any mention of COVID-19 on the death certificate. An 'underlying cause of death' refers to the main cause of death, whereas a cause being 'mentioned on the death certificate' means that it might be the main reason or a contributory reason to the cause of death.[1] These should be distinguished from the daily Covid-19 death reports produced by Public Health England (PHE), which are deaths following a positive test result confirmed by a PHE or NHS laboratory. The registered deaths reported by ONS will include deaths that are not included in the PHE definition, because they had no positive test result, but the doctor thought it likely that the person had COVID-19. They may also exclude cases that are included in the PHE definition because although the patient had a positive test for COVID-19 this was not mentioned on the death certificate. However, in general, the numbers of deaths reported by ONS will be larger than those included in the PHE definition.[2] The most recent data suggest that the cumulative number of Covid-19 deaths reported by PHE (by 29 May) understate the corresponding death registration number by about 20%. All figures are for England and Wales only.
The impact of the Covid-19 pandemic on mortality could be more, or less, than the number of deaths in which Covid-19 was mentioned on the death certificate. There are different ways to calculate ‘excess deaths’ corresponding to a disease. First, the timeframe of the calculation matters because deaths may be brought forward or postponed, say during a year, and so the excess deaths for the year is probably a more meaningful figure than the same calculation over a shorter time frame. Second, there must be a ‘counterfactual’; i.e. what would deaths have been had the disease been absent? Using death registration data up to 29 May for England and Wales, Figure 1 shows excess mortality for each of the first 22 weeks of 2020 using average deaths during the previous 5 years in the corresponding week as the counterfactual. It indicates negative excess deaths from mid-January to mid-March. After 13 March, excess deaths rose sharply, as did Covid-19 deaths, peaking in the week ending 17 April, during which registered deaths were 113% above average and remaining high the following week (110% above average). Excess deaths fell sharply after 24 April, so that during the week ending 29 May excess deaths were 20% above average and lower than Covid-19 deaths during that week. It is possible that some of the excess deaths since 13 March are ‘deaths postponed’ from earlier in the year. For instance, during the pre-epidemic period up to 13 March, 65% of the shortfall in deaths relative to the 5-year average are accounted for by 3,200 fewer deaths from respiratory causes. Also, the peak of the pandemic in mid-April may have ‘brought deaths forward’. To take better account of such intertemporal substitution we measure excess deaths as deaths during the first 22 weeks of 2020 above the 5-year average for these weeks. On this accounting, excess deaths in England and Wales were 53,140, or 22% above average. Covid-19 deaths during the same period were 45,510. Thus, in addition to the Covid-19 deaths during the period there were another 7,630 deaths above average, some of which may be misdiagnosed Covid-19 deaths. If we had only looked at the period since 13 March, as many newspapers report, we would obtain 58,030 excess deaths, or 53% above average for those weeks.
Figure 1: Weekly Excess deaths and Covid-19 deaths during 2020, England and Wales
Alternative counterfactuals
Instead of average deaths over the previous 5 years for each corresponding week, here we use data on monthly deaths since January 2006 to construct the counterfactual. Apart from a trend adjustment and the longer time period, the approach here is in principle similar to the earlier one: i.e. to obtain excess deaths we deduct an average of deaths based on past mortality experience. Because the approach is regression based, computation of standard error of the estimate of excess deaths is straightforward.
The analysis allows for a cubic trend over the 172 months from January 2006 (date=1) through April 2020 (date=172) and for fixed effects for each month of the year (April being the reference month). Excess deaths are estimated from dummy variables for March 2020 and April 2020, the two months of the Covid-19 epidemic included in the available monthly data.[3] The baseline model estimates are shown in the first column of Table 1, in which the trend is estimated using all the data.[4] The second column estimates the trend coefficients with only data up to January 2020 and then adjusts the dependent variable by subtracting the predicted trend for each month. (the trend component is shown in Figure 2). In both models, the estimates of excess deaths (in bold type) indicate about 2,200 excess deaths in March 2020, and about 43,400 in April 2020, the latter with a 95% confidence interval of 41,250 to 45,664 (model 1) or 41,632 to 45,199 (model 2). The third column introduces a new variable ‘Covid-19’ which combines the two epidemic months by assuming that it takes the value 0.5 in March 2020 (there were only 5 registered Covid-19 deaths in the period up to 13 March) and the value 1.0 in April 2020. It suggests 36,500 excess deaths over the two months together, but with a wide confidence interval (20,215 to 52,800).
We can compare the monthly excess death estimate for April 2020 with one using average deaths during the previous 5 years in the corresponding weeks as the counterfactual for the 4 week- period 3 April to 1 May (excess=39,401), and multiplying it by 30/28 to get a rough estimate for the month of April of about 42,200, which is very close to that in the first two columns of Table 1.
Model 4 expresses deaths in terms of their natural logarithm. According to this model deaths in April 2020 are 98% higher (exp(0.683)-1)) than expected (implying excess deaths of 34,210) and 5% higher (1,780 deaths) in March. In terms of model ‘fit’ there is not much to choose between models 1 and 4, but the former is preferred because of the easier computation of robust standard errors.
Finally, model 5 ignores the longer-term trend and essentially uses the average deaths for each month in the 14 years since 1 January 2006 as the counterfactual. It produces similar estimates for April 2020 and larger ones for March 2020 than the models 1 and 2, but the standard errors are incorrect because of the presence of serial correlation in the residuals, doubtless a reflection of excluding the cubic trend.
Table 1: Regression estimates of Excess Deaths in March and April 2020
Figure 2: Cubic trend in monthly deaths, January 2006 to April 2020.
Excess Death rates and Deaths by Age, Gender and Region
Death rates from Covid-19 reflect the product of the probability of being infected and the probability of dying if infected. Before moving on to excess death rates, Figure 3 shows death rates per 10,000 population (as of mid-year 2018) from Covid-19 and non-Covid causes during the first 22 weeks of 2020. In addition to the usual feature of higher death rates among men than women, what stands out is the steeper increase with age in death rates from Covid-19 than death from other causes. Figure 4 illustrates the high excess death rates among men aged 85 and over, and also to a less degree among men aged 75-84. Among women, not only are their excess death rates lower, but excess death rates exceed Covid-19 death rates by more for men than for women. The reasons for the latter gender difference are unclear. Is there more misdiagnosis among males than females? Or is it the result of more delayed care for other illnesses among males?
Figure 3: Death rates from Covid-19 and from other causes during first 22 weeks of 2020
Figure 4: Excess Death (solid bars) and Covid-19 Death Rates (hatched bars) per 10,000 population in the first 22 weeks of 2020 by Age and Gender
Figure 5 combines death rates and the age-sex distribution of population to show the incidence of excess deaths across ages and gender. Among men, who contribute 63% of excess deaths, excess deaths at ages 45-64 and 65-74 are a more common feature than among women (e.g. 16% cf. 11% of all excess deaths at ages 65-74).
Figure 5: Excess Deaths in the first 22 weeks of 2020 by Age and Gender
Figure 6: Excess Deaths and Covid-19 Deaths per 100,000 population in the first 22 weeks of 2020 by Region
Regional dimensions. Figure 6 shows excess deaths and Covid-19 deaths per 100,000 population by standard regions. London, the West Midlands and the North West have experienced proportionately much more excess mortality than elsewhere. These three regions account for 45% of all excess deaths, but they constitute only 37% of the population of England and Wales. The West Midlands and the East have a particularly large number of excess deaths per 100,000 population not accounted for by Covid-19 deaths (the difference between the bars in Figure 6), which are likely to partly reflect misdiagnosis, particularly when testing is low.
Figure 7 shows excess deaths per 100,000 population by standard regions outside London and its strikingly strong relationship with the region’s population density. The reason for the strong positive relationship appears to be the higher chance of Covid-19 infection in denser populations. For instance, it contrasts with a significant negative correlation between population density and overall regional mortality in 2018. The North West has the highest excess mortality rates and density, and Wales the lowest. The South East (outside London) fares relatively better than might be expected from its high population density. At a finer spatial grain, across 299 local authorities, the correlation between the age and sex standardised death rate and density is 0.56.
Figure 7: Region’s Excess Deaths per 100,000 population in the first 22 weeks of 2020 and Region’s Population Density (correlation coefficient=0.80)
Nursing Homes and Excess Deaths
Press attention has been attracted by the increase in the proportion of deaths occurring in care homes as the pandemic evolved. Figure 8 illustrates the changing location of deaths, distinguishing Covid-19 deaths, other deaths and all deaths. In the period leading up to 17 April, when ‘Covid-19 deaths’ (i.e. deaths with Covid-19 on the death certificate) as a percentage of all deaths increased sharply to its peak (39%), a rise in the proportion of non-Covid deaths occurring in care homes contributed by far the most to the rise in deaths in care homes, with the rise in the percentage of Covid-19 deaths among all deaths actually reducing the rise, because a lower percentage of them were reported to occur in care homes at that stage. Subsequently, the proportion of deaths in care homes fell to more normal levels of around one-quarter by week ending 29 May, and this fall was mainly driven by the decline in the proportion of non-Covid deaths in care homes. Thus, it was the development of reported non-Covid deaths in care homes that drove the rise and fall in the overall proportion of deaths there. This may reflect misdiagnosis of cause of death in the earlier period (e.g. no Covid-test was done), which was gradually corrected by testing or enhanced awareness among doctors. Recent analysis by ONS of excess deaths not associated with Covid-19 up to 1 May 2020 strongly suggests that misdiagnosis plays an important role: ‘The largest increases in non-COVID-19 deaths compared to the five-year average are seen in deaths due to “dementia and Alzheimer disease” and “symptoms, signs and ill-defined conditions”…’ (ONS, Analysis of death registrations not involving coronavirus (COVID-19), England and Wales: 28 December 2019 to 1 May 2020, 5 June 2020), particularly in care homes. ONS (ibid.) go on to say: ‘This suggests that undiagnosed COVID-19 is a likely explanation for some of non-COVID-19 excess deaths observed in this setting [care homes], because of the increased vulnerability of this population and increased likelihood that symptoms would be hard to identify in addition to existing comorbidities. The additional difficulty in communicating symptoms for some people with dementia and Alzheimer disease could explain why there is a large increase for this disease in particular.’
Figure 8: Proportion of Covid-19 deaths, non-Covid deaths and all deaths in Care Homes, weekly during 2020, England and Wales
Delayed care might also play a role in pattern of non-Covid deaths in Figure 8 (e.g. a tendency not to transfer seriously ill care home patients to hospitals when there was high bed-occupation by Covid-19 patients there). The ONS (ibid.) suggests that delayed care may have affected deaths related to asthma and diabetes, which were occurring at a significantly higher rate compared to the five-year average in the period up to 1 May. ONS states that ‘This could indicate that some people suffering from these conditions are not receiving care fast enough to prevent death occurring. It is also plausible that some of these deaths are because undiagnosed COVID-19 had exacerbated the pre-existing condition.’ …This is particularly important as diabetes and asthma are known to be common pre-existing conditions in deaths due to COVID-19…’
Care homes have, of course, elderly populations with comorbidities, making them particularly vulnerable to Covid-19, which helped bring about a higher proportion of deaths in care homes during May than in mid-March. It appears that more protection of care home residents from Covid-19 would have reduced deaths there.
[1] Coding of deaths by cause for the latest week is not yet complete. [2] The PHE Covid-19 death reports are based on three sources, which are linked to the list of people who have had a diagnosis of COVID-19 confirmed by a PHE or NHS laboratory. This is to identify as many people with a confirmed case who have died as possible (https://coronavirus.data.gov.uk/about#process-from-29-april-2020.) These data became available on 29 April, and they replace the previous series of confirmed cases dying in hospital reported by the NHS. This new series of deaths includes all deaths previously reported by NHS England, but also includes other deaths of patients who were confirmed cases (i.e. a positive test result), whether they died in hospital or elsewhere. [3] This procedure is equivalent to estimating the model without the dummy variables up to February 2020 and then taking excess deaths as the difference between actual deaths and the ones forecast by the model for March and April 2020. [4] In computing the standard errors of the parameter estimates it is important not to constrain the variance of the residual to be constant across observations. The estimates of the standard errors are adjusted to allow for such heteroskedasticity using the Huber/White/sandwich estimator in Stata. E.g. in model 1, the Breusch-Pagan/Cook-Weisberg test for heteroskedasticity has a p-value<0.00001, and in model 4 it is p= 0.012.
