Thoughts on Estimating Excess Mortality from COVID-19

June 18, 2020

Carl Heneghan, Jason Oke


To estimate the excess mortality, a count of deaths from all causes relative to what would normally be expected is required.

To estimate the excess mortality we can use three different approaches to make comparisons:

  1. The 5-year average (Doesn’t take account of any  changes in the population over time)
  2. The linear trend using data back to 2010 (tries to fit the best average fit to all the past data)
  3. The harmonic trend back to 2010 (tries to account for the patterns in disease occurrence)

And then we can  consider two times points

  1. Estimating the excess deaths during the COVID-19 outbreak, or
  2. Take the whole year up to week 23 and compare the excess.

1. Estimating the excess deaths during the COVID-19 outbreak (weeks 10 to 23)

The start of the excess can be predicted by a single point above one standard deviation above the expected: week 13 when there were 1,011 excess deaths in England  & Wales.

In the 10 weeks covering the COVID pandemic, there were 168,396 deaths in 2020.  Using the three different approaches leads to similar estimates of the excess (55,587 to 58,711)

Approach The predicted number of deaths Excess
Five-year average 109,685 58,711
Linear model 112,809 55,587
Harmonic trend 112,528 55,868

2. Estimating the excess deaths in 2020 so far

Using the three different approaches for 2020 so far gives substantially different estimates of excess.

The linear trends estimate is close to the 47,387 deaths involving COVID-19 registered in England and Wales so far.

One of the reasons for the discrepancy is the considerable variation year to year when including the winter period. Some years have lower counts that are then followed by more deaths in the next year. This is particularly the case in 2014/15 and now 2019/20.

This variation in excess deaths is greatest for the over 85s

In summary, the excess can be calculated in different ways. Taking a simple average does not account for changes in the structure of the population as well as the variations from year to year.

We can expect to see variations in mortality and these will be greatest in those over 85 years of age.

And accounting for the yearly variations suggests there are 31,417 excess deaths so far in 2020, which is much less than the 54,000 predicted using the 5-year average.


AUTHORS

Jason Oke is a Senior Statistician at the Nuffield Department of Primary Care Health Sciences and Module Coordinator for Statistical Computing with R and Stata (EBHC Med Stats), and Introduction to Statistics for Health Care Research (EBHC), as part of the Evidence-Based Health Care Programme.

Carl Heneghan is Professor of Evidence-Based Medicine and Director of Studies for the Evidence-Based Health Care Programmes (Full bio and disclosure statement here)

Disclaimer the article has not been peer-reviewed; it should not replace individual clinical judgement, and the sources cited should be checked. The views expressed in this commentary represent the views of the authors and not necessarily those of the host institution, the NHS, the NIHR, or the Department of Health and Social Care. The views are not a substitute