Making sense of the ONS COVID-19 death count

The Office for National Statistics (ONS) has just published their latest release for deaths registered weekly in England and Wales (provisional) for the week ending 3rd April. The main point from this release is that the provisional number of deaths registered in England and Wales in the week ending 3rd April 2020 was 16,387, the highest number of deaths since official weekly statistics began 15 years ago. This period in question doesn’t even cover the last week, which has seen the highest number of deaths being published by the Department for Health and Social Care. What should we take away from these figures?

Firstly, this figure of 16,387 represents an increase of 5,246 deaths registered compared with the previous week and 6,082 more than the five-year average. Of these deaths, 3,475 (21.2%) mentioned “novel coronavirus (COVID-19)” and is comparable to 539 (4.8%) in the week previous. What is interesting here is that we have seen a spike in the number of COVID-19 related deaths, but also an astounding spike in the number of non-coronavirus deaths, an excess of 2,607 when compared to the five-year average. What could be behind this spike in non-coronavirus deaths? The statistics unfortunately don’t tell us this, but there are several reasons which could all be plausible: 

  1. Some of the reported non-coronavirus deaths should actually have been recorded as COVID-19. It could be that unless doctors are 100% certain that the person had coronavirus (i.e. they had been tested for it), they remain reluctant to record that as a cause of death. ONS figures are based on registered causes of death, this isn’t always necessarily the true cause of death.
  2. Some deadly conditions that would normally have been picked up are being missed as resources are focussed on the coronavirus. Perhaps there’s an over cautious attitude to seeking help. Many operations and treatments that would usually have gone ahead have now been postponed due to COVID-19; could this be causing an increase in excess deaths?
  3. Perhaps this is collateral damage of lockdown. There are a number of media outlets that have commented on the mental health consequences of a prolonged lockdown, so should we expect to see a spike in non-coronavirus deaths because of this?

It’s likely that not just one of these reasons is correct, but that it is a combination of a number of these factors.

These numbers from ONS are 15% higher than the NHS numbers that we have been seeing as they include all mentions of COVID-19 on the death certificate, including deaths in the community. But just what is the distribution of coronavirus deaths when we look at this by registered place of occurrence. The vast majority of deaths involving COVID-19 occurred in hospital, 90.2%, with the remainder occurring in hospices, care homes, or private homes. But in general, the vast majority of all deaths occur in hospital so should we be surprised? Looking at all-cause mortality for the same period, 46.1% of all deaths occurred in hospital. It doesn’t take a mathematician to realise that 46.1% is a lot smaller than 90.2%. If we look at care homes, for example, 22.5% of all deaths occur in these locations, but only 5.3% of COVID-19 deaths. Additionally, COVID-19 deaths make up only 0.6% of all deaths in care homes (compared with 4.8% in hospitals). We may expect these numbers to be different and we may expect more COVID-19 deaths to occur in hospitals than anywhere else, but should we expect these figures to be so different, or perhaps could there be a number of COVID-19 care home deaths not being registered as such?

Why are there so many different death counts? This has to be one of the main questions that I get asked. And why are these ONS figures almost two weeks old? The ONS remains the gold standard as they must be certified by a doctor, registered, and processed. But this also makes them the slowest. The daily counts that we have become accustomed to hearing in the Government updates only include deaths in hospital of those who have tested positive for COVID-19. This makes them much quicker to collect, but also less accurate. The COVID-19 pandemic is a rapidly changing situation and there will always be a balancing act between getting access to the most accurate information as quickly as possible. Any decisions made should, of course, be based on evidence and the data we have at hand. But those decisions should be constantly updated as we receive more accurate and reliable data. This leads me onto a rather more philosophical question: What has COVID-19 taught us about statistics? This is the first time that I have ever experienced such a public interest in data and statistics. This is a huge opportunity for statisticians to not only have access to large amounts of data for interesting analyses, but also to educate the public in how best to make sense of the data that they are seeing. For example, how useful are the total number of cases that we are seeing? Cumulative counts are useless when trying to spot trends and daily counts are much more useful. But as a statistician, I know that granular data at a daily level is volatile and subject to random fluctuations. World in Data uses a three-day moving average which may be more useful in bringing out underlying trends. I’m a statistical ambassador for the Royal Statistical Society, and we wrote a useful article which tries to make sense of coronavirus case numbers. For example, the number of cases is only the number of people who have tested positive for coronavirus and so may be a very poor reflection of the number of people who have actually been infected. How much of a poor reflection depends heavily on the number of people who have been tested.

And finally, I’m going to repeat what I said at the very start of this article: the provisional number of deaths registered in England and Wales in the week ending 3rd April 2020 was 16,387, the highest number of deaths since official weekly statistics began 15 years ago. What can we do to bring down this death count? The way to bring down the number of COVID-19 deaths is to bring down the number of COVID-19 cases, and that is something that we can all contribute to, every single one of us. By adhering to social distancing guidelines, we should hopefully see a reduction in the number of COVID-19 cases and a reduction in the number of COVID-19 deaths should follow a few weeks later.


Hear Professor Rogers discussing these figures on BBC Radio 5 Live here at 28:52.