COVID-19 Report 4: Massive UK analysis of deaths from virus

COVID-19 Report 4: Massive UK analysis of deaths from virusImage: Trinity Care Foundation/Flickr
Comment & Analysis

This is the fourth issue of COVID-19 Report. We point you to the latest quality science on the pandemic. If you come across unfamiliar terms, there is a glossary at the bottom of the article.

Rich data from UK study

A massive study in the UK shows who is dying from COVID-19. The OpenSAFELY analysis led by well-known science writer and doctor, Ben Goldacre, analysed the health records of over 17 million adults for the period between 1 February to 25 April. They compared the characteristics of the 5 683 deaths attributed to COVID-19 in that time in the UK.

They found that dying from COVID-19 was strongly associated with being male, older age and deprivation, uncontrolled diabetes, severe asthma and various other medical conditions. Black and Asian people had double the risk of dying than white people. Hypertension does not appear to be a big risk in this study.

The study also casts further doubt that smoking is protective against COVID-19. Here’s an excellent explanation (free registration required) of it.
The study isn’t peer-reviewed yet and we are reluctant to give too much coverage to so-called pre-prints. Too many poor but sensational preprint COVID-19 studies have received too much media coverage. But the group that conducted this study is as reputable as it gets.

Spanish prevalence study

The Spanish public health institute, Carlos III, has published a large survey of people who have been infected with SARS-CoV-2 (the coronavirus that causes COVID-19). It found that 5% of the population has been infected.
The study used an antibody test, so people who are no longer infected will still test positive. In other words, this survey tried to determine who has at any point till now been infected (even if they’ve since recovered) with SARS-CoV-2.

The official document is in Spanish, but Reuters has a useful English summary.

This was a large study conducted by a reputable institution. But as with all antibody studies, its accuracy depends in large part on the quality of the test used. A Financial Times analysis of excess deaths in Spain puts it at about 27 600 during the time of the survey. This indicates a real death rate (infection mortality rate in the official jargon) of approximately 1.1%. Not all these excess deaths are necessarily COVID-19. It is possible that some people with other diseases might be dying because they are squeezed out of the health system, but that nevertheless is a consequence of the pandemic.

If we use this mortality rate for the South African population, then this comes to very roughly 30 000, 155 000, 310 000 and 470 000 excess deaths if 5%, 25%, 50% or 75% of the population gets infected, respectively.

But we must be cautious.

First, all this depends on the quality of the Spanish study. If it turns out to be junk, then the mortality estimate is wrong. Second, South Africa has a much younger population than Spain, suggesting we would have fewer deaths. On the other hand, matched for age, our population is much less healthy and our health system much worse than Spain, which may push mortality higher. Also, if the infections take place over a long period of time (i.e. a very flat curve due to social distancing measures), the health system may not be overwhelmed, treatments might improve, and outcomes might be better.

More hydroxychloroquine setbacks

Two new studies published in the British Medical Journal show no significant benefit for COVID-19 patients taking the drug hydroxychloroquine. One was a randomised controlled trial of 150 hospitalised patients in China with mild to moderate or severe illness and the other was an observational study of 181 hospitalised patients in France requiring oxygen. It might yet turn out that hydroxychloroquine alone or in combination with another drug benefits some sub-group of people with COVID-19, but so far the evidence is very discouraging.

Sensitivity of PCR tests

The type of test currently used to diagnose COVID-19 (a RT-PCR test) sometimes produces false negatives. A study in AIM estimated (based on previous studies) that the probability of a false negative in an infected person is 100% on the first day of infection, after which it drops to 38% five days later (when symptoms typically appear), reaches a low of 20% three days later, and then rises again. The estimates are quite uncertain (wide confidence intervals in statistical jargon) – but the message that people with COVID-19 can test negative early in the course of their infections is clear.

Vaccine hope?

An experimental vaccine being developed by Moderna Inc and NIAID made headlines this week as interim data from a phase I trial of the vaccine was made public (explained nicely in this New York Times report).

The vaccine appeared to be safe in the eight healthy study participants and there were signs of an immune response, but this is still extremely early data and a tiny number of people. We will only know whether the vaccine is safe and effective once large randomised controlled trials have been conducted. Vaccines for the treatment of HIV have also often started off with promising results and then disappointed, so don’t get your hopes up too high.

Even though an effective treatment or vaccine for COVID-19 is still not in sight, the speed at which COVID-19 science has progressed is unprecedented. This is apparent from a timeline of the efforts to solve the protein structures of SARS-CoV-2 published this week in Nature.

From South Africa

The NICD continues to make more useful data available. In previous COVID-19 Reports we noted the weekly COVID-19 epidemiological reports (latest one here). The NICD has also started publishing daily COVID-19 reports (report for 17 May), and a useful summary of COVID-19 testing. One interesting thing to note is the graphs showing cases by date of specimen collection rather than by date of test result. This makes the data less likely to be distorted by long turnaround times at labs.

Another useful NICD publication is titled Impact of COVID-19 intervention on TB testing in South Africa. It shows that daily TB testing volumes declined sharply after commencement of the lockdown. The average number of TB tests per week dropped from 47,520 prior to the lockdown to 24,574 during the lockdown period (a 48% decline).

The MRC continues to update the weekly number of deaths. The latest figures go up to 12 May. As has been the case since late March, deaths remain lower than expected because of reduced homicides and vehicle accidents. There is no noticeable uptick in natural deaths yet.


AIM (Annals of Internal medicine): A leading medical journal
• BMJ (British Medical Journal): A leading medical journal

• Hydroxychloroquine: Drug used to treat malaria, rheumatoid arthritis and other conditions.

• MRC (Medical Research Council): South African state research institution.

• Nature: Leading science journal

NIAID (National Institute of Allergy and Infectious Diseases): One of the leading medical research institutions in the world and one of the 27 institutions that make up the US National Institutes of Health (NIH).

• NICD (National Institute for Communicable Diseases): South African state institution responsible for dealing with infectious diseases like TB and Covid-19.


SIGN UP for our newsletter