8 April 2020
Most of the Covid-19 numbers we have paint only part of the picture and can easily be misinterpreted. Here are some tips on understanding the numbers.
Every day Minister of Health Dr Zweli Mkhize provides the total confirmed cases and the increase in confirmed cases since the previous day. But how much can we really tell from these numbers?
Confirmed cases are only a fraction of the total cases in the country. Maybe the real number of cases is twice or maybe it is ten times the reported cases; we simply don’t know. As testing scales up, the picture may become a bit clearer. Unfortunately, in the past ten days or so the number of tests has declined from a peak of 8,066 on 27 March to 1,225 on 6 April. (Hopefully this will improve in the next few days if the government successfully rolls out the 45-minute GeneXpert testing machines.)
The number of confirmed cases depends on the number of tests done and how well those tests are targeted. All other things being equal, a sudden increase in testing capacity might result in a sudden increase in confirmed cases. This would not mean that there is definitely explosive growth in the number of cases; only that we would be finding more of them.
Similarly, it can become easier or harder over time to know who to test. Most of South Africa’s early Covid-19 cases were people travelling back from overseas, where they probably contracted the virus. It made sense to focus testing on people with Covid-19 symptoms and who had travelled to affected countries. South Africa’s travel ban came into effect on 18 March – effectively turning off this source of new infections. Now, as recent overseas travel has become a less important indicator of possible infection, it is becoming harder to know who to test.
All this probably explains why the daily increases haven’t been that high in recent days. Everyday from 23 to 27 March, over a hundred new cases were reported in South Africa – from 28 March to 6 April the daily increases were under a hundred.
Apart from all this, there will also be random fluctuations in the data. By chance you may find 80 confirmed cases one day and 60 the next day. In itself a decrease in confirmed cases like this means nothing.
For all these reasons it’s best to be cautious about how you interpret day-to-day changes in confirmed cases.
Calculating the death rate is not as simple as dividing the number of deaths by the number of cases. For example, as of 7 April, 13 deaths in 1,749 cases were reported in South Africa - giving a death rate of 0.74%. But it is much more complicated than that. For one thing, most of the 1,749 people confirmed to have Covid-19 on 7 April were still infected and some of them may still die.
By the time South Africa had its 1,000th case, the country had no Covid-19 deaths. All our deaths have come subsequently. Similarly, Germany also had no deaths when it had its 1,000th case. As time has passed, the death rate in that country has crept up. It’s now about 2%.
The death rates usually quoted in the media, and those above, are case fatality rates (CFRs): the number of deaths divided by the number of cases (multiply by 100 to convert to a percentage). Since the CFR only uses known cases and known deaths, it is heavily influenced by how well our Covid-19 testing and cause of death systems work.
The UK CFR rate is extraordinarily high, over 10% at the time of writing. Germany is only at 2%. This tells us more about how widely each country has tested for the virus than about the deadliness of the virus in the two countries. The UK has left wide-scale testing very late.
Testing differences between countries have been well-reported, but accurately tracking Covid-19 deaths (the other part of calculating CFR) is also not straight-forward. The Economist reports that in parts of Europe the number of expected deaths over the past few weeks has increased dramatically, and not all of these deaths have been attributed to Covid-19. This could mean people are dying of Covid-19 without being diagnosed with it (perhaps even at home), or that Covid-19 is putting so much pressure on health systems that people with other health problems who would normally be saved by the system are dying. It will be a while before we understand this properly.
While some differences in CFR are about how well cases and deaths are being measured, there are also real differences between how deadly Covid-19 is in different settings. Countries with older populations will likely have more Covid-19 deaths. But so might countries with younger populations but poorer health systems or less healthy populations. Treatment capacity also matters, because people who may otherwise be saved might not be saved if the capacity to provide intensive care is not available.
We don’t know. But the available data suggests this is a very infectious virus. At the beginning of the epidemics in China and Europe, the average infected person was infecting at least two, maybe even three or four, other people before recovering or dying. And they in turn were infecting two, three or four people. This resulted in the epidemic growing very quickly.
What we need to do to wipe out the epidemic is reduce the average number of times an infected person infects another person to below one. Then the epidemic will fade away.
The rate of new infections depends on three things: how many people the average person comes into contact with daily, how long people remain infectious, and how easily the virus transmits between two people in contact.
Let’s say you are infected and you come into contact with ten uninfected people daily and the risk of infecting any one of them is 5% (these numbers are completely made up; we have no idea what they really are). Multiply these two numbers together and the risk that you infect any one person a day is 50%. If you are infected for four days, you will infect two people (it’s more complicated than that in reality but this is close enough).
The point of the lockdown is to reduce the average number of people we come into contact with. If the same person instead of being in contact with ten people a day is in contact with only two per day, then over four days that person might not infect anyone.
The point of getting people to wash their hands often and reducing how often they touch their faces is to reduce the risk of infection per contact.
If we reduce both the number of contacts and the risk per contact enough, we may bring the epidemic under control.
A reader has pointed out that a better definition of CFR is deaths divided by the sum of the deaths and recoveries. Although this is a sensible definition, the recovery estimates vary dramatically from country to country.