18 August 2020
Eight months into the Covid-19 pandemic, how well has South Africa done at limiting infections, and how big a deal is the pandemic anyway? To answer that, we need to think about how the data actually works and doesn’t work.
Dashboards like Bing and Worldometer report that South Africa has the fifth highest number of infections globally. By the measure of deaths per million people, which more directly expresses how dangerous it is to live somewhere, we are ranked 31st - so nowhere near the podium. Both of these rankings, though, require a pinch of salt. There are many countries that almost certainly have more infections — see the grey box below.
South Africa has performed more tests than most (16th place, as far as we can tell) and we track births and deaths about as well as some developed countries. These two factors, of which we can be proud, unfortunately also make us look worse than many countries that have little capacity to test for Covid, or to count deaths.
Officially, we have had about 12,000 “confirmed” Covid-19 deaths. However, the Medical Research Council (MRC) reports that there were 27,500 more deaths (from all causes) between 6 May and 4 August, than we have seen in that time window in recent years. That number is probably a conservative estimate of the true Covid-19 death toll. Here’s why:
Firstly: The number of confirmed Covid deaths only started increasing dramatically in June, and is now on the decline. This aligns impressively with the MRC mortality trends. ‘Collateral’ deaths - non-Covid deaths brought on by generally reduced access to health care - would be expected to follow the severity of the lockdown, which has been slowly tapering off due to relaxation of rules and fatigue.
Secondly: Because of the lockdown, deaths from “unnatural” causes (particularly homicides and accidents) have dropped dramatically, even as the total has gone up.
Thirdly: Deaths from other infectious diseases which we usually see at this time of the year have also dropped dramatically. There has been no significant flu season. In children, there have been very few Respiratory Syncytial Virus (RSV) infections and other respiratory tract infections that usually cause deaths during winter, and less TB than usual.
Fourthly: Taking into account the previous two points, the MRC produced a best estimate of excess “natural” deaths over the same period - and that estimate runs to 33,500.
Fifthly: The death count gap, comparing the officially confirmed Covid deaths and the total excess deaths, displays a sharp provincial pattern. In the Western Cape, the official Covid death toll is about 70% of the MRC’s excess death estimate. Next is Gauteng, where official Covid deaths are a mere 24% of the estimate’s excess deaths. In four provinces, the official Covid-19 death toll is just one tenth of the excess deaths. This is what we would expect if the excess deaths are mainly direct Covid deaths, but provincial data collection systems are not equally effective at picking them up. To believe that a Covid death in all provinces is roughly equally likely to be recorded as such, would leave us with a strange and unexplained distribution of “collateral” deaths.
Some of the excess deaths are presumably not Covid disease, but the result of difficulties accessing health care, owing to lockdown restrictions, fear of going to a clinic, or over-burdened health systems. Given the preceding points, these are likely to be a relatively small part of the excess deaths.
Reliable long-term disease death predictions are simply not available. Nevertheless, it now seems safe to say that by the end of September there will be over 40,000 actual Covid-19 deaths, though these will not be reflected in the official count.
The epidemic is in decline, but it is far from over and it could surge again if lockdown relaxes and if we collectively drop our guard with physical distancing, face masks and hand hygiene - that is what is currently happening in many countries.
In South Africa, annual deaths have been declining since 2006, mainly thanks to antiretroviral treatment for HIV, and progress against TB. There are also about 50,000 unnatural deaths annually, but barring an outbreak of extreme violence, there will be far fewer this year.
There were about 450,000 recorded deaths in 2017, the latest year for which official numbers are available. Covid will increase our death toll this year by 10% or more. That’s huge.
TB is the most frequently recorded cause of death in South Africa, and there were under 30,000 such deaths in 2017 - fewer than any serious estimate of Covid deaths in 2020. Even though the real TB number is likely to be significantly higher due to how death certificates get filled in, we can reasonably expect more Covid deaths this year than TB deaths.
South Africa’s most informative AIDS model (Thembisa) estimated that there will be 63,000 AIDS deaths this year. That was before we had social distancing and all the anti-Covid measures which have reduced transmission of respiratory infections - a major immediate cause of death for people living with HIV.
But taking a longer term view, Covid in South Africa doesn’t even come close to the HIV epidemic. In 2006, about 300,000 people died of AIDS here, almost 6,000 people per week, sustained throughout the year. At the height of the just-passed Covid peak, it was about as bad - more than 6,000 excess natural deaths in the week of 15 July. Because many people get very sick from Covid and recover with oxygen support, some hospitals were for a few weeks hit harder by it than they were ever hit by HIV.
It’s clear that the measures taken since 16 March have saved lives. Lockdowns work. China, where measures were highly enforced, and testing was pushed as hard as possible, had a tiny first peak compared with European countries. We know it worked here because we see the reduction in other infectious diseases. Perhaps the lack of international traffic helped prevent the flu season, but it doesn’t explain the drop in other infections in children which do not depend on international travellers.
The restrictions on inter-provincial travel, coupled with masks, physical distancing, hand hygiene, and a massive reduction in public events and large gatherings have likely played their part.
Had we continued life more or less as normal after 16 March, what would have happened? It’s impossible to produce credible hard numbers, but our guess is that several times as many deaths have been prevented; we have probably avoided a devastating flood of deaths.
Yes, planning for the surge of cases, in some provinces, was abysmal. The ban on cigarettes made no sense; nor did the restrictions on exercise hours. Over-zealous restrictions on workplace activity have left the economy reeling, but a comparison of other countries (including Sweden) shows that epidemics can easily stall economies even if one does not explicitly decree a halt to most activities. Stealing protective equipment contracts, and other opportunistic corruption, is indeed abominable. Despite all this, it could easily have been so much worse. The government has not done too badly in its stated main aim: to reduce Covid-19 deaths.
Further reading: MRC analysis of excess deaths
Worldometer says 777,076 people have died of Covid-19 as of 17 August. Except for clerical errors and some minor technical exceptions, every single one of those deaths links back to a specific person, certified by a health care worker as having probably died of the disease. (Bing and the WHO also tally the daily deaths with slight differences.) This is an astonishing accomplishment, and a first in history. Yes, there have been gimmicky daily counters of global AIDS and other disease deaths on some websites, but these are at best rough estimates, based on sparse direct data, spread thin by mathematical models.
Yet far more people have died of Covid-19 than are recorded on the global dashboards. There are several reasons why people who die of the disease don’t get counted in the official stats. Many people die at home or in care institutions, without being diagnosed. Even if they’ve been tested, their results often come back long after a doctor has filled in the death certificate, and PCR tests miss many positive cases.
The New York Times (among others) tracks “missing deaths” in 28 locations. This is estimated by subtracting the excess deaths in the location from the official Covid-19 death toll. As of 31 July there were 161,000 missing deaths in these locations. This means that in about eight months at least 938,000 people have died of Covid-19 (i.e. adding specifically counted deaths from WHO, Worldometer, Bing etc to the missing deaths). Some of these excess deaths may be due to health systems being overrun or people being too scared to go near health facilities, but they are nevertheless deaths due to the pandemic.
But in fact the real death toll is much greater than this. Those 28 locations being tracked by the Times don’t even account for half the world’s population. Missing are China, India, Pakistan, Bangladesh, Iran, most of Russia, most of Indonesia, all of Africa except South Africa, and many other places. The places where excess deaths are being counted are generally those with the best systems for keeping vital statistics, and counting Covid deaths directly.
Although current data collection systems are far better than ever before, we really don’t know the total number of Covid deaths. It is definitely over a million; it could well be over two million already; and even three million is not implausible.
About 60 million people die globally annually. So the Covid death toll, a brand new way of dying upon all the other ways we die, is a significant extra few percent, even using the minimum possible estimate. It’s also important to account for the ages at which people die, and while covid fatality is highest at older ages, that’s true of death generally, and covid is certainly striking down many otherwise healthy people who had reasonable expectations of living out many more years.
This lack of certainty is not new; we should treat all estimates of disease tolls with caution. Although it’s commonly said that 50 million people died in the Spanish flu pandemic of 1918-19, this is just intelligent guessing; it could be wildly wrong.
But there are some things that we can say with great confidence.
This is unequivocally the worst short-duration pandemic since 1919. In bad years, standard seasonal flu kills no more than 650,000 people globally, but usually much fewer than that (about 400k). The 1957 and 1968 flu pandemics were exceptional, killing about a million people (estimates, not known for sure). What’s more, most countries in the world, and most people, have taken exceptional measures to avoid contracting Covid-19. Things would be much, much worse had we carried on as people did in 1957 and 1968; perhaps as bad as the Spanish flu, which is estimated to be the worst pandemic in history in terms of total number of people killed - though far from the worst in terms of the proportion of the total human population.
Compared to the slow-acting infectious pandemics, like AIDS and TB, Covid-19 will kill more than both those diseases this year, but cumulatively, HIV and TB have been far worse.