South Africa has experienced the Covid-19 epidemic rather differently from many other countries and there is much interest and speculation why this might be. It appears to be different in two ways. First, the total number of infections presenting with illness or discovered through limited community testing is relatively low at this point (4,361 on 25 April). Compared with the United Kingdom, or with some comparable developing countries, such as Brazil and Mexico, there are far fewer cases per million population, and South Africa finds itself in the company of countries such as Finland, Egypt and Argentina.
The first explanation for these lower rates is simply that the initial infectious cases only arrived in South Africa a month or more after it started in the UK and US. One way of marking the “take-off” point would be the date on which the number of cases exceeded 100. In SA that was 18 March. In the UK and US, it was 4 and 5 March, respectively. The longer period of community transmission in some countries is certainly part of the explanation.
But in Brazil and Mexico the take-off dates were 13 and 19 March respectively. The fact that Mexico, at 13,800 and Brazil at 59,300 are so much higher than SA even though the epidemics took off at roughly the same time, and that the number of cases in the UK and US is far greater than can be accounted for by the two-week lag, suggests that something else is at play. This also challenges another speculative explanation — that warmer temperatures, as found in South Africa in summer, inhibits Sars-Cov-2 transmission. While other coronaviruses show marked winter seasonality, and there is laboratory evidence that the Covid-19 virus has decreased survival in high humidity and warmer temperatures, there is plenty of evidence, even as illustrated with the few examples here, that transmission may be high in warm climates.
There was initially a concern that SA might be missing most cases since testing had been done by private laboratories accessible only to middle-class patients, with the state laboratories initially lacking the capacity to test anyone other than those with contact history and symptoms. This presumed undercount would also explain the apparent skewed distribution of cases predominating in middle-class areas. But with the extension of testing to poorer communities, no hidden mass of infections has yet been discovered.
So something else has happened. This is well illustrated by the kink in the SA epidemic curve at 27 March (see Figure 1). The curves of SA and UK cases track each other for the first 10 days after “take-off” but then there is a sudden decline in the rate of new cases. This is most likely due to the declaration of a “State of Disaster” two weeks earlier, though effecting social distancing would only have had a significant impact on symptomatic infections after some 10 days.
At the time when restrictions were imposed, the infections were all but restricted to the middle-class suburbs from which most business and recreational travel originate and which are tourism magnets. The lockdown, in particular, reduced the spread to rural areas, townships and overcrowded informal settlements, reflecting the ongoing legacy of apartheid geography. The infections have scarcely taken hold in these areas, where the majority of the population resides, and are still so few that it will still take some weeks before we see the explosive growth in numbers.
The second difference between South Africa and most other countries is the low death rates (see figure 2). Covid-19 deaths per 100,000 population as of 22 April were: UK 27.3, US 14.2, Germany 6.4, Brazil 1.39, South Korea 0.46, China 0.33, Egypt 0.28, and SA 0.11. In other words, the rate is nearly 300 times higher in the UK than SA.
In fact the population death rate tells us nothing, since the number of deaths are related to the number of cases, and that in turn is strongly dependent on where each country is in terms of when the first cases occurred, and how many cases (usually infected travellers) seeded the local epidemic.
More useful will be the case fatality ratio (CFR), ie the cumulative number of deaths compared with the cumulative number of cases. Here too, however, it appears South Africa is an outlier. SA’s CFR is 1.8% compared with, for example, the UK at 13.5%, Brazil at 6.4%, or South Korea at 2.2% (see figure). Expressed another way, if there were a similar number of infections in each country, for every death in South Africa, there would be 1.2 in South Korea, two in Germany, four in Brazil and Egypt, six in the US and eight in the UK.
This confounds predictions for several reasons.
First, it is generally thought that populations with high levels of comorbidity, as is found in South Africa, will have higher mortality rates. SA has about 7.7 million people living with HIV (13% of the population) of whom about 2.5 million are not on treatment and therefore very vulnerable to infections. Tuberculosis rates are very high too, with about a third of a million new infections annually, and 63,000 deaths annually.
Second, it is increasingly being found that death rates are related to risk factors, including poverty, nutrition, overcrowding and exposure, access to health services and so on. This has been amply demonstrated in the US where, for example, while black Americans represent only about 13% of the population in the states reporting racial/ethnic information, they account for about 34% of total Covid-19 deaths in those states. In the UK, death rates per 100,000 population were 23 for white British, 27 for Asians and 43 for black people (as categorised by NHS England and the Office for National Statistics). In South Africa, with very high levels of inequality and deprivation, one might expect similarly raised death rates.
A third factor that determines the CFR and the differences in CFR between countries is the adequacy of the health service to treat those who become severely ill – hence the emphasis on flattening the curve to avoid the number of severe cases exceeding the country’s hospital capacity. SA is poorly endowed with hospital beds, personnel and ventilators compared with most of the other countries quoted here, and so this ought to lead to a higher death ratio, rather than the low CFR we observe. However, the general state of our national health services may not be relevant at this stage for two reasons.
First, the significant flattening of the curve means that we have had relatively few seriously ill patients, well within the capacity of our public and private health services. Once infections go up and exceed that capacity, the death rate may well increase.
A further factor is that while our public hospitals’ services are patchy in their capacity to support patients in intensive care, the majority of sick cases have occurred within the middle-class population with medical aid and access to private hospitals with excellent ICU facilities. This will have helped to keep the CFR low thus far. Thus the weaknesses in the SA health system will only become a contributing factor to death rates once the total demand on the health service exceeds its capacity, and that has not happened yet.
Another factor that may explain differing CFRs across countries is the proportion of the population above, say, age 55. In SA that proportion is 13%. In the UK it is 31%. The higher proportion of older people will result in more deaths for the same number of infections. I do not think this explains the low death rate here for two reasons. First, compared to countries like Egypt, Brazil and Mexico, which also have relatively young populations, this argument would not be sufficient since they have much higher death rates. Second, the epidemic in SA has, thus far, been concentrated in geographic communities that have older age distributions. The relative youth of the population may be expected to be a factor in the future, but cannot explain the pattern thus far.
One theory for the lower Covid-19 death rates in SA and other developing countries relates to BCG, the anti-TB vaccine. The theory is that South Africa, and many other developing countries and also Portugal, have high BCG vaccination levels and low Covid-19 infection and death rates, whereas the reverse is true for the US and many European countries including Portugal’s neighbour, Spain. However, the data to support this is suspect.
Many countries with low BCG rates now had compulsory BCG a few decades ago, such that almost all the population over 60 are in fact vaccinated – yet they are the most vulnerable age groups in those countries. Furthermore, at the individual level, there does not appear to be any correlation between those with severe illness and their BCG status. Moreover, there are many exceptions to the inverse correlation of country BCG rates and Covid-19 rates. Well-designed studies will still be required to see if there is any protective effect.
Probably the most important reasons for the variation in death rates, at least between countries that are similar in many relevant respects, is how widely each country has tested for the virus rather than relying only on severely symptomatic cases that come to hospital as the measure of infections. If the infections are undercounted, this will naturally appear to increase the death rate. Yet in SA, until the last 10 days, testing was also restricted to people who were symptomatic. If anything, there has been less testing here than in many other countries which should overstate the CFR. In fact as testing has expanded, the CFR has gone up slightly.
So we do not yet have a good enough explanation of South Africa’s lower death rates. The predominance of cases in middle-class areas and the effect of the lockdown in limiting the spread to more vulnerable populations is probably the most significant factor. A second is the fact that at this stage the health services are coping with the relatively small numbers. No doubt there might be other factors such as immunity related to prior exposure to other infections, average infectious dose exposure (often higher when people are in close quarters such as in cold weather), and genetic factors, but these remain to be demonstrated.
We may yet be lucky that our population is more resilient to the infection than elsewhere, and a relatively small older population will be helpful – it is too early to tell. In particular, with only 87 deaths (as of 27 April), and almost none in HIV or TB patients, we do not know how this will impact on the CFRs.
What may we expect ahead? The easing of restrictions on movement and work from 1 May will likely rapidly increase the flow of infected individuals between middle-class areas, industrial clusters, and the dormitory cities surrounding them. The impossibility of social distancing in informal settlements, housing several million people, will likely lead to a rapid spread of Covid-19 in those areas.
We have no reason to think the rate of transmission would be any different from other countries once the lockdown is eased, and it may well be worse given the high rates of HIV and TB mentioned above. We can only hope that the six-week delay in the exponential rise of cases will have been enough to increase testing and field hospital capacity, the supply of protective equipment and the inculcation of social distancing habits and lifestyle to hold the epidemic from its worst-case forecasts. DM
Dr Max Price is the former vice-chancellor of the University of Cape Town. He is a medical doctor. A version of this article first appeared in the London Sunday Times.
"We spend the first year of a child's life teaching it to walk and talk and the rest of its life to shut up and sit down. There's something wrong there." ~ Neil deGrasse Tyson