‘We are planning under the assumption that 80% of the population will be affected and with a mortality rate of 1% there are likely to be up to 800,000 deaths’
A typical report over the last few weeks, a reflection of comments from health ministers and politicians in various countries. But what does it mean?
Basically, this is primary level mathematics. The population of this hypothetical country is 100 million. 80% of people being infected is a number picked out of the air and 1% whilst not randomly chosen is far from hard data.
We therefore have 80% of 100 million x 1% and bingo we have a death number. But is this prediction valid and reliable? No, is the unsurprising answer.
Let's assume the population is fixed and ignore the fact that some people will die (and be born). Let us consider 100 million to be hard data.
What about that 1% mortality rate? The mean age of death in the first published Italian cohort was 81. It was actually 83.4 for females and 79.9 for men and there was a 4 to 1 ratio of males to females. Two thirds of patients had other disease or were smokers. The Italian cluster were slightly older than the studies from China but otherwise the general trends are similar. So maybe this illness will have the greatest impact on older people with other illnesses and especially men? That seems to be the case. In the first 44,000 cases in China the case fatality over 80 years of age was 14.8%. This also means that 85.2% of elderly patients who became ill enough to go to hospital with COVID-19 recovered.
1. The generalizability issue:
Disease does not always behave in the same way in different populations. I have described in a previous article becoming interested in this concept as a young doctor working with HIV/AIDS. Being told that I had a 25% chance of being dead within 15 years (actually higher as a health care worker) naturally made an impression. It was a prediction that laid the foundation of a personal interest in epidemiology. There are predictable reasons why such reductionist projections are invariably wrong early in epidemics. I described viral load as a factor in the severity of illness of HIV/AIDS. I suspect it will also be a factor in COVID-19. Age, male sex, co-morbidity and smoking certainly seem to be factors. It is a common and frequently repeated error to take mortality rates from single clusters and extrapolate to fundamentally different populations.
2. The denominator issue:
I have explained the dilemma of predicting mortality early in epidemics previously. The mortality rate is calculated by a simple formula:
Number of deaths from COVID-19 / Number of cases of COVID-19
There is a problem with both the numerator and the denominator early in epidemics. The denominator is the bigger issue as it is harder to establish until widespread antibody tests become available. How many mild cases are there in the community for every diagnosed case? Where are all the children and young people in this cohort? Until we get more accurate testing we will not know how many mild cases have never been diagnosed. COVID-19 has a current crude mortality rate of around 4% (this is the crude fatality rate of the China and combined rest of the world data). But what about inter cluster variability? Mortality in Italy is 9.5%, but in Korea it is 1.3%. For Germany it is 0.4% and in Singapore there have only been 2 deaths with > 500 cases. The role of health care systems in these discrepancies is overstated. The biggest difference is the number of tests performed and hence the number of milder cases recognized. If we eventually do widespread antibody testing and find that there are 10 mild cases for every diagnosis made that changes the mortality in our imaginary country from 1% to 0.1% which changes 800,000 deaths to 80,000 even before the factors above influence our target population.
So what about the 80%? Again this is a simplification. This disease is clearly very infectious. We know that above a certain threshold the immunity of the population stops an epidemic from being able to take hold. There is a lot of talk about a ‘worst case scenario’ assuming this to be 80%. In the real world this number will be impacted by unpredictable factors in the epidemic including the public health measures, climate and the potential natural attenuation (weakening) of the virus with time. As a rule of thumb an infectious illness with R0 of 2 will begin to die as 50% of the population acquire immunity. For R0 of 3 the number rises to around 66%. This is the concept of herd immunity which forms a part of the UK strategy. This is not however the whole strategy. The UK is advocating targeted social distancing measures to flatten the curve, isolate vulnerable and elderly patients whilst maintaining some degree of social mixing such that the less vulnerable members of the community develop immunity. At the same time shifting resources to mitigation and management rather than testing and aggressive containment.
Essentially, we have an academic debate in public health circles which is playing out on the world stage. The WHO have clearly adopted a policy based upon testing, isolation and quarantine to contain the disease. This has been successful in China, Hong Kong, Singapore and Korea and provides a model for reducing the total epidemic size. This buys time to get more information on the best containment and mitigation strategies, develop testing, research treatments and work on a vaccine whilst also hoping that natural environmental factors help in reducing the epidemic. The counter argument is that once the disease has ‘escaped’ the containment phase the populations in the locked down communities will not be able to stay protected forever without the isolation having a significant impact on social and economic life. Unless the disease dies or a vaccine is developed the non-immune populations will eventually need to acquire immunity. The results of these different strategies in addition to the differing methods of social distancing and isolation will provide useful information in managing not only this epidemic, but future epidemics of new disease.
Whether the statements by politicians about predicted mortality are intended to manage expectations or scare behavioural change is not certain. We know from previous epidemics that scare tactics may produce short term impact but they quickly fatigue. In the early stages of the HIV/AIDS epidemic politicians advocated a variety of policies including abstinence. You didn’t need to be a public health expert to predict how that would work as a health policy.
Management of infectious disease is about education and information. Strong political leadership which recognizes uncertainty, honestly acknowledges challenges and risks whilst challenging misinformation and managing anxiety by providing calm reassurance and public health messaging is the ideal. This epidemic has produced some good (and some not so good) examples of political leadership through a public health crisis.