The advanced economies can keep the vast majority of their workforce working normally allowing the economy to continue without major disruptions, while a small subset of the older population and the ones with underlying health issues stay in strict lockdown mode. This exposes the working population to a risk that is of the same order of magnitude as the risk of driving a vehicle during a year, while enabling the resources to keep the economy from collapsing and potentially developing a group immunity to the disease quicker.
Note for those who won't read further: This strategy is only applicable after this first wave of existing infections is controlled. Right now, in most countries, there are too many infected already in the system for this strategy to work because there is currently no spare health system capacity. (Full )
Here is the breakdown
I am going to show the Case Fatality Rates based for different cutoff ages of the working population. (The population below 60 years old has a case fatality rate of about 0.2% and represents about 90% of the work force). Many people are infected but are not detected because they have no symptoms. This reduces the real fatality rate by about 50-70% (given that about 50% to 75% of infected have no symptoms and are not detected). Even more than age, the primary reason for death in Covid is pre-existing health conditions. In Italy 98.8% of deaths had a pre-existing health condition. Strictly isolating this group further reduces the death rate by another 80-90% - this is the major driver in reducing the death rates! Compliance rates with that mandatory isolation is critical for the system to work. I'll adjust the risk rates for the whole working population assuming that most of the population will eventually get infected. (Only need about 70-75% of the population to become infected for the epidemic to stop naturally) Finally, I will compare that death risk rate with a risk of driving one year in order to show that it is comparable.
What is the risk of death from the novel Corona Virus?
The risk of death depends primarily on the age of the infected and the existence of previous health conditions in the patient. For people above 80 years old the death rate is more than 200 times higher than that for infected people below 50 years old for example.
So, the government could choose to lockdown only those with high risk of death, and allow younger people to work and keep the economy going.
To calculate the risk of such a strategy, I used data from South Korea.
Why South Korea?
It is the country that is testing the highest number of people per population and is also an advanced economy with good health care. It also didn't apply drastic social distancing. Hence a good model for Europe and the US if most of the population would continue working.
South Korea: Infections and deaths per age group
Death rate per confirmed case
: Deaths by age group in South Korea as of April 7th. Note: different countries have different death rates - .
Workforce by age group
Ideally, you want to keep as much population as possible working while reducing the overall risk for those at work.
Raw rate of death per confirmed infected
The working population changes a lot by country. I use the . If we want to keep the vast majority of the working force going to work normally (and the others in lockdown still able to join via remote work), then a sensible age cutoff for the lockdown would be about 60 years old (because that would represent about 10% of the active population). In this dynamic section we calculate the death rate per 100,000 for the ages under : years old. (You can change this slider to try other numbers - and all the rest of the document will change accordingly). Number of Deaths for that age was and had confirmed cases (based on data from Korea), or %. Therefore that the Death rate is per 100,000 confirmed infected cases. You can see the same statistics for Spain and Italy . (Spain is about 2.8X higher, and Italy is about 3.8X higher. But aggregating data from Germany, Spain, Italy, Korea shows a death rate of % below 60 years old).
Infected but not detected
Besides the detected cases in Korea, there are still many asymptomatic people who don't get tested but are infected. This number is difficult to estimate, but there are a couple studies where 100% of a small population was measured. And hence the true number of asymptomatic cases is known for those populations.
The head of the civil protection agency in Italy said that "It is credible to estimate that there are 10 positive case for every one officially reported."(). This would mean a 1 to 10 ratio of detected to infected. This is either because they have mild symptoms aren't tested, or are completely asymptomatic. One study shows data from a town in Italy that tested everyone in town (3,000 people), and that for each patient that is infected and has symptoms, about 1 to 3 are infected and completely asymptomatic ( and ). It could be that the completely asymptomatic in this sample also contain people who only later developed symptoms (pre-symptomatic). Based on the study from that town, Andrea Crisanti (professor of microbiology at the University of Padua, Italy) and Antonio Cassone (former director of the department of infectious diseases at the Italian institute of health) write in the Guardian that the "asymptomatic or quasi-symptomatic subjects represent a good 70% of all virus-infected people" This would mean a ratio of 1 to 2.33 for people infected but unlikely to be detected. On the other hand, a study from 536 evacuees from Wuhan to Japan, estimated that the asymptomatic rate on the infected is about 31% of all the infected among evacuees. Meaning for each infected that is symptomatic, there is about people who are not symptomatic.(). That suggest a lower rate than the Italian town, but it is also a smaller and more biased population than a complete town.
South Korea, the country with the highest test rate per population, has only tested about 0.5% of their population and it is very skewed towards suspected people. A randomized sample of the population would give the real statistics - but so far no results that I'm aware exist. Therefore, we need to model based some assumption of the true ratio of infected but not detected.
Now for our model let us assume that for each confirmed infected there is at least person that is infected but not known (asymptomatic or quasi-symptomatic and not detected).
Taking into account these infected but not symptomatic cases:
There are death per 100,000 infected cases of age below years old.
Patients with previous health issues
An absolutely critical factor for the mortality with Covid-19 is the existence of pre-existing health conditions in the patients. This factor seems to be even more important than age.
Data from Italy on pre-existing conditions, shows that 98.8% of the Covid19 deaths had a pre-existing condition. And the average number of diseases was 2.7 per death (: Summary report Italian Covid-19 Surveillance Group: 20th March). Similarly, in Spain 90% of those who died had at least one pre-existing condition (: Table 7: Enfermedades y factores de riesgo, 3rd of April).
If people with pre-existing conditions are in a strict lockdown, you can see the huge effect this will have on reducing the overall mortality.
For younger ages, the effect of pre-existing conditions is also critical. It is likely to be even more critical than for older ages.
According to the Italian Covid Surveillance group, out of a sample of 3200 positive patients that died (up to March 20th) only 9 of these were younger than 40 years old, and of those 7 had serious pre-existing pathologies (cardiovascular, renal, psychiatric pathologies, diabetes, obesity) and for the remaining 2 patients no clinical information was available. (: Summary report Italian Covid-19 Surveillance Group: 20th March). ( from 7th of April, shows what out of the 36 patients younger than 40 years old that died in Italy (and who had information), there were 28 (=78%) who had serious pre-existing pathologies and 8 had no major pathologies). In France the numbers also show a very high number for pre-existing conditions on the deaths from coronavirus in hospitals: 91% deaths of age <64 years old in ICUs had co-morbidities (and also for 92% between 65-74 years old). (: Table 3 - Reanimation service in Hospitals from Coronavirus between 16/03 au 29/03/2020 French Institute of Public Health - April 2nd) Another data point is the final report of the WHO joint Mission to China ( and ): "Patients who reported no comorbid conditions had a CFR (Case Fatality rate) of 1.4%, patients with comorbid conditions had much higher rates: 13.2% for those with cardiovascular disease, 9.2% for diabetes, 8.4% for hypertension, 8.0% for chronic respiratory disease, and 7.6% for cancer." This means a large increase in CFR of about 7-8X for relatively common issues like cardiovascular disease or diabetes. And this data is probably underestimating it, because in severe cases that lead to a death, it is quite possible that a full history of the patient is unknown, and hence it would count as no-comorbid. According to Lee Riley, chair of the division of infectious disease and vaccinology at the University of California, Berkeley: "Where you find severe cases or even death in young people, we don’t really have full information on these patients" ().
Previous conditions correlate with age, so the data in the table conflates both age and higher prevalence of underlying conditions. Also the same person can have multiple comorbid diseases. I have not yet come across a data source that has age group and previous health conditions broken down by disease outcome.
Given this data, we should assume that previous health conditions is a major reason for deaths in any age bucket including the young.
For the model, a key parameter is the percentage of people in lower age groups that die and have a previous health conditions within the whole age bucket: % With this assumption, it means that there are death per 100,000 people of age below years old and previously without underlying health conditions.
Whole population (infected and not infected)
Mathematical models show that at an infection rate of around 70% of the population the virus stops spreading naturally because the remaining infected people that are still infectious would naturally only meet other people who have previously been infected.
Secondly, the summer is also approaching, and it is likely that this virus will slow the infection rate due to higher temperature and UV light from the sun, like the other Coronaviruses do.
Let's assume that in the end the rate of actual infected is % of the population (below years old). If that is the case, the risk of death for the population below age and no previous health condition is equal to per 100,000 people
Comparing to other known risks
To compare the risk of death with other known causes, the motor vehicle death rate in the US was deaths per 100,000 population (). This means that the death rate from Covid for the whole population of people below years old and without previous underlying health issues is about of the risk of motor vehicle death in the US in one year.
Governments can restrict the lockdown to the population above 60 years old and also to those with pre-existing health conditions (which is about 10% of the workforce) and thereby protecting the economy from collapsing and at the same time incurring a risk of death for the working population that is lower than the existing risk of death by motor vehicles, even if you reached virus saturation of 75% in that population - which would create group immunity and prevent the harm from future outbreaks.
About myself and this model
I am an entrepreneur who worked at Google on Machine Learning models and AI Research for the last 8 years after my startup got acquired by them. I am not an expert in the subject matter of public health, and this is just an attempt to put some data together to compare the risk of this novel and terrible disease to other risks out there while minimizing the impact to the economy. So pardon me any glaring mistakes that I might have made. I do try to get the data and sources accurately, so please do add any comments with criticism, better data sources that are unknown to me and improvement suggestions.
"The hospitalization rate for the younger population and without pre-conditions (even if they don't die) could still be high, and overwhelm the existing healthcare system. " I need to get some more data on the hospitalization rate for this subgroup of users that don't have pre-existing health conditions and are younger - and also remove hospitalizations that are precautionary rather than absolutely necessary (non ICU). Remember that it can very well be the case that 95%+ of the hospitalizations are driven by older people (or younger but with pre-existing conditions). But without data on whether the younger people were previously very healthy vs having some known underlying condition, even the ICU utilization rate for younger groups would not be telling. It is quite possible that many young people and up in the ICU because they had previous health issues (from the data on deaths - this is the case). Diabetes or a previous respiratory infection that reduced their immune system, or cardiovascular difficulties due to obesity are relatively frequent even in younger age groups. Remember that in Italy only 1.2% of the death had no pre-existing conditions (). The key assumption in this model is that those high risk groups irrespective of their age would be under a strict lock-down. The case fatality rate without these cohorts is a small fraction of the values reported in other popular studies (0.0125% vs 0.9% best case in ). That is a factor of 72x lower rate by excluding these cohorts! In any case, I'll use the epidemic calculator () to model the dynamic effect of such lower rate on the healthcare system capacity once I find some more time. "The motor vehicle deaths occur throughout the year, not concentrated over a few months." That is generally true. I have to get data for the vehicle deaths over time too, but I assume there are also significant clusters of accidents over the holiday season for example. On the other hand, the vehicle deaths occur every year, whereas the Covid-19 (without a vaccination) will likely only occur this year. My main point though was to use a known risk that most people engage with to compare with the risk of death under this thesis. The point is the order of magnitude, not exactly if one is 2-5X higher or lower. I could also look at the death rate over a period of 5 or 10 years, and then the Covid-19 would likely reduce even more in significance. "There will be no group immunity if the virus changes" or "there are re-infections already happening in the population that got the virus". I am not an immunologist nor health care practitioner, so can't evaluate the risk of a substantial mutation of this virus that could render the group immunity (or even the individual immunity) useless. But the group immunity is only a side benefit from this strategy and not it's primarily goal. The primary goal is to avoid having the major economies on full lock down mode for a year (or longer) until the successive waves of infections from countries at different stages come and go. As for the re-infections
South Korea: Infections and deaths per age group