What the Imperial COVID-19 Report Means for Asia

A report released by Imperial College a two days has shaken many countries out of COVID-19 complacency. The report isn’t particularly long, but we will summarize the main findings of the report and the Ferguson epidemiological model that informed it. The Ferguson Model looked at the US and the UK, both of which share enough base similarities that they can be grouped into a single category with much of Southeast Asia.

Three Categories of COVID-19 Countries

Before we explain the report and its model, we think it’s important to separate three broad categories of countries based on their history to date with COVID-19. The COVID-A countries are China and South Korea, which experienced uncontained outbreaks that spread all over the world. COVID-B places bordered China or had high levels of Mainland Chinese traffic, had a collective memory of SARS, and seemed to have identified most imports and contained First Wave community outbreaks.

COVID-C outbreak sites are generally places where surveillance discovered a few imported cases from COVID-A countries in January or early February. By February, they had either travel restrictions or quarantine for Chinese passengers (and then Korea). What we know now is that, as a rule of thumb, only one in four imports are caught by public health surveillance. Of the three-quarters of undetected imports, about a third of those went on to infect about a thousand more people in the next 45 days. 

The epidemiological model for the Philippines, for instance, starts by assuming just ten undetected (or discovered too late) imports in late January. That became the seed for an estimated 3-4K cases in the Philippines today. Two hundred imports in the United States at about the same time period seeded what estimated as about 20k infections by last week. Over the past week, it has become apparent that COVID-C imports the primary Second Wave threat to COVID-A and -B countries which currently have community outbreaks under control. 

AChina and South Korea Places that experienced uncontained outbreaks directly from Wuhan in January and February. 
Currently under control.
BHong Kong, Macau, Taiwan, Vietnam, and Singapore

Likely: Australia, New Zealand
Places at direct risk of Chinese imports, took early decisive action, and seem to have detected community outbreaks early and stopped them. 

At risk of Second Wave imports from Europe and the US, so are currently closing borders. So long as they can catch community outbreaks from 2nd Wave early, they can wait out the epidemic in isolation until a vaccine arrives.
CUS, UK, Italy, Spain, Germany, Philippines, Malaysia

Likely: Thailand, Indonesia, Myanmar, Laos, Cambodia, Latin America

Possible: Japan
Countries where a relatively small number of imports from China were undetected in January and February. Two months later, they are now dealing with a community outbreak that has been building for nearly two months with a 4-5 day doubling time.

The Ferguson Model in the Imperial College report mostly concerns COVID-C countries, though has possible implications for COVID-A and -B. Over the past few days, COVID-A and -C countries have been announcing policies to either quarantine and now ban all international arrivals. The model likely does not apply if these policies can stay in place until a vaccine arrives and insure Second Wave COVID-C imports are discovered and contained. That every country needs a strategy to suppress and contain COVID-19 until a vaccine arrives is the primary takeaway. 

The report was written to wake British, European, and American policy makers up to the fact that unmitigated disaster awaits COVID-C countries that think they can ‘control’ COVID-19 via “herd immunity” or otherwise stay asleep at the wheel. Many leaders, including Boris Johnson, seemed to have resigned themselves to believing there wasn’t much that could be done about the outbreak at this point. The model found that a “do nothing” approach to COVID-19 would lead to an outbreak peak in the UK in late May/early June. 80% of the country would be infected and 510k people would die (2.2m in the United States).

The Ferguson Model

The Ferguson model assumes a few things from the limited data that we have. First, it believes about half of infections are asymptotic which, in turn, are overwhelmingly clusters among the young. The overall infection fatality rate (IFR), then, is much lower than the current confirmed case numbers suggest. They believe the IFR is .9%, but 4.4% with cases severe enough to require hospitalizations. Of those that need hospitalization, 30% require critical care in the form of oxygen ventilation (EMCO). Based on the Italian experience, 50% of that sub-group will die. If ICU capacity is overwhelmed, patients requiring ventilation who don’t receive it will die.

The Ferguson Model looks at four interventions across four potential COVID-19 R0’s (ranging from 2 to 2.6) and five different ‘trigger’ thresholds. Lack of testing and suspicion of the accuracy of Chinese numbers means we’re still uncertain of COVID-19’s R0, though the best guess now is 2.4. There is a chance warmer weather and humidity might lower the R0 as well, though neither this report or a recent model for the Philippines assumes that. The four interventions the model looked at were combinations of the following:

The report assumes COVID-C countries, or any lapsed -A or B countries, will trying to suppress COVID-19 for the next 18-24 months. Because all four interventions can’t be maintained until a vaccine arrives, suppression efforts will need to be turned off and on at different trigger thresholds. These triggers mean “enact X policies when Y number of COVID-19 patients are currently in ICU beds in the country,” and range from 60 to 400 for the UK in the report. 

The model believes school and work closures have a lot more impact than banning mass gatherings because of the short length of time in contact with other people. It also finds the current policy, found nearly everywhere now, of isolating individual cases doesn’t go nearly far enough. Entire households need to be placed in house arrest-style home quarantine when an infection is discovered. 

The modeling found that all four interventions would bring total deaths in the UK down from 510k to 8,600 if the COVID-19’s R0 is 2.4 if the trigger was set to 60 current cases in ICU. The death toll would go up to 39k with the same interventions but with a trigger of 400 current ICU patients. Even with the lowest trigger, case isolation + general social distancing + home quarantine would lead to 85k deaths. Setting the trigger at 100 and implementing all four interventions would look like this:

Successfully ‘bending the curve’ would have the peak number of ICU cases come early, followed by rolling cycles of relaxing suppression interventions until the trigger was reset. In the best case scenario, COVID-C countries will be fighting back waves of outbreaks roughly every two months until a vaccine arrives. 

What the Model Means for Asia

COVID-19 will likely be far more lethal in the -C countries we identify here. Put simply, the baseline critical care capacity of hospitals is likely much lower in most of Southeast Asia than it is in either the US or UK. Where the Ferguson model assumes a 1:1 correlation between ICU beds and oxygen ventilation (EMCU) equipment, we noted yesterday that the number is 1 ventilator per 10 ICU beds in the Philippines. Critical care capacity rapidly degrades the further from Tier 1 cities one gets. Absent extreme interventions, the mortality for COVID-19 will likely look like this in countries with very low EMCU capacity.

The report is more skeptical of China and Korea’s suppression model than we are. In Hong, we braced for a wave of imported cases from China that would overwhelm our ability to treat or even diagnose on time. The high-geographic resolution maps OSINT HK was producing showed that while the overwhelming majority of cases were in Hubei, COVID-19 had spread to every city in China. HKU epidemiologists determined these had already turned into community outbreaks.

In retrospect, an important variable that saved Hong Kong from the fate we feared in the First Wave was Guangdong acting as an COVID-19 firewall. Rather than just looking at Wuhan, Asian countries with low ICU capacity but with uncontained outbreaks should look at lockdowns in places like Shenzhen and Guangzhou as models to follow now. They followed all the interventions this report models and describes, with the exception of isolating population sub-groups at the highest risk of COVID-19.

Chinese provinces in late January are similar to where the Philippines, Malaysia, and Indonesia are today. All they knew was that they had an uncontained outbreak that would remain undetected until testing capacity was scaled up. The lockdowns were eased after two months when testing finally reached sufficient scale to find most, if not all, of the community outbreaks and isolate them. 

The report also didn’t account for what we’re increasingly seeing in COVID-A and -B countries: sealing themselves off from Second and Third Waves by quarantining or outright banning new international arrivals. The Imperial report seems to assume international air traffic continues to re-infect countries as soon as they get one wave under control. This is why we strongly advise international arrival quarantines now and Restrictions on all non-urgent international travel anywhere.

Locking down most of the country until testing capacity is increased to where it needs to be will prevent a lot of deaths in places like the Philippines. It won’t work, however, if new infections keep arriving. This is going to batter economies and we need to be extremely concerned about food security and the poor. Right now, however, it’s this draconian partial cure or just accepting the millions of deaths that come with letting 60-80% of the population get infected. 

10 Replies to “What the Imperial COVID-19 Report Means for Asia”

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.