As the month of May greets us, many countries have already begun early actions to reopen their economies post stringent lock-downs endured during March and April. Once again, we see a general lack of detail in articulated plans for this phase (at least from what is made available to the general public); reminiscent perhaps of how lock-down plans were rolled out in the first place - rushed and lacking in detailed coordination. And once again, in spite of different circumstances from country to country, there is oddly and coincidentally, a common timeline across countries; first to reopen was Italy, Spain, Austria, then US, then India, Malaysia, and so on. The domino effect has started, and it is certain many more countries will follow suit over the coming weeks. Meanwhile, we feel the right approach is to deeply deliberate the critical factors that distinguishes the epidemiological characteristics of one country from the next; and from there, evaluate and determine what course to take. These factors are namely: (A) initial growth/transmission rate, (B) current case detection/testing delays, (C) asymptomatic prevalence, (D) relationship between mobility and transmission. Each of these vary by country, and their combination has a bearing on viable options each can take.
Let's first explore these four factors a bit deeper;
(A) Initial growth rate
Our study across countries indicate an initial transmission factor/rate between 1.14-1.4, or equivalently, an initial doubling time of 2.2 days through to 5.7 days. Hard hit European countries as well as the US have been on the lower end of this scale, whereas Southeast Asian countries on the opposite end. The shorter the doubling time, the more active the virus is at spreading in this uncontrolled initial phase of growth spread. And as we have seen in previous articles this has tremendous impact on the surge in infections and peak of infections figures as the country enters into lockdowns. As yet we do not have a view of what makes this factor vary so much from one country to the next, neither are we fully certain if and how these factors would change in future situations of outbreaks.
(B) Case detection / testing delays
To recap, case detection delay here refers to the mean time between real infection and case detection. Delays vary widely, from 0 days to 20+ days depending on country. Surprisingly, a number of European countries such as Sweden, Germany, Belgium and Austria are doing relatively rather well on this front (0 to 6 days), compared to others such as Italy, UK and Spain (9 to 11 days). Meanwhile, Asian countries such as India, Philippines, Indonesia and Malaysia are doing far worse at 10 to 16 days. It should be noted, with median time between infection and death of ~14 days, some countries are in fact detecting cases beyond the death timeframe. Detection delays are so crucial when viewed against the backdrop of exponential growth rates; for example a 10 day test delay and 2.5 days to double growth rate would mean the infected figure would have increased from 1 to 2^3 = 8 by the time the test results are tallied.
(C) Asymptomatic transmission
This factor is a bit if a wild card, or in other words, a highly uncertain unknown at this point in time. Narrow studies have indicated widely varying estimates from 20% to 88% (of total infected persons), but the scientific community mostly already agree that pre-symptomatic, mildly symptomatic and asymptomatic prevalence is very evident in COVID-19 spread; in part due to long incubation of the virus with significant shedding even during these periods, but also other factors that have not been fully understood. Regardless, a view on asymptomatic prevalence needs to be staked as these silent carriers would be a very significant vector within an uncontrolled environment where the population is free to roam. Our own (less scientific) estimates place this range between 33% and 92% based on computation from a range of countries. We describe the method of computation for estimating asymptomatic prevalence here. We should warn that these figures are unlikely to be the most accurate estimates of the asymptomatic prevalence, but suffices to give the reader a directional sense from country to country. Note: a 33% asymptomatic prevalence means that one in three infected people will never show any symptoms during their infectious phase, and will therefore infect others who are in close contact.
(D) Relationship between mobility and transmission
Our study of mobility changes during lockdowns gives us an idea of how effective this tactic can be from country to country, based on what has transpired over the past 2 months. Knowing this will be important as lockdowns in essence form an emergency cord that countries can pull if all else fails, that is on-going efforts on improving personal hygiene, and test, track and trace activities do not prevail. By knowing the strength of lock downs, for example, how low mobility changes can decline, how quickly, and what effect this has on virus transmission will be crucial to design a future set of control interventions. For example, an on and off lockdown strategy over a longer period of time - which inevitably seems to be where most the world will head to over the coming 12-18 months. We see a general range for the relationship between transmission factor and mobility (in %), i.e. the correlation gradient, of between 0.4 to 0.65. So for example, for an 80% reduction in mobility and a correlation gradient of 0.5, the transmission factor reduces by 0.4 from baseline growth. For the case of Italy with initial transmission 1.355, this reduces to 0.955 i.e. outside of epidemic levels (<1.0).
Figure: How the four factors can be visualised to drive infections and how they are measured via case detection statistics;
The table below summarises the four factors for a range of countries. Note this data is not without gaps and simplifications by way of averages, but should suffice to give the reader a cursory view of the different factors and how they vary across countries.
Let's start with some observations across a select number of countries.
First up, Germany. While its initial growth rate was high at 2.8 days to double, a short test delay of only 2 days enabled it to keep on top of case growth; its lockdown was relatively effective too at 0.63 correlation gradient to mobility, and as a result its death toll is only one quarter that of Italy and the UK.
The UK and US on the other hand, were likely overwhelmed by surges in cases and deaths; this and visible constraints in testing capacity led to testing delays of 11 days and 9 days respectively. Lockdowns were weaker to effect change in transmission compared to Germany, and both countries were thus far less able to overcome the spread. Asymptomatic prevalence estimated at 80-90% meant however that lockdowns were really the only choice they had. However, this may also have contributed to the fact that infections and deaths have continued to grow even under lockdown as residual mobility is still sizeable at ~30% from baseline for the UK and ~45% for the US.
Indonesia and Philippines have lengthy estimated test delays of 16 and 14 days respectively. While initial growth is far more muted compared to the UK and US, at 5 and 3.4 days to double respectively, their asymptomatic prevalence is high at >80%. Their only choice to avert a surge in infections and deaths was to lock-down, which they did mid-end March. Deaths have been contained to less than 1,000 respectively, however a burning question is how they will be able to restore normalcy without first improving testing. Neither country is in the clear, even under lockdown, with transmission rates still hovering around the 1.0 mark, evident from their daily infection and death statistics.
For both South Korea and Singapore, we have estimated testing delay at ~0 days, i.e. they are able to detect cases less than a day from actual infection. Coincidentally, both countries have managed to keep their economies outside of lockdown for far longer than others; in fact South Korea has yet to implement a lockdown, whereas Singapore, due to a surge in cases amongst foreign workers has put in place a form stay at home order as of the last month or so. Mobility levels in South Korea are some 70% from baseline levels, while in Singapore ~40%. Likewise, Sweden, even with a potential 95% of asymptomatic prevalence, has kept mobility levels at 65-70% and the spread relatively contained - deaths are still less than half that of Germany.
Compare Thailand and Malaysia. Each started with relatively low initial growth at 5-7 days to double. However Thailand's testing delay is 5 days compared to Malaysia's 10 days. While both countries' lockdowns have quickly stemmed further deaths and infections, mobility in Thailand has been maintained at 40-50% compared to Malaysia at ~25% levels. For reference, the two countries have comparable asymptomatic prevalence at 33-38%.
Austria, with testing delay of 2 days and initial growth of 2.85 days to double, is a peculiar case study. See below an excerpt from our daily projection report on the country, followed by a rolling cross-correlation analysis of case detection with mobility changes. Notice in the top right hand line-chart how the case detection curve does not follow the modelled logistic curve fit (derived from death statistics), unlike other countries we track. Assuming reported death figures are not in error, it would suggest that Austria had significantly augmented its test regime within the last week of March resulting in much faster turnaround in testing. The cross-correlation chart confirms this, whereas testing was around the 6 day mark prior to April, after that case detection registers between 0 to 3 days [notwithstanding errors from the 'dirty' fit of this approach]. Apart from stringent lock-downs and general advice to avoid hospitals, Austria had rolled out a mobile testing unit that visit individuals calling in to a hotline with symptoms within their own homes - it seems this has proven a worthy exercise.
To round things off, we look at Egypt and India with test delay estimates of 12 days for both. In spite of lockdowns, driving mobility to around 50% and 25% levels respectively; both countries show infections and deaths continue to creep up albeit at a lower growth rate compared to their initial growth phase. Both countries by the way enjoyed a low initial growth rate of 7 and 5 days to double respectively, however still struggle to contain continued infections.
So what can we summarise from these observations? Here are some simplified takeaways and a brief discussion;
Lock-downs are really the only choice if initial growth is high, suspected asymptomatic prevalence is high, and testing delays are significant. Examples are UK, Italy, Spain, France, the US and many others
Reducing testing delays and speeding up detection is an extremely effective tool to combat the virus spread. South Korea, Singapore, Germany have all fared relatively well in spite of their different starting positions, simply by this measure alone
Even with high asymptomatic prevalence, e.g. Germany (76%) and Sweden (95%), boosting test speeds can do wonders to overcome the need for stringent lockdowns - testing delays of 2 and 0 days respectively. Sweden has yet to lock down to this day, and maintains high levels of mobility. While infections and deaths have outpaced many Asian countries, they are still within reasonable levels and show a visible taper (i.e. reaching transmission rate <1). Likewise, South Korea is similar to Sweden in this regard, with limited controls and no lockdown it has managed to keep the spread at bay with only some 250 deaths to date.
Asymptomatic carriers will be a big problem for certain countries where prevalence is high and infection growth is rapid. In these instances, lockdowns may be the only choice. Rapid testing may overcome this challenge somewhat (as Sweden has done), however it is more likely that a change in test regimes is required. Whereas a lot of test regimes are focused on symptom based testing, asymptomatic prevalence may require conducting mass scale testing across the population. In this case it would no longer be about speed of testing from infection, rather the sheer volume of tests administered to capture the infected in the pre-symptomatic and asymptomatic state.
Lockdowns are a blunt tool, but an assured guarantee to effect change rapidly in absence of any other choice. In our view, they should remain as back-pocket aces that countries can pull as and when needed, but not as the primary strategy to be used for an indefinite period - as the projected economic impact would be unimaginable.
Beyond lockdowns and testing, there are likely other factors such as personal hygiene, and other precautionary measures taken by the population that contribute to a reduction in transmission. Thailand, Malaysia, Japan, Indonesia namely all exhibited initial growth rate greater than 5 days to double, compared with a number of European countries at <3 days; something we would attribute to earlier awareness of the on-going outbreak that started in Wuhan, as we have explained here. Notwithstanding biological differences in how the virus effects different people, it is likely that these countries had been better prepared to face the virus spread.
Exploring potential scenarios
We explore some hypothetical scenarios; hopefully to provoke some thought on what could be the right or wrong way to approach the question of relenting on lockdowns and reopening economies.
The absolute worse thing to do. For high initial growth countries with sizeable proportion of asymptomatic carriers (say 70% or more), to reopen their economies when testing delays still are still at 10 or more days. Slow testing, and the risk of being blindsided by asymptomatic infected persons will inevitably result in high exponential growth from the reopening point, and the only choice a month or two at most down the road would be another lockdown. And at that point, a great number of deaths would again likely be incurred, possible a few tens of thousands. Countries like the US, UK and Spain risk this happening if they are not careful.
The lower middling path. For low to medium initial growth countries with a sizeable proportion of asymptomatic carriers (50-80%), to reopen their economies while testing delays are still at 10 days or more, and there is no mass testing regime in place (testing everyone, not just symptomatic persons). Either one of late testing, or a high prevalence of asymptomatic carriers would continue growing the numbers of the infected, resulting in no other choice than hard lockdowns down the road. At which point a few thousand people would have died. Turkey, Brazil and Philippines possibly risk this.
The other (slightly better) middle path. For low to medium initial growth countries with a more reasonable proportion of asymptomatic carriers (<50%), to reopen their economies while testing delays are still at 10 days or more. Mass testing may not be the most pertinent agenda, however, speed of testing needs to be greatly improved. While the number of deaths would naturally be contained by virtue of supportive, muted growth levels - possibly a few hundred at worst; to avert a higher number of deaths it is highly likely further lockdowns will be necessary in the future. Much of the gains from the earlier lockdown would be lost and the timer reset back to the start. Malaysia, Turkey and Bangladesh are here.
The right thing to do. Regardless of starting point; the uncontrolled growth rate should be reduced as much as possible by improving personal protection & hygiene practices, and inculcating, no, institutionalising, good practices around social distancing. Maintain rules on gatherings, international travel and transit hubs, or wherever else large crowds may congregate. Do this while lockdowns are active, and allow controlled mobility so, importantly, habit forming and role modelling can happen. Further and even more crucially, use the time to bring down testing delays to low single digit levels, say 2-5 days maximum. Try everything from app based contact tracing, mobile roving test units, walk-in self testing, subsidised testing costs, cluster/pooled testing, sewage testing - the world is now replete with good examples to pull from and replicate. Only once all this is in place can the economy be reopened with confidence.
Resetting is okay if all else fails, but doing great means we should make sure to reset to a better starting point. Most may think that reopening the economy following a point of lockdown resets the timeline back to before lockdowns start. However, that is not how viral epidemics work. Exponential growth continues from the last point of infections prevalent in the population, and exponential decline from lockdowns can easily take 3-6 months to bring ambient infection levels back to pre-lockdown levels (lockdown transmissions have ranged between 0.95-0.99) - and no lockdowns have lasted that long. Reopening the economy restarts the exponential growth from that new starting point, i.e. a higher point than before. It is therefore critical to be extremely clear what is different going into a reopening phase; if very little is, then it risks just repeating the cycle all over again.
A crucial balancing act for governments and populations to get right. In closing; we should acknowledge the fact that, underlying the motivation to hastily reopen economies are valid concerns around the deeply negative economic impact that the virus has caused thus far. This is indeed a careful balancing act that needs to be struck well, and a strategy that simply follows the next country's may not be up to the task. We do believe a scientifically driven approach can be taken, by weighing up the different factors at play, each unique to the circumstance of any one country. A robust plan that then invites clear commitments from all stakeholders across the value chain can then be developed, syndicated and stress tested. Only then is there better hope to overcome the pandemic.