C-19: Early ideas on how different countries could be managing the pandemic

Updated: May 11, 2020

COVID-19 has to date hit more than 150 countries all over the world. As it continues unabated in absence of a viable vaccine, many countries have frantically been battling the spread of the virus through crude measures like travel bans, social distancing, movement controls and population lock-downs. We will attempt here to provide brief analysis of what has transpired across different country experiences and offer commentary on what this could mean for planning their futures.

Visit https://www.agility.asia/covid for background information on analysis conducted by Agility to yield the data referred to here.

(May 1st 2020): We have updated our views on lock-down exit strategy here:


Outline of analysis: a deeper look at initial growth

The table below, developed with data available as at and up to 9th April, and based on the last 3 weeks of reviewing daily trend patterns of death statistics across the listed countries summarises the key statistics for each countries' COVID-19 growth patterns along three phases: (1) initial growth as virus first spreads unabated, (2) growth taper through lock-down/movement controls, and (3) eventual late stage latent growth as it reaches an equilibrium point between mitigation and continued spread. Naturally not all the data is yet conclusive at this point, and where appropriate, early estimates are highlighted in yellow.

When analysing their Phase 1 behaviour (initial growth), there are three broad groups;

  • Very high initial growth countries, where the surge in deaths have been rapid at ~2.2-2.8 days to double (time taken for deaths to double). These countries include Spain, Italy, Netherlands, Belgium and UK

  • Medium initial growth countries, where growth is reduced at ~3-4 days to double. Countries include Brazil, Switzerland, Sweden and Indonesia

  • Low initial growth countries, where growth is much more contained at 4.5+ days to double. Countries include India, Malaysia and Philippines

Figure: tabulated model fitting estimates for countries under study. Visit: https://www.agility.asia/covid for latest statistics

To further study the effects of just this first Phase three generalised infection (or death) growth models were set up with parameters as follows;

  1. Starting n=1 / Lock-down at day 20 / starting growth=2.2 days to double [e.g. Spain] / -1% growth decline at lockdown / 0.93 long term transmission factor

  2. Same as (1) but with starting growth=3.1 days to double [e.g. Switzerland]

  3. Same as (1) but with starting growth=5.7 days to double [e.g. Malaysia]

The three are shown in the next 3 charts with colours blue, red and orange. To remind the reader; the three differ only in their initial growth rates.

Discussion and key findings from analysis;

  1. The assumptions taken to generate the three charts are based loosely on actual empirical data generated through observation of the 20 countries under study. Most countries did in fact implement lock-down and/or partial lockdown within 15-30 days of their first COVID-19 deaths, and lock-downs have generally resulted in daily reduction in transmission factor within order of -(0.5-3%) per day.

  2. As can be observed, the effect of initial growth on the eventual trajectory is tremendous: the height of the peak in Model 3 is 1% that of Model 1, and 10% that of Model 2. For reference, Italy and Spain registered peak deaths around 1,000 per day (Model 1); while for Malaysia and Japan it was 8 and 14 respectively (Model 3).

  3. It is not clear why different countries have shown widely varying initial growth rates, and we believe in time this will investigated further by relevant parties. It is sensible to assume that as social interaction amongst the population is different from country to country, representing different social norms, so too would be transmissibility of the virus. It is also possible that although lock-downs had not started, authorities had been vigilant and already started tracking and tracing detected cases, leading to an inhibited growth rate even during the initial phase. It is also not certain at this point how different physiology or even the weather contribute to transmissibility of COVID-19.

Key implications and early thoughts on a way forward

  1. Exponential growth is often understated; likewise we also contend that comparing exponential growth across different growth rates is not something that is easily understood without an excel sheet. We have seen here how the effect of initial growth rate on COVID-19 growth significantly alters the pathway of infections and deaths, up to 2 or 3 order of magnitudes across the range observed in 20 countries under study; ranging from Spain through to China.

  2. Blunt measures such as lock-downs have now been deployed pervasively across most countries, in absence of any other means to abate the surge of COVID-19. However, while European countries have struggled to avert a large number of deaths, the South and Southeast Asian countries appear to have fared far better, with only <100 deaths amongst their populations, to date, much to the relief of billions of people across these countries. While the worst is not yet over, there is much to be hopeful about.

  3. These measures have however had a dire side effect; an almost totally dormant economy, streets devoid of people, business premises laying in wait, with little to no certainty on when things will go back to normal. Governments meanwhile scramble to introduce stimulus packages, focusing primarily on savings jobs and providing some semblance of a social safety net for those affected. These often follows a blanket decision for a 15, 30 or 60 day social lock-down that moves people off the streets and shutters businesses overnight.

  4. But what if there was a better playbook for how to manage ourselves through this ordeal? What if we could apply our knowledge and understanding of how fast COVID-19 spread before and after mitigation was introduced, and make better, more informed decisions on continued lock-downs and social controls? What if we could assess different scenarios of control, and weigh up pros and cons on infection of the population, healthcare resources, and society at large? These questions are not at all impossible to address given the data in hand.

  5. It is reasonable to assume that if a country locked-down early, and exhibited low intrinsic initial growth that lock-down measures can and should be removed.

Thought experiment: pulsing controls on and off

A plausible scenario of a way forward could be to sequentially pulse control measures on and off through time; i.e. allow for a period of opening up the economy after some time of locking-down, and doing so repeatedly until the point a vaccine or treatment is found. We set up the following experiment to explore this (see table below for summary);

  1. The model is based broadly around the case of Malaysia (a low initial growth country), where the assumed Phase 1-3 input assumptions were generated from the empirical study of real-world case and death statistics to present day

  2. Each wave is represented by; (i) an open period, where the virus grows exponentially at a certain transmission factor, and (ii) a locked down period, where taper effects kick in to bring the infected/death figures down. Up to 6 waves of switching controls/lock-downs on and off were created, lasting a total of some 50 weeks (i.e. up to one whole year)

  3. Transmission factor for the virus in each open period was assumed to decline (Phase 1 growth factor), representing a 'learning population' where better personal hygiene, social distancing practices would be better institutionalised after some time

  4. The taper decline due to lock-down (Phase 2 growth) was assumed to reduce in its efficacy in each subsequent wave. This represents a potential reduction in stringency by authorities in the controls they put in place, e.g. roadblocks, penalties and fines for non-adherence etc.

  5. Phase 3 growth is assumed to keep at a transmission factor of 0.9 throughout - this is the residual effect of the lock-down in absence of reopening the economy in the next wave

  6. The control scenario is a hypothetical case where the transmission factor can be reduced to a steady state of just 1.003 after the initial 2 weeks of infection spread. This represents the model whereby a light/moderate lock-down but with stringent control of the population persists indefinitely. This scenario is simply to compare the final results of the wave method

Modelled daily flow and cumulative deaths under the pulsing model:

Although this is admittedly a relatively simplified and rudimentary modelling of such a concept, it does establish that it is possible to implement such a strategy in order to open up the economy. The model suggests that it is possible to achieve a total open period of 35 weeks, while maintaining a closed/lock-down of only 12 weeks, including the initial outbreak phase; resulting in a total of an average of 10 deaths per day due to COVID-19.

In theory Model 3 (low initial growth) and Model 2 (medium growth) countries may well be able to leverage such a strategy. Unfortunately Model 1 countries would require a longer period of time to taper of average daily deaths and infections and thus require a far more elongated lock-down period particularly after the initial/first surge.

Visit https://www.agility.asia/covid for more reading.


Recent Posts

See All