With lockdown measures starting to be moderated, the first phase of the pandemic outbreak: initial wave of infections (and deaths), has now passed. As crude as it may sound, the death tolls tallied up indicates clearly which countries were winners and which were losers.
We have now entered the second phase: reopening economies and navigating the path to normalcy. To date, many hard hit European countries have reopened while keeping transmission contained at non-epidemic levels. That is, transmission <1.0 despite mobility levels rising to some 80% of pre-outbreak baseline levels.
Some other countries have not been so fortunate. From our consideration set, we expect that countries like Mexico, Bangladesh, Egypt, Brazil, India, Indonesia and Philippines will continue to face a very tough time reopening. Even today, transmission fleets back and forth across the epidemic threshold; with mobility levels at just 40-60%, i.e. far from baseline level. It will be a painful journey and prolonged crisis for these countries unless practices are further changed accordingly from where they currently stand today.
1. Defining a target objective for the second phase: daily life as close to normalcy as possible
The following is crucial to realise: in this second wave the responsibility to 'manage' the spread of the virus is effectively transferred from the government through various control procedures (e.g. lockdowns, movement control and shelter in place orders); and into the hands of general populous, through better practices (e.g. personal hygiene, social distancing, and other precautionary measures taken at the individual level).
A target objective within, or rather, for the end of this phase is to reach mobility levels of 100% across all major mobilities within as (reasonably) short a time as possible. That is to bring back previous levels of mobility in, across retail, workplace, transit stations, walking and driving activities, while maintaining transmission sufficiently and strictly below epidemic levels (i.e. 0.9-0.95 at maximum) in order to breathe much needed life back into already ailing economies.
Our analysis indicates that this is already very attainable within a period of the next 1-2 months (e.g. Switzerland, Austria, Netherlands, Germany, France), or up to 3-4 months (Spain, Belgium, Italy) - crucially, without invoking epidemic transmission rates. Refer to our daily reports for supporting data on this.
For the record, this in reality is still far from 'real' normalcy. 100% mobility does mean that at an aggregate level, physical movement would be back to pre-outbreak baseline levels. However it does not mean life would be the same as before, as it would be, by definition, defined by a 'new normal'. Each person will need to maintain caution in their daily interactions, and accordingly apply a changed behaviour in daily practices to avoid contracting or further spreading the virus. Three months ago this may have needed further explaining, but today we believe most of us know what this means.
2. A balancing act critical to succeed in the second phase: maintaining non-epidemic transmission rates, against rising mobility and physical interactions
Critical to delivering on this objective is to maintain transmission below epidemic levels, or <1.0. The transmission (factor or rate) in loose terms represents the ratio of new infections or deaths in any given day, compared to that of the previous day. For reference, a transmission of 0.95 would imply a period of 13.5 days to halve an initial number of infections or deaths [ -log(2)/log(0.95) ]. Likewise, a transmission of 1.05 would require about 14 days to double the figure.
A key point to highlight (as it may be initially unintuitive): the goal posts have categorically changed for this second phase. While the initial phase was about containment of deaths and infections, this second phase is about ensuring non-epidemic transmission. And importantly, the resultant toll the virus has taken during Phase 1 (e.g. cases, recoveries and deaths) has little to no bearing on how this current phase will develop. With that a pivotal rule needs to change or be bucked: we want to reverse the physical restriction of the previous lockdown phase, but not invite the increased spread of infection that would naturally come with it.
This challenge is not a consistent one from country to country; some will just struggle more. Given at the very least, differing intrinsic social norms and practices developed thus far. See below for an illustration contrasting two countries - Netherlands and Brazil - to further illustrate this point.
Figures: contrasting Netherlands and Brazil. Both countries locked down decisively around 2nd week March); however Brazil never reached transmission <1.0 in spite of its efforts to bring mobility down to ~35-40% of baseline levels. Today it risks going further above this threshold and driving additional outbreak waves.
3. A worthwhile discussion on the effect of mobility on COVID-19 transmission, and quantifying the effects of non-physical distancing interventions. What works, and what may not.
To date, it is still surprisingly uncommon to see the use of mobility data as means to help forecast COVID-19 case detection and deaths (for example, often cited popular forecasts from IHME - US, Imperial College - UK). Notwithstanding some limitations of the approach, for example, unweighted proportionality across mobilities, and unquantified overlap/cross-correlation between them, we have been using this approach for some time and have found it to yield robust and credible results - see figures (A) below for a few examples.
We will not go into an elaborate discussion of modelling methods, specific assumptions and potential caveats; suffices to say there is visible and strong correlation between mobility and transmission, which can yield credible forecast projections of COVID-19 spread. Ultimately, this is also further guided by a known and certain eventual trajectory of mobilities for all countries: mobility levels back to 100% and with that, daily life returning back to normal. The only questions remaining are what path and timeline each country takes to get there.
A rhetorical question that the reader may ask: if increased mobility leads to more transmission, how do we avoid merely repeating the past - i.e. returning to higher mobility levels and higher transmission of the virus? Enter non-distancing interventions, which can be studied indirectly through measuring the effects of changes in mobility on measured transmission rates since lockdown. Specifically, we model this as a hysteresis to the transmission-mobility relationship, where upwards movement of mobility has a lower correlation to transmission compared to that of downward movement. This reflects how much more cautious and conscientious behaviours and daily practices are amongst the general population, leading to lower transmission potential.
Figures (A): predictability of COVID-19 transmission by utilising 'crude' mobility data, by using examples of Turkey and Egypt. No other type of modelling yields as accurate a modelling of daily death statistics as this method, given discontinuous nature of transmission changes.
In Figure (B) below, we analyse the net change in transmission factor since the bottom most mobility point during start of lockdown to present day, and attribute that change to a net change in mobility; lets call this the 'net'-transmission-to-mobility factor. Take for example, Egypt at +0.034, this means a +10% net change in mobility over the period leads to a +0.034 increase in transmission. For Spain at -0.011, a +10% net increase in mobility leads to a reduction in transmission by -0.011. It is important to understand that this does is not a direct and linear correlation, rather it is an accrued impact over a period of cycling mobility upwards and downwards over the period.
Circling back to non-distancing interventions. In our set of three countries (in figures part B), it seems Spain has done well here, followed by Switzerland, and finally Egypt. Computed hysteresis obtained through fitting actual data with the model suggests hysteresis of 40%, 29% and 1% respectively for the three countries, which indicates the relative degree to which upwards mobility changes affect transmission compared to downwards changes. To further clarify, at a transmission-mobility correlation of 0.5, and a hysteresis of 40%: transmission would reduce by -0.05 for every -10% decrease in mobility, while upwards increase in mobility by +10% would result in a transmission increase of +0.02 (or 40%).
Spain stayed in lockdown for quite some time; for 5-6 weeks since mid March where it held the population at very low levels of mobility of just ~15-20% vs. baseline level. Mobility increased noticeably only towards the final week of April, and today has already reached ~60-65% levels. The country has for a while advised the wearing of masks, and on 21 May mandated them as obligatory by law in public spaces. In addition, it still has controlled/locked down areas across a number of major cities, and its parliament approved a fifth extension to its state of emergency at the end of May despite strong opposition from at least two non-ruling parties. Read more about this here and here.
Egypt, with an equivalent case detection delay to that of Spain at 11 days, on the other hand only reached the lowest level of mobility of 50% vs. baseline by the end of March. Mobility has immediately shown a steady increase since then, peaking at similar levels to that of Spain, at 60%, just before the Eid break on 24-25 May. Mask wearing is prevalent, although not obligatory. It has a system of night time curfews in place, and tourist hotels are allowed to operate under strict SOPs.
On paper alone it is hard to tell the practices of these two countries apart; notably if, how and why they have led to different outcomes. The results are starkly different - Spain is on a clear road to recovery, whereas Egypt will struggle for some time.
The Oxford Government response stringency tracker measures quantitively government policy and control responses in combating COVID-19, such as school closures, workplace closures, border controls, public transport use across a wide range of countries. It assigns a stringency score of 84 to Egypt, higher than that of Spain at 81; the higher the score the more stringent the policies. Egypt has in fact totally closed almost all public transportation, while Spain has only reduced theirs. Both restrict workplace access only to a number of essential services and sectors, and have still shuttered all schools until present day.
Figures (B): Example results from analysis of Egypt [+0.034], Spain [-0.011] and Switzerland [+0.0027]. [x] indicates the net change in transmission with 10% net change in mobility since lockdowns to present day (late May). For reference;
Egypt: transmission-mobility % correl. 0.4 and hysteresis 1%
Spain: transmission-mobility % correl. 0.25 and hysteresis 40%
Switzerland: transmission-mobility % correl. 0.38 and hysteresis 29%
4. Countries in trouble: expect a prolonged crisis, and a long-drawn continued struggle; likely resulting in an even deeper economic crisis and strong uncertainty on economic outlook
In the introductory section of this article we listed a number of countries where the struggle to reach 100% mobility will be challenging, and that may be putting things lightly. These were Mexico, Bangladesh, Egypt, Brazil, India, Indonesia and Philippines. A few relevant country report excerpts are included below for reference; the reader is encouraged to review our daily reports for more detailed and up-to-date information.
As illustrated by the earlier example of Egypt, these countries have struggled to reach and maintain transmission to below epidemic levels. And as lethargy from lockdown sets in and economies inevitably are forced to reopen, transmission is expected to rise again. Future options are limited to more lockdowns, or continued limiting of physical movement to current or even lower levels. Either scenario will destroy any hope for burgeoning economic activity in the near terms - a bleak prospect after having endured a long hard campaign of some 2 to 3 months of restricted livelihoods for the vast majority of urban populations.
For reference the Oxford University stringency index for these countries as at today are Mexico 82, Bangladesh 94, Egypt 84, Brazil 81, India 79, Indonesia 58 and Philippines 72. For comparison, the index for 'better-performing' countries are; Thailand 69, Malaysia 75, United Kingdom 67. Also, at present, almost all countries still maintain full restriction on international border travel, as well as on social and private gatherings. In short, governments are already doing as much as is reasonably expected of them, through a range of policy and control measures. And to make matters worse, sustaining those measures over a longer period of time may not be possible, from an economic standpoint or a social and psychological one.
We will not elaborate too much on what can be done here in light of this fact. However, it seems obvious that it is now even more incumbent on the population of these countries themselves to institutionalise the right behaviours and daily practices, as they continue the good fight against the virus.
Perhaps the governments role to improve awareness and education should be highlighted even more, changing slogans from 'staying home' to more practical how to guides and advice on what each person needs to do differently. They could step in where resources are genuinely lacking, for example provision of clean water, better sanitation, and supply of masks and other basic protective equipment. These challenges are likely unique to each country and interventions may likely vary from one to another depending on circumstance.
Lastly, there should really be a concurrent improvement to testing regimes, given that case detection delays for countries on this list range from 9 days to 22 days. The speed of case detection needs to be sped up if the countries are to stand a fighting chance, and with that, higher volumes of testing rolled out, and more aggressive isolation and tracing activities. We should warn though, that there is no visibly strong evidence to suggest that faster detection alone would lead to a definitively lower transmission of the virus. Presumably, smarter and more targeted testing is really what is needed as the holy trinity of tracking, isolation and tracing would, in reality, need to precede the wave of the spread instead of constantly chasing behind it.
Figures: excerpt reports for Indonesia, Bangladesh and India. In addition to Egypt and several other countries, these countries exhibit one common attribute: a struggle to reach and maintain transmission below epidemic levels while allowing mobility to rise beyond 60-70% vs. baseline levels.