A cautious case for reconsidering the Covid 19 response narrative

In the fight against the Covid 19 pandemic, most countries have gone down the lockdown route, with varying intensities of enforcement. The alternative path of herd immunity or even calibrated spread has fallen by the side, except in some exceptions like Brazil, Sweden, and Netherlands.

Given the overwhelming expert opinion consensus in favour of lockdowns, blanket ones at that, governments in developing countries cannot be faulted at all for having walked the course. Indeed, lockdowns have been the only politically acceptable choice before them.

But it needs to be borne in mind that right now, developing countries are unquestioningly applying a narrative generated from the spread of Covid 19 in developed countries. However, granted that lockdowns may be the only alternative available for Europe and US, is it the wisest option for developing countries like India?

Though I have been sympathetic to an alternative view, it has taken couple of conversations with people who know better to make me confident enough to bite the bullet on this.

From the perspective of developing countries, what do the facts inform? Three in particular deserve attention.

1. The number infected is completely unknown and will remain so for the foreseeable future. It will be all the more so for developing countries. Given resource and state capacity constraints, and practical problems, there are only so many you can test when it comes to these countries. At best, with testing, we can get higher numbers of those identified as positive, but cannot come anywhere close with the actual infected. The denominator in the infection rate calculation will therefore remain heavily suppressed and unknown.

But we do know the numerator, or the deaths, with reasonable certainty. Respiratory infection deaths are very painful and easily observed, and it is impossible for even small outbreaks to go undetected for long in any country.

This means that the graphs of the detected cases that feature in all the several Covid 19 trackers are conveying almost nothing. In fact, they are deeply misleading. Instead, a more relevant decision-support and informative graphic should be one that tracks the progression of deaths from the date of the first death. This generated from here by a friend is what Covid 19 trackers should be reporting.

In my limited searches, I have not been able to see any graphic that assesses the pandemic from this perspective. Actually, even this is deceptive since it does not discount for the higher natural mortality rates in developing compared to developed countries. So death cases are, as of now, perhaps the only reliable metric for assessment of the definitive medical cost of Covid 19. 

2. What do the absolute numbers tell us?  

One can scan the numerous Dashboards around and easily come to the conclusion that the vast majority of cases appear to be concentrated in a band of 35 to 50 degree North latitudes. Something like over 90% of the cases come from this band. Actually, someone could do an analysis of the Pandemic based on the latitudes, and likely find that the vast majority of cases are concentrated within an even smaller latitude band.

Take the most vulnerable Greater China area. The progress of the epidemic, especially in terms of the deaths reported, in countries like Taiwan, Vietnam, Philippines, Indonesia, Thailand, Malaysia etc which routinely receive large pool of Chinese visitors have been strikingly small relative to the  countries like Italy. It is most certain that many of these countries have received more visitors from Wuhan and other affected places in China during the same period, perhaps by orders of magnitude, than either Italy or Spain. Further, the conditions in terms of population density, medical facilities etc in these countries make them far more vulnerable to the spread of a disease than Italy or Spain. Much the same applies to the countries bordering Iran, especially those like Pakistan.  

On the other hand, in case of countries within the latitude band, unlike China and South East Asia, the spread of the disease across neighbouring countries has been consistent. The spread from Italy to Switzerland, which too has similar flow of people, is an illustrative example. In fact, even the small city-states like Luxembourg, San Marino, Andorra etc have not been spared. Even a cursory glance at the data here (say, deaths per million) reveal the point.

Even in Canada, the cases have been in the southern provinces bordering the US. Russia and Scandinavia too have a disproportionately smaller number of cases. 

Take Africa. Thanks to the Belt and Road Initiative, many African countries have large Chinese populations and also receive large numbers of Chinese business visitors. And in terms of physical conditions for the spread of the virus, these countries are among the most hospitable. Besides, all these countries even started taking the problem seriously since only the last few days, leaving enough unfettered runway for the spread of the virus. But we do not have even one example of an eruption comparable (not in absolute terms, but in proportionate terms) to that happening in any country within the band. 

In all the countries outside the band, the overwhelming majority of cases are those who have either themselves returned from the hotspots or are relatives or close contacts of the former. Also, the death cases are predominantly the old and with multiple medical conditions. Preliminary evidence on virulence is encouraging. In fact, this has been very pronounced in the band countries. Less than 4% of the Italian deaths have been below 60 years and the median age has been over 80, and 99.2% of cases had prior illness.

It's just that the data points are so overwhelming in pointing towards a very significant location bias for the virus' infectiousness and virulence. In fact, if you throw data on deaths from all the countries into a graph, it will most likely reveal a clearly diverging group of two trends.

3. If these are the absolute numbers, what about the trends, or the deltas? Unlike couple of weeks back, most countries, including India, are today at a steady state in the pipeline of infections and mortalities. But even with the low baseline flow of cases (there are most certainly far more symptomatic and asymptomatic cases), we are not seeing anything close to a proportionate rate of  consequent fatalities as elsewhere.

Further, even in the many countries outside the band where the first cases were detected as early as late January, the progression has been remarkably muted. Again, in contrast, in many European countries, despite the cases beginning to be detected only in late February, the progression has been exponential. Of course, all this despite the living conditions for spread being more favourable in the former than the latter. 

Furthermore, in countries like India, there has been no perceptible increases in patient load, much less those with respiratory problems, into hospitals. Nor has there been any anecdotal observations about significant rise in such cases. It is very reasonable to argue that, atleast for now, the novel coronavirus has been far less infectious in countries outside the band. 

All these above are stark facts (atleast for now). They stand alongside the models which use data from the countries within the band and form the basis for policy choices by developing countries.

This post will confine to the headline emerging data, and will not go deep into the scientific evidence or economic considerations. For reference, this, this, this, this, this, this, and this are scientific papers which point to a correlation with temperatures of 7-15 degree centigrade and latitude above 30 degrees North. They also point to coronaviruses as a family always exhibiting seasonality, mainly December-March, and falling after April as temperatures exceed 15 degree centigrade. In the mainstream media, the scientific case has been been covered by the likes of Eran Bendavid and Jay Bhattacharya and Vikram Patel. On the economic costs, please do read this analysis from Pensford Financial (HT: Ananth). This is even more relevant to the developing countries. 

Finally, what's the end-game to a lockdown?

Developed countries will always have end-games. Given the pace at which the transmission has been happening, the virus too will play itself out in the not so distant future. Also, they would have tested enough and identified existing and potential cases, and would also be in a position to isolate them and get on with life, and also do 'whatever it takes' to restore economic normalcy. But this strategy is impossible for developing countries to even think of executing. Not to speak of bearing the exorbitant fiscal costs of relief and recovery.

If we are driven by the risk of uncontrolled community transmission, it is hard to believe that in three weeks or even couple of months countries will be in a position to be able to make a definitive technical choice on this issue. But by then the economic costs and human suffering from a lockdown will start to far outstrip the medical costs. Discontent would have erupted. People will start to feel that it may be better to die of Covid 19 than starvation. That will perhaps finally force the delayed political choice.

It is easy for experts to suggest these measures, which by the way is a staple feed for every bureaucrat worth their salt, but to implement them effectively for several weeks at a national scale is too daunting a challenge in most contexts.

The most likely scenario is that 3-4 weeks into the lockdown, governments in developing countries will find that the trends on infection and death are far smaller than those in the likes of Italy and US. This will be used to phase out the lockdown. But, as the aforementioned analysis shows, this assessment could be made with more or less the same level of confidence/objectivity right now or over the coming week itself. 

This brings to the point of this post. At the outset, the objective is not to dismiss scientific concerns and argue against the lockdown. It is only to highlight that there is atleast as much an objective rationale to argue in favour of alternatives. And it should be a consideration going forward.

The poverty of thought among the so-called experts and opinion makers has been laid bare in this crisis. This is not to blame the scientists and epidemiologists. They are supposed to be scientific and only provide the inputs for decision-makers (like the now famous Dr Anthony Fauci). But even they should take some of the blame for presenting narrow models based on very skewed data as universally applicable models of infection and mortality. But their interpreters (opinion makers and thought leaders) cannot be given any benefit of doubt for mindlessly disseminating information from these models.

Note that this criticism is relevant only for modellers and experts focused on developing countries. The models may hold with a fair degree of accuracy for the countries within the latitude band.

The entire narrative has been constructed on the course that the pandemic has taken in North East Asia, Europe, and US. In the absence of data from elsewhere, the models naturally extrapolate from the trends from these areas. Also this debate in developing countries is being framed narrowly in terms of medical costs, and that too from models with evidence drawn from an arguably different context.

Never mind, even within the band, the example of the Imperial College study illustrates the folly of relying on such models with several limitations to in critical national policy decisions. In fact, today there are so many models that one can find atleast a model to fit any kind of priors you have about the infectiousness of Covid 19. Even mathematicians and physicists have their share of models. Therefore, how is this line of limited models-based reasoning any more objective or scientific than that outlined above making the case for exploring alternative options?

Unfortunately domestic opinion makers and their amplifiers in the media have been co-opted into blindly following the prescriptions of experts and their counterparts from the developed countries. It needs to be borne in mind that for countries in the latitude band, who originate most of the research, the lockdown choice is less under dispute. As with any such public issue nowadays, television talking heads, WhatsApp messages and social media posts have exacerbated the problem. Political leaders have been frightened into submission with dire predictions. The space for weighing the facts and exercising judgement, how decisions of all kinds get made, has almost disappeared. 

If the trends outlined persist by the end of this week, developing countries should consider a recalibration of their response. But, in this environment of fear and universal scaremongering,  amplified by expert opinion, coupled of course with the nightmare scenario of uncontrolled community transmission, governments will be loath to change course on their own. Opinion makers and thought leaders will have to contribute to triggering a debate which engages with these emergent facts and associated questions, thereby creating the conditions for governments to consider making alternative choices. This will help politicians take a call on how much evidence is enough to pivot away from the lockdown approach currently being followed.

To conclude. For now, developing countries have decided on a course, and let's focus on it. But let's also keep track of the trends and reasoning as discussed above. Let opinion makers and thought leaders engage on the issue to bring a greater balance into the mainstream debates on responding to the pandemic so that governments are in a position to take a more objective view and not be dictated by the narratives from developed countries. Given the stakes involved, it is important for governments across developing countries to be ready to revise their priors based on emerging trends from their own contexts.

Update 1 (31.03.2020)

Highlighting the unreliability of the infected numbers, 

John Ioannidis, a professor of epidemiology at Stanford University, has branded the data we have about the epidemic “utterly unreliable”. “We don’t know if we are failing to capture infections by a factor of three or 300,” he wrote last week. If thousands more people are surviving than we know about, then current mortality rate estimates are too high — perhaps by a large margin.

Similar problems exist with death numbers too,

In the UK, about 150,000 people die every year between January and March. To date, the vast majority of those who have died from Covid-19 in Britain have been aged 70 or older or had serious pre-existing health conditions. What is not clear is how many of those deaths would have occurred anyway if the patients had not contracted Covid-19... Professor Neil Ferguson, director of the MRC Centre for Global Infectious Disease Analysis at Imperial College London, said it was not yet clear how many “excess deaths” caused by coronavirus there would be in the UK. However, he said the proportion of Covid-19 victims who would have died anyway could be “as many as half or two-thirds”.

Sweden and Brazil are two good outlier examples among countries outside the latitude band who have been largely doing business as usual. However, compared to the countries in the band, despite their business as usual approach, their death numbers have been very small. Although when compared to their neighbours Norway, Finland etc and Argentina, Mexico etc respectively, who all have lockdowns or some form of stricter enforcement, they have higher deaths (even if not significantly high). This also means that some form of social distancing (not lockdown) is perhaps appropriate.

Update 2 (15.04.2020)

Ananth points to this study by ETH Zurich. The graphic has something to tell.

These are the updates from ETH. This is the version for April 15, 2020.

Update 3 (15.04.2020)

FT has this long read on epidemiological models,

There was a similar dispute after the 2009 swine flu outbreak when advice based on Imperial’s model was made public by ministers. This described a “reasonable worst-case scenario” in which there could be 65,000 deaths. In practise, there were only 457.

Update 4 (18.04.2020)

It was all along known that the prevalence of infected cases was by orders of magnitude higher and alarmist predictions of infection and death rates were therefore badly off the mark. Highlighting the point comes new evidence from seroprevalence data from Santa Clara County in the US,

We measured the seroprevalence of antibodies to SARS-CoV-2 in Santa Clara County. Methods On 4/3-4/4, 2020, we tested county residents for antibodies to SARS-CoV-2 using a lateral flow immunoassay. Participants were recruited using Facebook ads targeting a representative sample of the county by demographic and geographic characteristics. We report the prevalence of antibodies to SARS-CoV-2 in a sample of 3,330 people, adjusting for zip code, sex, and race/ethnicity. We also adjust for test performance characteristics using 3 different estimates: (i) the test manufacturer's data, (ii) a sample of 37 positive and 30 negative controls tested at Stanford, and (iii) a combination of both. Results The unadjusted prevalence of antibodies to SARS-CoV-2 in Santa Clara County was 1.5% (exact binomial 95CI 1.11-1.97%), and the population-weighted prevalence was 2.81% (95CI 2.24-3.37%). Under the three scenarios for test performance characteristics, the population prevalence of COVID-19 in Santa Clara ranged from 2.49% (95CI 1.80-3.17%) to 4.16% (2.58-5.70%). These prevalence estimates represent a range between 48,000 and 81,000 people infected in Santa Clara County by early April, 50-85-fold more than the number of confirmed cases. The population prevalence of SARS-CoV-2 antibodies in Santa Clara County implies that the infection is much more widespread than indicated by the number of confirmed cases. Population prevalence estimates can now be used to calibrate epidemic and mortality projections.

And this

Based on this seroprevalence data, the authors estimate that in Santa Clara County the true infection fatality rate is somewhere in the range of 0.12% to 0.2%—far closer to seasonal influenza than to the original, case-based estimates.

Update 5 (23.04.2020)

Very good interview of Johan Giesecke, a globally renowned Swedish epidemiologist who is engaged with the Swedish strategy. The summary

• UK policy on lockdown and other European countries are not evidence-based

• The correct policy is to protect the old and the frail only

• This will eventually lead to herd immunity as a “by-product”

• The initial UK response, before the “180 degree U-turn”, was better

• The Imperial College paper was “not very good” and he has never seen an unpublished paper have so much policy impact

• The paper was very much too pessimistic

• Any such models are a dubious basis for public policy anyway

• The flattening of the curve is due to the most vulnerable dying first as much as the lockdown

• The results will eventually be similar for all countries

• Covid-19 is a “mild disease” and similar to the flu, and it was the novelty of the disease that scared people.

• The actual fatality rate of Covid-19 is the region of 0.1%

• At least 50% of the population of both the UK and Sweden will be shown to have already had the disease when mass antibody testing becomes available

Update 6 (06.05.2020)

Neil Ferguson is a remarkable man. He has successfully scare-mongered four times in the last twenty years - mad-cow disease in 2001 (50-150000 human deaths due to BSE predicted, millions of cattle culled, actual human deaths less than 200), bird-flu in 2005 (upto 200 million deaths, a few hundred actually died), swine-flu in 2009 (65,000 deaths in UK, actual was 457), and now Covid 19 (250,000 deaths in UK predicted). His respectability and credibility survived all of them. Amazing!

The 2017-18 flu-season in the US killed over 80,000 people.

Sample this very good cautionary note on the use of models,

When governments design policy based on epidemiological forecasts, their choice of the model to use could be the difference between a mild mitigation strategy and a large proactive intervention, such as the mass slaughter of livestock in the case of Mad Cow Disease or aggressive and wide-scale societal lockdowns in the case of COVID-19. That choice, often made amid severe data limitations, is often presented to the public as an unfortunate but necessary action to forestall an apocalyptic scenario from playing out. But we must also consider the unseen harms incurred when politicians base decisions on a modeled scenario that is not only unlikely but also wildly alarmist and likely exaggerated by the dual temptations of media attention and gaining the ear of politicians. Given the high uncertainties revealed by statistical scrutiny of epidemiological models including among other medical experts, the presumption should go the other way instead. What is warranted is not bold political action in response to speculative models generated with little transparency and dubious suppositions, but rather extreme caution when relying on the very same models to determine policy.

Ananth has a great summary here.

Update 7 (08.05.2020)

Interview of Hendrik Streeck of University Bonn, a virologist who conducted a study of a representative sample population within Germany, 

The headline result is that 15% of that population was infected, which implies an Infection Fatality Rate of 0.36%. This would put him somewhat in the middle of the previous experts we have spoken to. Professor Streeck was keen to point out, however, that he still believes this is a conservative estimate, and thinks it may be closer to 0.24-0.26% and may come down further still as we know more. He published the higher number to err on the side of caution: “it is more important to have the most conservative estimate and see the virus as more dangerous than it is,” he said.