Mumbai: While indicators such as case fatality and doubling time are being analysed to understand the spread of the COVID-19 pandemic, the number of deaths per million people is a good indicator of both containment of contagion and of case management, says K. Srinath Reddy, president, Public Health Foundation of India. A slow increase in this number, combined with a fall in new deaths per day, indicates good control, he adds.
India’s case mortality and deaths per 100,000 population are among the lowest in the world, the Union Ministry of Health and Family Welfare has said. Even as the number of infections and recoveries provide insights to the spread of the virus, the objective of India’s overall response is to contain the deaths in the population, Reddy explains.
Variations in testing affect case fatality and doubling time, Reddy says: Increased testing leads to increased detection of cases, including among asymptomatic people, thereby decreasing the case fatality rate. Similarly, such detection will also lower the doubling time.
An increase in doubling time is a “positive sign”, says Reddy, an adjunct professor of epidemiology at Harvard University and formerly the head of department of cardiology at the All India Institute of Medical Sciences.
One set of numbers can instill blind panic, while another can give a sense of relief--and that is happening to many of us on almost a daily basis as we are trying to look at numbers. So, how do we manage and calibrate our fear?
Well, there are a number of figures that are being published, and I admit that this is certainly confusing. But when you look at what’s happening in Europe, for example--whether it is Italy or Spain or even the United Kingdom--the number they are now focusing on in order to start relaxing the lockdown is the number of deaths per day. And when they say, “for the first time, the number of deaths per day is coming down and is consistently going down, [so] now we can come out of the lockdown”, that is a clear indication that ultimately the objective of the overall response is to try and contain the deaths in the population.
And that is fully understandable, because we know that this virus does kill a fraction of the people infected. Therefore, knowing just the number of people infected is not going to be helpful because even on testing widely, we recognise that there are a number of asymptomatic people [who] could be carrying the virus and may have remained undetected because they are not going to be tested.
Given that situation, we do recognise that ultimately the public health objective is to reduce the deaths. We may underestimate the number of deaths, because we may not be counting them well. But still, the proportion by which these are undercounted will remain generally steady over a period of time.
So, if we can actually track the decline in the rate of death, that will be helpful. But our question is, do we track it as case fatality rate or as deaths per million. Case fatality rate depends upon the number of cases who are diagnosed and labelled as cases. That depends upon the testing rate--the more tests that you perform, your denominator will keep expanding, and quite often with [a] lower number of very sick patients--because the more people you test, the more mild cases you are likely to capture. Even otherwise, the more tests you do, your denominator increases and only a fraction of them will end up with deaths that is in the numerator. So, if we do widespread testing, we will have a lower case fatality rate. So, the case fatality rate is not a particularly helpful indicator.
Then comes the question of what we are going to do about the so-called recovery rate. That’s a little odd statistic, because [a] majority of the people will recover. It depends upon when you are tracking. If you have tracked them one month after the day of the diagnosis, most of them would have recovered. If you are reporting on the fifth day after diagnosis, then many of them would still be in an active recovery phase. So, that’s not a particularly helpful statistic, unless you are looking at the closed cases and seeing how many have died and how many have recovered. That is a better way of looking at it.
We are left with deaths per million. Why am I taking deaths per million? Because ultimately our response is not only limited to salvaging the cases who have been diagnosed and admitted in the hospital or being isolated at home. Our response is also to contain the spread of the epidemic, [to ensure] that more uninfected people and more uninfected regions are not going to be infected. So, deaths per million population gives you a combined measure of both your containment success as well as your case management success. And if you are seeing that number rising very slowly and your deaths per day or deaths per week--whichever way you want to calculate--coming down, then that means you are able to get a good control of the epidemic.
You are talking about deaths per million triggered by the coronavirus, and not deaths per million overall?
Yes. We are talking about deaths per million triggered by the coronavirus. Now that raises a question: How are you sure that you are not undercounting? Of course, we could be undercounting but as I said, if you are actually looking at the trend, that may only create a little noise but it's not going to ultimately disturb the trend too much.
But certainly, we want to know how many people died of the coronavirus. It is possible that out-of-hospital deaths--particularly those who have not been tested--are going to create an artefact. So, we need to do what’s called a “verbal autopsy technique”. This has been employed in a number of other conditions where there are out-of-hospital deaths. People interview family members, friends or close companions, and then, based upon the symptom list, they find out if the person had any symptoms. We now know that there are seven symptoms suggested of coronavirus infection. We can run through that symptom list and based on that we can say “okay, there is a high probability of this person having died of coronavirus”.
We can even get an idea of the out-of-hospital or the undocumented deaths, but we have to remember that when we are estimating this, we have to remove the road-accident deaths because during the period of lockdown, the deaths due to the road accidents would have come down markedly. So, take the total deaths, remove the road accident deaths and then see whether there has been an increase or a decrease compared to the previous year. That is one way of looking at it, but in order to get a better estimate of the absolute number of deaths, have the certified deaths as well as the verbal autopsy documented deaths.
A lot of people are bashing around numbers, for instance, on the number of days it takes for the infection to double; and within that, we have the slice which is recovered, or admitted to hospitals and in severe to moderate cases. Does this doubling make any difference eventually--except to understand how it is spreading?
Well, doubling actually looks at it as an exponential rate of growth. That is helpful to some extent in understanding how the virus is behaving in terms of its growth pattern, but it doesn’t necessarily tell us about how well we are controlling--because the more tests you perform, the more the number of cases you are likely to detect, and therefore your doubling time actually can shorten, whereas if you reduce your number of tests, the number of new cases being added is going to be lower.
Secondly, also in terms of doubling time, you are adding the old cases--many of whom would have recovered and have actually become closed cases--because you are looking at total count doubling.
On the other hand, we may be really interested in finding out how many tests are being performed per day, and of them, how many are actually turning out to be positive. Is that actually decreasing daily, assuming that the testing numbers are standardised and testing criteria are standardised? That gives us an idea of active advancement of the epidemic and the response that the epidemic is having to the public health containment measures. But having said that, even the conventional doubling rate--with increased testing numbers--if it is actually increasing, that’s a positive sign.
The number that you have mentioned is 2 deaths per million. That figure obviously looks good when compared to other countries. To what extent should we even be comparing with other countries at this point of time? Secondly, can it be unpacked into what is the major cause of disease, as we know now looking back and all the deaths that we have seen in the last 60 days?
Well, the deaths per million is going to rise week by week, because there will be more deaths. The denominator, the total population, is going to remain fixed. The question is, is it going to reach the above-300 figure that some of the European countries have reached, the above-250 figure that the United States has reached, or the above-three or above-four figure that some of the Asian countries have reached? Or is it going to remain below five or below 10? That’s what we need to see ultimately.
For cities like Mumbai, it would be much higher, wouldn’t it?
Yes, it would definitely be higher. We are taking for the whole of India, because the whole idea here is if you are using it as a tracker, you are also trying to see how well you are succeeding in containing it. You could unpack it province by province. If your deaths have not reached a high proportion let’s say in Jharkhand or in Meghalaya and Mizoram, then at least you are reasonably happy that, at this point in time, our containment measures are succeeding apart from our case management methods. So, combining both of them gives you an idea of, overall, how the response has been--not just in terms of case management.
If you were to now go one step forward, how should we be responding when it comes to our public health system? Knowing the data that you have now, and using this as a starting point, what else can we be doing, or should we be doing?
I think our public health response in terms of containment has to depend a lot on testing of symptomatic cases and their isolation. If they are mild, they can actually stay at home. If they are moderately severe or severe, they should certainly be hospitalised. And now the question, of course, has come in as to how long to keep them hospitalised and whether to test them or not before discharge.
Data from South Korea and other places clearly show that once a person is treated and becomes asymptomatic, their chances of their infecting are very low and the viruses that may be found by chance on repeat testing are dead viruses. And therefore, the idea is that you may not necessarily want to keep them for too long once they have become asymptomatic. So, that reduces the pressure on the hospital beds.
But certainly, all our containment measures must take into account the need for both home isolation as well as hospital isolation, and increase the facilities for both--in the sense that you must actually educate the people how to isolate at home, and secondly, increase the number of hospital beds as much as possible, as much as needed, particularly in places like Mumbai where the cases are certainly mounting every day.
Amongst active cases requiring critical care in India, only 0.45% need a ventilator, 2.94% need intensive care and 2.94% need oxygen support, the ministry said. So, in all, only 6.3% of active cases require critical care. How does this figure look, and should we take it as something that we should be comfortable with and/or as a source of relief?
I think it's definitely a matter of comfort, because when lockdown started, prior to that, there was a huge amount of anxiety, if not panic, as to whether we will have enough intensive care beds and ventilators, because the experience was actually being extrapolated to the SARS epidemic--SARS 1--where a lot of people had requirement of ventilators. And here, we also saw what was happening in places like the United States and the UK where they were running short of ventilators despite being high-income countries. It appears [that], even in those countries, the need for ventilators is now being reviewed because firstly, the pathology is not uniform. It's not always acute respiratory distress syndrome; it appears to be multiple clotting disorders.
Secondly, interestingly, it has been observed that even high-flow oxygen is adequate; and something as simple as proning--turning the person onto the belly and letting the person sleep on the belly rather than on the back--itself seems to markedly improve the oxygenation. So, we are discovering a lot of things about the management also in this particular case and therefore, the need for [a] ventilator is now considered to be far less than what it was envisaged to be.
Recently, the Director General of ICMR said that there are only 100 patients now across India on a ventilator. It doesn’t mean that we need not have actually arranged for the ventilators, because there was an apprehension at that time. But what we really now require is to strengthen our first level and second level health services in primary healthcare and district hospitals, because that is where most of the case management will have to be done.
A report in the Harvard Business Review said that you should be careful about which data to trust--you should look at transparency, thoughtfulness (whether there is regard for privacy while the data are being collected), expertise and open platforms. So, to come back to my original question, what is the one thing that we should be focusing on now?
Well, I would say, because there is a lag between the actual occurrence of symptoms and death, what we are talking about deaths today is actually telling us what the situation was about 10-15 days back in terms of infections. That is something we need to look at. But despite that, I would say deaths per day are the best indicator for us to find out whether we are actually getting control over the epidemic or not. Then deaths per million, as I said, is a better summative indicator of both containment of spread and of improved case management. So, these are the things I would be looking at.
But I would also be looking at the number of new tests performed each day, and the number of new cases detected from those, as a fraction, to see whether that rate is coming down. That will be another way of looking at it.
Plus, I would also supplement it with syndromic surveillance of household visits by frontline health workers, at least on a weekly basis, to find out influenza-like illness, the clinical assessment, and if need be testing. So, there are multiple measures that we need to undertake. But ultimately, at the end of the day, if you ask six months later or one year later, how did India do in the whole epidemic, we have to go back to the number of deaths and say that these are the kind of deaths that occurred in India, compared to other countries, and we managed to keep it pegged down, because that is the most certain measure of success that we would have achieved.
We welcome feedback. Please write to firstname.lastname@example.org. We reserve the right to edit responses for language and grammar.