Dilemma of an epidemiologist

AM Zakir Hussain

WE KNOW the number of the people in Bangladesh who are suffering from the noticeable features of COVID-19 as they are identified through tests every day although it does not offer a full picture. But do we know what the actual volume of the problem is? We do not.

Process of caseload estimation

THE number of people suffering from an acute disease is estimated in terms of incidence or attack, provided the exact date when it started occurring is known, while that of a disease that is long running and the beginning of which is not discernible is measured in terms of prevalence. The examples of the first type are pneumonia, COVID-19, injury, etc and the examples of the second type are cancer, tuberculosis, etc. The first case of COVID-19 in Bangladesh was reported on March 8. More than three months have gone, but we do not have information on the COVID-19 caseload, more specifically its rate in population. The incidence or the attack rate gives more specific information, particularly for comparison with other diseases and/or comparison of counts of the same disease in space, time and person — the disease are occurring in different areas at different times and among people of different characteristics. Do we know these facts for COVID-19 patients? No, we do not, except for age and sex of the deceased.

So what do we know about the disease then? On a daily basis, the government gives out the number of people tested and the number of infection cases and death. This does not give information on the burden or the size of the disease. To answer them, we need random sample-based tests among people in general. What is so special about the random sampling process? Before answering this, let us first see how the tests are carried out. These are obviously not random and are, rather, purposive, ie tests are done for people who show noticeable clinical features of COVID-19. The tests obviously do not include all those who might also have been infected but do not exhibit clinical features. So we are not getting information on the incidence of the disease or its rate. A rate is estimated by dividing those who were tested and found positive (numerator) with all those who were tested (denominator). The result is then multiplied by a constant, which is 100, if the rate is to be estimated in percentage, or 1,000, eg for child death or 100,000, eg for maternal death — the higher the events, the smaller is the constant. While the tests now conducted give a particular type of rate, which does not show the actual infection rate as not all those who should have been in the denominator are in the denominator, many in the denominator should not have been in the denominator. This weakens the rate estimation that we are provided with now.

 Utility of COVID-19 tests done now

THE present tests give a trend of the occurrence of the disease. But we need to understand the result as it comes. As the number of tests varies on a daily basis, a count of 100 on different days may not mean the same magnitude of the disease on these different days, ie while the numerator was the same, the denominators varied and hence the rate also varied. This means that a higher number on a particular day may in fact be lower than a lower number obtained on another day. These tests, however, give information on the efficiency of the test besides showing the trend of the problem over time.

But these tests have some other problems that relate to the identification of the tested people. There is a possibility that because of the stigma attached to the disease, many tested do not give their true address or residential locations. There was some mix-up of names and locations, eg, one belonging to a district was recorded as belonging to another district because of the similarity of names of people being tested. What about those who are tested several times? Were they accounted as a single person or as many as the tests were? Epidemiologically, the effect will be a poor estimation of rates. For example, they will be figured in the denominator but they will be negative as numerator, dragging the test positivity to the lower side and, hence, reducing the test efficiency and wasting resources, time and taxing on the needs of others, queuing up for days for the same service. On top of that, there are people who get themselves tested only because of their inquisitiveness.

The issue of the use of the available daily data is that it does not represent people or disease in general, as the tests are functions of the number of sample collection booths, number of sample testing laboratories and number of people who are interested to get tested. It will not capture those who are not interested in getting tested, which may vary by location, by literacy, by economy, by age, etc. So, a conclusion on the volume or size of infection in a particular location, based on the present tests, will be misleading. It is, therefore, inferred that the zoning of areas as green, yellow and red, which signify the expanse and gravity of the disease, probably misses some accuracy. As the zoning relate to the potential of transmissibility of infection, and as zoning is based only on the number of identified positive cases and as many who are infected but not tested and are asymptomatic, zoning appears to create a gap in the tightening of the rope.

 Need for random sampling for health outcome

WHY do epidemiologists bet only on random sample-based estimates? How many people in Bangladesh have become infected with COVID-19? To know this, apparently, everyone will have to be tested. That is a preposterous idea. So, what is the alternative? The answer may be obtained even if a much smaller size of population, who can truly represent the total population, is taken and tested. Random sampling is the only technique which ensures an accurate infection rate as this also ensures inclusivity of every type of people — suspected, not suspected, symptomatic and asymptomatic. For region or district-wise estimates, region-wise and district-wise random sampling will be necessary. 

Dilemma of lockdown

LOCKDOWN does not prevent the transmission of infectious diseases completely. It delays and slows transmission. It helps the virus to die out in the infected and to stop further spread of the disease. The other benefit of lockdown is that it ensures an appearance of fewer cases at any given time and, thus, reduces the burden on hospitals at any given time. But on lockdown withdrawal, the simmering number of cases will start infecting others although the reproduction number (R0) will fall below the number when the infection surges. This is why no country has been able to completely wipe out the scourge yet.

Not everyone agrees that the application of lockdown is a good strategy. Many argue that more people die of many other diseases than of COVID-19. So, why create a barrier to people’s lives which will push more people below the poverty threshold? The global economy is already feared to be on the verge of a recession and many countries, including the United Kingdom and the United States, are apparently staring at the recession. The economic meltdown in these countries will severely affect the export-based Bangladesh economy, which is also bleeding with local economic causes. Many, therefore, convincingly claim that more people will die from the lockdown effect than of COVID-19. So, before enforcing lockdown, we need to consider pros and cons of the impact lockdown.

Smaller zones will be inefficient because it will be difficult to enforce and will be costly to manage as it will require more intense deployment of law enforcers, placing them at greater risk of infection. The possibility of the transmission of infection will also go beyond small zones. So, there should be lumping, rather than clustering in small sizes. The aim should, therefore, be a zoning based on social similarity and interaction among people of a locality.

Lockdown in itself is an isolation process but en bloc. The people who are within a lockdown zone are thought to have already been substantially infected. Lockdown actually aims at preventing infection of others from the people who are residing within a lockdown zone. As it has been assumed that most or majority of the people have been already infected in the red zone, it will be a wastage of resources to test them for putting them in isolation. Isolation within isolation will not be a pragmatic step. Cautious people in lockdown zones will take their own preventive measures while others should be compelled to adopt preventive measures meticulously. Other measures will be inefficient.

Conclusion

IF CONTROLLING the spread of infection is the aim of zoning, which it is, the people who are likely to spread the infection should be identified first in a zone before lockdown or colouring it into red, yellow or green. Who are they? They are the unsuspected ones who are infected but unknown to themselves, unknown to others and unknown to all as infected. The people who are already bed-ridden are the least likely to transmit COVID-19.

AM Zakir Hussain is a former director, Primary Health Care and Disease Control, former director of IEDCR, DGHS, former regional adviser of SEARO, WHO, former staff consultant, Asian Development Bank, Bangladesh and working group member of  Bangladesh Health watch (BHW)