There is a debate on the number of deaths in India caused by COVID-19. The official number as on August 26 was 59,449. Nobody believes this number to be true, yet nobody has an alternative number. So, when two young scientists conjured up a specific estimate of 1,88,939 deaths on July 31 based on an ingenuously simple approach (multiply the reported deaths by five) in an article in these pages, they were criticised. There is also a campaign by activists asking the government to put COVID-19 deaths-related data in the public domain. Before we discuss the merit of these two arguments, we need to be clear about the urgency of the need of this estimate. While knowing the number of COVID-19 deaths is a good indicator of burden, this alone is not of much use in planning our current public health strategies, for which case fatality rates or infection fatality rates are far more useful. Also, as the pandemic is far from over, this estimate becomes outdated almost the instant it is generated. Despite this, there is no question that such an estimate needs to be generated, if only for academic purposes. It also gives us an opportunity to be better prepared for such estimations in future.
There are two kinds of challenges in estimating the number of deaths due to COVID-19 – the first is related to weaknesses in our mortality surveillance system; the second to COVID-19-specific issues. It might also be relevant to point out that even for other diseases, estimating the number of deaths is a problem.
Counting deaths
In an ideal world, we would be able to count all deaths and have a certified cause of death in all the cases. In the real world, we have neither. According to the Civil Registration Scheme report for 2018, 86% of all deaths were registered, with 34% being through hospitals. However, this would not be a good source for cause of death as it is as reported by the next of kin at the time of registration.
The second source of cause of death is the medical certification of cause of death (MCCD). As per the report released by the Office of the Registrar General of India (ORGI) in 2018, 1.45 million deaths (21% of all registered deaths and hence, multiplication by five suggested by the scientists) were medically certified. MCCD has an urban bias and there are also issues related to quality of death certification by doctors. The scientists mentioned above ignored the fact that COVID-19 has largely been an urban phenomenon till now, where medical certification coverage is much better, and also did not address COVID-19-specific issues.
The third and final source for cause of death is the Sample Registration System. Here good quality mortality surveillance is carried out in a small representative population. All identified deaths are noted by a trained worker using standard procedures to generate information which is reviewed by a doctor to arrive at a cause of death. This occurs in a small number of deaths (40,000-45,000) nationally per year. However, due to delays in the process, this information is available in the public domain only till 2013.
Coronavirus | 444 deaths added to Chennai’s COVID-19 toll
Impact of the pandemic
In addition, COVID-19 has adversely impacted death registrations due to lockdowns, travel restrictions as well as social stigma resulting in people hiding these deaths. As COVID-19 moves to rural areas, these challenges are going to worsen. Also, as many of these systems are not yet fully electronic, there will be a delay in the availability of these data even after registration. Thus, it is quite likely that the data available with the ORGI for 2020 will need evaluation for its usefulness for estimation procedures, even if made public. I have always supported efforts to put publicly funded data in the public domain. Yet, I would caution against uncontrolled access in this case.
Identifying COVID-19-specific mortality is a challenge across the globe. An approach of using reported death data would be applicable if all those who developed COVID-19 were tested and their deaths certified. However, there are serious limitations of test availability due to restrictive criteria for access and use, even for people with symptoms. Without testing of all suspected cases, healthcare providers have to rely on clinical features to label a death as being due to COVID-19. Our inability to differentiate these from other causes of death may result in misclassification. Without testing, it is going to be difficult for a doctor to differentiate between COVID-19 and, say, influenza. The second issue is related to defining indirect deaths caused by COVID-19. Due to societal and health system disruptions, the pandemic is likely to contribute to deaths from other causes — suicides, non-communicable diseases, tuberculosis. Also, think of a person who had a compromised heart which was worsened when she suffered a COVID-19 infection, precipitating a heart attack, which otherwise would not have occurred now. In a sense, this is also a death caused by COVID-19 as it initiated the final sequence of events leading to death. However, a doctor would certify this as a cardiac death as one cannot ever be sure that COVID-19 initiated the process. How do we include this in our estimate? All these factors will have to be considered when we estimate COVID-19-related deaths.
So, what is the way out? Almost all the issues mentioned for COVID-19 are also applicable to influenza. For the swine flu pandemic of 2009, an estimate published in The Lancet in 2012 indicated that the real mortality was 15 times higher than the globally reported laboratory-confirmed cases. For many years, influenza-associated mortality has been estimated using an indirect modelling approach based on the concept of ‘excess mortality’. An estimate for influenza mortality in India using the same approach based on above data sources has been generated by my team. Such exercises always give you a range and not a single estimate.
Excess mortality
Even countries with near-complete registration of deaths in normal times have acknowledged that their officially recorded COVID-19 death tolls might be underestimates and they are now focusing on excess mortality – the number of deaths above that expected during ‘normal’ times. This approach first statistically defines the ‘expected’ number of deaths in a given period in a given population using past mortality records and then uses the 2020 mortality data to estimate increase in rates of death beyond what would be expected if SARS-CoV-2 had not circulated. The ‘excess death’ approach would include deaths directly (pneumonia or other respiratory conditions) as well as indirectly caused by the pandemic or its interventions (cardiac conditions). It will have to be acknowledged here that lockdowns may have had a preventive effect on some other deaths, notably road accidents and pollution-linked deaths. Depending on the availability of data and analytic capacity, this can lead to a real-time measurement of excess mortality or it could be an exercise done at the end of the epidemic.
Thus, the estimation of ‘true’ mortality due to COVID-19 is likely to pose significant methodological and systemic challenges. However, sufficient technical capacity exists in the country to do so. It should be noted that deficiencies in mortality surveillance have existed for a long time. There is no denying that these need to be addressed urgently.
Anand Krishnan is a Professor at the Centre for Community Medicine at the AIIMS, New Delhi