A view of IIT Kanpur. Photo: iitk.ac.in
- The recently published report is authored by Manindra Agrawal, well known before the pandemic for his work on complexity theory.
- The document is easy to dismiss as political propaganda – but it deserves careful scrutiny given the wide media coverage and the eminence of its author.
- Its most striking weakness is that it fails to ask, in any meaningful way, how many people died in Uttar Pradesh as a result of the COVID-19 epidemic.
A recent IIT Kanpur report, entitled ‘COVID War, UP Model: Strategies, Tactics, Impact’, showered praise on the Uttar Pradesh government for its handling of the COVID-19 crisis. The full report, over 100 pages long, is available in English and Hindi.
The report is authored by Manindra Agrawal, well known before the pandemic for his work on complexity theory. He has also been a regular pandemic commentator, making widely reported, but incorrect, claims in early 2021 that India had reached herd immunity and that there would be no second COVID-19 wave.
(I have discussed these claims before in the context of how weak and politically-sponsored science likely fed into India’s second wave devastation.)
The IIT Kanpur report is deeply flawed in many ways, and it would be easy to dismiss it as mere political propaganda. But it deserves careful scrutiny given the wide media coverage and the eminence of its author.
Its most striking weakness is that it fails to ask, in any meaningful way, how many people died in Uttar Pradesh as a result of the COVID-19 epidemic.
Deaths, counted and uncounted
Apart from a passing reference to “media reports on bodies floating in the Ganga”, the report finds no space for the numerous ground reports on the state’s devastating second wave – which described the overwhelmed crematoria, the panic, the confusion, the lack of medical care, the oxygen shortage, and many probable COVID-19 deaths that went unchecked and unrecorded.
One particular group that the IIT report doesn’t mention is Uttar Pradesh’s school teachers. According to the Uttar Pradesh Primary Teachers Association 2,046 teachers died in Uttar Pradesh following election duties during the massive second COVID-19 surge. According to the state government, this number was three.
This mismatch is extreme but not entirely surprising. Several studies have now shown that India has been very hard hit in terms of mortality, with official data hugely underestimating epidemic deaths. States comparable to Uttar Pradesh in terms of development, such as Madhya Pradesh and Bihar, stand out for having excess deaths more than 20-times higher than their official death tallies.
The IIT report doesn’t discuss any of this contextual information on mortality nor does it refer to any mortality studies. Yet its author finds space for a remark about China: “the reported deaths per million from China is a flat line almost on X-axis, thereby raising doubts about the sanctity of its data.”
The first wave
While most of its attention is focused on official fatalities, the IIT report does briefly discuss all-cause mortality data, but in order to make a misleading claim. On page 100, we find the following para (reproduced verbatim):
“Is there a way to analytically estimate excess number of deaths in the state due to the pandemic? … The most robust method is to use Civil Registration System (CRS) that records deaths across the country. At present, data for the year 2020 is available and one can use it to estimate deaths caused by the first wave in the state. As the CRS data shows, the increase in number of deaths was around 3% in 2020 over 2019 which matches with the average growth rate during 2009-19 for the state. This leads to the conclusion that the actual number of deaths due to Covid during the first wave in the state was not significantly higher than the reported number, otherwise the increase would have been significantly more than 3%…”.
There are no references or links to external sources, but this para could refer to data obtained via an RTI application by investigative journalist Saurav Das, and reported in Article 14. The total number of death registrations for 2020 in this data is, indeed, 3% higher than for 2019.
The striking thing about this data, obtained from the Uttar Pradesh government’s chief registrar, is its many glaring anomalies. There are unexplained zeros, and large and unlikely fluctuations in the numbers from some districts. The totals of registrations for 2019 in many districts are not even close to the district totals in the 2019 CRS report.
But despite these discrepancies, the data deserves a closer look if it is being used to make claims about underreporting. In fact, it shows a major surge in death registrations. The 12-month period from April 2020 to March 2021 saw 16% more registrations than 2019, and 22% more than during April 2019-March 2020.
Basically, there were large drops in registration during March-May 2020, and large increases in late 2020 and early 2021. We know that the epidemic and the lockdown widely disrupted death registrations across India, and that registration often occurs with major delays in Uttar Pradesh. These factors might explain the pattern in the data.
There is certainly no reason to claim, on the basis of this data, that there was a low number of COVID-19 deaths during the first wave in Uttar Pradesh. A surge of 16-22% in registrations over 12 months is comparable to the first-wave surges in badly hit states like Karnataka, Tamil Nadu, Maharashtra and Andhra Pradesh.
The second wave
At the time of writing, there appeared to be no publicly available CRS data for Uttar Pradesh beyond April 2021. So we are forced to try and estimate the state’s second-wave COVID-19 mortality in other ways.
In earlier work, this author and others collected numerous reports on probable COVID-19 deaths in rural areas of North and Central India that appeared in the Hindi press during the first three weeks of May 2021. From these reports, we extracted 61 case-studies, each describing what happened in a village or small cluster of villages. Twenty-six of these case-studies were from Uttar Pradesh.
They tell a tale of epidemics rapidly sweeping through villages, leaving a trail of death in their wake. The pattern was consistent: testing and access to healthcare were minimal or absent, and almost none of the deaths were officially recorded as COVID-19 deaths.
Can we use the data from such reports to estimate Uttar Pradesh’s second-wave mortality? Not directly. It is likely that such reports focused on the hardest hit villages, so a direct extrapolation would risk overestimating total deaths in the state. We can, nevertheless, use this data to estimate fatality rates and thus estimate COVID-19 deaths more indirectly.
Suppose, for example, a village of 5,000 had 17 deaths over three weeks1. Based on rural death rates in the state, we would expect around two deaths in a village of this size over three weeks. So this village saw 17 – 2 = 15 excess deaths; this is an excess mortality of 15/5,000 = 0.3%.
If all residents of the village were infected as COVID-19 swept through, and this resulted in 15 deaths, this would mean a COVID-19 infection fatality rate (IFR) in the village of 0.3%. The true IFR is likely to have been higher since it is unlikely that all residents were infected (ergo the denominator could be lower).
Based on such calculations, we arrived at a conservative estimate of second wave IFR in rural Uttar Pradesh of 0.25% (the median value of excess mortality in reports from the state), and a more likely value closer to 0.5%. Indeed, six out of the 26 rural clusters in Uttar Pradesh saw excess deaths amounting to more than 0.5% of their population in short periods of time.
Limited seroprevalence survey data suggests that the state had, perhaps, 100 million COVID-19 infections during its second wave – amounting to a little under half of its population being infected. An IFR of 0.25-0.5% would then imply between 250,000 and 500,000 second-wave deaths. This amounts to 1.1-2.2 deaths per 1,000 population, or around 18-36 times the state’s recorded second-wave COVID-19 deaths.
Is this toll plausible?
Madhya Pradesh has a similar age structure and a similar human development index as Uttar Pradesh. This suggests that the two states are likely to have similar values of COVID-19 IFR. Estimates based on Madhya Pradesh’s relatively good quality CRS data suggest it had over 200,000 excess deaths during March-May 2021, which is around 2.6 excess deaths per 1,000 of its population. Unless there is good reason to believe that Uttar Pradesh fared much better – there isn’t – the estimates for Uttar Pradesh don’t appear to be far-fetched.
The huge mismatch between estimated deaths and official deaths is also not unexpected. Indeed, we found similar ratios in Andhra Pradesh, Bihar and Madhya Pradesh as well.
What can we conclude?
Data about COVID-19 mortality from Uttar Pradesh is limited, and official data is of low quality. But whatever is available is consistent with high mortality – comparable to the levels in other badly hit states. It is also likely that there has been extremely poor recording of COVID-19 deaths in the state.
If CRS data of decent quality becomes available, it will provide clues about the true scale of Uttar Pradesh’s pandemic deaths. But given the low and fluctuating civil registration of deaths in Uttar Pradesh, the picture will still be incomplete. Ultimately, the best way to find out the scale of epidemic mortality in Uttar Pradesh will be a large-scale mortality survey.
Now, did the IIT Kanpur report make any serious effort to assess the mortality impact of the COVID-19 epidemic in Uttar Pradesh? The answer is clear and unambiguous: no. Given that its foreword thanks the state government for “taking this very progressive step of providing detailed pandemic data for the purpose of analysis”, surely the authors could have requested complete CRS data from the state government. Perhaps they could even have encouraged the government to undertake its own mortality survey.
But by sidestepping the question of pandemic mortality, the IIT Kanpur report ultimately trivialises the outbreak’s impact on the state’s more badly-hit communities, and immensely disrespects all those families who lost loved ones.
Murad Banaji is a mathematician with an interest in disease modelling.
This was typical.↩