Representative photo: People wait to cremate their relatives who died of COVID-19 at a crematorium in New Delhi, April 23, 2021. Photo: Reuters/Danish Siddiqui
Even as accounts of overflowing crematoria, hasty riverine burials and enormous suffering dominated the news in India in April and May 2021, the true death toll from COVID-19 in India remains a matter of dispute. Official figures are widely understood to be vast underestimates. There are several reasons for this, as Aashish Gupta, Dhanya Rajendran, and Rukmini S. wrote in The Wire: home deaths of COVID-positive people may be missed or misrecorded as non-COVID deaths, as might the deaths of those who had COVID-19 but were not tested, quite apart from intentional suppression.
Because there is reason to believe that all of the above are happening to various degrees, demographers have argued that it is important for researchers and reporters to estimate all-cause ‘excess mortality’ during India’s epidemic. In essence, the goal of such an exercise would be to ask how many more people died in a certain period than might have been the case if COVID-19 hadn’t been in the picture.
Since this counterfactual number cannot be observed, it must be estimated – based on past death records or by calculating expected deaths based on non-pandemic death rates.
In the US, for example, the Centres for Disease Control (CDC) publishes daily updated estimates of excess mortality due to COVID-19, noting that “provisional death counts are based on death certificate data received and coded by the National Center for Health Statistics”. Even here, as the CDC notes, there may be a lag in receiving, collating, coding and tabulating data, so the most recent weeks’ data may not be complete. Nonetheless, universal or near-universal civil registration systems, and timely releases of data from these systems, allow us to estimate all-cause mortality from death registration data.
Recently, researchers used India’s Civil Registration System data for various jurisdictions to perform this exercise for some cities and states. For example, for Kolkata, Rukmini, a data journalist, found that there were 17,587 certified deaths from April 1 to May 27, 2021, compared with an average of 9,353 deaths for the corresponding period in the pre-pandemic years of 2015-2019. Thus, there were approximately 1.88x deaths in Kolkata in April-May 2021 as the expected number – or an excess mortality of 8,234.
Most recently, Rukmini also calculated excess deaths in Madhya Pradesh. She found that MP had reported 3.3 lakh deaths in 2021 from January to May 2021. The average number of deaths registered during the corresponding months of 2018-2019 was 1.6 lakh. So there were 2.06x deaths in 2021 (until May) as would have been expected. The number of excess deaths also shot up in April and May.
As the researchers wrote in The Wire, several journalists have extensively surveyed crematoria and burial grounds to arrive at alternative measures of true mortality, but there is reason to believe that this data is incomplete.
However, in India, incomplete death registration also means the CRS data in some states are also undercounts themselves. A recent estimate for MP put the fraction of deaths registered in the state at just over 60%. This figure was likely lower during India’s second COVID-19 wave. It can also take several months, or even years, for death registration data to be collated and made available, as Rukmini noted.
So as such, it may be useful to compare CRS-based estimates to estimates of excess deaths using other data sources, where available, to get a sense of how useful these alternative data sources are.
This article explores such a calculation for MP, using funeral counts for April 2021 for a set of urban areas in the state, following up on the painstaking work of Dainik Bhaskar reporters.
Also read: Covaxin: Why Antibody Levels After Vaccination Aren’t the Same As Efficacy
The numbers below were calculated as follows. First, I used the April funeral counts by district; these are only for the urban areas of each district. This yields an estimate of the total all-cause mortality in urban areas in these districts in April 2021.
To estimate the expected number of deaths – i.e. the number that would have been expected to occur in the absence of COVID-19 – I used the latest available Sample Registration System data for urban MP to arrive at an expected count. The 2018 SRS estimate is 5.5 deaths for 1,000 urban inhabitants per annum, which means that on a typical day, an urban area in MP would expect to have 15.0685 deaths per million. This translates to 452.055 deaths/million in April.
Then, data from the November 2019 report of the Government of India’s technical group on population projections, is used to estimate the growth rate of MP’s urban population between 2011 and 2021. Tables 9 and 12 of this report provide estimates for March 1 and July 1 of each year. Using the 2021 figures, it would appear the state’s urban population has grown 20.9% in the 10 years up to July 1, 2021, and 21.1% up to March 1, 2021. I went with 21% overall.
Applying this to the 2011 Census gives us the estimated urban populations for each district in 2021. Together with the expected deaths/million numbers from the SRS data, let’s calculate the expected number of deaths in the urban areas of each district in Dainik Bhaskar‘s reports.
For a concrete illustration, consider Bhopal. Daink Bhaskar reported that there were 5,433 funerals in April. The 2011 Census placed Bhopal’s population at 1.91 million. A decadal growth of 21% would make its population in 2021 equal to 2.31 million. And the SRS estimate of 5.5 deaths per 1,000 urban inhabitants gives us 1,405 deaths in April 2021 – if it had been a ‘normal’ year.
The excess deaths are therefore 5,433 minus 1,045, which is 4,388. So almost as many people died in Bhopal in April 2021 as would have been expected to die in January, February, March and April. And Bhopal officially reported only 109 COVID-19 deaths in April.
Findings
For April 2021, I have data on urban areas in 26 of MP’s 50 districts, and the numbers imply that there were 22,170 deaths in a month when these areas would normally have expected 7,986. These districts accounted for 14.6 million of MP’s total urban population in 2011 of 20.6 million. So we can naively extrapolate to all of urban MP to estimate that where urban MP would have reported 10,972 expected deaths in April, using the multiple of 2.78 obtained from the calculations above suggests 30,461 people may in fact have died in April in urban MP alone.
Also read: India Is Undercounting Its COVID-19 Deaths. This Is How.
For context, all of MP reported 8,207 official COVID-19 deaths during both waves, until June 3, 2021. This works out to a death rate of 15.29 deaths per 1,000 population, compared to the 2018 SRS estimate of 5.5 deaths per 1,000 for urban MP.
According to Rukmini’s calculations, MP had 28,000 deaths on average in April 2018 and April 2019 – but had 40,761 excess deaths in April 2021. The ‘multiple’ – i.e. total deaths divided by expected deaths – for this data works out to 2.45 for all of MP, while our estimate of a multiple of 2.78 is based on funeral data for around 75% of the state’s urban population alone. And these estimates are strikingly similar.
Where available and possible, therefore, we should carefully collate funeral data – because it could provide a useful handle on excess deaths during COVID-19, at least until such time more formal data sources become widely available.
Saugato Datta oversees anti-poverty work at ideas42, a behavioral science firm. He has a PhD in economics from MIT and has worked at The Economist and the World Bank.