News
Mint Explainer | New covid death estimate: Does Indian govt's rebuttal add up?
Summary
- The Indian government has countered a new study that estimated that the country saw 1.19 million excess deaths in 2020. But such estimates, using indirect methods, are only trying to fill a void created by insufficient and sketchy official data
A new study by researchers at several universities, including the University of Oxford, estimates 1.19 million ‘excess deaths’ in India in 2020, the first year of covid-19, compared to the preceding year. The Indian government promptly rejected the report’s methodology and findings.
The study, published in the Science Advances journal on 19 July, says that life expectancy at birth was 2.6 years lower and mortality 17% higher in 2020 compared to 2019. This is for a subsample of 14 states and Union territories, but the authors said that if other states followed the same trends, it would add up to 1.19 million excess deaths across India in 2020. The study found that life expectancy declined more for women than for men, and also more for children and for marginalized groups. Mortality was particularly high in the last four months of 2020.
Mint explains how to read the numbers, and understand the nuances of official estimation of fatalities in a country where many deaths go unregistered.
First, what are excess deaths?
Governments will naturally fail to capture the true toll of a pandemic like covid-19: Not all patients are tested, and defining a “covid death" is complicated. This left the “excess deaths" approach the best-placed method to estimate the toll.
"Excess deaths" approach t is the difference between the number of deaths in an unusual period (such as 2020 and 2021) and the number of deaths in a normal year. This can give a sense of the direct and indirect toll of a pandemic like covid-19.
How did the new study reach its estimates?
The latest estimate is eight times India’s official death count attributed to covid-19 in 2020, and 1.5 times the World Health Organization’s excess deaths estimate. It relies on the last National Family Health Survey (NFHS), fieldwork for which was conducted between June 2019 and April 2021. The survey asks questions that can be used to assess life expectancy and history of recent deaths in a family.
The study used data for a subsample of NFHS-surveyed households that were interviewed in the four months of 2021, which amounts to 23.2% of the overall sample and is spread over 14 states. To explain the credibility of this method, the study first establishes that mortality estimates derived in this manner from NFHS for 2018 and 2019 (based on households’ recall of recent deaths) mirrored the estimates from other official data, such as the Sample Registration System (SRS, which uses a nationally representative sample of households to estimate death rates, among other statistics, every year). But for 2020, it wasn’t so.
How has the government countered this?
The Indian government, in its rebuttal, said that 14 states could not be extrapolated to the entire country. It also said that there could be reporting and selection biases in the chosen subsample since the data was collected at the peak of the pandemic. The Centre said the study failed to note that SRS data for 2020 showed no increase in death rates and no reduction in life expectancy.
It said the report’s finding that more women and more children died was at odds with official data, but did not comment on the claim of higher mortality among marginalized groups.
Tell me more on how else one can estimate excess deaths.
In India, the Civil Registration System (CRS) counts the number of deaths registered in a given year. But it’s not sufficient, since it misses the deaths that weren’t registered. For this, the CRS for past years used to give both the registered deaths and the total estimated deaths (the latter relied on SRS). It shows that 92% of the deaths in India were registered in 2019, up from 85% in 2018.
The CRS report for 2020 (the last year for which it’s available, released in early May 2022, a day ahead of the scheduled release of the WHO’s estimates) only reported the number of registered deaths (8.1 million, up 6.2%) that year. One cannot estimate excess deaths without a number for estimated deaths, too, which tends to be much higher. That estimate was released curiously by the Press Information Bureau a few days later, while countering the WHO estimates.
It claimed that 99.9% of deaths had been registered in 2020, and hence the 8.1 million number was somewhat an accurate measure of the actual number of deaths that year. This effectively meant that the official estimate for deaths (and not just registered deaths) in India was actually lower in 2020, at 8.1 million, down from 8.3 million in 2019.
Why are researchers trying to use indirect methods to calculate excess deaths?
One, because the 99.9% registration figure for 2020 is too good to seem true (particularly for a country like India in a year affected by lockdowns. For context: NFHS reported that only 71% of deaths were registered in the survey period). Second, because crucially, the CRS for 2021, the year in which covid was at its peak, is still awaited. Such a delay is unusual, and makes it difficult to have an official estimate of ‘excess deaths’.
This is where research like the new one, trying to derive indirect estimates from other sources, comes into the picture, as it fills the void created by official data. The WHO estimates were based on incomplete official estimates from 17 states and Union territories.
Has the study addressed the points raised by the Centre?
Yes, in many ways. The study admits that the analysis of 14 states may not be sufficient to understand the national trends. However, it says that the changes in mortality trends in the subsample may not have been too different from the rest of India. This claim is based on a national serosurvey, which had shown that the subsample states had similar spread of the virus.
The study also says that its estimates held true even in hypothetical subsamples within the subsample, lending credence to the extrapolation. However, to account for any likely flaw in the extrapolation method, the study reports a bigger estimate range, with 1.19 million being the midpoint. The study is 95% certain that the pan-India figure was between 730,000 and 1.65 million.
The study also says in states with high rates of death registration, the CRS data showed excess mortality patterns that were broadly consistent with what it derived from the NFHS.
