Number Theory: What does CRS data tell us about India’s Covid deaths?

Niti Aayog member (health) VK Paul recently said the latest CRS numbers show that claims of India’s official Covid-19 death toll being gross underestimates are untrue. An HT analysis shows that this conclusion is best reserved.

The official Covid-19 death toll in India as on May 3 is 522,676. Does this capture the actual mortality due to the pandemic in the country? Some experts believe that the actual death toll is likely to be significantly higher, primarily a result of Covid-19 deaths not being classified as such (See here). People in charge of India’s response to the pandemic believe that the reported number is accurate. The World Health Organization (WHO) is expected to release its Covid excess mortality data this week.

Workers put on personal protective suits before carrying the body of a Covid-19 victim for cremation in Delhi in September 2020 (AP File Photo)

The only piece of empirical, although piecemeal, evidence which has been produced in favour of the argument of underestimated fatalities is an abnormal increase in registered deaths during the pandemic, especially during the second wave which peaked on May 9, 2021. Because official data on registered deaths was not available when these state or even city level estimates were being published by various media organisations, a holistic evaluation of this claim was not possible.

On May 3, the ministry of home affairs published the official birth and death registration numbers for 2020. The annual report of the Civil Registration System (CRS), which contains these statistics, shows that registered deaths increased by 6.2% (474,806 in absolute terms) compared to their 2019 level. This is less than the increase in registered deaths between 2018 and 2019 (690,469 or 9.9%) and also between 2017 and 2018 (486,828 or 7.5%).

Commenting on these numbers, VK Paul, member (health), Niti Aayog, said that the latest CRS numbers show that claims of India’s official Covid-19 death toll being gross underestimates are untrue. “Nobody should use modeling anymore, and come up with exorbitant multiples based on those modeling exercises. Apart from Covid-19 additionality there are other factors such as increase in population, increase in reporting (registrations), more awareness, etc.” he said.

An HT analysis shows that this conclusion is best reserved until the 2021 CRS and 2020 and 2021 SRS numbers are published. Here are four charts which explain this.

Improvement in death registrations need not mean a rise in total number of deaths

How much of the increase in number of registered deaths in 2020 a result of Covid-19 and how much of it is due to factors such as better registration and an increase in population? The latter is a factor because it acts as a base for a given death rate in any country. Past data can offer some insights here. Death registrations, as per the estimates given in the CRS, have improved significantly in India over the last decade. The CRS makes this estimate by using data from the Sample Registration System (SRS), which is published by the same agency which releases CRS. Unlike the CRS, the SRS is based on a state-level representative survey.

A comparison of number of estimated (not registered) deaths from the CRS reports and share of registered deaths in total deaths shows that the former has not increased as much as the latter. In fact, number of estimated deaths has been falling continuously since 2013. This is basically a result of an increase in life expectancy as seen in falling crude death rate (CDR) levels. CDR is the number of deaths per thousand people and its estimates are published by the SRS. Falling CDR is a common phenomenon as countries make economic progress.

But there is a caveat here: the pandemic made people tardy about registering births (as indicated in the 2020 numbers in CRS), and it might have done the same thing about deaths too, perhaps reversing trend of increasing registrations (with the fact that the number went up despite this pointing to the impact of the pandemic).

2020 CRS death numbers had two mutually opposite trends

The headline number of registered deaths in 2020 hides behind it two mutually opposite trends. While the pandemic is likely to have increased the number of deaths – India’s official Covid-19 death toll in 2020 is 149,036 – the 68-day long hard lockdown which began on March 25, 2020, is expected to have generated headwinds for death numbers as fatalities due to traffic accidents or heat strokes etc came down.

There is at least one set of numbers – the Accidental Deaths and Suicides (ADSI) report of the National Crime Records Bureau (NCRB) under the MHA – that suggests this is possible. As HT reported on October 29 last year, deaths due to traffic accidents (which account for around 40% of all accidental deaths in NCRB data), fell by 19% in 2020, a likely result of the lockdown. Overall accidental deaths also fell by 11%. To put these numbers in context, the number of deaths due to accidents came down from 421,104 in 2019 to 374,397.

Had CRS published monthly death registration numbers, it would have been easier to notice these two trends. The number of Covid-19 infections and deaths was significantly lower during the lockdown period and surged around September 2020 which is when the first wave of infections peaked. This trend can be seen in a comparison of monthly death registrations in 2020 and earlier years in the state of Assam, which was published first in Scroll.in .

But monthly numbers could be contaminated by delay in registrations

To be sure, even if monthly death registration statistics were available, they might not capture the immediate mortality scenario in a given region. This is because India also has a problem of delayed registrations along with deaths not being registered at all. Assam, the data for which has been cited above, had a major problem on this count in 2020. While the share of within 30-day death registrations – the CRS records this explicitly as late registrations carry a monetary penalty – increased in 15 out of 29 states which gave this data, Assam was not one of them. In fact, the share of within 30-day death registrations in Assam decreased from 86.6% in 2019 to 36.3% in 2020.

CRS data also shows Covid-19 official death toll being an underestimate is not because of some conspiracy

While there have been anecdotal accounts of deaths of Covid-19 positive patients being attributed to other causes, this is not the only factor which could have created a mismatch between the official Covid-19 death toll and estimated number of deaths. There is a possibility that some people who died due to Covid-19 never received any medical attention and hence were not included in either the total number of infections or deaths. The CRS data supports such an argument as it shows a sharp rise in share of recorded deaths where no medical attention was received. Once again, the data has been patchy even in the pre-pandemic period, as some states do not report this. However, the 2020 share is quite high by historical standards.

What does this mean for the debate on India’s Covid-19 death toll?

Even the official death toll suggests that 64% of India’s Covid-19 deaths took place in 2021, largely a result of the second wave driven by the delta variant of the Covid-19 virus. As and when the 2021 CRS data is published, we will know whether the number of recorded deaths, parts of which were published by various media organisations, were true. To be sure, we will also need SRS data for 2020 and 2021 to get an idea about the estimated number of total deaths and whether the pandemic led to a one-time reversal in India’s death rate.