Coronavirus | Spike in excess deaths during pandemic

Lessons from India’s all-cause mortality data

Close up of a COVID-19 text tag on paper

Close up of a COVID-19 text tag on paper

The scale of devastation caused by India’s COVID-19 epidemic is gradually becoming clearer. This is thanks to the efforts of journalists, The Hindu included, who have been gathering all-cause mortality data from around the country .

The mortality data, from State and city civil registration systems, paint a grim picture of a major increase in deaths across the country during the novel coronavirus pandemic. Very few of these additional deaths have been recorded as COVID-19 deaths.

Cautious estimate

We can try to understand the scale of the tragedy via a simple question. How many extra deaths have occurred, over and above those expected in normal times? The data suggest an approximate answer: during 15 months from April 2020 to June 2021, there were 3.5 million-3.7 million “excess deaths” nationwide. This amounts to 35% more deaths than expected.


This estimate is cautious, and likely to increase as more data come in. Data for June and beyond are very limited, and so the story is incomplete.

Before we examine the numbers it is important to understand the context. There are several reasons why estimating a surge in mortality is difficult. We need to know how many deaths have occurred during the pandemic, and how many to “expect” in normal times. The idea is to carefully reconstruct these numbers from death registrations and survey-based estimates of pre-pandemic mortality.

Data in the pandemic period

But death registration data for the pandemic period are limited. It is unavailable for some States, and incomplete in others, for example coming from online systems which do not log all death registrations. Some data are organised according to date of death, and some by date of registration. Moreover, there are uncertainties about death registration prior to the pandemic. In some States, official estimates of levels of registration, which we use, appear to be overestimated.


Compounding the difficulties, registered deaths show complex trends in some States — for example, gradually increasing prior to the pandemic, but dropping sharply around the time of the national lockdown before the pandemic deaths start to show.

To arrive at the estimates here, we examined data from 12 States where partial or complete civil registration data are available for at least January 2018 to May 2021: Andhra Pradesh, Bihar, Haryana, Himachal Pradesh, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Punjab, Rajasthan, Tamil Nadu and West Bengal. These States comprise roughly 60% of the national population.

During April 2020-May 2021 we found six million death registrations in the data; that is 1.3 million more than expected from 2019 data. If we assume — perhaps optimistically — that the deaths which were not captured in these registration systems, including unregistered deaths, rose proportionately, we arrive at an estimate of around 1.7 million excess deaths in these States up to May.


If these 12 States reflect the national picture, then India saw around 2.8 million excess deaths nationwide during April 2020-May 2021. This is 8.5 times the official COVID-19 death toll of 3,32,000 over the same period.

Global comparison

Using limited data for June (currently available only for Andhra Pradesh and Punjab), or assuming that the ratio of excess deaths to official COVID-19 deaths does not change rapidly, we estimate 3.5 million-3.7 million excess deaths nationwide by the end of June. Over a 15 month period, for every three expected deaths, there was a further “pandemic death”.

This places India among the harder hit countries in the world. It would mean that relative to baseline, India’s surge in mortality is lower than that of Mexico, similar to that of Brazil and South Africa, and considerably higher than in the United States, the United Kingdom and most of western Europe.


Moreover, the estimates here are conservative. More up-to-date data will push up the numbers. There are also hints that disruption may have prevented — and not merely delayed — many death registrations. For instance, we see significant drops in birth registrations during 2020 in some States where this data are available, most noticeably in Kerala and Andhra Pradesh. Moreover, there are good reasons to believe the mortality surge may have been greatest in marginalised communities where death registration is weaker.

We cannot be sure how many of India’s excess deaths were from COVID-19. According to the latest national serosurvey, around 60%-70% of people in India may have been infected with the virus by June. If so, international data on fatality rates suggest we should expect two million to four million COVID-19 deaths. So, it is quite plausible that the majority of India’s excess deaths have been from COVID-19. But we cannot rule out a significant surge in non-COVID deaths too.

Individual States

In individual States, all-cause mortality data paint diverse pictures. Kerala, Punjab and Himachal Pradesh stand out for having somewhat lower excess mortality than expected, even after we adjust for possible disruptions to registration. Andhra Pradesh and Madhya Pradesh, on the other hand, saw considerably more deaths than expected.


Overall, around two-thirds of the excess deaths took place during a shocking mortality spike around May 2021. But the time-course varies in different States. Madhya Pradesh’s explosive second wave accounted for 80-90% of its excess deaths. By contrast, Maharashtra saw more even surges, with over 40% of its excess deaths during its first wave.

There are striking variations in the ratio of excess deaths to recorded COVID-19 deaths. In Maharashtra excess deaths up to May 2021 are roughly four times recorded COVID-19 deaths, or less if we factor in reconciliations of COVID-19 deaths during June and July. By contrast, in Madhya Pradesh, excess deaths are an astonishing 25-30 times recorded COVID-19 deaths.

Exploring the stories behind these variations is important for understanding the pandemic and disease surveillance in India.

Editorial | Counting the dead: On measuring excess deaths

Considerable gaps remain. Some civil registration data are available for Uttar Pradesh up to April 2021, and these appear to show a major surge in mortality; but there are huge fluctuations in registrations, and unexplained discrepancies with historical data which make it hard to use this data with any confidence. Where civil registration data are of poor quality or unavailable, large-scale mortality surveying could help to fill the gaps.

A perspective

Could the sharp rise in death registrations reflect not a surge in mortality but improvements in death registration? This claim has been made, but is not credible for several reasons. If we accept the estimate that 92% of deaths were registered in 2019, higher registration coverage could not cause a 35% surge in death registrations.


In fact there is little evidence for improving death registration during the pandemic. During the relatively quiet period between the two COVID-19 waves (January-March 2021), we see death registrations return close to 2019 baseline levels. And throughout the pandemic period, we see a very strong association between monthly excess deaths and official COVID-19 deaths, strongly suggesting these are pandemic-related excess deaths, and not a reflection of underlying trends in registration.

There are no easy ways to explain away or deny the scale of the catastrophe. Yes, there are uncertainties, and details will change as more data become available. Most likely, the numbers will increase. One thing is clear: during the COVID-19 pandemic, India has witnessed a surge in mortality on a scale not seen since Independence.

Murad Banaji is a mathematician at Middlesex University London. Aashish Gupta is a demographer, and a research fellow at the research institute for compassionate economics

Our code of editorial values

This article is closed for comments.
Please Email the Editor

Printable version | Apr 2, 2022 5:26:19 am |