Over the last month, Indian journalists from across the country have accessed and reported on State-level all-cause mortality from the Civil Registration System (CRS), currently confidential and closed to the public. While India’s official death toll from COVID-19 has been suspect from the very beginning — for reasons that are partly institutional and partly new and specific to the pandemic — it was the second wave that made the scale of devastation hard to ignore.
States and data
Using the example set by those attempting to capture missed COVID-19 deaths in the United Kingdom, South Africa and Peru, journalists from a range of news organisations accessed all-cause mortality data from the CRS for Madhya Pradesh, Andhra Pradesh, Tamil Nadu, Bihar, parts of Uttar Pradesh, Kerala, and other States as well as for several Indian cities. This data has been able to show a large rise in excess mortality in the surge months of the second wave of April and May 2021 in particular, far in excess of the official COVID-19 toll for the same period. The increase in mortality ranges from reported deaths climbing to five times the usual monthly data in Madhya Pradesh and Andhra Pradesh (between 30-40 times the official COVID-19 toll for the same period), to more modest increases in Tamil Nadu and Kerala.
Sources of error
While this data has added an important dimension to the question of undercounting of COVID-19 deaths in India and remains the best source of data on all-cause mortality, its continued use will require careful work on account of some known sources of error that exist with the data.
The first is that these numbers are underestimates of total mortality. The Union government last published annual Civil Registration System data for 2019. This data shows that India registered 92 of every 100 deaths as of 2019, but there was large variation between States. Bihar, for instance, registered just half of all deaths. Additionally, the data currently available on State-level CRS portals (behind a login) miss even more deaths. The CRS moved online relatively recently (between five and three years ago in most States), and State-level portals now display data for every day from January 1, 2018 to the present day. For all States for which data is available except Kerala, the online portals show lower numbers than what was published by the Union government for 2018 and 2019. What this implies is that the data currently being used for analyses are an underestimate of total deaths, but should not by itself cause an inflation of the excess mortality estimates for the pandemic (since the online portal’s data was used for past and current years).
What could, however, alter the magnitude of calculated excess mortality would be a second possible source of error — if mortality was either naturally increasing over time, or if registration was getting better, or if both were taking place. If any of these three phenomena were taking place, the magnitude of excess mortality in 2020 and 2021 could be moderated by these processes.
On the first count, total estimated mortality (registered deaths plus those that are not registered) has actually been declining since 2013 (although there was an uptick in some States in 2019) according to the CRS. So, in the normal course of events (if there had not been a pandemic) we would not expect an increase in mortality
Registered deaths, on the other hand, have been steadily increasing in four of the five States that we have detailed mortality data for (but fluctuating in Kerala). While this will mean that some of the excess mortality in 2020-21 will be accounted for by improved registration, this is not a complete explanation. The increase in registered deaths will not continue at the same pace in all States — once States begin to achieve full registration, increases should be more modest. In Madhya Pradesh, for instance, the level of death registration improved by over 21 percentage points between 2017 and 2019 to reach 89% in 2019. It is simply not mathematically possible for the level of registered deaths to continue to increase at the same pace for the years after 2019.
For the most conservative estimate, let us assume that Madhya Pradesh achieved 100% registration by May 2021 (although we know from State level data that this was far from the case). Given the flattening of mortality in the State, let us assume that the total mortality in 2021 in ‘normal’ times would have been about the same as the total estimated deaths for 2019 — 5.53 lakh deaths. Even then, the numbers for Madhya Pradesh show over 1.18 lakh excess deaths in 2021, which is over 26 times the official COVID-19 death toll for the same period. The other States too show excess mortality in 2021 even assuming full registration.
Moreover, single-month spikes in months like the one seen in May 2021 cannot be explained away by the gradual increase in registrations alone. Nor are these deaths that occurred months before. From Andhra Pradesh, for instance, we know that 98% of the deaths registered in May 2021 took place within the previous 30 days.
An imperfect resource
Taken together, these potential sources of error indicate that while the total number of deaths in India during the pandemic reported using Civil Registration System data might be an underestimate, a part of the increase in mortality can be explained by improved registration. However, the scale of excess mortality during the second wave in particular is undeniable, especially compared to the low number of COVID-19 deaths officially recorded in the same period.
The Civil Registration System is an imperfect system, yet one that journalists are having to turn to in the face of the continuing refusal of the Union government to engage with the issue of the true extent of mortality. Acknowledging the gaps in the system will help journalists, but does not minimise the magnitude of the crisis.
Rukmini S. is an independent journalist