Why there’s no agreement on India’s Covid death figures

As accurate death statistics by causes is hard to come by — around 70 per cent of deaths take place at homes even in normal times — attempts have been made to estimate the Covid deaths by identifying “excess” deaths over what would have been the number otherwise. There are conceptual issues involved since deaths due to starvation, returnee migrants miseries, those linked to the lockdown, unemployment, and other ailments not receiving medical help due to pressure on the system, are not due to Covid, but are part of excess deaths. Also, the Supreme Court’s decision to consider all deaths within three months of Covid infliction as Covid deaths resulted in a surge in reporting. Similarly, the lives saved due to the lockdown and measures adopted by people to protect themselves from the virus and those who would have died due to other causes had there been no Covid, would have to be taken on the positive side, increasing the discrepancy between Covid and excess death.

Notwithstanding this, the figures of “excess mortality” over and above what would have been the number without Covid can be computed without much hassle in the absence of detailed data on the causes of death. Understandably, this would be different from the numbers reported as Covid deaths by the civil registration system (CRS). The differences are not due to conceptual coverage alone. A large part of it is due to the inadequacy of the registration system, particularly in a period of medical and economic emergency. Understandably, independent journalists and health statisticians who had sourced registration data and fragmented ground-level information have raised this concern. All this has made experts within the country and in the concerned global institutions sceptical regarding the authenticity of the statistics of the Ministry of Health and Family Welfare resulting in the WHO placing India in the category for which the Covid death figures are to be generated through modelling and not based on the officially reported numbers.

The CRS used to be a poorly administered data collection system but the coverage has increased significantly, thanks to greater awareness and enhanced field level surveillance and follow-ups. There has been a jump in registrations starting from 2018. Such jumps are not infrequent, caused by procedural changes and backlog reporting.

Given this, the selection of the base years to be used for estimating “normal deaths” becomes extremely important. If one chooses to use, say the average deaths during 2015-2019, one will get a higher “excess death” estimate compared to using only 2018 and 2019 for the baseline comparison. This appears to have happened in the excess death estimation by independent journalists and researchers who had forced the CRS data out of state governments through RTI and other means. These data have also been used by many researchers to arrive at Covid mortality.

The Sample Registration System (SRS) was initiated in the Sixties as a continuous large-scale demographic survey to overcome the deficiencies of CRS. Its most recent bulletin, released in January 2022, provides estimates for 2019, and hence any increase in SRS death rates from Covid would be available sometime in 2023 or 2024 only. It is evident that a smaller increase in the number of deaths in 2020 compared to 2019, despite a larger population base and Covid, is due to an improvement in registration in 2019. It is still impossible to defend the official claims of 99.9 per cent registration in 2019-20 compared to 84.6 per cent in 2018. The WHO figure of 0.83 million of Covid deaths in 2020, however, may not be difficult to reconcile if one considers the total deaths reported by CRS to be 0.81 million and the fact that there were serious quality and coverage issues in the registration system in 2021. The pandemic and the strategy to combat its spread through lockdowns etc in 2020 are known to have affected all routine administrative activities of public agencies. The latest National Family Health Survey reveals that only 70 per cent of the deaths were registered by the households during 2019-21. And yet, the ministry has not made any attempt to reconsider the CRS estimate of 0.52 million Covid deaths till early May 2022 against the WHO figure.

In an otherwise rare instance, the ministry makes the claim that “modelling, more often than not, can lead to over-estimation and on few occasions, these estimates may stretch to the limits of absurdity”. This reminds one of a statement by late B S Minhas that international agencies often give income estimates for countries for which they do not have population figures. The limitations of such modelling exercises are well known. For instance, the NITI Aayog had predicted Covid infections to hit zero in July 2020 itself. There are, however, adequate grounds to question the information released through the CRS. In most states, the compensation claims for Covid deaths have been several times the officially recorded deaths.

Clearly, there is an urgent need to have robust data from local government agencies collated scientifically and also validated through external mechanisms like field surveys. Unfortunately, this is not happening. Data analysis is also constrained by the fact that most departments present data in dashboards with the underlying data inaccessible to researchers. Till then claims and counterclaims will remain.

This column first appeared in the print edition on May 11, 2022 under the title ‘Number wars’. Mohanan is chairperson of Kerala Statistical Commission and Kundu is senior fellow at the World Resources Institute, New Delhi