Preliminary Analysis of Excess Mortality in India During the COVID-19 Pandemic

ABSTRACT. We studied all-cause mortality during the COVID-19 pandemic in 19 Indian states (population 1.27 billion). Excess mortality was calculated by comparison with years 2015 to 2019. The known COVID-19 deaths reported for a state were assumed to be accurate, unless excess mortality data suggested a higher toll. Data from one state were excluded due to anomalies. In several regions, fewer deaths were reported in 2020 than expected. Areas in Andhra Pradesh, Delhi, Haryana, Karnataka, Madhya Pradesh, Tamil Nadu, and West Bengal saw spikes in mortality in Spring 2021. The pandemic-related mortality through August 31, 2021, in 18 Indian states was estimated to be 198.7 per 100,000 population (range 146.1–263.8 per 100,000). If these rates apply nationally, then 2.69 million people (range 1.98 to 3.57 million) may have perished in India as a result of the pandemic by August 31, 2021.


INTRODUCTION
Because SARS CoV-2 testing and death registrations are often incomplete, the full impact of the COVID-19 pandemic is unknown in many regions. 1 Excess mortality, defined as the increase in all-cause mortality relative to the expected mortality, is assumed to result from infection with the SARS CoV-2 virus, or from the indirect effects of health system overload or social responses to the pandemic. 2, 3 We sought to integrate published mortality figures from Indian regions to estimate the national pandemic impact.

METHODS
We used mortality figures published by regional governments and Indian journalists, often from Right-to-Information (RTI) requests (Supplemental References and Supplemental  Tables 1 and 2). Most data were archived by the Local Mortality project 2,4 or the Development Data Laboratory. 5 These were supplemented by government hospital data, funeral counts, and handwritten death registers. The Methods are detailed further in the Supplemental Appendix and were approved by the Virginia Commonwealth University Office of Research Subjects Protection.
We acquired mortality data from 19 states ( Table 1, Supplemental Table 1). Some smaller states were excluded because relevant mortality data are unavailable. The baseline Uttar Pradesh data contained districts with zero deaths for numerous months and were therefore excluded from the top-line model.
Mortality attributed to COVID-19 has been tabulated by the Johns Hopkins University Center for Systems Science and Engineering (CSSE, Supplemental Table 3). 6 The CSSE obtains data from governmental health authorities 7 and links to the webpage for the Health Ministry of India. 8 Indian health authorities were reportedly not attributing deaths to COVID-19 if the patient had comorbidities or SARS-CoV-2 tests were not performed. 9 We assumed reported COVID-19 mortality reflected the annual pandemic-related mortality in a state, unless excess mortality suggested a higher toll.
For some states, several data sources were available, and we presented the median, lowest, and highest estimates of per capita excess mortality. We calculated regional excess mortality by comparing the mortality for 2020 or 2021 with the value expected based on linear regression over the years 2015 to 2019. 2 The Indian government has estimated registration completeness by dividing the death registrations by the state mortality from the Sample Registration System, an annual survey (Supplemental Table 2). 10 We adjusted for registration completeness as outlined in the appendix.
The national mortality rate was estimated by summing the pandemic-related deaths for the states analyzed and dividing by the population. Indian state populations were taken from the Johns Hopkins dataset. 6 Raw data are presented in Supplemental Table 2.
To graph mortality timing, mortality rates (deaths divided by population) were calculated for 14 states with monthly data from January 2019 through May 2021. After May, per capita mortality was calculated from five states.

RESULTS
We studied 19 states, with a combined population of 1.27 billion ( Table 1, Supplemental Table 1). Reported per capita COVID-19 mortality (per 100,000) was low in the 19 states studied: 10.9 in 2020 and 20.4 in 2021 as of August 31, 2021 (Supplemental Table 3).
In 2020, there was a slight increase in all-cause mortality from August through October compared with 2019 ( Figure 1). The most prominent rise in all-cause mortality was evident by April and May of 2021 ( Figure 1). After May 2021, available data suggest that mortality registrations continued to rise through June but began to wane thereafter ( Figure 1).
For six states, the excess mortality based on registered deaths was actually negative for 2020 (Supplemental Table 1). At the other extreme, the per capita excess mortality (per 100,000) in 2020 was 137.0 in Mumbai (in Maharashtra) and 121.5 in Andhra Pradesh (Supplemental Table 1 The estimated COVID-19 mortality can also be expressed as a fraction of the baseline mortality, taken from the 2019 national vital statistics reports (Supplemental Table 4). The estimated COVID-19 mortality represented an increase over the baseline annual mortality of 10.70% (range 7.91-13.21%) in 2020 and 22.40% (range 16.44-30.73%) in 2021 (as of August 31). Pandemic-related deaths in the final 4 months of 2021 will increase these values. If the data from Uttar Pradesh are included, then the estimated per capita pandemic-related mortality (per 100,000) was 165.3 through August 31, 2021 (range 122.8-217.9, Table 1 and in Supplemental Table 4), corresponding to a mortality from COVID-19 of 2.24 million people (range 1.66-2.95 million) during the pandemic in India.

DISCUSSION
This analysis of excess mortality found that 2.69 million people (range 1.98-3.57 million) may have perished from COVID-19 in India as of August 31, 2021.
Data from Uttar Pradesh contained anomalies and had an estimated excess mortality lower than other regions, in part because Uttar Pradesh data from after April 2021 were unavailable. However, if the Uttar Pradesh data are included, the estimated pandemic-related mortality in India through August 31, 2021, was 2.24 million (range 1.66-2.95 million).
This mortality level is well above the reported COVID-19 mortality of 438,560 in India as of August 31, 2021. 6 The Institute for Health Metrics and Evaluation (IHME) at the University of Washington estimated that excess mortality in India was 1.23 million persons on August 31, 2021. 11,12 COVID-19 patients who likely would have died during the study period due to their comorbidities even if they had not been infected should not result in excess deaths.
Our analysis used the COVID-19 mortality reported by Johns Hopkins University, unless the excess mortality data suggested a higher pandemic-related toll. Our method of determining the expected baseline by making a projection using linear regression 1 year (for 2020) or 2 years (for 2021) into the future was conservative. Our definition of the baseline by linear regression was used previously in the World Mortality Dataset. 2 Other groups have estimated Indian pandemic-related mortality to be 4.9 million as of April 2021 based on a household survey 13 ; 2.2 million by late May 2021 based on well-defined populations 14 ; and, by June 2021, to be 3.2 million 15 to 3.4 million 13 based on excess mortality, 2.7 million based on a governmental hospital informatics system, 15 3.1 to 3.4 million based on a telephone survey, 15 and 4.0 million based on seroprevalence. 13 Our analysis has limitations. The data are incomplete for some regions and times, which might lead to underestimation of mortality. Death registrations may be delayed. Data from regional government websites, central government compilations, and RTI requests are not in complete agreement (Supplemental Table 2). Early data may contain errors. Mortality can be underestimated because of incomplete registration, health system underreporting, and data management errors. Health system overload, delays in patients seeking unrelated healthcare, 2,3,16,17 and social changes such as lockdowns may produce excess mortality. Excess mortality might also be observed due to other diseases, war, or environmental factors.