Making sense of excess mortality
Open access data on excess mortality across 132 districts, 8 states
There is consensus that reported COVID-19 mortality across India are the tip of a rather grim iceberg. However, the degree of the undercounting is far from settled. The degree of underestimation also potentially varies drastically across cities, urban and rural India, and state administrations. Vital statistics data maintained by districts are sporadically becoming available.
Data journalists across news-houses and RTI activists have done incredible work uncovering crucial data on registered deaths for states and districts across 2019–2021 over the past few weeks. Development Data Lab collaborated with a pioneering group of data journalists and researchers to bring these all-cause data to the public, fully linked to standard district identifiers, aggregated to both state and district levels. Specifically, we are very grateful to Rukmini S, Anurabh Saikia (Scroll), Vignesh Radhakrishnan and Srinivasan Ramani (The Hindu), Mariyam Alavi (NDTV), Saurav Das (RTI Activist, Article 14), Dhanya Rajendran (TheNewsMinute) and Chinmay Tumbe (IIM-A) for contributing district and state-level mortality datasets toward this process.
The data can be found on our COVID open access platform, where they can be directly linked to the wide range of district-level data that already exist on the platform — daily reported COVID infections and deaths, vaccination data, demographic data, health infrastructure and hospital capacity data, and so on.
These data have just been uncovered in the past weeks, so work remains to analyze them thoroughly. The graphs below show an early analysis of excess mortality in 2020 and 2021 compared to previous years. It is immediately clear that the excess mortality during the second wave (April-May 2021) was substantially higher than that of the first wave of the pandemic (September 2020) across all states, and that excess mortality has been far higher than reported COVID deaths anywhere.
There are some caveats. India’s data collection systems have been under extreme stress during the crisis, and we don’t yet know whether deaths are missing, erroneously coded, or backlogged. It is possible that some of the spiking deaths in May 2021 could reflect a backlog from earlier months in the second wave. Several states also show a fall in deaths in April 2020 — we don’t yet know whether this is reduced mortality due to the lockdown, or simply that deaths were not being accurately recorded in that period. Two additional caveats should be noted. Place and date of death registration may vary from actual place and date of death. Coverage of the death registration system may vary substantially across states.
The graph for AP shows exactly what we would expect to find in an excess mortality graph: mortality in 2020 tracks earlier years until rising substantially during the summer COVID wave. The second wave in 2021 is far, far worse, with more than three times as many excess deaths as in the first wave. But even a single month of the 2020 summer wave shows far more excess deaths than the 12,000 cumulative confirmed COVID deaths reported to date by the government.
The low average number and steady rise in reported deaths in 2018 suggests that there may have been a data issue in that year. 2020 and 2021 show a similar pattern to AP, but with a 2020 first wave that rises later in the summer.
The first pandemic wave in Karnataka peaked in September 2020, according to excess mortality statistics. In Karnataka, the first wave is more similar in size to the second wave than in previous states. We also see a big fall in excess deaths in April — we do not know whether this is actual reduced mortality or just reduced recording of deaths.
Total mortality in Kerala shows a much more seasonal pattern than in the other states, with deaths rising in the summer of all years. The first wave of the pandemic peaks in October, but excess deaths are much lower than in the other states. The second wave of the pandemic in 2021 remains severe.
Tamil Nadu follows the pattern of earlier states, with a long summer pandemic wave and a more severe 2021. Tamil Nadu’s excess deaths in 2020 are closer in scale to what was being reported for COVID deaths. Tamil Nadu reported about 4000 confirmed COVID deaths in August 2020; the figure suggests that excess deaths in August were around 24,000. This is 6x higher than the reported death count, but in fact a much smaller multiplier than in many other states.
The data for Uttar Pradesh appears to be unreliable. We were able to obtain only one year of pre-pandemic data, and in both 2020 and 2021, excess mortality falls substantially when COVID waves were at their most severe. It is possible that deaths were not correctly recorded and dated during some of these months of severe illness.
Madhya Pradesh shows little excess mortality in 2020, but a huge departure from the trend in 2021.
Details on the data available on the COVID platform in addition to the district-level all-cause mortality data can be found here.
— Sankalp Sharma, Aditi Bhowmick, Paul Novosad & Sam Asher