Exploratory analysis of the suitability of data from the civil registration system for estimating excess mortality due to COVID-19 in Faridabad district of India

BACKGROUND & OBJECTIVES: Due to shortcomings in death registration and medical certification, the excess death approach is recommended for COVID-19 mortality burden estimation. In this study the data from the civil registration system (CRS) from one district in India was explored for its suitability in the estimation of excess deaths, both directly and indirectly attributable to COVID-19. METHODS: All deaths registered on the CRS portal at the selected registrar’s office of Faridabad district in Haryana between January 2016 and September 2021 were included. The deaths registered in 2020 and 2021 were compared to previous years (2016-2019), and excess mortality in both years was estimated by gender and age groups as the difference between the registered deaths and historical average month wise during 2016-2019 using three approaches – mean and 95 per cent confidence interval, FORECAST.ETS function in Microsoft Excel and linear regression. To assess the completeness of registration in the district, 150 deaths were sampled from crematoria and graveyards during 2020 and checked for registration in the CRS portal. Agreement in the cause of death (CoD) in CRS with the International Classification of Diseases-10 codes assigned for a subset of 585 deaths after verbal autopsy was calculated. RESULTS: A total of 7017 deaths were registered in 2020, whereas 6792 deaths were registered till 30 September 2021 which represent a 9 and 44 per cent increase, respectively, from the historical average for that period. The highest increase was seen in the age group >60 yr (19% in 2020 and 56% in 2021). All deaths identified in crematoria and graveyards in 2020 had been registered. Observed peaks of all-cause excess deaths corresponded temporally and in magnitude to infection surges in the district. All three approaches gave overlapping estimates of the ratio of excess mortality to reported COVID-19 deaths of 1.8-4 in 2020 and 10.9-13.9 in 2021. There was poor agreement (κ<0.4) between CoD in CRS and that assigned after physician review for most causes, except tuberculosis and injuries. INTERPRETATION & CONCLUSIONS: CRS data, despite the limitations, appeared to be appropriate for all-cause excess mortality estimation by age and sex but not by cause. There was an increase in death registration in 2020 and 2021 in the district.