Figure 3.

Socioeconomic attributes predicting differences in excess mortality across communities of Chennai

We illustrate differences in excess mortality for the first 4 weeks of the country-wide lockdown, the first wave, the second wave, and the total pandemic period, associated with a 1 SD increase in each measure of community-level socioeconomic deprivation (detailed in appendix 2 p 6). Additionally, we relate pandemic-associated mortality to the first two principal components derived across all 13 socioeconomic variables, alongside measures of Spearman's correlation (ρ) between each principal component and community attributes (additional details on the principal component analysis are presented in appendix 2 p 23). Time series of relative mortality across deciles of community socioeconomic status, as measured by the first principal component, are presented in appendix 2 (p 33). For all panels, lines denote 95% uncertainty intervals based on regression models fit across draws from the distribution of excess mortality measures, with (log-transformed) ratios of observed to expected deaths defined as the outcome variable (further methodological details are presented in appendix 2 p 4). Values are corrected for lagged reporting of deaths based on 2019 observations (appendix 2 p 27). Numerical estimates for all age groups corresponding to values presented in this figure are presented in appendix 2 (pp 21, 24), as well as estimates for individuals aged 50 years or older (pp 22, 25). Deaths are aggregated by PIN code areas; analyses include PIN code areas for which 1000 deaths or more were registered over the study period. Measures of excess mortality by PIN code area are presented in appendix 2 (p 32). PC=principal component.