A steady trickle-down from metro districts and improving epidemic-parameters characterized the increasing COVID-19 cases in India

Background By mid-September of 2020, the number of daily new infections in India crossed 95,000. We aimed to characterize the spatio-temporal shifts in the disease burden as the infections rose during the first wave of COVID-19. Methods We gathered the publicly available district-level (equivalent of counties) granular data for the 15 April to 31 August 2020 period. We used the epidemiological data from 186 districts with the highest case burden as of August 31, 559,566 active cases and 2,715,656 cumulative infections, and the governing epidemic parameters were estimated by fitting it to a susceptible-asymptomatic-infected-recovered-dead (SAIRD) model. The space-time trends in the case burden and epidemic parameters were analyzed. When the physical proximity of the districts did not explain the spreading patterns, we developed a metric for accessibility of the districts via air and train travel. The districts were categorized as large metro, metro, urban and sub-urban and the spatial shifts in case burden were analyzed. Results The center of the burden of the current-active infections which on May 15 was in the large metro districts with easy international access shifted continuously and smoothly towards districts which could be accessed by domestic airports and by trains. A linear trend-analysis showed a continuous improvement in the governing epidemic parameters consistently across the four categories of districts. The reproduction numbers improved from 1.77±0.58 on May 15 to 1.07± 0.13 on August 31 in large metro districts (p-Value of trend 0.0001053); and from 1.58±0.39 on May 15 to 0.94±0.11 on August 31 in sub-urban districts (p-Value of trend 0.0067). The recovery rate per infected person per day improved from 0.0581±0.009 on May 15 to 0.091±0.010 on August 31 in large metro districts (p-Value of trend 0.26\times10^-12); and from 0.059±0.011 on May 15 to 0.100±0.010 on August 31 in sub-urban districts (p-Value of trend 0.12\times10^-16). The death rate of symptomatic individuals which includes the case-fatality-rate as well as the time from symptoms to death, consistently decreased from 0.0025±0.0014 on May 15 to 0.0013±0.0003 on August 31 in large metro districts (p-Value of trend 0.0010); and from 0.0018±0.0008 on May 15 to 0.0014±0.0003 on August 31 in sub-urban districts (p-Value of trend 0.2789). Conclusions As the daily infections continued to rise at a national level, the ``center'' of the pandemic-burden shifted smoothly and predictably towards smaller sized districts in a clear hierarchical fashion of accessibility from an international travel perspective. This observed trend was meant to serve as an alert to re-organize healthcare resources towards remote districts. The geographical spreading patterns continue to be relevant as the second wave of infections began in March 2021 with a center in the mid-range districts.