Predictions, role of interventions and effects of a historic national lockdown in India's response to the COVID-19 pandemic: data science call to arms

Debashree Ray; Maxwell Salvatore; Rupam Bhattacharyya; Lili Wang; Shariq Mohammed; Soumik Purkayastha; Aritra Halder; Alexander Rix; Daniel Barker; Michael Kleinsasser; Yiwang Zhou; Peter Song; Debraj Bose; Mousumi Banerjee; Veerabhadran Baladandayuthapani; Parikshit Ghosh; Bhramar Mukherjee

We made projections of the cumulative number of cases over a time horizon to assess the shortterm impact of lockdown as well as the long-term impact of lockdown and post-lockdown activities. For the short-term forecast on April 30, we assumed lockdown is implemented until April 14 with either a 1-or a 2-week delay in people's adherence/compliance to lockdown restrictions. We compared these projections with two hypothetical scenarios: (A) no nonpharmaceutical intervention (i.e., a constant disease transmission rate over time since the first case was reported in India), (B) a moderate intervention with social distancing and travel bans only (i.e., a decreased transmission rate compared to no intervention). For the no intervention and the moderate intervention scenarios, we chose the transmission rate and the removal rate such that the means for the prior distribution of the basic reproductive number R0 (the expected number of cases generated by one infected person assuming that the whole population is susceptible) are 2.0 and 1.5 respectively [the change in R0 was created based on what we saw in Wuhan 16 ]. The value of 2.0 was estimated based on the early phase data in India. For the current scenario of lockdown, our chosen mean for R0 prior starts with 2.0 during the period of no intervention, drops to 1.5 during the period of moderate intervention, and further drops to 0.8 during the 21-day lockdown period, and moves back up to 1.5 after the lockdown ends as described in Figure 3 (assuming a gradual, moderate resumption of daily activities).