Direct and indirect effects of the covid-19 pandemic and response in South Asia

Model parameters

Model parameters were populated using a combination of model calibration for a small subset consisting of four key parameters and choice of plausible values for the remaining ones (Supplementary Table 2). The four parameters chosen for calibration include the daily average number of exposure events involving symptomatic individuals, the probability of transmission by symptomatic cases, the daily hospitalization rate of symptomatic cases and the daily case fatality rate.

Weekly totals of case fatalities, as recorded on the Covid-19 page of the Our World in Data (OWID) website (15) were the basis for the calibration, using the mean squared error

i

where F model (θ) is the total number of deaths simulated

for the ith week using the quadruple of parameters θ for simulation, i=1,…,K, K=14 is the number of weekly totals between June 1st and August 31st, 2020 (the time period

i

used for calibration) and F obs is the corresponding observed

number of deaths recorded. Calibrated values for the four key parameters are then chosen to minimize the MSE,

namely where minimization was carried out using sequential Bayesian optimization based on the Expected Improvement criterion proposed by Jones et al (16).

© UNICEF/India/UNI342622/Panjwani/2020