ELiXSIR is a code to simulate COVID19 infection. I have developed this code within the INDSCI-SIM collaboration. It solves 9 compartmental extended SEIR model with age stratification and migration. The code is publicly available at https://gitlab.com/dhirajhazra/eSIR_INDIA
Here is a schematic diagram of the connections between the compartments of the INDSCI-SIM model. The model largely follows the one discussed in http://covid-measures.stanford.edu/.
Here the differential contacts between age groups are used. Several types of lockdown mechanisms can be tested to get a guideline for policies. Effects of migration can be explored with Gravity algorithm and with a generic structure.
ELiXSIR can be easily combined with samplers for parameter estimation. In the INDSCI-SIM analysis, ELiXSIR is integrated with CosmoChord.
INDSCI-SIM preprint is available here: https://www.medrxiv.org/content/10.1101/2021.06.02.21258203v1
Given a set of parameters, system is set up for ELiXSIR. In this set-up the population, number of age groups and fraction of population in these age groups are initialized. The fraction and rates of transition between compartments that are runtime constants, are fixed. Lockdown dates are mentioned based on which the code switches from lockdown to unlock modes switching contact matrices.
After the initial set up, the priors, initial starting value of all parameter and widths are provided. Daily data of reported infection and deaths are supplied to the code with the dates.
The CosmoChord integrated with ELiXSIR is then run in several processors (MPI) in the cluster. The samples within the parameter volume are drawn and sent to ELiXSIR for the time-series. Daily infection and deaths are computed from the compartments. Bias evolution is generated according to the parametric form. After scaling the theoretically obtained daily infection numbers by the bias, the prediction of daily infected is then compared with the reported daily infection. Predicted daily deaths are directly compared with the daily deaths data.
After termination of the sampling, getdist is used to generate posterior distributions. The samples are then supplied to ELiXSIR directly to generate the bounds on the timeseries.
Here is the Delhi COVID19 data analysis with ELiXSIR and CosmoChord.
(a) Fit to the reported infection (b) and deaths.
(c) Prediction of actual infection.
(d) Bias factor between actual and reported infections.
(e) Evolution of infection fatality ratio.
(f) Our prediction of effective reproduction number R(t) against an independent estimation.