Modeling the Bias between Number of Confirmed Cases and Actual Number of Infections in COVID-19 Pandemic

The preprint is available for download in https://www.medrxiv.org/content/10.1101/2020.06.22.20137208v1

Authors: Xingang Chen, Institute for Theory and Computation, Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138, USA; Dhiraj Kumar Hazra,The Institute of Mathematical Sciences, HBNI, CIT Campus, Chennai 600113, India

Information regarding the article

The actual number of COVID-19 infections is a key to understanding the transmission of the disease. It is also a critical factor for policy-making during the pandemic. However, the number of confirmed cases in viral tests is only a biased probe of this number. A general method of predicting the actual number of infections from the measured variables, such as the total tests, confirmed cases, hospitalizations and fatalities, is unavailable. In the article (https://www.medrxiv.org/content/10.1101/2020.06.22.20137208v1), we achieve this by developing a model that motivates a method of data analyses of these variables. By comparing with the historical data of the USA in the past few months, we find a simple formula relating these four variables. Among a few applications, we show how the model can be used to predict the number of actual infections and the optimal test volume. This repository is to provide the data and code we have used for our analysis.

Brief information about the supplied data

Here we provide the relevant data used in our analysis. Note that we have collected all the data from the website https://covidtracking.com. Thereafter we correct for the gliches in the data using the methods described in the Data collection section of the preprint. We bin the data using binning method described in Some data analyses details and plots section. Here we provide the unbinned data in the input_data folder. Unbinned data for each states are provided where first column is T/H_N and second column is p_+/H_N. We fit T/H_N as a function of p_+/H_N. Files do not have headers.

Brief description of the code

We also provide the code to obtain the best fit to the curve given a particular function. Here we supply the example of Tangent function. But the code can be used to fit other functional forms as well. It uses NLOPT (Steven G. Johnson, The NLopt nonlinear-optimization package, http://github.com/stevengj/nlopt) package for the optimization

In order to compile and run the code:

make clean

make all

make run

The code will read binned data from input_data folder (binned_data.txt). This file has a header. After optimization the code will print the best fit values of f0 and b and the asymptotic value of p+/H_N. The code will generate the logfiles (Log-Tan.txt) and best fit function (Bestfit-Tan.txt) within plot_data folder.

Requirements

1. The code needs Intel fortran compiler (ifort: https://software.intel.com/content/www/us/en/develop/tools/compilers/fortran-compilers.html) to compile and run.
2. The code also needs the NLOPT software for optimization. The user must download the inclide and library files from NLOPT website (https://nlopt.readthedocs.io/en/latest/) in the local directory.

• Developed and maintained by Dhiraj Kumar Hazra, IMSc, Chennai, India and Xingang Chen, ITC, Harvard-Smithsonian CfA, Cambridge, USA
• for bugreport and suggestions, please write to [email protected] and/or [email protected]