Background

For my project, I use a simple SEIR-like model I previously worked on to simulate cases of severe acute respiratory syndrome coronoavirus 2 (SARS-CoV2) in Georgia. State-level shelter-in-place orders began on March 16th where schools, work places and leisure areas were closed. At the time, I was subsequently interested in the impact of different social distancing strategies on infection transmission. The goal of my model was to address the following primary research question (from early May): how will the epidemic differ if current shelter-in-place orders are relaxed with a 50% return to work and no additional intervention versus a 100% return to work and either 1) additional measures to social distance at home; 2) isolation of 20% of pre-infectious cases identified from contact tracing or 3) isolation of 50% of pre-infectious cases identified from contact tracing?

Methods

We adapt code from a previously published age-structured deterministic susceptible-exposed-infected-removed (S-E-I-R) model using difference equations with no births and no deaths. The figure below details the structure and transitions of the model. Briefly, our model consists of six compartments, susceptibles (S), pre-infectious (E), pre-infectious and isolated (EI), infectious (I), infectious and hospitalized (H), and recovered (R). Each compartment is further stratified into 16 five-year age groups and the model tracks the number of individuals in S, E, EI, I, H and R compartments per age group at each time step. The subscript i denotes the age-group of the stratified compartment. The age-group stratification allows for heterogeneous mixing patterns between and across age groups and an understanding of infection and disease rates per age group.

Parameters

The table below details the parameters used to parameterized the model

Parameter	Description	Value	Source
beta	Transmission prob per contact} 0.035 Estimated[1]
alpha	Latent period	5.2 days	[6]
gamma	Recovery rate	7 days	[5]
delta	Hospitalization rate per time step	0.0075	Estimated[2]
Initials
Initial infections	Initial infections	1 per age group (16 total)	Determined
Initial population	Initial population	10.5 million	[9]
Interventions
tau	Prop. isolated per time step	20%/50%	Determined

Intervention scenarios

The table below details the intervention scenarios explored in our model

Scenario	Do Nothing	Indefinite Shelter in Place	Relax Shelter in Place & Some return to work	Relax Shelter in place & all return to work & home distance
Shelter-in-place start	Never	March 16th, 2020	March 16th, 2020	March 16th, 2020
Shelter-in-place relax	NA	Never	May 1st, 2020	May 1st, 2020
Home contacts	0% reduction	0% reduction	0% reduction	0% reduction/50%reduction
School contacts	0% reduction	100% reduction	Closed (100% reduction)	Closed (100% reduction)
Work contacts	0% reduction	90% reduction	90% reduction/50% reduction	90% reduction/0% reduction
Other contacts	0% reduction	90% reduction	90% reduction/50% reduction	90% reduction/50% reduction

Execute the analysis using docker

1. Download the Docker image:

docker pull cliu369/dockerhub:info550_project

2. Mount to the output directory and run:

docker run -v/output/path:/project -it cliu369/dockerhub:info550_project

3. This will generate the report report.html that will be created in your local working directory.

Execute the analysis using make

To execute the analysis, from the project folder you can run the following command:

Rscript -e "rmarkdown::render('4_info550_project.Rmd')"

This will create a file called info550_project.html output in your directory that contains the results.

GNU Make

You can also use the following make commands to install packages and execute the analysis:

make install

make report.html

Package install

To analyze the data you will need to install some R packages. The required packages can be installed using R commands (if not using make commands or docker)

installed_pkgs <- row.names(installed.packages())

pkgs <- c("ggplot2", "reshape2","kableExtra")
for(p in pkgs){
	if(!(p %in% installed_pkgs)){
		install.packages(p)
	}
}