This software supports the results in the paper: https://www.medrxiv.org/content/10.1101/2020.11.26.20152520v1

The algorithm simulates a COVID-19 epidemic for Austin, TX, and determines triggers to enact social distancing orders to avoid exceeding hospital capacity. Chosen triggers attempt to minimize the total number of days that a city is in strict levels.

All codes are tested on MacOS but should work on Windows and Linux.

• Download and unzip the code to a local path (e.g., .../COVID_Staged_Alert)
• Add both /COVID_Staged_Alert and /COVID_Staged_Alert/InterventionsMIP to your $PYTHONPATH
• The following packages are required:

• matplotlib
• pandas
• numpy
• scipy

• The installation should take less than a minute.

• Create new branches to test new features
• Create a pull request to merge with master

• Main module to launch the search
• Functions to execute the search and find optimized thresholds to enact lock-downs.
• Iterators for traing and testing
• Calendar generation (ad-hoc for Austin instance)

• Main module to run the search
• Functions to perform chance-constrained check and select the optimal policies w/o ACS setup

• Simulator engine
• Parallelizarion functions
• Calander utils class (SimCalendar)

• Class EpiSetup characterizes the simulation and recompute contact matrices as needed.

• Class intervention defines its properties and used in the simulator.
• Helper function to create multiple interventions

• Class MultiTierPolicy and MultiTierPolicy_ACS defines policy given the thresholds and will have functions to obtain corresponding tier given trigger statistics
• Functions to build a list of trial policies

• Construct the objective function for a simulation path

• Timing function
• Rounding functions

• Module summarizing the input of the simulator.
• Creates an instance of EpiSetup

• Least-squares fit

Related input files:

instances/austin/

• austin_real_icu_lsq.csv
• austin_real_hosp_lsq.csv
• austin_hos_ad_lsq.csv
• transmission_Final_lsq.csv
• setup_data_Final_lsq.json

instances/houston/

• houston_real_icu_lsq.csv
• houston_real_hosp_lsq.csv
• transmission_Final_lsq.csv
• setup_data_Final_lsq.json

• Donwsampling procedure (algorithm 1)
• Call downsampling.py

Related input files:

instances/austin/

• tiers5_ds_Final.json
• austin_test_IHT_ds.json

• script to obtain optimal trigger policy
• Call policy_search_functions.py

Related input files:

instances/austin/

• tiers5_opt_Final.json
• transmission_Final.csv

• script to obtain optimal trigger policy and ACS setup threshold
• Call policy_search_functions.py

instances/austin/

• tiers5_acs_Final.json
• transmission_Final_orange25_yellow75.csv

• Plot the figures used in the manuscript

• ICU, general wards heads-in-beds
• Total daily admission
• Stacked plots

• Obtain statistics based on simulation results in .p file

• generate pdf file by filling in statistics in the .tex template and compile

Run the bash file test.sh in Demo folder to obtain an optimal trigger and generate visualization and a report

• The .p file (data file) will be generated in /output
• The plots will be generated under /plots
• The report will be generated under /reporting/reports
• The run time should be within a minutes