Skip to content

Latest commit

 

History

History

CCF

README.md

covid

COVID-19 tool development

An open collaboration between the Cleveland Clinic and SAS Institute.

What this Code Does

This code takes a set of input parameters and uses them in infectious disease models (SIR & SEIR). Model output is used to calculate useful metrics for each day of an epidemic, such as the number of hospitalizations.

Documentation

In addition to the information shared within this readme and the commenting within the code, you can also review the documentation on the implementation of the models in SAS.

For a good understanding of the SIR and SEIR model approaches, we recommend this very well written blog post from the folks at TRIPLEBYTE.

User Interface: COVID-19 Epidemiological Scenario Analysis

Visually interact with this code to investigate your scenarios using a user interface by getting access here: COVID-19 Epidemiological Scenario Analysis

Getting Started with the Code

  • PREPARE
    • Download COVID_19.sas and fit_input.csv to your SAS environment
      • Also, download the run_scenario.csv file for an example submission file to run many scenarios in batch
  • SETUP
    • Edit line 10 to a local directory you want to save datasets with model output and scenario information to
    • If you do not have SAS/ETS then edit line 16 to 'NO'
      • This option causes the SIR/SEIR models to run with a SAS Data Step version only
      • If you are unsure then you can run PROC PRODUCT_STATUS; run; in SAS and view the log for this information
    • If you have the latest analytical release of SAS, 15.1, then set line 17 to YES
      • This option swaps out PROC MODEL for PROC TMODEL
      • If you are unsure then you can run PROC PRODUCT_STATUS; run; in SAS and view the log for this information
  • RUN
    • Make calls to the macro %EasyRun. Example scenarios are at the end of the file.
    • Submit many scenarios in batch by using an input file. An example file, run_scenarios.csv, is provided. Each row of this file will feed individual calls to the %EasyRun macro.
  • REVIEW
    • All model output for each call to %EasyRun saves in the dataset STORE.MODEL_FINAL
    • All of the parameters that lead to the results in STORE.MODEL_FINAL save in STORE.SCENARIOS, and all inputs to the macro also save to STORE.INPUTS. The variable SCENARIOINDEX links these files.
  • ADJUST inputs to your population
    • Change the input parameters to match the population you are working with. If their is a wide range of scenarios you want to run then use the run_scenarios.csv method to easily submit all the combinations.
  • CREATE your data feed for the fitting method
    • The example code reads input cases from fit_input.csv.
    • This file can be replaced with a similar .csv or the section of the code can be updated to read your data feed for case counts. Search COVID_19.sas for /* START: STORE.FIT_INPUT READ */ and you will find the section that does this import which has helpful commenting on this adaptation.
    • An example of tying this section to a data feed from a US State website can be found in /examples/fit_input_ohio.sas

Fitting to Actual Case Counts

The /examples/ folder contains two example approaches for using PROC TMODEL to fit a curve to actual case counts in a data set stored as STORE.FIT_INPUT:

  • Ohio_Data_Intervene_Point.sas - Fit a curve with a sudden, point change in social distancing on a single, estimated date
  • Ohio_Data_Intervene_Smooth.sas - Fit a curve with a smooth, gradual change in social distancing spread out over an estimated window of dates

Explore the Inputs, Outputs and Details in the Wiki

Explore the wiki

Development Notes

  • The current locked version of the project is in COVID_19.sas.
  • Progress towards the next locked version is in the /progress folder
  • the COVID_19.sas file is built from modular parts in /build/parts into the /build/public folder by /build/build.py and then copied here

Example Scenario Visuals

With the option plots=yes, the COVID_19.sas program will create diagnostic visuals for each scenario. Some example of these visuals follow. The output data is also available in SAS Visual Analytics with user interface to drive the running of scenarios is in the works.

All Approaches Fitting Approaches

Preferred Reference for Citation

Cleveland Clinic and SAS COVID-19 Development Team. Developer Documentation [Internet]. 2020. Available from: https://github.com/sassoftware/covid-19-sas