team_name: Johns Hopkins ID Dynamics COVID-19 Working Group
model_name: CovidScenarioPipeline
model_abbr: JHU_IDD-CovidSP
model_contributors: Joseph C. Lemaitre (EPFL),  Juan Dent Hulse (Johns Hopkins Infectious Disease Dynamics), Kyra H. Grantz (Johns
  Hopkins Infectious Disease Dynamics),  Joshua Kaminsky (Johns Hopkins Infectious
  Disease Dynamics),  Stephen A. Lauer (Johns Hopkins Infectious Disease Dynamics),  Elizabeth
  C. Lee (Johns Hopkins Infectious Disease Dynamics)<elizabeth.c.lee@jhu.edu>,  Justin
  Lessler (Johns Hopkins Infectious Disease Dynamics)<justin@jhu.edu>,  Hannah R.
  Meredith (Johns Hopkins Infectious Disease Dynamics),  Javier Perez-Saez (Johns
  Hopkins Infectious Disease Dynamics),  Shaun A. Truelove (Johns Hopkins Infectious
  Disease Dynamics), Claire Smith (Johns Hopkins Infectious Disease Dynamics), Lindsay T. Keegan (University of Utah),  Kathryn Kaminsky,  Sam
  Shah,  Josh Wills, Pierre-Yves Aquilanti (Amazon Web Services),  Karthik Raman (Amazon
  Web Services),  Arun Subramaniyan (Amazon Web Services),  Greg Thursam (Amazon Web
  Services),  Anh Tran (Amazon Web Services), Claire Smith (Johns Hopkins Infectious
  Disease Dynamics)
website_url: https://github.com/HopkinsIDD/COVIDScenarioPipeline
license: mit
team_model_designation: primary
methods: County-level metapopulation model with commuting and stochastic SEIR disease dynamics
  with social-distancing indicators.
team_funding: State of California and US Government
data_inputs: JHU CSSE (confirmed cases; reported fatalities), US Census (population; mobility)
citation: https://www.medrxiv.org/content/10.1101/2020.06.11.20127894v1
methods_long: "This is a county-level metapopulation model with stochastic SEIR disease\
  \  dynamics. This model incorporates uncertainty in epidemiological parameters \
  \ and the effectiveness of state-wide intervention policies on social  distancing\
  \ and stay-at-home orders. Infections are seeded stochastically  with roughly 10\
  \ times the number of confirmed cases in the first five days  of a county epidemic\
  \ and R0 and the duration of the infectious period were  drawn stochastically for\
  \ each county and simulation. Disease follows SEIR dynamics and county commuting\
  \ data modulates the spread of disease within counties. \nStay-at-home orders and \
  \ subsequent social distancing interventions are updated at the state-level according\
  \ to recent policy documents. Ongoing interventions
  continued through \
  \ the remainder of the simulation period. Intervention effects are inferred\
  \ where  possible.\nTo model deaths and hospitalizations in the population, we incorporated\
  \  realistic but fixed time delays from infection to symptom onset to hospitalization to \
  \ ICU to ventilator use to death and used age-specific demographics to perform \
  \ a county-level age standardization of health outcome risk. Currently, we  assume\
  \ a 0.5% infection fatality rate and a 3.5:1 ratio of incident hospitalizations\
  \ to deaths. \nFor each county, we estimate the seeding time and amount, the baseline\
  \  reproductive number, the case confirmation to infection ratio, and the  effectiveness\
  \ of interventions (from closing to phased reopenings). We use an  MCMC-like inference\
  \ procedure that calibrates model outputs to weekly county  aggregations of incident\
  \ confirmed cases and deaths as reported by USA Facts. For the inference of the\
  \ baseline reproductive number and the case  confirmation to infection ratio, hierarchical\
  \ grouping terms enable sharing  of strength among counties in the same state.\n\
  Our procedure forecasts incident and cumulative deaths by incorporating the last\
  \ reported data point (pre-forecast date) from USA Facts.\nThe estimates reported\
  \ by this model incorporate uncertainty in both epidemiological parameters (e.g.,\
  \ R0, infectious period, time delays to health outcomes) and the effectiveness\
  \ of state-wide intervention policies  on social distancing, stay-at-home orders,\
  \ and phased reopenings. We submit a single calibration to forecast deaths, cases,\
  \ and hospitalizations."
