metadata-PSI-PROF.txt

team_name: PSI
model_name: PROF
model_abbr: PSI-PROF
model_contributors: Ben-Nun M (Predictive Science) <mbennun@predsci.com>, Turtle J (Predictive Science) <jturtle@predsci.com>, Riley P (Predictive Science) <pete@predsci.com>
website_url: https://predsci.com/usa-flu-hosp/
model_version: "1.0"
methods: For each state our scenario projections are generated using an age/risk stratified mechanistic model with S[Sv]E[Ev]I[Iv]HR compartments, where 'v' subscripts indicate vaccinated. The model is calibrated to the 2023-24 data, and extrapolated through the 2024-25 season. 
license: mit
modeling_NPI: "Not applicable"
compliance_NPI: "Not applicable"
contact_tracing: "Not applicable"
testing: "Not applicable"
vaccine_efficacy_transmission: "One half VE against hospitalization"
vaccine_efficacy_delay: "Not applicable"
vaccine_hesitancy: "Not applicable"
vaccine_immunity_duration: "Waning against hospitalization: 270 days (mean). Immunity escape: scenario specified"
natural_immunity_duration: "Calibrated to each state"
case_fatality_rate: "Not applicable"
infection_fatality_rate: "Not applicable, but IHR is calibrated to each state"
asymptomatics: "Not applicable"
age_groups: "0-17, 18-64 low risk, 18-64 high risk, 65+"
importations: We assumed 10 importations per day per state seeded randomly.
confidence_interval_method:  "Not applicable"
calibration:  "Not applicable"
spatial_structure: "Not applicable"
methods_long: "We use the population and vaccine time series provided by the Scenario Hub and 
  \ assume that the infection-susceptibility of vaccinated individuals accounts for one half \
  \ of vaccine effectiveness against severe disease. The model is calibrated to 2023-24 season adult \
  \ and pediatric COVID hospital admissions, employing a flexible time-dependent transmission term. \
  \ Calibration assumes that last season fell somewhere between scenarios E and F. \
  \ Additionally, calibration includes a preference for periodic/quasi-periodic trajectories. \
  \ Calibration is a two-stage process.  The first stage uses a traditional optimizer to find a \
  \ centroid for the hyper-cube used in the second stage. \
  \ The national projection profiles are calculated as \
  \ the sum of states using semi-randomly ordered state level profiles. "\