updating abstract and metadata with information about deaths submissi…

…on 5-31-2024

data-processed/CFA-Scenarios/2024-04-28-CFA-Scenarios-Abstract.md

@@ -50,14 +50,14 @@ Vaccination reduces the force of infection for a (partially) susceptible individ
 
 ## Describe the process used to set or calibrate disease severity, ie P(hosp given current infection) and P(death given current infection) details. What are the datasets used for calibration of the death targets?
 
-We infer the age group-specific infection hospitalization ratios separately for each state during the fitting procedure. Specifically, we fit the model to the weekly hospitalization data with an assumption that hospitalization lags infection by a week. Since we do not project death incidence, we did not calibrate the infection fatality ratios.
+We infer the age group-specific infection hospitalization ratios separately for each state during the fitting procedure. Specifically, we fit the model to the weekly hospitalization data with an assumption that hospitalization lags infection by a week. Deaths are calibrated at the national level by estimating the hospitalization fatality ratio for each age group based on data from March of 2022 through December 2023. Simultaneously, we infer a gamma delay distribution between hospitalizations and deaths.
 
 ## Seasonality implementation, e.g., whether seasonality varies by geography and which datasets are used to fit seasonal parameters
 
 Seasonality is inferred by state, with a peak in seasonal forcing typically in late December, and amplitudes ranging between 4% and 15%.  We assume forcing follows a sine function with a 1-year period. Seasonality is fit at the same time we fit all other model parameters, using hospitalization, serology, and variant prevalence data.
 
 ## Details about modeling of age-specific outcomes, including assumptions on age-specific parameters (e.g., susceptibility, infection hospitalization risk or fatality risk, VE)
-We assume age-structured mixing of the population, and infer infection-hospitalization ratios that are state- and age- specific.  We do not assume that there are inherent differences in susceptibility or immune competance based on age.
+We assume age-structured mixing of the population, and infer infection-hospitalization ratios that are state- and age- specific. We infer hospitalization fatality ratios that are age- but not state-specific. We do not assume that there are inherent differences in susceptibility or immune competance based on age.
 
 ## Details about modeling of high-risk individuals, e.g., susceptibility and infection hospitalization risk or infection fatality risk, VE
 High-risk individuals are not modeled explicitly.  Because we infer the IHR, it is not possible to tease apart high vs low risk IHR while keeping them coherent across fitting and projection periods. We use the fraction of high-risk individuals by state to determine vaccine uptake in the 65Boo scenarios.

data-processed/CFA-Scenarios/metadata-CFA-Scenarios.txt

@@ -16,12 +16,12 @@ vaccine_efficacy_delay: 0
 vaccine_hesitancy: Not applicable
 vaccine_immunity_duration: Wanes approximately logistically to half of the starting value in 6 months
 natural_immunity_duration: Wanes approximately logistically to half of the starting value in 6 months 
-case_fatality_rate: Not applicable
-infection_fatality_rate: Not applicable
+case_fatality_rate: Age-specific hospitalization fatality ratio is infered from national data: [0.023, 0.077, 0.193, 0.545]
+infection_fatality_rate: Varies by state depending on inferred IHR
 asymptomatics: Presence/absence of symptoms is not explicitly modeled
 age_groups: [0-17, 18-49, 50-64, 65+]
 importations: Represented as a small, transient new-variant infection hazard for 18- to 49-year-olds, over a one-month period.
 confidence_interval_method:  100 replicates are sampled from fitted posterior distributions
 calibration:  Hamiltonian Monte Carlo with No U-Turn sampling, using 1000 warm-up iterations and 1000 samples on 4 chains.  28 parameters are fit to the past two years of COVID-19 hospitalization, seroprevalence and strain prevalence data.
 spatial_structure: "Not applicable"
-methods_long: We created state-specific COVID-19 burden projections using a deterministic, modified SEIS model with additional stratifications and partial immunity. Each infection state is stratified by age, immune history, vaccination history, waning status (for Susceptibles), and infecting strain (for all other compartments). Immunity is determined by immunogenic events (infections and vaccinations) and time since the most recent event. Protection against infection is strain-specific: past infection with a more similar variant, or vaccination with a better-matched vaccine, provides a higher level of protection against the challenging strain.  External introductions of new variants are represented by introducing a small, transient new-variant infection hazard for 18- to 49-year-olds, over a one-month period. We assume that new variants are introduced on average three months apart, with intrinsic infectiousness values sampled from the posterior distribution for previous strains for each state.  State-based models are fit to 26 months of COVID-19 hospitalization, seroprevalence, and strain prevalence data. The U.S. model is an aggregate of the state models. Hospitalizations are estimated after simulation by applying locale- and age-specific infection-hospitalization ratios that are estimated during the fitting process. 
+methods_long: We created state-specific COVID-19 burden projections using a deterministic, modified SEIS model with additional stratifications and partial immunity. Each infection state is stratified by age, immune history, vaccination history, waning status (for Susceptibles), and infecting strain (for all other compartments). Immunity is determined by immunogenic events (infections and vaccinations) and time since the most recent event. Protection against infection is strain-specific: past infection with a more similar variant, or vaccination with a better-matched vaccine, provides a higher level of protection against the challenging strain.  External introductions of new variants are represented by introducing a small, transient new-variant infection hazard for 18- to 49-year-olds, over a one-month period. We assume that new variants are introduced on average three months apart, with intrinsic infectiousness values sampled from the posterior distribution for previous strains for each state.  State-based models are fit to 26 months of COVID-19 hospitalization, seroprevalence, and strain prevalence data. The U.S. model is an aggregate of the state models. Hospitalizations are estimated after simulation by applying locale- and age-specific infection-hospitalization ratios that are estimated during the fitting process.