UVA-EpiHiper submission for Disparities Round Phase 1: abstract

model-abstract/UVA-EpiHiper/2024-06-25-UVA-EpiHiper.md

@@ -0,0 +1,48 @@
+## Summary of results given model assumptions? 
+In EpiHiper we model racial/ethnic heterogeneities explicity at individual level in demographics and activities. We also calibrate R/E specific transmissibility and R/E specific infection fatality rate using serology, cases, and deaths data sets.
+
+For California, we find that both target deaths and target cases are well captured for white, asian, latino, and overall in our projections, but are under-projected for black and other. In our projections deaths and cases decrease slower after peak than in the target.
+
+For North Carolina, we find that target deaths are well captured for white, asian, other, and overall in our projections, but are slightly over-projected for black. Target cases are well captured for white and other in our projections, but are slightly over-projected for overall and significantly over-projected for black and asian. The peak timing is also well captured by our projections.
+
+## Distribution of susceptibility at the start of the projection period? 
+Susceptibility of each racial/ethnic group at the start of the projection period depends on race/ethnicity specific prior infections, which are derived from serology data and imputed R/E specific cases data.
+
+## Which disease datasets (serology, cases, cases with imputed race/ethnicity information provided by coordination team, hospitalizations, deaths) were used for calibration?
+We use serology, imputed race/ethnicity specific cases, and race/ethnicity specific deaths data sets provided by coordination team for calibration.
+
+## How were the transmission (P(infection)) and severity (P(death|infection) risks estimated across racial/ethnic groups? 
+The transmission risks across racial/ethnic groups are partly modeled in our contact networks, as results of R/E specific demographic distributions and activity/mobility patterns, and also estimated in our calibration of R/E specific transmissibility parameters. The severity risks across racial/ethnic groups are estimated through calibration of R/E specific IFR using serology and deaths data sets.
+
+## How was the suppression of deaths handled in calibration? 
+We added "value" and "min_suppressed" columns in the target data to get deaths for calibration.
+
+## Details about calibration of race/ethnicity showing zero deaths throughout the calibration period (e.g., Others in CA and Asian in NC)? 
+We used "value + min_suppressed" from the target data and did not have any race/ethnicity group with zero deaths.
+
+## How was contact mixing between racial/ethnic groups characterized? Was a contact matrix (either population-level or setting-specific) used? If so, describe the process used to apply the contact matrix.)
+We do not use an explicit contact matrix. Instead, the contact mixing between R/E groups comes from our synthetic population and network modeling, where we assign R/E to each individual based on R/E specific distributions and assign activities and activity locations to each individual based on their R/E (and other demographic attributes).
+
+## Was mobility data used throughout the calibration/projection period? 
+Mobility data was used to derive a relative adjustment factor for scaling the transmissibility throughout the projection period.
+
+## Besides transmission, severity risk and contact matrices, did any other parameters vary by race/ethnicity (e.g., case reporting, NPI compliance)? 
+Case ascertainment rate varied by race/ethnicity.
+
+## How was the introduction of more transmissible variants modeled? 
+We assumed that when the new variant emerged, 0.5% of the infections were new variant; and calibrated the emerging time towards 66% prevalence of new variant in mid-April 2021.
+
+## How were NPIs implemented? 
+We implemented (1) school closure with 65% compliance which removes in-school contacts of those who comply, (2) mask wearing outside of the household with 85% compliance which reduces transmission risk of compliant people by 60%, and (3) mobility changes in the projection period as a scaling on the transmissibility parameter.
+
+## What was assumed about immunity/protection following infection? 
+Natural immunity (from infection) provides 100% protection against infection. Vaccinal immunity (from vaccination) provides no protection against infection but reduces risk of symptomatic infection.
+
+## How was vaccination implemented? What was assumed about immunity/protection following vaccination? 
+To implement vaccination, we move a vaccinated node to a vaccinated state V1 after one dose or V2 after two doses, which has the same susceptibility to infection, but if infected has reduced probability (50% reduction with one dose, 95% reduction with two doses) of becoming symtomatic. This subsequently reduces risk of severe outcomes including IFR.
+
+## Was age modeled in addition to race/ethnicity? Was there consideration for differences in the age structure across racial/ethnic populations? 
+We model age (and race/ethnicity) explicitly at individual level in our agent-based model. The joint distribution of age and race/ethnicity is reflected in our synthetic population and heterogeneities between these demographic groups are reflected in our contact network. Our disease model has age and race/ethnicity stratifications.
+
+## Was there importation of cases (e.g. from another external state)? 
+We do not consider importations.