UNCC SMH Round #7 submission

data-processed/UNCC-hierbin/metadata-UNCC-hierbin.txt

@@ -0,0 +1,24 @@
+team_name: UNCC
+model_name: HierBin
+model_abbr: UNCC-hierbin
+model_version: 2021-07-13
+model_contributors: Shi Chen (UNC Charlotte Department of Public Health Sciences & School of Data Science) <schen56@uncc.edu>, Rajib Paul (UNC Charlotte Department of Public Health Sciences and School of Data Science) <rajib.paul@uncc.edu>, Daniel Janies (UNC Charlotte Department of Bioinformatics and Genomics) <djanies@uncc.edu>, Jean-Claude Thill (UNC Charlotte Department of Geography and Earth Sciences and School of Data Science) <jean-claude.thill@uncc.edu>
+website_url: NA
+license: cc-by-4.0
+methods: Because of the complexity of socio-epidemiological system of COVID-19, we developed a statistical model to track and predict the cumulative case in 2021, assuming it follow certain asymptotic curves. We then projected cumulative case based on the developed model which served as the baseline scenario. The difference of daily cumulative case would be the baseline incident case. Vaccination was modeled as a time-dependent function to reduce incident case. Hospitalization was modeled as a binomial process of incident case and no overload of healthcare system was considered in 2021. Hospitalization rate was estimated as a function of vaccination rate. We assumed that only severe condition patients required hospitalization, and vaccination would reduce the prognosis from non-severe to severe. Similarly, death was modeled as a binomial process of hospitalization, assuming all deaths occurred in hospitalized patients in severe condition. Deaths in long-term care facility were not considered. Death rate was also another function of vaccination rate. To model B1617 (delta variant), we assumed an initial prevalence of 40-45% across all states as of the end of June (though this can be relaxed to reflect more specific state-level variation in prevalence of B1617). We modeled B1617 as a exclusively competitive strain, where it would take over the entire COVID-19 strain "population" anywhere in 10-30 days (high variant) or 20-40 days (low variant), respectively. Increased transmissibility is then modeled as a time-dependent function of the invasion of B1617 strain. The maximum increase would be capped at 20% (low transmissibility) or 60% (high transmissibility) when B1617 was the dominant strain. Otherwise, increased transmissibility would be modeled as a mixture of current strain (no change) and B1617.
+contact_tracing: NA
+testing: NA
+vaccine_efficacy_transmission: Efficacy of vaccine to reduce transmission was modeled as a state- and time-dependent function based on existing data in 2021.
+vaccine_efficacy_delay: NA
+vaccine_hesitancy: Vaccine hesitancy was implicitly modeled as an asympototic curve of vaccination rate in each state.
+vaccine_immunity_duration: For this project we assumed no leaking immunity from vaccine, i.e., much longer than 6mo.
+natural_immunity_duration: For this project we assumed no leaking immunity from previous infection, i.e., much longer than 6mo.
+case_fatality_rate: CFR was estimated as a function of time (implicitly with vaccination rate) in each state with data in 2021.
+infection_fatality_rate: NA
+asymptomatics: NA
+age_groups: All age groups.
+importations: NA for current version.
+confidence_interval_method: Several assumptions were made to address uncertainty, e.g., system stochasticity from random noise, state-level variation in vaccination rate, state-level variation in lifting NPI mandate, etc.
+calibration: 2021 cumulative case data by state were used to model and calibrate the initial model.
+spatial_structure: By state.
+data_inputs: JHU case and death data, HHS hospitalization data, NYT vaccination data.