metadata-Karlen-pypm.txt

team_name: Karlen Working Group
model_name: python Population Modeller
model_abbr: Karlen-pypm
model_version: "1.0"
model_contributors: Dean Karlen (University of Victoria and TRIUMF) <karlen@uvic.ca>
website_url: https://pypm.github.io/home/
license: cc-by-4.0
methods: A network of populations connected via delay distributions calculated with
	 discrete-time difference equations. The general framework allows for calculation
	 of expectation values and for simulation of data that include stochastic variation.
modeling_NPI: A model for the case history for each state, consisting of relatively
	 long periods of constant behaviour. The observed growth rate within those periods
	 determine the transmission rate parameter corresponding to the NPI in force at each
	 time. For round 6, NPI is held constant, since model has fit case data well for past
	 8 weeks, despite general relaxation of measures.
compliance_NPI: Variance of transmission rate is included for the interval construction. The
		standard deviation of the transmission rates arise from the standard deviation of the
		estimator for current growth rate and from translation to growth rate to new variant.
contact_tracing: The removal delay time distribution for individuals from the circulating contagious
		 population is fixed through the entire history. The effect of changes to
		 contact tracing policy is absorbed in the inferred transmission rate.
testing: The testing fraction was adjusted for the period March-August in order to improve agreement
	 with hospitalization and death data as well as to bring the model into agreement with
	 the Nationwide Commercial Laboratory Seroprevalence Survey for each state. Testing delays
	 have assumed to be constant.
vaccine_efficacy_transmission: The efficacy for transmission is assumed to be identical to the
			       efficacy for immunity (specified in the scenario descriptions).
vaccine_efficacy_delay: Delay distributions are defined to roughly match the two dose
			schedule defined in the scenario descriptions. Gamma distributions are
			used with mean and standard deviation of 20 and 15 days respectively.
vaccine_hesitancy: As prescribed. 1st dose vaccination rate is assumed to continue at current rate, which
                   results in many states not reaching the goal by end of scenario period.
vaccine_immunity_duration: Starting in round 6, vaccine immunity wanes according to the CDF of 
                           a gamma distribution with mean of 365 days and standard deviation of 50 days.
			   For round 7, vaccine immunity mean wane delay changed from 1 year to 2 years.
natural_immunity_duration: Same assumption as for vaccine immunity duration.
case_fatality_rate: The infection fatality rate is the fundamental model parameter, and is estimated
		    from case and death data. The inferred case fatality rate is therefore
		    different for each state.
infection_fatality_rate: This is estimated from case and death data from each state. The rate is 
                         therefore different for each state.
asymptomatics: 0.1 is the assumed proportion for purposes of defining the connection to testing and
	       hospitalization. A single delay distribution is assumed for removing individuals from
	       the circulating contagious population, which captures many aspects of the infection
	       cycle, including the asymptomatic fraction. Effectively the asymptomatic fraction is
	       absorbed into the inferred transmission rate.
age_groups: Starting in round 6, a homogeneous model is used since vaccination age profile has
            flattened considerably.
importations: Not applicable
confidence_interval_method: Historical case/hospitalization/death time series data are used to estimate
			    model parameters and give standard deviations of the estimators. To produce
			    confidence intervals for time series forecasts, simulations are repeated
			    hundreds of times with parameter variation and include the underlying
			    probabilistic description of the infection process (negative binomial)
			    and connections between populations (typically binomial).
calibration: Case, hospitalization, death, and seroprevalence data are used to estimate model parameters
	     to ensure the model predictions are consistent with observations. Now that variant strains
	     are dominant, genomic data no longer play a role in projecting future growth rates.
spatial_structure: Each state treated independently.
data_inputs: JHU CSSE case and death data, HHS hospitalization data, Google open-data vaccination data,
	     Nationwide Commercial Laboratory Seroprevalence Survey
citation: https://arxiv.org/abs/2007.07156