Epidemiological COVID-19 model for Los Angeles County

Epidemic dynamic models have provided estimates of key epidemic parameters important for policy planning throughout the evolving COVID-19 pandemic. However, currently implemented models have failed to investigate how these parameters differ across regions based on the prevalence of various combinations of key risk factors in the population, and how these differences in prevalence affect the dynamics of the epidemic in terms of impact on healthcare and spread of infection. Importantly, these models have also failed to investigate how the epidemic dynamics differ across racial and ethnic populations, across age, sex, and across cities.

 

To address this critical gap, we have been developing an epidemiological modeling framework that integrates estimated key epidemic quantities and model dynamics from a stochastic epidemic model for Los Angeles County (LAC) with risk estimates and population-specific prevalence for risk factors of severe illness. In the absence of direct estimates of conditional risks, we apply a statistical technique commonly used in genetics to obtain these risks from reported marginal estimates and the correlation structure between the single risk factors, accounting for age, existing comorbidities, BMI, and smoking status. The model estimates the probability of severe illness given infection and given combinations of risk factors in LAC; the fraction of the overall population that is hospitalized, admitted to ICU, and deceased by each risk group; epidemic parameters including the risk factor stratified case fatality rates (CFR) and infection fatality rates (IFR), the time-varying reproductive number, and the illness ascertainment rate; and the impact on healthcare system and deaths for specific risk groups if various interventions had been implemented. The current approach is a way to use commonly available community-level data to model multiple groups in a single population, and then project for different groups. While this model allows us to make some key inferences, a major limitation is that it assumes a single, homogenous population with equal mixing of all individuals in Los Angeles. As the observed Los Angeles County COVID-19 data clearly demonstrates, the COVID-19 pandemic has had very different dynamics across ethnic and racial groups in terms of infectious rates, spatial patterns and time periods. This also includes differences across age groups, sex, and across cities within LA County. Thus, the assumption of a single population is not appropriate for making inference into the different dynamics across ethnic/racial groups.

 

We are also working to expand our current epidemic model to a multiple populations model and to incorporate a hierarchical modeling framework to borrow information across population sub-groups to obtain estimates of key epidemic quantities for combinations of ethnicity/race, age, sex, and city strata. Moreover, as we have previously done, we can incorporate our risk factor model to further infer the distribution of risk profiles (i.e. combinations of risk factors within an individual) within each combination strata. This model will allow us to estimate key epidemic quantities across these strata, compare how the pandemic has differentially impacted populations, and model the impact of alternative interventions that take specific populations and regions into consideration. Specifically, the model will provide estimates for:

 

Time-varying reproductive number by ethnicity/race, age, sex, and city combinations.

Probability of severe illness given infection, including estimates of the number of individuals for the key model components of hospitalized, admitted to ICU, and deceased by ethnicity/race, age, sex, and city combinations.

Case fatality rates (CFR) and infection fatality rates (IFR) by ethnicity/race, age, sex, and city combinations.

Within stratum, further estimate the proportion of individuals in key model components with certain risk profiles or combinations of risk factors.

Counterfactual projections of how the epidemic dynamics will change with varying policy implementations and how those policy decisions can mitigate or aggravate disparities.