COVID-19 healthcare demand projections: Arizona

Abstract

Beginning in March 2020, the United States emerged as the global epicenter for COVID-19 cases with little to guide policy response in the absence of extensive data available for reliable epidemiological modeling in the early phases of the pandemic. In the ensuing weeks, American jurisdictions attempted to manage disease spread on a regional basis using non-pharmaceutical interventions (i.e., social distancing), as uneven disease burden across the expansive geography of the United States exerted different implications for policy management in different regions. While Arizona policymakers relied initially on state-by-state national modeling projections from different groups outside of the state, we sought to create a state-specific model using a mathematical framework that ties disease surveillance with the future burden on Arizona's healthcare system. Our framework uses a compartmental system dynamics model using a SEIRD framework that accounts for multiple types of disease manifestations for the COVID-19 infection, as well as the observed time delay in epidemiological findings following public policy enactments. We use a compartment initialization logic coupled with a fitting technique to construct projections for key metrics to guide public health policy, including exposures, infections, hospitalizations, and deaths under a variety of social reopening scenarios. Our approach makes use of X-factor fitting and backcasting methods to construct meaningful and reliable models with minimal available data in order to provide timely policy guidance in the early phases of a pandemic.

Fig 1. Cumulative confirmed COVID-19 cases in Arizona, between March 4 to June 7, 2020.

Fig 2. Cumulative COVID-19 related deaths in Arizona, between March 4 to June 7, 2020.

Fig 3. Depiction of the compartmentalized system dynamics model used to represent transmission and disease progression for State of Arizona projections.

Fig 4. Reported new cases and presumed exposure dates.

Fig 5. The W(s) function for the X-factor of 4 scenario, obtained by inflating the daily new cases.

Fig 6. fH+C+B(t) under assumed parameters.

Fig 7. 95% prediction bands for susceptibles; red dots show presumed susceptibles under 4X scenario.

Fig 8. 4X presumed exposures from data, and predicted exposures with the fitted β values under 4X.

Fig 9. Model predicted cumulative number of deaths between 3/31 and 6/7 with 95% prediction intervals; red dots are the reported COVID-19 deaths for the period in Arizona.

Fig 10. Susceptible, infected and recovered, 4X loading with fitted β and υ.

Fig 11. 1.67X exposures inferred from actual data (red dots) and projected by the model.

Fig 12. Cumulative number of deaths; actual data (red dots) and projected by the model.

Fig 13. Total infected projected by the model with fitted β and υ.

Fig 14. Hospitalized patients projected by the model with fitted β and υ.

Fig 15. Hospitalization projections under favorable summer effect scenarios for 1.67X.

Fig 16. Total infections under the baseline and five NPI initiation date options for 1.67X.