Ensemble²: Scenarios ensembling for communication and performance analysis

Area of Science:

• Epidemiology
• Public Health
• Mathematical Modeling

Background:

• Scenario modeling was vital for public health policy during the COVID-19 pandemic.
• Evaluating long-term pandemic scenario projections requires different criteria than short-term forecasts.
• The Scenario Modeling Hub (SMH) provided COVID-19 modeling data for the US.

Purpose of the Study:

• To propose a novel ensemble procedure for assessing pandemic scenario projections.
• To evaluate the performance of COVID-19 projections from the Scenario Modeling Hub (SMH).
• To synthesize potential epidemic outcomes without identifying the most plausible scenario.

Main Methods:

• Developed a novel ensemble procedure for assessing pandemic scenario projections.
• Defined a "scenario ensemble" for each model and an ensemble of models, termed "Ensemble²".
• Utilized SMH COVID-19 modeling results for the United States.

Main Results:

• The Ensemble² models were found to be well-calibrated.
• Ensemble² demonstrated better performance compared to the scenario ensemble of individual models.
• The ensemble procedure effectively synthesized a range of plausible epidemic outcomes.

Conclusions:

• The Ensemble² approach provides a robust method for assessing pandemic scenario projections.
• This ensembling strategy accounts for the full range of plausible outcomes, aiding policy decisions.
• The methodology can be extended to various scenario design strategies and refined over time.