Agent-based modelling for SARS-CoV-2 epidemic prediction and intervention assessment: A methodological appraisal

Area of Science:

• Epidemiology
• Computational modeling
• Public health

Background:

• Assessing epidemiological agent-based models (ABMs) for SARS-CoV-2 pandemic is crucial due to limited evidence on non-pharmaceutical interventions.
• Previous coronavirus outbreak data lacks external validity for SARS-CoV-2 due to differing contagiousness.
• Epidemiologists use various models, including ABMs, to address COVID-19 policy questions, but face criticism regarding simplifications and empirical support.

Purpose of the Study:

• To methodologically appraise epidemiological agent-based models (ABMs) of the SARS-CoV-2 pandemic.
• To address criticisms regarding the simplifications and empirical support of these models.
• To evaluate the utility of ABMs in informing policy decisions regarding COVID-19 mitigation.

Main Methods:

• Utilized AceMod (model of the COVID-19 epidemic in Australia) as a case study for methodological appraisal of epidemiological ABMs.
• Examined how empirical results are used as inputs for model assumptions and rules.
• Employed calibration techniques to validate model adequacy against benchmark variables.

Main Results:

• Epidemiological ABMs, despite inherent simplifications, sufficiently represent key characteristics of social interactions and SARS-CoV-2 spread.
• Models accurately incorporate empirical data into assumptions and agent rules.
• Calibration ensures model adequacy by comparing outputs to benchmark variables.

Conclusions:

• The best epidemiological ABMs function as models of actual mechanisms, providing both mechanistic and difference-making evidence.
• These models can adequately describe the effects of potential public health interventions.
• Discussion includes limitations of ABMs and policy recommendations for their use.