Abstract
Governments around the globe introduce public health (nonpharmaceutical) interventions aiming at slowing down the spread of the SARS-CoV-2 pandemic. Despite their significant impact, the measures are not supported with high-quality evidence.
Conducting RCT is not feasible for both ethical and practical constraints. One of the many ways to address the issue concerning the impracticality of conducting RCTs in the context of an ongoing pandemic is through scientific modeling, namely epidemiological modeling.
Here we focus on the so-called agent-based modeling (ABM) approach that differs from the more traditional epidemiological modeling in several ways. ABMs are a form of computational modeling strategy where agents are described as entities interacting with each other and their environment in a locally defined fashion.
A set of rules defines the behavior of agents, and the dynamics is then computed, allowing for simulating complex patterns and understanding how these patterns arise. Unfortunately for the assessment of healthcare interventions based on this type of epidemic models, the standard evidence hierarchies either exclude such studies or locate it at the lowest level of hierarchy.
The exclusion of ABMs from evidence hierarchies can be explained by the novelty of agent-based modeling, and the limited trust of EBMers in theoretical and mechanistic reasoning. The causal claims supported with agent-based models have been interpreted inconsonantly: in line with the potential outcome approach (POA), as delivering theory-driven understanding and mechanistic evidence.
Below, we show, using as an example the ABM of SARS-CoV-2 epidemic in Australia, that these interpretations are right because best contemporary ABMs bear a resemblance to the actual mechanisms and therefore allow for the counterfactual assessment of intervention efficacy but also deliver the understanding of phenomena of interest. Our argument proceeds by (1) discussing a case study of model representing the SARS-CoV-2 epidemic, (2) showing that best ABMs represent actual mechanisms despite the presence of isolations and idealizations, and (3) considering the limitations of using ABMs as evidence for clinical and policy decisions.