The global spread of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), and the coronavirus disease 2019 (COVID-19) that the virus causes, began in approximately December 2019.[1] On January 30, 2020, the World Health Organization (WHO) declared the emergence of COVID-19 as a public health emergency of international concern[2] and a pandemic on March 11, 2020.[3] COVID-19 caused an infectious disease crisis not seen since the 1918 H1N1 flu (aka the “Spanish flu”) pandemic. The SARS-CoV-2 and COVID-19 crisis generated thousands of published quantitative model-based analyses, as mitigation strategies were urgently needed despite the uncertainties about SARS-CoV-2 and COVID-19.

As background, in healthcare, models are frequently employed in situations when it is too slow, too disruptive, too expensive, and/or unethical to experiment with the actual system. The COVID-19 pandemic exemplifies such a situation. Nevertheless, these models, in their role to inform public and policy debate, have been criticized.[4] Some of this criticism can be ameliorated by revisiting the purpose of such models, which is not to provide “the” answer, but serve as a means to inform the answer. A model is a structured representation of a decision process that allows the performance of a decision analysis. To paraphrase Howard Raiffa of Harvard University: Decision analysis is just the systematic articulation of common sense. Any decent decision maker reflects on alternatives, is aware of uncertainties, modifies judgments on the basis of accumulated evidence, balances risks of various kinds, considers the potential consequences of his or her decisions, and synthesizes all of this in making a reasoned decision that he or she decrees right for the persons affected. All that decision analysis is asking the decision maker to do is to do this a lot more systematically and in such a way that others can see what is going on and can contribute to the decision process.”[5]

In keeping with the theme of models being a structured representation of a decision process, we performed a random review of the structures of published COVID-19 epidemiological models on PubMed. We randomly selected 100 publications of 3,518 hits from a PubMed search using the following search string: “(covid[Title/Abstract]) AND (model*[Title/Abstract]) AND (simul*[Title/Abstract])”. Nearly all of the 100 publications had a compartmental structure that was derivative of the classic susceptible (S) – infectious (I) – recovered (R) infectious disease model structure. What primarily differentiated the publications was whether they modeled populations as homogenous cohorts or as heterogenous individuals (agent-based simulations or microsimulations). With homogenous cohorts the information about the patient is determined by what health state they are in, e.g., “infectious, not yet detected.” With heterogenous individuals the information about the patient is contained as attributes in a simulated entity, e.g., 50 year old, Hispanic male, with cardiovascular and diabetes. The homogenous cohort approach can be employed when the distribution of population characteristics is not known; the cohort is then composed of a hypothetical number of representative or “average” individuals. The heterogenous approach is more conducive to simulating real-world populations, assuming there are data (demographic, treatment efficacy, etc.) to support such a heterogenous analysis. Eighty-three percent (83%) of publications simulated patients as homogenous cohorts and 17% heterogenous individuals. These percentages were from publications from 2020 to 2022. Analyzing the subset of 2022 publications alone (42 publications), the percent of models with heterogenous individuals rose to 27%, possibly reflecting the increasing recognition of the value of agent-based-/micro- simulations.

Information provided by Harry J. Smolen, President – ISPOR United States Midwest Region Chapter, and President & CEO – Medical Decision Modeling Inc., USA