The global pandemic created by SARS-COV2 has dramatically altered the landscape of contemporary life, bringing personal and economic hardships to many and profound loss of life. Scientists, health care workers and others have rapidly responded to address the challenge of this pandemic, and our insight into the structure and function of the virus, how it attacks the body, treatment and prevention approaches, and epidemiological models have evolved at lightning speed. While the world awaits a vaccine to prevent SARS-COV2 infection, we rely heavily on epidemiologists to devise models of disease transmission to understand the course of infection and progression of disease, to prevent further spread and to assign resources for effective treatment.
Epidemiologists use a variety of models to estimate disease transmission, susceptibility and severity. In addition to the use of “model organisms” suspected to have similarities with SARS-COV2, epidemiologists use a variety of statistical models to describe and predict the behaviour of this disease. Each of these models involves assumptions that affect the validity and interpretation of findings. We invite our readers of the Annals of Human Biology to consider these issues in this commentary from Dr. George Ellison. This commentary details what epidemiological models can (and cannot tell) us about COVID-19. Dr. Ellison reviews numerous assumptions and sources of bias such as those arising from unmeasured variables in statistical models, “collider bias”, causal inference, extrapolation and simulation techniques, and how these affect the validity and interpretation of their findings. This commentary is longer than usual, but we believe it is an important contribution to provide our readers with a deeper and more nuanced understanding of the strengths and limitations of epidemiological analyses of the COVID-19 pandemic.