This Viewpoint discusses the challenges of accurately modeling the COVID-19 pandemic and reviews principles that will make some models more useful than others, such as use of granular local data when available, regular updating and revision, and specification of uncertainty around estimates.

Numerous mathematical models are being produced to forecast the future of coronavirus disease 2019 (COVID-19) epidemics in the US and worldwide. These predictions have far-reaching consequences regarding how quickly and how strongly governments move to curb an epidemic.

The author of this viewpoint presents reasons why predictive models are problematic, including because they: often fail to account for factors like accuracy of diagnostic tests; whether immunity will wane quickly; if reinfection could occur; or population characteristics, such as age distribution, percentage of older adults with comorbidities, and risk factors (eg, smoking, exposure to air pollution). Other critical variables such as the reproductive number and social distancing effects, can also change over time. For large countries, such as the US, predictive models are even more problematic because they aggregate heterogeneous subepidemics in local area and cannot account for wide variations of in individual risk factors that occur across the population.

Hi also discusses how the best, most useful models need to: be dynamic; be transparent about assumptions; report ranges (such as CIs);incorporate measures of accuracy as additional data become available; be properly reported to the public to avoid misinterpretation; and seek to use the best possible data for local predictions.