Challenges in specifying parameter values for COVID-19 simulation models

In a recent paper in Lancet Public Health, Bartsch et al.¹ used an age-stratified transmission model to simulate the coronavirus disease 2019 (COVID-19) epidemic in the US and predicted the cost-effectiveness of maintaining face mask use until a high vaccine coverage (70–90%) is achieved. Their simulations showed that continued face mask use is generally cost-effective and even cost-saving in many of the scenarios considered. Such model-based economic analyses along with epidemiological evidence have the potential to guide policymakers in a timely manner.

One of the biggest challenges in modelling studies is how to reliably choose parameter inputs as their misspecifications can substantially affect the conclusions.² Bartsch et al.¹ chose over 80 parameter inputs in their model, one of which represented the effectiveness of continued face mask use. The authors referred to a study that analysed temporal changes in the effective reproduction number (R_t) around the introduction of state-level mask mandate policies in the US in the summer of 2020 to find an 18% reduction in R_t associated with the policies.³ Bartsch et al. chose this 18% for their effectiveness parameter; however, we need to be careful because this choice implicitly produces an assumption: the effect of face mask use in 2021 through 2022 is the same as that in summer 2020, when stringent interventions including a stay-at-home order and school closure were in place.

This assumption may need to be revisited. COVID-19 frequently spreads over social contacts in settings where people do not wear masks, e.g. dining at restaurants, drinking at bars and social gathering with friends and relatives,⁴ and the stringent interventions in 2020 aimed to restrict contacts in these settings for outbreak control.⁵ With an increased proportion of contacts in these settings after the lifting of restrictions, public space mask mandates alone may not be able to easily achieve an equivalent R_t reduction of 18%.

Conversely, there are also changes from summer 2020 that likely favour the effect of facial mask use, e.g. improved supply of better-quality masks.⁶ In sum, the mask effectiveness in 2021–2022 is more uncertain than considered in Bartsch et al. Rigorous sensitivity analysis on this parameter (e.g. between 5% and 50%) is warranted to provide a balanced view on this important policy question.

Data availability

No data are associated with this article.