(a–c) Observed (bars) and estimated (crosses and error-bars) proportions of women among authors of 1,893 US papers on COVID-19 and 85,373 papers published in the same journals in 2019. The bars show differences in the observed proportions of women in the first-author position (a), the last-author position (b), and any author position (c), for papers published in 2020 COVID-19 papers (blue bars) versus papers from the same journals in 2019 (orange bars). All three panels suggest a decrease in the observed proportion of women. The crosses and error bars show the adjusted means and 95% confidence intervals (CIs) derived from mixed regression models with scientific journal as random effect parameter. (d–f) Adjusted means (crosses) and 95% CIs (error bars) derived from mixed regression models for the proportion of women in the first-author position (d), last-author position (e) and any author position (f), for papers published in 2019 (left-most crosses and error bars in each panel), papers published in March and April 2020 (middle), and papers published in May 2020 (right). For all models, there is a drop in March and April, followed by a partial resurgence in May. However, the uncertainty of the estimates make these comparisons inconclusive. See Supplementary file 1 for details of the mixed regression models used to estimate adjusted means and 95% CIs.