Vaccine Effects on In-hospital COVID-19 Outcomes

Abstract

Here, we posit that studies comparing outcomes of patients hospitalized with COVID-19 by vaccination status are important descriptive epidemiologic studies, but they contrast two groups that are not comparable with regard to causal analyses. We use the principal stratification framework to show that these studies can estimate a causal vaccine effect only for the subgroup of individuals who would be hospitalized with or without vaccination. Further, we describe the methodology for, and present sensitivity analyses of, this effect. Using this approach can change the interpretation of studies only reporting the standard analyses that condition on observed hospital admission status-that is, analyses comparing outcomes for all hospitalized COVID-19 patients by vaccination status.

Keywords: Causal inference; Disease progression; Post-treatment variable; Potential outcomes; Principal stratification; Vaccination.

Figure.

Sensitivity analyses. The figure shows principal effects (the different colors indicate the magnitude of $V{E}_{in-h}$ in the “doomed” stratum; scale is shown on the right) in different scenarios that assume 90% vaccine efficacy against COVID-19 hospitalization and the odds ratio, $\beta$, values of 1 (left panel) and 5 (right panel). X- and y-axes correspond to proportions of unvaccinated and vaccinated patients with the more severe in-hospital outcome, Pr(Y=1 | H=1, V=0) and Pr(Y=1 | H=1, V=1). Calculations used equations 8, 15, and 28 in the manuscript by Hudgens and Halloran (see main text). Notice that in this figure, rather than using data from a single study, we perform calculations over a range of plausible values for the frequencies of the in-hospital outcome. The color scale only ranges from 0 to 1, and negative principal effects in the “doomed” stratum are represented with the same color as no effect (dark red). In the eAppendix; https://links.lww.com/EDE/C244 we present sensitivity analyses for different values of $\beta$ (eFigure; https://links.lww.com/EDE/C244), and present additional information on the calculations performed to generate these figures.