Study population and setting
This case-control study included individuals who were hospitalized at one of 21 hospitals in 18 US states between March 11 and August 15, 2021. Cases were individuals 18+ years hospitalized with a clinical syndrome consistent with COVID-19 and a positive antigen or molecular SARS-CoV-2 test within 10 days after symptom onset. Controls, selected from the same hospital within two weeks of case hospitalization, included one “test-negative” control per case (hospitalized with symptoms consistent with COVID-19 but with a negative SARS-CoV test result) and one “syndrome-negative” control per case (hospitalized for other conditions with a negative SARS-CoV-2 test result). Demographic, clinical, laboratory, and vaccination history data were collected through standardized patient (or proxy) interviews and medical record reviews. Vaccination status was collected after enrollment. Participants who experienced symptom onset 14+ days after their second mRNA vaccine dose were considered fully vaccinated (or 4 days before hospital admission for syndrome-negative controls). Participants who had not received any COVID-19 vaccines were considered unvaccinated. All other individuals (those less than 14 days after the second mRNA vaccine, between doses, or vaccinated with a non-mRNA vaccine) were excluded. A composite outcome of death or invasive mechanical ventilation (i.e., severe disease progression) was also considered. The study assessed the association between prior vaccination and COVID-19 hospitalization adjusted for admission date with a random effect for the study site. They also assessed the potential for effect modification by immunosuppressive status, time between vaccination and symptom onset, SARS-CoV-2 lineage (Alpha or Delta), and vaccine product (Pfizer or Moderna). Finally, among patients hospitalized through July 14, 2021, the study assessed the relationship between prior vaccination and progression to high disease severity adjusted for medical comorbidities. In both models the authors used multivariable logistic regression adjusting for age, sex, and self-reported race and ethnicity.
Summary of Main Findings
The study population included 4,513 individuals: 1,983 cases, 1,359 test-negative controls, and 1,171 syndrome-negative controls (2,530 total controls, combined for the analyses). The median age was 59 years (IQR 45-69), 48.8% were female, 23.0% identified as non-Hispanic Black, 15.9% identified as Hispanic, and 20.1% had an immunocompromising condition. Those with one of the 142 breakthrough cases (COVID-19 cases among participants 14+ days from their second mRNA vaccination at symptom onset) tended to be older than unvaccinated controls (median age 67 vs 53 years), were more likely to self-identify as non-Hispanic white (64.0 vs. 43%), and were more likely to have an immunocompromising condition (40.8% vs. 11.5%). Cases were less likely to be fully vaccinated (15.8%) than controls (54.8%). The predominant SARS-CoV-2 strain shifted from the Alpha variant to the Delta variant in mid-June 2021.
Full vaccination was associated with 85% lower odds of hospitalization with COVID-19 (adjusted odds ratio [aOR]: 0.15, 95% confidence interval [CI]: 0.13, 0.18) compared to controls. Immunocompromised vaccinated individuals (aOR: 0.49, 95% CI: 0.35, 0.69) were less protected than immunocompetent vaccinated individuals (aOR: 0.10, 95% CI: 0.09, 0.13). Moderna-vaccinated individuals (aOR: 0.11, 95% CI: 0.08, 0.14) were more protected than Pfizer-vaccinated individuals (aOR: 0.19, 95% CI: 0.16, 0.23), likely due to waning protective effects in Pfizer vaccinated individuals 120+ days after vaccination. Vaccination was similarly protective against infection with the Alpha variant (aOR: 0.10, 95% CI: 0.06, 0.16) and the Delta variant (aOR: 0.14, 95% CI: 0.10, 0.21) of SARS-CoV-2. Finally, among individuals hospitalized for COVID-19, vaccination was protective against progression to death or invasive mechanical ventilation (aOR: 0.33, 95% CI: 0.19, 0.58). Again, immunocompromised individuals were less protected (aOR: 0.54, 95% CI: 0.21, 1.38) than immunocompetent individuals (aOR: 0.29, 95% CI: 0.14, 0.60).
This study included a large, diverse patient population. It also collected data from a variety of complementary data sources and provided sufficient details about their statistical methods to assess for risks of bias. Finally, the authors included relevant likely confounding variables in their statistical models.
The authors dichotomized time since vaccination at 120 days, which may have resulted in residual confounding. Furthermore, it is difficult to determine how the emergence of the Delta variant of SARS-CoV-2 may have impacted vaccine effectiveness, given that it approximately coincided with participants being 120 or more days from full vaccination. Controls included patients both with (test-negative) and without (syndrome-negative) COVID-19-like illnesses. While a test-negative design improved the likelihood of including cases and controls from the same source population (a critical prerequisite for validity in case-control studies), it did not guarantee this. Vaccination status may have influenced the probability of being hospitalized for a non-SARS-CoV-2 infection (e.g., if vaccination made it more likely for individuals to have close contacts with others), which could have led to bias. Similarly, if vaccination status was correlated with seeking care relative to COVID-19 (e.g., if vaccinated people were less inclined to seek care upon becoming sick), the results could be biased. Finally, as an observational study, unmeasured confounding is a concern.
This study adds to the literature on vaccine effectiveness against COVID-19 hospitalization. In particular, it provides useful comparative information in immunosuppressed versus immunocompromised individuals and on vaccine effectiveness with the emergence of the Delta variant.