Comparison of test-negative and syndrome-negative controls in SARS-CoV-2 vaccine effectiveness evaluations for preventing COVID-19 hospitalizations in the United States

Caitlin Turbyfill, Katherine Adams, Mark W. Tenforde, Nancy L. Murray, Manjusha Gaglani, Adit A. Ginde, Tresa McNeal, Shekhar Ghamande, David J. Douin, H. Keipp Talbot, Jonathan D. Casey, Nicholas M. Mohr, Anne Zepeski, Nathan I. Shapiro, Kevin W. Gibbs, D. Clark Files, David N. Hager, Arber Shehu, Matthew E. Prekker, Anne E. FroschMatthew C. Exline, Michelle N. Gong, Amira Mohamed, Nicholas J. Johnson, Vasisht Srinivasan, Jay S. Steingrub, Ithan D. Peltan, Samuel M. Brown, Emily T. Martin, Adam S. Lauring, Akram Khan, Laurence W. Busse, Caitlin C. ten Lohuis, Abhijit Duggal, Jennifer G. Wilson, Alexandra June Gordon, Nida Qadir, Steven Y. Chang, Christopher Mallow, Carolina Rivas, Jennie H. Kwon, Natasha Halasa, James D. Chappell, Carlos G. Grijalva, Todd W. Rice, William B. Stubblefield, Adrienne Baughman, Jillian P. Rhoads, Christopher J. Lindsell, Kimberly W. Hart, Meredith McMorrow, Diya Surie, Wesley H. Self, Manish M. Patel

Research output: Contribution to journalArticlepeer-review

3 Scopus citations


Background: Test-negative design (TND) studies have produced validated estimates of vaccine effectiveness (VE) for influenza vaccine studies. However, syndrome-negative controls have been proposed for differentiating bias and true estimates in VE evaluations for COVID-19. To understand the use of alternative control groups, we compared characteristics and VE estimates of syndrome-negative and test-negative VE controls. Methods: Adults hospitalized at 21 medical centers in 18 states March 11–August 31, 2021 were eligible for analysis. Case patients had symptomatic acute respiratory infection (ARI) and tested positive for SARS-CoV-2. Control groups were test-negative patients with ARI but negative SARS-CoV-2 testing, and syndrome-negative controls were without ARI and negative SARS-CoV-2 testing. Chi square and Wilcoxon rank sum tests were used to detect differences in baseline characteristics. VE against COVID-19 hospitalization was calculated using logistic regression comparing adjusted odds of prior mRNA vaccination between cases hospitalized with COVID-19 and each control group. Results: 5811 adults (2726 cases, 1696 test-negative controls, and 1389 syndrome-negative controls) were included. Control groups differed across characteristics including age, race/ethnicity, employment, previous hospitalizations, medical conditions, and immunosuppression. However, control-group-specific VE estimates were very similar. Among immunocompetent patients aged 18–64 years, VE was 93 % (95 % CI: 90–94) using syndrome-negative controls and 91 % (95 % CI: 88–93) using test-negative controls. Conclusions: Despite demographic and clinical differences between control groups, the use of either control group produced similar VE estimates across age groups and immunosuppression status. These findings support the use of test-negative controls and increase confidence in COVID-19 VE estimates produced by test-negative design studies.

Original languageEnglish (US)
Pages (from-to)6979-6986
Number of pages8
Issue number48
StatePublished - Nov 15 2022


  • COVID-19
  • Case-control study
  • Control groups
  • Research design
  • Test-negative
  • Vaccine effectiveness

ASJC Scopus subject areas

  • Molecular Medicine
  • General Immunology and Microbiology
  • General Veterinary
  • Public Health, Environmental and Occupational Health
  • Infectious Diseases


Dive into the research topics of 'Comparison of test-negative and syndrome-negative controls in SARS-CoV-2 vaccine effectiveness evaluations for preventing COVID-19 hospitalizations in the United States'. Together they form a unique fingerprint.

Cite this