Effectiveness of mRNA Vaccines Against COVID-19 Hospitalization by Age and Chronic Medical Conditions Burden Among Immunocompetent US Adults, March-August 2021

Abstract

Through October 2021, the coronavirus disease 2019 (COVID-19) pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) resulted in more than 2.3 million hospitalizations and 690 000 deaths in the United States [1]. In December 2020, the US Food and Drug Administration granted emergency use authorization for 2 messenger RNA (mRNA) COVID-19 vaccines, BNT162b2 from Pfizer-BioNTech and mRNA-1273 from Moderna [1].

mRNA COVID-19 vaccines elicit immunity against the SARS-CoV-2 spike protein [2] and have been shown in clinical trials [3, 4] and observational studies [5–7] to be effective in preventing severe COVID-19 requiring hospitalization. Real-world evaluations of vaccine effectiveness (VE) against COVID-19 hospitalization frequently pool multiple age groups and persons with and without preexisting chronic medical conditions. VE estimates by age group and comorbidity burden can complement large pooled VE analyses using heterogenous populations. Identifying the heterogeneity of VE against severe COVID-19 can also potentially inform decisions about targeting preventive measures to populations most likely to benefit from them.

Prior work has identified heterogeneity of VE by immunocompromised state, age, variants, and time since vaccination [5–7], leading to decisions in the United States to provide a third vaccine dose to immunocompromised persons and a booster dose to adults generally [8]. Despite some evidence that mRNA VE could differ by factors such as burden of chronic medical conditions, race/ethnicity, sex, and obesity (body mass index [BMI] ≥ 30 kg/m²) [9], these factors have not often been studied in detail. Therefore, we evaluated the effectiveness of mRNA vaccines against COVID-19 hospitalizations stratified by burden of chronic conditions, age, and other demographics.

METHODS

The Centers for Disease Control and Prevention (CDC) collaborates with the Influenza and Other Viruses in the Acutely Ill (IVY) Network to monitor effectiveness of COVID-19 vaccines against COVID-19 hospitalization. As previously described [5, 6], we prospectively enrolled adults ≥ 18 years of age at 21 IVY Network hospitals during 11 March to 15 August 2021. We have reported VE during this surveillance period [5, 6], but did not assess VE by burden of chronic conditions or other characteristics presented here. Using a case-control design, we assessed mRNA VE against COVID-19 hospitalization by comparing odds of prior vaccination in hospitalized case patients with COVID-19 versus control patients without COVID-19. We restricted analysis to immunocompetent persons because our objective was to inform VE within the general population.

Case or control status was determined by clinical testing results as well as centralized reverse transcription polymerase chain reaction (RT-PCR) testing of upper respiratory samples at Vanderbilt University Medical Center (Nashville, TN). Cases were hospitalized with a COVID-19-like illness (≥1 of fever, cough, shortness of breath, loss of taste or smell, use of respiratory support for the acute illness, pulmonary findings on chest imaging consistent with pneumonia) and had a positive RT-PCR or antigen test for SARS-CoV-2 within 10 days following symptom onset. Two hospitalized control groups that tested negative for SARS-CoV-2 by RT-PCR were enrolled: (1) test-negative controls with COVID-19-like illness and (2) syndrome-negative controls without COVID-19-like illness. The second control group was included due to imperfect diagnostic accuracy of SARS-CoV-2 tests, which could lead to misclassification of cases. Vaccination coverage and VE estimates were highly similar for individual control groups [7], and therefore these groups were combined to improve statistical power. Sites attempted a 1:1 enrollment ratio of cases to controls in each group, with controls admitted within 2 weeks of cases, although individual matching was not performed.

Vaccination status was determined primarily by review of electronic medical records (EMRs), state vaccine registries, provider or pharmacy records, and available CDC vaccination cards, with patient or proxy self-report of vaccination (including known date and location) during enrollment interviews considered as vaccinated in the absence of source documentation. Vaccination status was classified based on vaccine receipt before a reference date, defined as the date of symptom onset for cases and test-negative controls and days prior to admission for syndrome-negative controls. Participants were either unvaccinated or were fully vaccinated if 2 doses of a mRNA vaccine were received ≥ 14 days before the reference date; patients were excluded if they received ≥ 1 vaccine doses but were not fully vaccinated, received a non-mRNA vaccine, or received multiple COVID-19 vaccine products.

Trained personnel obtained information on participants’ preexisting chronic medical conditions associated with severe COVID-19 [10] through EMR review [7] using a standardized case report form (Supplementary Table 1). Individual conditions were grouped into 7 categories: cardiovascular, neurologic, pulmonary, gastrointestinal, endocrine, renal, and hematologic. Participants were classified by the number of categories in which conditions were documented (0, 1, 2 or ≥ 3).

Logistic regression models, with COVID-19 case status as the outcome and vaccination status as the predictor, were used to estimate overall VE adjusted for date of admission (biweekly intervals), age (continuous), sex, self-reported race and ethnicity, burden of underlying conditions (0, 1, 2, or ≥ 3 categories), and US Health and Human Services region of the admitting hospital. VE was calculated as: VE = (1 − adjusted odds ratio) × 100.

Interaction terms were introduced into the main VE model to evaluate VE by primary characteristics of interest: age group (18–64 years or ≥ 65 years) and number of chronic medical condition categories (0, 1, 2, or ≥ 3). Additional exploratory models were constructed with interaction terms between vaccination status and race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, all other races non-Hispanic, unknown), sex (female or male), and obesity (obese or not obese). To estimate VE by condition category within each age category, an additional regression model was fit to include the 3-way interaction among vaccination status, age group, and conditions category along with each 2-way interaction. Likelihood ratio χ² tests were used to estimate whether differences in VE across groups were significant at a threshold of P < .05.

Older adults and those with certain chronic medical conditions were targeted for early priority vaccination after mRNA vaccines became available and may be affected more by waning effectiveness than other groups that were vaccinated later, resulting in potentially lower VE. Therefore, we performed a sensitivity analysis that restricted vaccinated patients to those with a reference date within 120 days of their second vaccine dose, a period during which significant waning against severe COVID-19 is not expected [6]. An additional sensitivity analysis was included to explore differences in VE by conditions burden restricted to cases with hypoxemia (O₂ saturation < 92% or supplemental oxygen support) within 24 hours of admission to account for the possibility of patients with multimorbidity being more likely to be hospitalized for mild illness (ie, admission bias).

This activity was conducted as a public health surveillance activity, with waiver of informed consent, consistent with applicable federal law and CDC policy, eg, 45 C.F.R. part 46.102(l)(2), 21 C.F.R. part 56; 42 U.S.C. §241(d); 5 U.S.C. §552a; 44 U.S.C. §3501 et seq.

RESULTS

After excluding participants who did not meet inclusion criteria, 3619 immunocompetent adults (1669 case patients, 11% fully vaccinated; 1950 control patients, 54% fully vaccinated) were included. Fully vaccinated patients (compared with unvaccinated) were older (median age, 67 vs 53 years), and more likely to report being non-Hispanic white (63% vs 48%), have been hospitalized ≥ 1 time during the past year (48% vs 35%), and have ≥ 1 chronic medical condition (88% vs 69%), including cardiovascular disease (75% vs 52%), endocrine disease including diabetes (44% vs 30%), pulmonary disease (30% vs 22%), and renal disease (21% vs 10%), but were less commonly obese (45% vs 53%) (Table 1). A similar proportion of cases met our definition of hypoxemia across underlying condition categories (77%, 73%, 77%, and 71% for 0, 1, 2, and ≥ 3 condition categories, respectively).

VE differed by number of condition categories in a gradient, from persons with 0 categories of conditions (VE, 96%; 95% CI, 93%–98%) to persons with ≥ 3 (VE, 83%; 95% CI, 76%–88%; P < .001) (Figure 1); results were similar when restricted to hypoxemic cases (Supplementary Figure 1). VE was not statistically different for patients aged ≥ 65 years compared with patients aged 18–64 years (VE, 88% [95% CI, 84%–91%] vs 91% [95% CI, 88%–93%]; P = .142). The relationship between VE and number of chronic medical condition categories did not vary by age group (P = .903).

Figure 1.

Vaccine effectiveness by age group and number of chronic medical conditions. Adjusted vaccine effectiveness (VE) was estimated using logistic regression comparing odds of being fully vaccinated with an mRNA COVID-19 vaccine versus being unvaccinated, in case patients and control patients, using the equation VE = 100 × (1 − odds ratio). Overall VE by number of condition categories documented (0, 1, 2, or ≥ 3) was calculated by including an interaction term between vaccination status and number of condition categories. An additional model including a 3-way interaction between age, conditions, and vaccination status was included to calculate VE by number of condition categories within age groups. Models were additionally adjusted for date of hospital admission (biweekly intervals), US Department of Health and Human Services region of hospital, age (continuous), sex, and race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic of any race, non-Hispanic other, or unknown), and number of condition categories. VE point estimates are rounded to the nearest whole number. Error bars represent 95% confidence intervals.

Open in new tabDownload slide

Exploratory analyses showed that VE differed by sex (male, 88% [95% CI, 84%–91%] vs female, 92% [95% CI, 89%–94%]; P = .037) but not across race/ethnicity groups (P = .415) or by obesity status (obese, 91% [95% CI, 89%–94%] vs not obese, 88% [95% CI, 85%–91%]; P = .105; Supplementary Table 2). Results were similar in a sensitivity analysis limiting the analytical population to patients with illness onset within 120 days of full vaccination.

DISCUSSION

This analysis suggests COVID-19 mRNA vaccines are highly effective for preventing COVID-19 hospitalizations for heterogeneous immunocompetent adults, with some decline in VE with increasing burden of chronic medical conditions. VE was very high (96%), including among adults ≥ 65 years old (95%), among people without conditions, which make up half the US population [11]. In addition, adults aged 18–64 years have accounted for the majority of cumulative US COVID-19 hospitalizations, with hospitalization rates among the unvaccinated > 10 times those of the fully vaccinated [1]; they also tend to have lower vaccination coverage [12] and fewer conditions and compared with adults aged ≥ 65 years. As efforts continue to increase vaccination coverage in the United States, these data suggest that vaccinating younger persons without chronic medical conditions will substantially reduce COVID-19 hospitalizations.

Our findings also suggest that VE against COVID-19–associated hospitalization decreases roughly proportionally to the number of chronic medical conditions, both overall and stratified by age. The conditions included in our analysis align closely with those identified as risk factors for severe COVID-19 in the prevaccination era [10], suggesting that the same chronic conditions placing a person at high risk for severe COVID-19 are also associated with COVID-19 hospitalization among vaccinated persons. Prior studies have suggested that antibody response and immune protection after vaccination could be attenuated for persons with severe chronic medical conditions, including both immunocompromising conditions [13] and nonimmunocompromising conditions [14]; however, few show clear associations with lower VE. Our analysis shows that mRNA VE against severe disease is also lessened by chronic medical conditions among immunocompetent adults, independent of age. Booster vaccines for recipients of COVID-19 vaccines are now recommended for persons aged ≥ 18 years [15]. Our findings suggests that persons with a higher burden of chronic medical conditions may experience greater incremental benefit from additional vaccine doses.

Finally, our findings may have implications for interpretation of data from observational COVID-19 VE studies. Our finding of 96% VE against COVID-19 hospitalizations for healthy adults (ie, immunocompetent patients with no preexisting conditions) was similar to VE against severe disease observed in mRNA COVID-19 vaccine clinical trials [3, 4]. This similarity suggests that efficacy estimates from phase 3 COVID-19 clinical trials successfully approximated real-world VE for people without chronic medical conditions but overestimated effectiveness for people with chronic medical comorbidities.

Our findings are subject to some limitations. Enrolling sites were academic medical centers and may have inpatient populations with a higher burden of chronic diseases than community hospitals. While the number of categories of chronic conditions was considered, the severity of conditions within each category was not; for example, hypertension and severe heart failure were both quantified as 1 cardiovascular condition. We had incomplete capture of data for a few variables, such as height and weight used to determine BMI-based obesity (captured for 99% of patients). However, missing information was minimal and unlikely to bias VE estimates. Furthermore, persons with more chronic medical conditions may be more likely to be hospitalized even with milder COVID-19 illness, resulting in lower VE; however, VE was similar restricting analysis to cases with hypoxemia who would have an indication for admission. Multiple comparisons were made in this analysis, with the possibility of type I or type II error.

In conclusion, this analysis provides insight into the heterogeneity of VE by overall health status, beyond immunocompetency alone. Vaccination reduced the risk of COVID-19 hospitalization by > 95% in healthy adults without chronic medical conditions regardless of age, sex, obesity, or race/ethnicity. Continued efforts to vaccinate US adults, aided by the evaluation and reporting of VE for specific subpopulations, will likely have a marked impact on severe COVID-19.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

Financial support. This work was supported by the US Centers for Disease Control and Prevention (grant number 75D30121F00002) to W.H.S.; and the National Center for Advancing Translational Sciences (grant number UL1 TR002243 Clinical and Translational Science Award for the REDCap data tool). J. H. K. is supported by the National Institute of Allergy and Infectious Diseases, National Institutes of Health (grant number 1K23AI137321).

Potential conflict of interest. W. H. S. reports grants and consultant fees from Merck, and consultant fees from Aerpio Pharmaceuticals, all outside this work. A. A. G. reports grant support from NIH, Department of Defense (DoD), and investigator-initiated grant support from AbbVie and Faron Pharmaceuticals, all outside this work. J. D. C. reports a grant from the National Institutes of Health (NIH). D. C. F. reports consultant fees from Cytovale and membership on a Medpace Data Safety Monitoring Board (DSMB). D. N. H. reports salary support from Incyte Corporation, EMPACT Precision Medicine, and the Marcus Foundation. M. N. G. reports grant support from NIH and the Agency for Healthcare Research and Quality (AHRQ); and fees for participating on a DSMB for Regeneron and for participating on a scientific advisory panel for Philips Healthcare. D. J. H. reports consulting fees from Cytovale and Opticyte. I. D. P. reports grants from NIH and Janssen Pharmaceuticals; and institutional fees from Asahi Kasei Pharma and Regeneron. S. M. B. reports fees from Hamilton for chairing a DSMB, and institutional fees from Faron, Sedana, and Janssen; grants from Sedana, Janssen, NIH, and the DoD; book royalties from Oxford University and Brigham Young University; and personal fees from New York University for service on a DSMB. E. T. M. reports personal fees from Pfizer and a grant from Merck, all for unrelated work. A. K. reports grants from United Therapeutics, Johnson & Johnson, 4D Medical, Lung LLC, and Reata Pharmaceuticals. A. S. M. reports consulting fees from Sanofi-Pasteur and Seqirus. S. Y. C. was a speaker for La Jolla Pharmaceuticals and a Consultant for PureTech Health. J. H. K. reports grant support from NIH. M. C. E. reports talks on nutrition in COVID pneumonia at APEN conference sponsored by Abbott Labs. N. H. reports grants from Sanofi and Quidel. J. D. C. reports a grant from the National Center for Advancing Translational Sciences, NIH. A. S. L. reports consultant fees from Sanofi and fees from Roche for membership on a trial steering committee. C. G. G. reports consultant fees from Pfizer, Merck, and Sanofi-Pasteur; and grants from Campbell Alliance/Syneos Health, NIH, the Food and Drug Administration, AHRQ, and Sanofi. T. W. R. reports personal fees from Cumberland Pharmaceuticals, Inc as the Director of Medical Affairs; consultant fees from Cytovale, Inc; and DSMB membership fees from Sanofi. C. J. L. reports grants from NIH, DoD, and the Marcus Foundation; organizational contract fees from bioMerieux, Endpoint LLC, and Entegrion, Inc; and a patent issued to Cincinnati Children’s Hospital Medical Center for risk stratification in sepsis and septic shock. All other authors report no potential conflicts.

IVY Network Collaborators. Tresa McNeal, Shekhar Ghamande, Nicole Calhoun, Kempapura Murthy, Judy Herrick, Amanda McKillop, Eric Hoffman, Martha Zayed, Michael Smith, Natalie Settele, Jason Ettlinger, Elisa Priest, Jennifer Thomas, Alejandro Arroliga, and Madhava Beeram, Baylor Scott and White Health; Ryan Kindle, Lori-Ann Kozikowski, Lesley De Souza, Scott Ouellette, and Sherell Thornton-Thompson, Baystate Medical Center; Patrick Tyler, Beth Israel Deaconess Medical Center; Omar Mehkri, Meg Mitchell, Connery Brennan, Kiran Ashok, and Bryan Poynter, Cleveland Clinic; Nicholas Stanley, Emory University; Audrey Hendrickson, Sean Caspers, and Tyler Scharber, Hennepin County Medical Center; Jeffrey Jorgensen, Robert Bowers, Jennifer King, Valerie Aston, and Brent Armbruster, Intermountain Medical Center; Richard E. Rothman, Johns Hopkins University; Rahul Nair and Jen-Ting (Tina) Chen, Montefiore Medical Center; Sarah Karow, Emily Robart, Paulo Nunes Maldonado, Maryiam Khan, and Preston So, Ohio State University; Joe Levitt, Cynthia Perez, Anita Visweswaran, and Jonasel Roque, Stanford University; Trevor Frankel, Omai Garner, and Sukantha Chandrasekaran, University of California, Los Angeles; Jennifer Goff, David Huynh, Adit Ginder, David Douin, Kelly Jensen, Conner Driver, Michael Carricato, and Ian Chambers, UCHealth University of Colorado Hospital; Paul Nassar, Lori Stout, Zita Sibenaller, Alicia Walter, Jasmine Mares, Logan Olson, and Bradley Clinansmith, University of Iowa; Carolina Rivas and Hayley Gershengorn, University of Miami; E. J. McSpadden, Rachel Truscon, Anne Kaniclides, Lara Thomas, Ramsay Bielak, Weronika Damek Valvano, Rebecca Fong, William J. Fitzsimmons, Christopher Blair, Andrew L. Valesano, and Julie Gilbert, University of Michigan; Akram Khan, Olivia Krol, Zachary Zouyed, Emma Silverman, Genesis Briceno, and Emmanuel Mills, Oregon Health and Science University; Christine D. Crider, Kyle A. Steinbock, Thomas C. Paulson, and Layla A. Anderson, University of Washington; Christy Kampe, Jakea Johnson, Rendie McHenry, Marcia Blair, and Laura Short, Vanderbilt University Medical Center; Mary LaRose, Leigha Landreth, Madeline Hicks, and Lisa Parks, Wake Forest University; and Jahnavi Bongu, David McDonald, Candice Cass, Sondra Seiler, David Park, Tiffany Hink, Meghan Wallace, Carey-Ann Burnham, and Olivia G. Arter, Washington University.

All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References

Author notes