Early Outcomes of SARS-CoV-2 Infection in a Multisite Prospective Cohort of Inpatient Veterans

Abstract

The coronavirus disease 2019 (COVID-19) pandemic has profoundly affected the US population, with >1.1 million deaths recorded in the United States [1]. Among users of the Veterans Health Administration (VHA), one of the largest integrated health care systems in the United States, >870 000 severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections and 24 000 COVID-19 deaths have been reported [2]. The VHA's integrated electronic health record system (EHR) and COVID-19 Shared Data Resource have been used for several EHR-based analyses detailing COVID-19 incidence, outcomes, and vaccine effectiveness [3–7]. These analyses revealed temporal changes in COVID-19 risk factors and outcomes even within the first year of the pandemic, including declines in case fatality among VHA users hospitalized with COVID-19 before vaccine availability [3]. Additionally, analyses among VHA users have identified changes in sociodemographic risk factors for COVID-19 over time [8].

Individuals continue to be hospitalized with SARS-CoV-2 infection, despite the development of highly effective COVID-19 vaccines and widespread implementation of vaccination in health care systems including the VHA [6, 9, 10]. Contributing factors include SARS-CoV-2 viral evolution to more transmissible viral variants with higher rates of vaccine breakthrough infections, waning immunity over time following vaccination, and host comorbidities associated with severe illness after vaccination [6, 1–4]. The period beginning in December 2021, during which the Omicron variant of SARS-CoV-2 was dominant in the United States, was characterized by an unprecedented peak in infections but less illness severity than earlier dominant variants [15]. With evidence suggesting that SARS-CoV-2 variants are associated with different illness severity, there is a need to characterize outcomes over time to inform clinical care and health services planning [16]. Further, characterizing the relationships between vaccination, comorbidities, and illness severity is crucial to understanding mechanisms of protection from severe illness [17].

The Epidemiology, Immunology, and Clinical Characteristics of COVID-19 (EPIC³) study within the VHA aims to describe the epidemiology and natural history of SARS-CoV-2 among veterans and to longitudinally characterize host and viral factors associated with severity of infection and the development of immunity. Here we report outcomes of illness severity, hospital days, and need for rehospitalization within the first 30–60 days of SARS-CoV-2 infection, while considering vaccination status, over a study period characterized by different dominant SARS-CoV-2 variants.

METHODS

Study Design

This ongoing prospective, observational cohort study enrolled inpatient and outpatient veterans at 15 Department of Veterans Affairs medical centers (VAMCs) beginning in July 2020 (Supplementary Table 1) through September 2022. This analysis includes inpatient participants who became eligible for the study between February 1, 2021, and June 18, 2022, during which COVID-19 vaccines were available in VAMCs.

Inclusion and Exclusion Criteria

Veterans were eligible for recruitment if they were age 18 years or older, hospitalized at a participating VAMC, and had a SARS-CoV-2 PCR test result within 21 days before admission. Sites were instructed to enroll 1 SARS-CoV-2-negative participant for each SARS-CoV-2-positive participant enrolled, with hospitalization status (inpatient vs outpatient) as the only matching criterion. Among veterans approached for enrollment, 60% agreed to enroll in the study.

Study Procedures

Study data are collected on days 0, 3, 7, 14, 21, and 28 and months 3, 6, 12, 18, and 24 after enrollment. Trained study staff administer questionnaires by telephone or in person. At follow-up visits, study staff administer the FLU-PRO questionnaire and study-developed questionnaires to assess COVID-19 vaccination history and persistent symptoms [18]. Visits are conducted in person when feasible and remotely via telephone when the veteran is unable to come to the medical center.

Exposures

Participants were enrolled as SARS-CoV-2 positive or negative based on the result of the SARS-CoV-2 PCR associated with their enrollment eligibility. Within the group of participants enrolled as SARS-CoV-2 positive, we further stratified based on vaccination status. Participants were considered vaccinated with a primary vaccine series 14 days after receiving 2 doses of the Pfizer-BioNTech or Moderna/mRNA-1273 vaccine 21 and 28 days apart, respectively, or 1 dose of the Johnson and Johnson (Janssen) vaccine. Partially vaccinated participants who had received 1 dose of the Pfizer or Moderna vaccine series were grouped with unvaccinated participants for the analysis. Participant demographics, including age, sex, and self-reported race and ethnicity were ascertained using the study questionnaire or the electronic health record (EHR). We calculated participants’ Charlson comorbidity index (CCI), a composite measure of medical comorbidities, using the method detailed by Quan and colleagues, with data from the EHR from the 2 years before enrollment [19]. Participants were classified into 3 enrollment periods (February 1, 2021–May 31, 2021; June 1, 2021–November 30, 2021; and December 1, 2021–June 18, 2022) corresponding to periods of different dominant SARS-CoV-2 variants in the United States: Alpha (B.1.1.7 and descendent lineages), Delta (B.1.617 and AY lineages), and Omicron (B.1.1.529 and BA lineages), respectively [20].

Outcomes

We ascertained 3 outcomes among all participants. The first, illness severity, was the highest level of severity during 30 days after study entry; this was calculated using the Veterans Affairs Severity Index for COVID-19 (VASIC), a mutually exclusive, 4-category (mild, moderate, severe, or death) adaptation of the World Health Organization's COVID-19 severity scale utilizing the EHR data and validated by chart review [21]. Mild severity included participants who were hospitalized for ≤24 hours; moderate severity included participants hospitalized for >24 hours with or without low-flow oxygen therapy; the severe category included participants who received high-flow oxygen therapy, intubation, mechanical ventilation, extracorporeal membrane oxygenation, vasopressors, or new renal dialysis within 30 days of diagnosis; and death included deaths due to any cause within 30 days. The second outcome, hospitalization duration, included the total hospital days recorded within 60 days of study entry using 24-hour increments. We calculated the third outcome, rehospitalization among participants enrolled in the inpatient cohort, indicating whether participants were discharged and hospitalized again within 60 days of study enrollment, as a dichotomous variable.

Statistical Analyses

We compared participant characteristics including age, sex, CCI, enrollment period, presence of symptoms, and illness severity using frequencies, proportions, and means. We used multivariable regression models to estimate associations between all exposures and each outcome using multinomial logistic regression models for illness severity, negative binomial models for the duration of hospital days, and Cox proportional hazards models for rehospitalization. To account for death during follow-up for the hospitalization duration and re-hospitalization outcomes, we added an offset term and censored at death, respectively. To assess the risk estimates with and without major potential confounders selected a priori, we fit 3 separate models for illness severity, progressively entering additional terms for each model. The first model included only vaccination status as a covariate; the second added age, sex, and CCI as covariates; and the third added the calendar period of eligibility. We fit a single model for each of the 2 outcomes of hospitalization duration and rehospitalization. We used α = .05 for measures of statistical significance and complete case analysis for each of the regression models.

Patient Consent

Veterans or their legally authorized representative provided written informed consent for participation. The study was approved by the VA Central Institutional Review Board.

RESULTS

Participant Characteristics

The EPIC³ study consented and enrolled 542 inpatient participants during this analysis period (Table 1). Among them, 292/542 (54%) were SARS-CoV-2 positive, including 140/292 (48%) who were vaccinated and 152/292 (52%) who were not vaccinated. Further stratifying within the vaccination groups used in the analysis, 25/152 (16%) SARS-CoV-2-positive inpatients who were classified as unvaccinated had received 1 dose of vaccine, and 44/140 (31%) vaccinated SARS-CoV-2-positive inpatients had received both a primary vaccine series and ≥1 booster dose. Among participants enrolled as SARS-CoV-2 negative, 186/250 (74%) were vaccinated. Inpatients who were SARS-CoV-2 positive after vaccination were older than unvaccinated SARS-CoV-2-positive inpatients (median, 72.0 vs 61.0 years). More than 90% were male. High CCI scores (≥5) occurred in 61 (44%) vaccinated SARS-CoV-2-positive participants and 29 (19%) unvaccinated SARS-CoV-2-positive participants (Table 1).

Characteristics of the participants enrolled during different calendar periods of the study differed in several ways. The proportion of SARS-CoV-2-positive participants who were vaccinated increased between each calendar period, from 7/71 (10%) in the first period (Alpha) to 56/121 (46%) in the second period (Delta) to 77/100 (77%) in the third period (Omicron) (Supplementary Table 2). Additionally, 35% of participants enrolled during the latest calendar period (December 2021–June 2022) had a comorbidity index of ≥5 compared with 25% of those enrolled in the first period and 28% of those enrolled in the second period.

Illness Severity

We calculated illness severity among all SARS-CoV-2-positive participants (Table 2). Covariate data were complete for this outcome; 1.4% of vaccinated, SARS-CoV-2-positive inpatients did not report any symptoms on FLU-PRO surveys in the first 14 days, compared with 5.3% of SARS-CoV-2-positive unvaccinated inpatients and 5.2% of SARS-CoV-2-negative inpatients (Table 2). In unadjusted analyses, severe illness occurred in 20 (14%) vaccinated inpatients and 28 (18%) unvaccinated inpatients (Table 2). Also in unadjusted analyses, 9 (6%) of 140 fully vaccinated SARS-CoV-2-positive inpatients died from any cause, as did 3 (2%) of the 152 unvaccinated.

After controlling for age, sex, vaccination status, and calendar period of enrollment, a 1-unit higher CCI was associated with greater odds of severe illness relative to mild illness (aOR, 1.21; 95% CI, 1.01–1.45) among SARS-CoV-2-positive participants (Table 3). Vaccinated inpatients had lower odds of moderate illness compared with mild illness (aOR, 0.32; 95% CI, 0.12–0.86) when adjusting for other covariates, but there was no significant difference in odds of severe illness or death by vaccination status among inpatients with SARS-CoV-2 infection (Table 3). The point estimate for increased age and higher odds of death was >1, though the relationship was not statistically significant. An exploratory analysis (due to limited sample size) of the relationships between illness severity and calendar period did not show differences in severity by calendar period in multivariable analyses (Table 3).

Hospital Days and Rehospitalization

We calculated the outcomes of hospital days and rehospitalization among all participants, including those who tested positive vs negative for SARS-CoV-2 (Table 2). Data were complete for these outcomes. The median days of hospitalization (interquartile range [IQR]) was 5.5 (3–11) among vaccinated SARS-CoV-2-positive inpatients, 5 (3–8.25) among SARS-CoV-2-positive inpatients who were unvaccinated, and 3.5 (2–8) days among SARS-CoV-2-negative inpatients. In adjusted modeling, a higher number of hospitalization days was associated with vaccinated SARS-CoV-2-positive inpatients (compared with SARS-CoV-2-negative inpatients; adjusted incidence rate ratio [aIRR], 1.55; 95% CI, 1.26–1.91), older age (aIRR, 1.10; 95% CI, 1.03–1.17), and higher CCI (aIRR, 1.06; 95% CI, 1.03–1.10).

Sixty-one (44%) vaccinated SARS-CoV-2-positive inpatients required rehospitalization within 60 days, compared with 46 (30%) unvaccinated inpatients and 63 (25%) SARS-CoV-2-negative inpatients (Table 2). In adjusted analysis, the risk of rehospitalization within 60 days was higher among vaccinated SARS-CoV-2-positive participants compared with SARS-CoV-2-negative inpatients (aHR, 1.86; 95% CI, 1.28–2.69), and higher CCI was also associated with rehospitalization (aHR, 1.07; 95% CI, 1.01–1.12) (Table 4).

DISCUSSION

In a multisite, prospective longitudinal study among veterans, participants hospitalized with SARS-CoV-2 infection continued to experience severe outcomes across enrollment periods that coincided with SARS-CoV-2 surges of the Alpha, Delta, and Omicron viral variants. Participants with a higher comorbidity burden had greater odds of severe illness after accounting for other covariates. As widespread COVID-19 vaccination efforts reached millions of VHA users, an increasing proportion of our participants had completed a COVID-19 primary vaccine series before study enrollment, yet vaccinated participants who required hospitalization also experienced severe outcomes. A possible explanation for this finding is that vaccinated participants with SARS-CoV-2 were older and had higher comorbidity burden than unvaccinated participants with SARS-CoV-2. The VHA successfully prioritized vaccinating older and more comorbid (vulnerable) VHA users; these patients are generally in contact more with the VHA and therefore may have been more likely to be recruited into EPIC³.

Our findings extend evidence from other studies about the relationships among COVID-19 vaccination, demographic and clinical factors, and severity of SARS-CoV-2 infection over the course of the pandemic in the United States. Aligned with our findings, veterans with infection after vaccination were older and had higher prevalence of medical comorbidities than unvaccinated veterans with infection in a database analysis of all VHA users with confirmed SARS-CoV-2 infection between December 2020 and June 2021 [22]. Further, the Influenza and Other Viruses in the Acutely Ill Network also reported that vaccinated adults hospitalized with COVID-19 between March and August 2021 in the United States were older and had more chronic medical conditions than unvaccinated adults hospitalized with COVID-19 [23]. Analyses among vaccinated individuals have also shown a greater risk of severe COVID-19 in people with older age and/or multiple comorbidities [24] and older age and/or specific health conditions including immunocompromising conditions or chronic comorbidities associated with end-organ disease (eg, heart failure, dementia, and chronic kidney disease) [14]. In a retrospective analysis of VHA users who had received a primary vaccine series and booster dose, those with age >65 years, high-risk comorbidities, or immunocompromised condition each experienced higher incidence of COVID-19 pneumonia requiring hospitalization or death [25]. While a broader reduction in illness severity among vaccinated inpatients has been observed [26], we observed significant differences in severity outcomes for COVID-19 by vaccination status among inpatient participants only when comparing moderate to mild illness. Reasons for this difference could include the relatively high proportion of fully vaccinated participants with severe illness enrolled during the early period of our study and our smaller study sample size.

We found that participants with SARS-CoV-2 infection had more hospital days than uninfected participants even after adjusting for age and CCI; this highlights the complexity of caring for patients with COVID-19 and potentially infection-related barriers to discharge. Our study cohort had a higher frequency of rehospitalization within 60 days than prior reports [27, 28]. Namely, 3.6% of 29 659 adults admitted to US hospitals with COVID-19 between February 15, 2020, and June 9, 2020, were re-admitted within 30 days, and 9.3% of 68 236 patients were re-admitted over study periods that included 30-day rehospitalization windows and longer periods in a meta-analysis of 28 studies [27, 28]. Enrollment with SARS-CoV-2 infection following vaccination was associated with increased risk of rehospitalization in our study, aligning with the age and medical complexity of this group.

The strengths of our study are that it includes a racially diverse national sample, an enrollment period covering periods of different transmission intensity in the United States, and distinct dominant SARS-CoV-2 variants. Another strength is that we define the severity of illness using the VASIC scale, a validated and COVID-19-specific measure of the type of medical care received, as opposed to using the medical facility location where care was provided (eg, an intensive care unit) as a proxy [21]. Using the VASIC measure limits misclassification of the severity outcome due to the practice of caring for patients in an intensive care unit for the sake of infection prevention rather than the intensity of care required. Finally, we had complete data on participant characteristics and outcomes.

Our study has limitations. First, our study population may not be representative of the broader population of VHA users because of potential differences in the veterans who were approached for enrollment or who agreed to enroll. We took steps to improve the representativeness of our sample by including sites from across the United States and using a standard training protocol for all sites. We also supported recruitment of veterans who were unable to consent for their own participation but who had a legally authorized consenting representative; however, there were still difficulties recruiting veterans who were unable to consent for their own participation. Second, our study is based in a male-predominant veteran population, so we had limited power to assess for relationships between gender and study outcomes. Third, the study was too small to analyze subgroups of participants defined by vaccination status. Fourth, as we did not restrict study enrollment to participants seeking care for COVID-19, our study may include participants who accessed VHA care primarily for non-COVID care needs and were found incidentally to test positive for SARS-CoV-2. As analyses suggest, universal SARS-CoV-2 testing has become standard at the time of hospitalization, and a greater proportion of the US population has partial immunity to SARS-CoV-2 from infection and/or vaccination. Therefore, patients who are SARS-CoV-2-positive may have an alternate primary reason for hospitalization [25, 29], and some of the outcomes observed in our study may be due to causes other than COVID-19. Until an established or validated method exists to clearly identify patients seeking care for COVID-19 at time of hospitalization or through electronic health data during follow-up, discerning care sought specifically for COVID-19 remains desirable but very challenging.

CONCLUSIONS

In an ongoing longitudinal prospective cohort study of COVID-19 in hospitalized US veterans, participants who had completed at least a primary vaccine series were older and had more comorbidities than those who were not vaccinated. Higher comorbidity score was an independent risk factor for more severe illness within 30 days, more hospital days, and rehospitalization within 60 days, after adjusting for vaccination status, age, sex, and calendar period of enrollment. These results suggest that inpatients who are older or have a greater number of comorbidities, even if vaccinated, warrant close attention as they are at increased risk of worse outcomes after SARS-CoV-2 infection.

Supplementary Data

Supplementary materials are available at Open Forum 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.

Acknowledgments

The authors thank all of the veterans who generously volunteered to participate in the EPIC³ study and the many study staff members who carry out the study.

Members of the group author list of EPIC³ Investigators. Mary-Claire Roghmann, VA Maryland Health Care System; Karen Coffey, VA Maryland Health Care System; Les Katzel, VA Maryland Health Care System; Emily Wan, VA Boston Health Care System; Federico Perez, VA Cleveland Health Care System; Robin Jump, VA Cleveland Health Care System; Rohit Manaktala, VA North Texas Health Care System; Lindsay Nicholson, VA Rocky Mountain Health Care System; Micah McClain, VA Durham Health Care System; Christopher Woods, VA Durham Health Care System; Gary Wang, VA North Florida Health Care System; Amy Vittor, VA North Florida Health Care System; John Theus, North Arkansas VA Health Care System; Sheran Mahatme, Milwaukee VA Health Care System; Nathan Gundacker, Milwaukee VA Health Care System; Javeria Haque, Milwaukee VA Health Care System; Harman Paintal, VA Palo Alto Health Care System; Matthew Stevenson, VA Palo Alto Health Care System; Joshua Baker, VA Philadelphia Health Care System; Chris Pfeiffer, VA Portland Health Care System; Patrick Powers, VA Salt Lake Health Care System; Julia Lewis, VA Salt Lake Health Care System; Patrick Danaher, VA South Texas Health Care System; Antonio Anzueto, VA South Texas Health Care System; McKenna Eastment, VA Puget Sound Health Care System.

Financial support. This work was supported by project CSP #2028 from the US Department of Veterans Affairs, Cooperative Studies Program.

Disclaimer. The contents do not represent the views of the US Department of Veterans Affairs or the United States Government.

Author contributions. J.M.R., J.D.S., K.C., C.M.H., E.L., and J.S.L. conceived of the project idea. A.T. and K.M. completed the analysis, with guidance from A.K. J.A., C.L., and D.W. managed the project. All authors critically reviewed the study data and interpreted results. J.M.R. wrote the first draft of the manuscript. All authors reviewed the manuscript and contributed to subsequent revisions.

Availability of data. Final data sets and biospecimens from EPIC³ will be deposited into a VA data repository. Interested parties may contact the CSP #2028 Coordinating Center at [email protected] for guidance on accessing study materials.

References

Author notes