Clinical Epidemiology and Risk Factors for Critical Outcomes Among Vaccinated and Unvaccinated Adults Hospitalized With COVID-19-VISION Network, 10 States, June 2021-March 2023.
Feb 17, 2024DOI: 10.1093/cid/ciad505Journal ArticleCID
Written by: Eric P GriggsPatrick K MitchellVictoria LazariuManjusha GaglaniCharlene McEvoyNicola P KleinNimish R ValviStephanie A IrvingNoah KojimaEdward StenehjemBradley CraneSuchitra RaoShaun J GrannisPeter J EmbiAnupam B KharbandaToan C OngKarthik NatarajanKristin DascombAllison L NalewayElizabeth BassettMalini B DeSilvaMonica DickersonDeepika KonathamBruce FiremanKatie S AllenMichelle A BarronMaura BeatonJulie ArndorferGabriela Vazquez-BenitezShikha GargKempapura MurthyKristin GoddardBrian E DixonJungmi HanNancy GriselChandni RaiyaniNed LewisWilliam F FadelMelissa S StockwellMufaddal MamawalaJohn HansenOusseny ZerboPalak PatelRuth Link-GellesKatherine AdamsMark W Tenforde
This study was conducted within the virtual SARS-CoV-2, influenza, and other respiratory viruses network (VISION), a multistate network that captures EHR data from healthcare systems with integrated laboratory, clinical, and vaccination records. VISION has performed serial evaluations of COVID-19 vaccine effectiveness since 2021 [20]. This analysis included adult patients (≥18 years of age) hospitalized at 8 healthcare systems in 10 US states (including 254 hospitals) from 1 June 2021 through 29 March 2023 (Supplementary Table 1). The analysis was reviewed and approved by institutional review boards at participating sites or under reliance agreement with the institutional review board of Westat. All activities were conducted consistent with applicable federal law and Centers for Disease Control and Prevention (CDC) policy.We included adults hospitalized with COVID-19 during the period of Delta predominance (June 2021) through the post-BA.4/BA.5–predominant period including XBB sublineages (March 2023). Consistent with described methods [21], cases were patients with a COVID-19–associated hospitalization, defined as having 1 or more COVID-19–like illness International Classification of Diseases, 10th Revision (ICD-10) discharge codes and a positive molecular test for SARS-CoV-2 within 14 days before or up to 72 hours after admission. Pregnant patients were excluded due to differences in COVID-19 screening and testing practices and criteria for hospitalization. To characterize underserved populations, the CDC/Agency for Toxic Substances and Disease Registry Social Vulnerability Index (SVI) was included using geocoded data, as previously described [22].COVID-19 vaccination status was determined using local immunization information systems, EHRs, and insurance claims and categorized as primary series only or primary series with additional doses of monovalent or updated bivalent messenger RNA (mRNA) boosters. mRNA vaccine doses received starting 2 September 2022 were considered bivalent [23]. Primary series completion was defined as documented receipt of 2 doses of an mRNA vaccine (BNT162b2 [Pfizer-BioNTech] and/or mRNA-1273 [Moderna]) or 1 dose of Ad26.COV2 (Janssen/Johnson & Johnson [J&J]) ≥14 days before the admission index date, defined as the date of respiratory specimen collection associated with the most recent SARS-CoV-2 test result prior to hospitalization or admission date if testing occurred following admission. For patients with immunocompromising conditions, completion of a primary series was defined as having received 3 doses of an mRNA vaccine or 1 dose of J&J followed by 1 additional mRNA dose [20]. Encounters among patients who received 1 or more doses of a COVID-19 vaccine but had not completed a primary series were excluded, as were those who received any sequence of vaccine products outside of the recommended schedules.
The study period began on the earliest day the Delta variant accounted for ≥50% of sequenced isolates at each site, based on local surveillance data, and ended when another variant accounted for ≥50% of sequenced isolates (Delta date range: 1 June–29 December 2021). Each subsequent variant/sublineage-predominant period was defined similarly (Omicron BA.1: 16 December 2021–30 March 2022; Omicron BA.2/BA.2.12.1: 17 March–29 June 2022; Omicron BA.4/BA.5: 19 June–29 October 2022; post–Omicron BA.4/BA.5: 29 October 2022–29 March 2023).
The primary outcome was “critical” COVID-19 outcomes, defined as a composite of hospitalization resulting in intensive care unit (ICU) admission and/or in-hospital death. This endpoint was selected to reflect critical disease [24] while utilizing outcomes collected across all sites. Secondary critical outcomes, measured at a subset of partner sites, included any use of noninvasive mechanical ventilation (NMV, including bivalve positive airway pressure or continuous positive airway pressure), or invasive mechanical ventilation (IMV). Additional markers of critical illness described in the supplementary analyses included ICD-10 codes indicating COVID-19–associated complications such as acute respiratory distress syndrome (ARDS), respiratory failure, and all-cause pneumonia.
Baseline demographics, clinical characteristics, and outcomes were stratified by variant period and described using counts and percentages or medians and interquartile ranges (IQRs). Trends across periods were assessed using the Jonckheere-Terpstra nonparametric test [25] for continuous variables and ordered categorical variables with >2 groups, and the Cochran-Armitage [26, 27] test for binary variables.
Key characteristics (vaccination status, age ≥65 years, ≥4 underlying medical condition [UMC] categories) of patients with critical outcomes were visualized by period using Euler diagrams, with the size of ellipses representing the proportion of patients with each characteristic and overlap between ellipses representing co-presence of characteristics. Counts of COVID-19 hospitalizations and proportions with critical outcomes (primary outcomes of ICU admission and/or in-hospital death, secondary outcomes including NMV and/or IMV) were graphed by week.
We evaluated risk factors for critical outcomes overall (ie, across all study periods) and during each variant period separately for both unvaccinated and vaccinated (primary series or more) populations. Due to the similar characteristics and outcomes among patients with critical outcomes during the BA.2, BA.4/BA.5, and post-BA.4/BA.5 periods, we collapsed these into 1 post-BA.1 period. A modified Poisson regression approach through a generalized linear model with robust error variance [28] was used to estimate adjusted risk ratios (aRRs), accounting for potential confounders including patient characteristics (age, sex, race, Hispanic ethnicity, Medicaid status), UMCs (number of condition categories or by individual conditions), and facility characteristics (VISION site, hospital size, hospital type, urban/rural classification). Addition of calendar time was also considered but had little influence on risk ratio estimates. UMCs were defined using ICD-10 discharge codes for the index encounter and assumed to be baseline conditions predating hospitalization, although the chronicity of conditions or new-onset diagnoses associated with the index hospitalization could not be definitively determined. Condition categories included respiratory (asthma, chronic obstructive pulmonary disease [COPD]), cardiovascular (heart failure, ischemic heart disease, hypertension, congenital heart disease), cerebrovascular (stroke, other), neurological/musculoskeletal (dementia, Down syndrome, other), endocrine/metabolic (diabetes, other), hematologic, renal, hepatic, or immunocompromising conditions (Supplementary Table 2). Among vaccinated, risk ratios were also adjusted for whether the patient had completed a primary vaccine series or a primary vaccine series plus 1 or more booster doses; other specifications, including models also adjusting for time since most recent dose or number of booster doses, generated similar findings. Depending on the risk factor, 1 of 2 regression models was run to assess UMCs. In the first model, a categorical variable counting the number of system categories affected was included as an independent variable. For the second model, individual indicators of each UMC were included as independent variables. Analyses were conducted using R software (version 4.1.2; R Foundation).
RESULTS
Of 66 166 COVID-19–associated hospitalizations in nonpregnant adults during the study period, 5678 (8.6%) were excluded due to being partially vaccinated (n = 5382), receiving vaccines outside of recommended schedules (n = 294), or receiving a vaccine product other than Janssen, Moderna, or Pfizer (n = 2), leaving 60 488 eligible hospitalizations. Median age of patients was 67 years (IQR, 54–79 years), 31 230 (51.6%) were male, 41 893 (69.3%) were White, and 42 728 (70.6%) were non-Hispanic (Table 1). Of patients with geocoded information on place of residence, 19.5% were in the highest SVI quartile. The median number of UMC categories was 2 (IQR, 1–3), with 8021 (13.3%) patients having an immunocompromising condition.
Age at hospitalization increased across periods, with a median of 60 years (IQR, 48–72 years) during Delta and 75 years (IQR, 65–84 years) during post-BA.4/BA.5 (P < .001) (Table 1). The proportion of patients with ≥4 UMC categories increased from 13.4% during Delta to 31.6% during post-BA.4/5 (P < .001). The proportion vaccinated with at least a primary series increased from 18.2% during Delta to 70.1% during post-BA.4/BA.5 (P < .001).
Overall, 13 536 (22.4%) hospitalizations included critical outcomes (Table 2), with 11 408 (18.9%) ICU admissions and 5708 (9.4%) deaths (Supplementary Table 3). The proportion of patients with critical outcomes was highest during the Delta period (Supplementary Table 3; Figure 1). Overall, 4751 of 33 411 (14.2%) hospitalizations required NMV and 5171 of 48 441 (10.7%) required IMV. Demographic characteristics of patients with critical outcomes followed similar patterns as all COVID-19 hospitalizations. Compared with all hospitalized patients, a greater proportion of those with critical illness had ≥4 UMC categories (32.8% vs 23.0%) and immunocompromising conditions (17.9% vs 13.3%), and a lower proportion were vaccinated with at least a primary series (34.9% vs 41.9%).
The proportion of patients with critical outcomes who were older (≥65 years) and had ≥4 UMC categories increased across periods (P < .001), while the unvaccinated proportion decreased (P < .001) (Table 2, Figure 2). The overall proportion of ICU admissions and in-hospital deaths declined across periods (P < .001) (Figure 1; Supplementary Table 3). These trends were consistent across secondary outcomes such as ARDS, NMV, and IMV (P < .001) (Supplementary Table 3, Supplementary Figure 1). Hospital length of stay decreased across periods for patients with critical outcomes who survived (ie, admitted and discharged from ICU), with a median of 5 days (IQR, 3–10 days) during Delta and 4 days (IQR, 3–7 days) during post-BA.4/BA.5 (P < .001). However, length of stay also decreased among in-hospital deaths, with a median of 11 days (IQR, 6–22 days) during Delta and 7 days (IQR, 4–14 days) during post-BA.4/BA.5 (P < .001).
Adjusting for UMCs and other factors, during the Delta period age >49 years was associated with an increased risk of critical outcomes (Figure 3 and 4). During later periods, however, the adjusted risk of critical outcomes was lower in older versus young adults. During the Delta period, adults aged 18–49 years comprised 32.8% of unvaccinated and 6.9% of vaccinated patients, whereas during the post-BA.1 period young adults comprised 16.5% of unvaccinated and 7.0% of vaccinated patients. Female sex was associated with a reduced risk of critical outcomes across periods (aRR, 0.82 and 0.85 in unvaccinated and vaccinated, respectively).
Patients with ≥4 UMCs were at an increased risk for ICU admission and/or death across all periods in both unvaccinated (overall aRR, 2.27 [95% confidence interval {CI}, 2.14–2.41]) and vaccinated (overall aRR, 1.73 [95% CI, 1.56–1.92]) groups. Presence of 2–3 UMCs was also a risk factor across multiple periods for unvaccinated, but with a lower relative strength of association. Most categories of UMCs were associated with a modest increase in risk of critical outcomes (aRR >1.0) across periods in both unvaccinated (Figure 3) and vaccinated (Figure 4) adults.
In a secondary analysis evaluating risk factors for severe respiratory illness (NMV and/or IMV), similar results were observed as in the primary analysis, although the magnitude of effect size such as for multiple UMCs was higher and an association between older age and lower independent risk of critical illness was observed across periods (Supplementary Figures 2A and 2B).
Visualizing the relationship between categories of UMCs and critical outcomes, the proportion of patients with a critical outcome increased as the number of UMC categories increased across all periods (Supplementary Figure 3). The most common combinations of UMC categories for patients with ≥4 categories included cardiovascular, neurological or musculoskeletal, endocrine or metabolic, and renal (Supplementary Table 4).
DISCUSSION
Among adults hospitalized across 10 states between June 2021 and March 2023, the epidemiology of COVID-19 changed markedly. Over time, the overall proportion of ICU admissions, in-hospital deaths, use of noninvasive or invasive mechanical ventilation or ARDS, and length of stay decreased. Changes in baseline patient characteristics included an increase in median age of 15 years across the study period and a majority of those with COVID-19–associated hospitalizations now having been vaccinated against COVID-19 as population-level coverage expanded. Among both unvaccinated and vaccinated patients, a high burden of UMCs was the strongest predictor of ICU admission and/or death. A small number of factors, including vaccination status, high burden of UMCs, and advanced age, identified a vast majority of patients who experienced critical outcomes (90%–97% across periods), suggesting target populations for future public health and clinical efforts.
Adjusting for confounders, clinical risk factors for critical COVID-19 outcomes were similar over time between unvaccinated and vaccinated patients. Regardless of vaccination status, the presence of UMCs across multiple organ system categories was the strongest risk factor for critical disease. This suggests that overall frailty may play a larger role in susceptibility to critical disease than a singular diagnosis. Notably, we captured general condition categories, and within a single category (eg, immunocompromising conditions) there may be a gradient of risk. While VISION does not collect data on how well-controlled UMCs are, ongoing management of UMCs may minimize complications associated with COVID-19 [29]. Future studies should evaluate risk of critical outcomes accounting for severity or level of control of underlying conditions to provide greater understanding.
Various medical conditions and age have been consistently associated with an increased risk of severe COVID-19 illness from previous studies [30–32]. In our analysis, older adults were overrepresented among both total hospitalizations and severe hospitalizations across periods, and older age was associated with an increased risk of critical outcomes (ICU admission or death) during the Delta variant period. This association was reversed, however, during the post-BA.1 period, with older adults hospitalized with COVID-19 at lower risk for critical outcomes adjusting for UMCs and other confounders. These findings may be attributable to several factors, including a lower threshold for SARS-CoV-2 testing and admission for less severe disease in older adults, residual confounding based on types or severity of medical conditions or unmeasured factors, differential use of early antiviral therapies across ages, advanced directives or goals of in-hospital care in older adults, or differences in the clinical presentation of illness by age. For example, older adults may be more likely to be hospitalized due to exacerbation of chronic medical conditions associated with SARS-CoV-2 infection rather than primary viral pneumonia. Inclusion in this analysis was also conditioned on hospitalization, and older adults in institutional settings (eg, skilled nursing facilities) likely to have poor outcomes may not have been hospitalized. Regardless, the average age and proportion of older adults hospitalized with COVID-19 increased over time, highlighting that the greatest burden of critical COVID-19 outcomes is experienced among older adults.
Compared with all hospitalized patients, a lower proportion of patients with critical COVID-19 outcomes were vaccinated with at least a primary vaccination series; however, the severity of disease among both vaccinated and unvaccinated populations declined over time. These data support that COVID-19 severity declined during Omicron variant predominance [33, 34]. Risk of critical illness may have been attenuated by vaccination [9], the increased prevalence of immunity induced by prior SARS-CoV-2 infection [35, 36], changes in aspects of the predominant SARS-CoV-2 variant [37], or changes in clinical management [17–19].
This analysis was subject to limitations. First, as eligible encounters occurred across multiple variant periods, clinical management may have changed. Data on treatments such as antivirals were not captured by VISION partners throughout the study period, limiting our ability to assess the impact of changing therapeutic practices. Second, prior infection was unable to be reliably ascertained from available testing data, complicating interpretation of conclusions that aim to disentangle the contribution of vaccine-induced immunity and infection-induced immunity. Third, we could not reliably capture whether a patient with COVID-19–like illness and a positive SARS-CoV-2 test was admitted primarily because of COVID-19 or due to other contributing conditions. EHR data may also not fully capture all UMCs and could result in misclassification of higher-risk patients and underreporting of intensive care received, particularly during periods of high incidence of SARS-CoV-2 infections when non-ICU beds were used for critically ill patients [38, 39]. Fourth, these findings are from data collected in 10 states and may not be generalizable to the entire US population. Due to the absence of comprehensive national integrated systems across a representative population, networks like VISION rely on health systems that capture robust and integrated clinical, laboratory, and vaccination information. Furthermore, as our study population is patients already hospitalized with COVID-19, our findings are not intended to reflect a wider population outside the hospital setting. Fifth, our primary severity outcomes of ICU admission and/or in-hospital death are not specific to COVID-19. Sixth, chronicity of medical conditions was not assessed. While documented conditions such as diabetes and COPD are likely to represent chronic conditions, it is possible that some conditions included new-onset diagnoses or may have resulted from COVID-19. Encounter-based reporting of conditions may be incomplete without a lookback period predating admission; additional information including capture of all vaccine doses may also be incomplete.
In conclusion, from the Delta-predominant period to the post-BA.4/BA.5 period, the epidemiology of COVID-19 among hospitalized patients changed markedly. Over time, the proportion of ICU admissions, in-hospital deaths, use of NMV and IMV, and hospital length of stay decreased. However, in all periods the majority of hospitalizations resulting in ICU admission or death among both vaccinated and unvaccinated patients were among patients who were older (≥65 years) or had multiple underlying medical conditions. Continued research is needed to understand the changing epidemiology of COVID-19 hospitalizations with the emergence of new SARS-CoV-2 variants or sublineages, changes in population-level immunity, and development of new vaccines and therapies.
SG
Dr. Shikha Garg
Infectious Disease · General Infectious DiseaseAtlanta VA Medical Centershikhag