This analysis was conducted by the Investigating Respiratory Viruses in the Acutely Ill (IVY) network, a multisite inpatient network comprising 26 hospitals in 20 U.S. states. The IVY network employs a test-negative, case-control design to assess VE; methods have been described previously [22–25]. Briefly, site personnel prospectively enrolled patients aged ≥18 years admitted to IVY network hospitals who met a COVID-19-like illness case definition and received SARS-CoV-2 clinical testing. COVID-19-like illness was defined as ≥1 of the following: fever, cough, shortness of breath, new or worsening findings on chest imaging consistent with pneumonia, or hypoxemia (defined as oxygen saturation [SpO₂] <92% or supplemental oxygen use for patients without chronic oxygen needs, or escalation of oxygen therapy for patients on chronic supplemental oxygen).

Enrolled patients were tested for SARS-CoV-2 clinically at the local hospital and additionally had nasal swab specimens tested at a central laboratory at Vanderbilt University Medical Center for SARS-CoV-2, influenza, and respiratory syncytial virus (RSV) using RT-PCR (Supplementary Methods). In this analysis, case-patients were defined by a positive test for SARS-CoV-2 in local or central laboratories for a specimen collected within 10 days of symptom onset and 3 days of hospital admission. Control-patients were defined by a negative test for SARS-CoV-2. Case-patients who tested positive for influenza or RSV were excluded, while control-patients who tested positive for influenza were excluded due to correlation between COVID-19 and influenza vaccination behaviors [26]. Demographic and clinical data were collected through electronic medical record (EMR) review and patient or proxy interview.

This analysis included adults admitted to IVY network hospitals from October 18, 2023–March 9, 2024. The analysis period spans the first date on which a patient enrolled in the network was admitted with sequence-confirmed JN lineage infection to the last week during which a patient was admitted with sequence-confirmed XBB lineage infection.

These activities were determined to be public health surveillance with waiver of informed consent by institutional review boards at CDC and each enrolling site and were conducted in accordance with applicable federal law and CDC policy (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).

Verification of COVID-19 vaccination status was performed using hospital EMR, state immunization information systems (IIS), or patient or proxy interviews using a hierarchical approach. Evidence of vaccination from either EMR or IIS were preferentially used even if interview data were available. If EMR/IIS data were unavailable, plausible interview data with known or estimated dates of COVID-19 vaccine receipt were used. Patients were classified into two vaccination groups: 1) those who received an updated 2023–2024 COVID-19 vaccine dose ≥7 days before illness onset, and 2) those who did not receive an updated dose, comprising both patients who had received previous original monovalent and/or bivalent COVID-19 vaccine doses and patients who had never received a COVID-19 vaccine dose. We excluded patients if they received an updated COVID-19 dose <7 days before illness onset, received a non-updated dose <60 days before their updated dose, received a non-updated dose after September 10, 2023, received an updated dose before September 13, 2023, or received >1 updated dose.

Clinical severity of patients with JN and XBB lineage infection was characterized using severe in-hospital outcomes from hospital presentation to the first of hospital discharge, patient death, or hospital day 28 (Supplementary Methods) [23,24]. These outcomes included 1) supplemental oxygen therapy (defined as supplemental oxygen at any flow rate and by any device for those not on chronic oxygen therapy, or with escalation of oxygen therapy for patients receiving chronic oxygen therapy); 2) advanced respiratory support (defined as new receipt of high-flow nasal cannula, non-invasive ventilation, or invasive mechanical ventilation [IMV]); 3) intensive care unit (ICU) admission; or 4) a composite of IMV or death.

SARS-CoV-2–positive specimens were sent to the University of Michigan (Ann Arbor, Michigan) for whole-genome sequencing to identify lineage. Sequencing was performed using the Oxford Nanopore Technologies Midnight protocol on a GridION instrument. Lineages were called using PANGO (phylogenetic assignment of named global outbreak lineages) nomenclature (version 4.3.1). Sequences were considered adequate for lineage identification if they had a Nextclade (version 3.4.1) completeness score >80 and Nextclade quality control status of “good” or “mediocre”. Patients with sequences passing these criteria were categorized as either having JN lineage infection/hospitalization (BA.2.86 and its descendants) or XBB lineage infection/hospitalization (all other co-circulating lineages).

VE against COVID-19 hospitalization was estimated using multivariable logistic regression by comparing the odds of receipt of an updated COVID-19 dose versus no updated dose between case-patients and control-patients. Odds ratios were adjusted for age (18–49, 50–64, ≥65 years), sex (male, female), race/ethnicity (Hispanic or Latino, non-Hispanic Black, non-Hispanic White, non-Hispanic other race, other/unknown), U.S. Health and Human Services (HHS) region, admission date in biweekly intervals, and Charlson comorbidity index (0, 1–2, 3–4, 5–6, ≥7) [27]. VE against COVID-19–associated hospitalization was calculated as (1 – adjusted odds ratio) × 100%. VE estimates were calculated separately for case-patients with sequence-confirmed SARS-CoV-2 XBB and JN lineage infections. We conducted sensitivity analyses to evaluate robustness to adjustments for temporal confounding (adjusting for continuous or categorical time, matching on time, and restricting the analysis period) (Supplementary Results).

Multivariable logistic regression was used to estimate the adjusted odds ratio of each severe in-hospital outcome among case-patients with JN lineage hospitalizations compared to those with XBB lineage hospitalizations. Models were adjusted for age (18–49, 50–64, ≥65 years), sex (male, female), race/ethnicity (Hispanic or Latino, non-Hispanic Black, non-Hispanic White, non-Hispanic other race, other/unknown), HHS region, admission date in biweekly intervals, Charlson comorbidity index (0, 1–2, 3–4, 5–6, ≥7), and COVID-19 vaccination status (updated dose, no updated dose). In sensitivity analyses, estimates were stratified by updated dose receipt (in which COVID-19 vaccination status was not adjusted for in regression models) and COVID-19 treatment receipt, defined as receipt of COVID-19 antivirals, steroids, or immunomodulators, for strata with sufficient sample sizes. We conducted sensitivity analyses to evaluate robustness to adjustments for temporal confounding (adjusting for continuous or categorical time and matching on time) (Supplementary Results). R (version 4.1.3) was used to conduct all analyses.

During October 18, 2023–March 9, 2024, a total of 7,058 patients were admitted with COVID-19–like illness and enrolled in the IVY Network from 26 hospitals in 20 U.S. states. A total of 1,496 (21.2%) patients did not meet inclusion criteria and were removed from the analysis (Supplementary Figure 1). Identification of a SARS-CoV-2 lineage through viral whole-genome sequencing was successful for 62.5% (982/1570) of case-patients meeting all other inclusion criteria. Clinical and demographic characteristics were similar among case-patients with successfully sequenced specimens versus all case-patients (Supplementary Table 1). Among the 5,562 patients meeting all inclusion criteria, 585 (10.5%) were case-patients with sequence-confirmed XBB lineage infection, 397 (7.1%) were case-patients with sequence-confirmed JN lineage infection, and 4,580 (82.3%) were test-negative control-patients.

Among the 5,562 included patients, median age was 67 years (IQR = 55–76 years), 58.9% were non-Hispanic White, 49.9% were male, and 47.9% had a Charlson comorbidity index ≥5 (Table 1). Control-patients had a lower median age (65 years, IQR = 53–75 years) than case-patients with XBB lineage (72 years, IQR = 62–80 years) or JN lineage (71 years, IQR = 60–81 years) infection (P<.001); distributions of race/ethnicity and sex were more similar (Table 1). Control-patients had lower median Charlson comorbidity indices (4, IQR = 2–6) than case-patients with XBB lineage (5, IQR = 3–7) or JN lineage (5, IQR = 3–7) infections (P<.001). The prevalence of self-reported or documented previous SARS-CoV-2 infection was similar among case-patients with JN lineage infection (27.7%) and control-patients (31.0%), but higher than that among case-patients with XBB lineage infection (16.8%) (P<.001). A total of 53 (9.1%) case-patients with XBB lineage infection, 78 (19.6%) case-patients with JN lineage infection, and 844 (18.4%) control-patients had received an updated COVID-19 vaccine dose (Table 1). Documentation of updated COVID-19 vaccination status was based on hospital EMR or state IIS (N = 886; 90.9%) and patient or proxy interviews (N = 89; 9.1%).

The most common SARS-CoV-2 lineages among the 585 case-patients with XBB lineage infection were HV.1 (n = 157, 26.8%), FL.1.5.1 (n = 29, 5.0%), and XBB.1.16.6 (n = 26, 4.4%). Among the 397 case-patients with JN lineage infection, 6 (1.5%) were infected with BA.2.86 or BA.2.86.1, 382 (96.2%) with JN.1 or its descendants, and 9 (2.3%) with other descendants of BA.2.86. The proportion of case-patients admitted with JN lineage infection increased steadily beginning in December 2023, and JN lineages became predominant (>50% prevalence) by the week ending January 6, 2024 (Figure 1).

We estimated effectiveness of an updated COVID-19 vaccine dose against COVID-19–associated hospitalization due to XBB and JN lineage infection. VE against XBB lineage hospitalization in the first 7–89 days after receiving an updated dose was 54.2% (95% CI = 36.1%–67.1%; median interval since updated dose = 44 days [IQR = 22–67]), and VE against JN lineage hospitalization in the first 7–89 days was 32.7% (95% CI = 1.9%–53.8%; median interval since updated dose = 56 days [IQR = 31–74]) (Figure 2). VE estimates against XBB lineage hospitalization beyond 90 days from vaccination could not be calculated due to the shift in variant predominance from XBB to JN lineages by January 2024. VE against JN lineage hospitalization 90–179 days after receiving an updated dose was 23.4% (95% CI = -11.8% to 47.6%; median interval since updated dose = 118 days [IQR = 107–130]). Estimates were similar when adjusting for time using continuous or categorical calendar time, matching on time using conditional logistic regression, and restricting the analysis period to timeframes during which vaccination coverage was approximately stable (Supplementary Figures 2 and 3).

Among 585 case-patients with XBB lineage infections, 366 (62.6%) received supplemental oxygen therapy, 99 (16.9%) received advanced respiratory support, 87 (14.9%) were admitted to an ICU, and 44 (7.5%) received IMV or died. Among 397 case-patients with JN lineage infections, 245 (61.7%) received supplemental oxygen therapy, 76 (19.1%) received advanced respiratory support, 61 (15.4%) were admitted to an ICU, and 32 (8.1%) received IMV or died. The unadjusted proportions of these severe in-hospital outcomes were similar by lineage (Figure 3). The adjusted odds ratios comparing case-patients with JN versus XBB lineage infection for receipt of supplemental oxygen therapy (0.83; 95% CI = 0.54–1.28; P=0.40), advanced respiratory support (0.66; 95% CI = 0.38–1.12; P=0.12), ICU admission (0.80; 95% CI = 0.46–1.38; P=0.42), and IMV or death (0.69; 95% CI = 0.34–1.40; P=0.30) indicated no significant differences in clinical severity between lineage groups (Figure 3). Adjusted odds ratios were similar when restricting to patients who received COVID-19 treatment (n = 810, 82.5% of case-patients) and to patients who did not receive an updated vaccine dose (n = 851, 86.7% of case-patients) (Supplementary Figure 4). Estimates were similar adjusting for calendar time as a continuous or categorical variable and matching on time using conditional logistic regression (Supplementary Figure 5).