This retrospective study was conducted at a specialized hematology hospital in Tianjin, China, between January 2014 and December 2024. Patients were eligible if they were ≥14 years old, had at least one positive blood culture for Enterococcus spp., and had complete medical records with 30-day follow-up data. In cases of a second episode of EBSI occurring within 90 days of the initial episode, only the first episode was included for analysis; subsequent episodes were classified as recurrences. This study was approved by the Ethical Committee of the Institute of Hematology and Blood Diseases Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College.

Demographic and clinical data were retrieved from the electronic medical records, including type and phase of hematologic disease; chemotherapy or immunosuppressive therapy; allogenic or autologous HSCT (allo-HSCT, auto-HSCT); presence of grade II-IV acute graft-versus-host disease (aGVHD) or moderate-to-severe chronic graft-versus-host disease (cGVHD) (Jagasia et al., 2015; Schoemans et al., 2018); Charlson comorbidity index (CCI); absolute neutrophil count (ANC); duration of neutropenia before and after bacteremia; suspected infection source; site of acquisition; occurrence of septic shock; microbiological data (species, vancomycin and ampicillin resistance); follow-up blood culture results (recommended 48–96 h after antibiotic initiation (Lopez-Cortes et al., 2013)); antibiotic regimens; and clinical outcomes.

In accordance with current guidelines (Rosselli Del Turco et al., 2021), echocardiography was recommended at our center for EBSI patients meeting any of the following criteria: recent HSCT, community-acquired EBSI, prolonged fever or recurrent fever after initial defervescence, persistent or recurrent bacteremia, a predisposing condition for endocarditis (e.g., native valve disease, prosthetic valve, or any cardiac implantable electronic devices), clinical signs of endocarditis (e.g., embolic events, conjunctival hemorrhage, Janeway lesions and immunologic phenomena), or evidence of cardiac dysfunction.

The primary outcome was all-cause mortality within 30 days of the first positive blood culture. Secondary outcomes included 14-day all-cause mortality, infection-related mortality and 90-day recurrence. Death was considered infection-related if it occurred before resolution of signs or symptoms, or within 7 days of bacteremia onset, without other identifiable causes (Lopera et al., 2024). Recurrence was defined as a microbiologically confirmed Enterococcus-positive blood culture with the same species and resistance profile as the initial isolate, occurring within 90 days of the initial EBSI episode and after completion of antibiotic therapy (Peterson et al., 2009; Cattaneo et al., 2021).

The onset of EBSI was defined as the date the first positive blood culture sample was collected. Neutropenia was defined as an ANC < 0.5×10⁹/L, and severe neutropenia as ANC < 0.1×10⁹/L. Hematologic diseases were categorized into three groups: bone marrow failure syndromes (aplastic anemia [AA] and myelodysplastic syndromes [MDS]), acute leukemia (acute myeloid leukemia [AML], acute lymphoblastic leukemia [ALL] and mixed phenotype acute leukemia [MPAL]), and other hematologic diseases. Standard-risk disease was defined as acute leukemia in complete remission (CR); lymphoma or multiple myeloma in CR or partial remission (PR); MDS with <5% blasts; or untreated severe aplastic anemia (SAA). High-risk disease included newly diagnosed acute leukemia, induction failure or relapse, lymphoma or multiple myeloma with stable disease or progression, as well as MDS/SAA with transfusion dependence and no response to treatment (Wang et al., 2015; Zhang et al., 2023). Grade II-IV aGVHD or moderate-to-severe cGVHD was collectively defined as severe GVHD.

Acquisition of EBSI was classified as: (i) community-onset, defined as a positive blood culture obtained at or within 48 h of hospital admission; (ii) nosocomial, defined as a positive blood culture obtained ≥48 h after hospitalization. According to the Centers for Disease Control and Prevention (CDC) criteria, the source of bacteremia was categorized as mucosal barrier injury laboratory- confirmed bloodstream infection (MBI-LCBI), non–MBI primary EBSI, or secondary EBSI (Centers for Disease Control and Prevention, 2025). For patients with an eradicable focus, appropriate source control interventions (e.g., abscess drainage, excision, or catheter removal) were performed. Breakthrough EBSI was defined as the occurrence of EBSI in patients who were receiving systemic antimicrobial therapy for at least 48 hours prior to the collection of the index positive blood culture (Rangaraj et al., 2010). Septic shock was defined as systolic pressure <90 mmHg despite adequate fluid resuscitation or the need for vasopressor agents (Rhodes et al., 2017). Persistent bacteremia was defined as positive blood cultures for Enterococcus spp. persisting ≥72 h after initiating appropriate antibiotic therapy. Metastatic infection was defined as definite infective endocarditis fulfilling the modified Duke criteria (Li et al., 2000) or a secondary infection at a site distant from the primary focus. Polymicrobial bacteremia was defined as the isolation of Enterococcus spp. with one or more additional bacterial species from the same or another blood culture obtained within 24 h, meeting CDC criteria for bloodstream infection (Centers for Disease Control and Prevention, 2025). Appropriate antibiotic therapy was defined as administration of at least one in vitro-active agent against the Enterococcus isolate. For polymicrobial bacteremia, appropriate therapy required coverage of all identified pathogens.

Clinical samples were processed at the hospital microbiology laboratory using an automated VITEK 2 Compact system for species identification and susceptibility testing. Antibiotic susceptibilities were defined according to current Clinical and Laboratory Standards Institute (CLSI) criteria.

Categorical variables were compared using Chi-square test or Fisher’s exact test, as appropriate. Continuous variables were analyzed using the Mann-Whitney U test for nonparametric data. A two-tailed P value <0.05 was considered statistically significant.

Candidate predictors of 30-day mortality were identified using least absolute shrinkage and selection operator (LASSO) logistic regression. All candidate variables were restricted to those available within 72 hours after blood culture collection, as most Enterococcus-positive results are available within this timeframe. Predictors with non-zero coefficients at the minimum cross-validated lambda were retained. A multivariable logistic regression model including variables selected by LASSO was initially constructed. To enhance parsimony and clinical interpretability, nested models were subsequently compared. Variables whose exclusion did not adversely affect model discrimination or calibration were excluded. The final model was determined based on overall predictive performance (area under the receiver operating characteristic curve [AUROC], calibration, Brier score, decision curve analysis [DCA]) and model simplicity.

A clinical risk score was derived from the regression coefficients of the final multivariable model. The detailed calculation method is provided in the Supplementary Methods. Internal validation of both the regression model and the point-based score was performed using 2000-bootstrap resampling. The optimal cutoff value for risk stratification was determined using ROC curve analysis of the total risk score, applying the Youden index. Because the risk score was integer-based, the identified cutoff was rounded up to the nearest integer for clinical applicability. Data analyzes were performed using R software version 4.5.1.

From 2014-2024, a total of 192 patients (≥14 years) with hematologic diseases complicated by EBSI were included. Baseline clinical characteristics are summarized in Table 1. The median age was 43 years (interquartile range [IQR], 31.0–55.0), and 57.3% (n=110) were male. Acute leukemia was the most common underlying hematologic disease (n=139, 72.4%), including AML (n=96), ALL (n=39), and MPAL (n=4). Bone marrow failure syndromes accounted for 38 patients (19.8%), comprising MDS (n=18) and AA (n=20). The remaining 15 patients (7.8%) had other hematologic disorders, including multiple myeloma (n=3), Evans syndrome (n=1), lymphoma (n=6), and hemophagocytic lymphohistiocytosis (n=5). Ninety patients (46.9%) had high-risk hematologic diseases. Most patients (n=178, 92.7%) had received chemotherapy or immunosuppressive therapy within 1 month prior to EBSI. Within 100 days prior to EBSI, 44 (22.9%) and 6 (3.1%) patients had undergone allo-HSCT and auto-HSCT respectively. Fifteen patients (7.8%) had severe GVHD at the time of EBSI, including 13 with grade II-IV aGVHD and 2 with severe cGVHD. The 30-day mortality rate was significantly higher in patients with severe GVHD than those without (46.7% [7/15] vs. 20.3% [36/177], p=0.043).

Enterococcus faecium was the predominant pathogen (n=137, 71.4%), followed by Enterococcus faecalis (n=40, 20.8%) and other Enterococcus species (n=15, 7.8%). The temporal distribution of species is shown in Figure 1. Polymicrobial bloodstream infection was identified in 34 patients (17.7%). The detaileddistribution of the accompanying pathogens is summarized in Supplementary Table 1. At EBSI onset, 162 patients (84.4%) had neutropenia, of whom 145 had severe neutropenia. The median duration of neutropenia prior to EBSI onset was 12.0 days (IQR: 4.0–19.3). Most infections were nosocomial (n=184, 95.8%). The predominant infection type was MBI-LCBI (n=140, 72.9%), followed by secondary EBSI (n=34, 17.7%), most commonly of gastrointestinal origin (n=15). Ninety-three patients (48.4%) had pneumonia (of any etiology) at the time of EBSI onset, which was associated with significantly higher 30-day mortality compared to those without pneumonia (30.1% [28/93] vs. 15.2% [15/99], p=0.013). Among these patients, microbiologically confirmed pulmonary pathogens were identified in 12 cases, whereas the remaining cases were diagnosed based on clinical and radiological findings. The identified pathogens were heterogeneous, involving Gram-negative bacteria (including Acinetobacter baumannii, Pseudomonas aeruginosa and Stenotrophomonas maltophilia) and fungal pathogens (Candida species and molds). Regarding antimicrobial susceptibility, 67.7% of isolates (n=130) were resistant to ampicillin, including 126 E. faecium and 4 other Enterococcus species; all E. faecalis isolates were susceptible. Vancomycin resistance was rare, detected in only 3 isolates (1.6%), all E. faecium, indicating a low prevalence of vancomycin resistance in this cohort.

A total of 138 patients (71.9%) developed breakthrough EBSI while receiving antibiotics, with a median antibiotic exposure of 9.0 days (IQR, 5.0–12.0). The antibiotics administered at the time of breakthrough infection and their duration are summarized in Supplementary Table 2. Breakthrough infections occurred most frequently during carbapenem therapy (n=71, 51.4%), with a median duration of 9.0 days (IQR, 5.5–12.5), followed by cephalosporins (n=29, 21.0%) with a median duration of 9.0 days (IQR, 7.0–12.0). Metastatic infection developed in 9 patients (4.7%), involving soft tissues (n=5), lungs (n=3) and the abdominal cavity (n=1). Echocardiography was performed in 44 patients as clinically indicated, and no cases of metastatic endocarditis were identified. Persistent bacteremia was observed in 15 patients (7.8%), and septic shock in 12 patients (6.3%). Among patients with septic shock, 11 died within 30 days whereas only 1 survived, demonstrating a significant association with 30-day mortality (91.7% [11/12] vs. 17.8% [32/180], p<0.001).

The vast majority of patients (92.2%, n=177) received appropriate antibiotic therapy within 72 hours of blood culture collection. The antibiotics used for treatment are summarized in Supplementary Table 3, with linezolid and vancomycin being the most frequently administered agents for EBSI in our center. The median duration of antibiotics was 12.0 days (IQR: 8.0–17.0).

All-cause mortality rates at 14 and 30 days were 13.5% (n=26) and 22.4% (n=43), respectively. Infection-related mortality was 17.7% (n=34), and the 90-day recurrence rate was 3.1% (n=6). In univariate analysis, factors significantly associated with 30-day mortality included age ≥50 years (p=0.010), type of hematologic disease (p=0.042), severe GVHD (p=0.043), polymicrobial bacteremia (p=0.047), pneumonia at the time of EBSI (any pathogen) (p=0.013), septic shock (p<0.001), and duration of antibiotic therapy (p<0.001).

LASSO logistic regression identified age ≥50 years, severe GVHD, septic shock, pneumonia, and type and stage of hematologic disease as candidate predictors with non-zero coefficients at the minimum cross-validated lambda. These variables were subsequently entered into a multivariable logistic regression model. To achieve a more parsimonious and clinically interpretable model, nested models were further evaluated. The final model included age ≥50 years, severe GVHD, septic shock, pneumonia, and high-risk hematologic disease, achieving an optimal balance between predictive accuracy, clinical interpretability, and model parsimony.

In the final model, age ≥50 years (aOR=2.29, 95%CI 1.05-5.10, p=0.038), severe GVHD (aOR=6.06, 95%CI 1.86-19.96, p=0.003), septic shock (aOR=30.01, 95%CI 6.49-291.53, p<0.001) were independently associated with 30-day mortality. The model demonstrated good discrimination, with an AUROC of 0.79 (95% CI: 0.705–0.867) after 2000-bootstrap internal validation (Figure 2A). The Hosmer–Lemeshow goodness-of-fit test indicated excellent calibration (X²=0.994, df=5, P = 0.963), which was visually supported by the calibration plot (Figure 2B).

A simplified risk score was derived from regression coefficients: age ≥50 years (1 point), GVHD (2 points), septic shock (4 points), pneumonia (1 point) and high-risk hematologic disease (1 point) (Table 2). In bootstrap validation (B = 2000), the risk score achieved an AUROC of 0.76 (95% CI: 0.676-0.839), comparable to the original model (Figure 2C, 2D). DCA was conducted for both the multivariable logistic regression model and the simplified risk score. The DCA curves demonstrated potential clinical usefulness of the prediction tools (Figure 3). Using the optimal cutoff, patients were stratified into low-risk (<2 points) and high-risk (≥2 points) groups. The high-risk group had significantly higher 30-day mortality compared to the low-risk group (39.0% vs. 11.3%, P<0.001) (Figure 4).

In univariable analysis (Table 1), duration of appropriate antibiotic therapy was significantly associated with 30-day mortality (p<0.001). To minimize immortal time bias, patients who died within 10 days of EBSI onset or received less than 5 days of appropriate antibiotic therapy were excluded, leaving 166 patients for further analysis. Multivariate logistic regression analysis showed that duration of appropriate antibiotic therapy was not significantly associated with mortality (OR = 0.95, 95%CI 0.88-1.00, p=0.076) (Supplementary Table 4).