Abstract
Background
Escherichia coli bloodstream infection (E. coli-BSI) is a common cause of sepsis and may progress to septic shock. However, the association between the albumin–bilirubin (ALBI) score and septic shock in patients with E. coli-BSI remains unclear.
Methods
This retrospective cohort study included adults with E. coli-BSI admitted to Beijing Luhe Hospital between 2019 and 2024. The ALBI score was calculated from serum albumin (ALB) and total bilirubin measured within 24 hours of the index blood culture. Associations between ALBI and septic shock were assessed using multivariable logistic regression, restricted cubic splines (RCS), subgroup and sensitivity analyses. Receiver operating characteristic (ROC) analysis assessed discrimination. Exploratory LASSO regression and SHAP analyses evaluated variable contributions.
Results
Among 334 patients, 97 (29%) developed septic shock, with the proportions increasing across ALBI tertiles (T1: 5.4%; T2: 27%; T3: 54.5%; p < 0.001). Higher ALBI scores were independently associated with septic shock (adjusted OR, 5.12; 95% CI, 2.37–11.05; p < 0.001), with higher odds in T3 than T1 (adjusted OR, 7.40; 95% CI, 2.63–20.83; p < 0.001). The association remained consistent across subgroup and sensitivity analyses. ALBI showed moderate discrimination (AUC, 0.796; 95% CI, 0.745–0.847), which was inferior to SOFA (AUC, 0.909; 95% CI, 0.877–0.941; p < 0.001), but comparable to ALB alone (AUC, 0.816; 95% CI, 0.766–0.865; p = 0.22). Exploratory LASSO and SHAP analyses identified ALBI as an important contributor.
Conclusion
Higher ALBI scores were independently associated with septic shock in adults with E. coli-BSI. ALBI showed moderate discrimination but was inferior to SOFA and not significantly better than ALB alone. These findings indicate an association between ALBI and septic shock but do not establish its predictive or clinical utility. Prospective multicenter studies are warranted to determine its clinical value for risk assessment.
Keywords: albumin–bilirubin score, ALBI, bloodstream infection, Escherichia coli, septic shock
Plain Language Summary
Why was this study done?
Escherichia coli bloodstream infection can progress to life-threatening septic shock. We investigated whether the albumin–bilirubin (ALBI) score, based on two routine available blood tests, serum albumin and total bilirubin, was associated with septic shock.
What did the researchers do and find?
Among 334 adults with E. coli bloodstream infection, 97 (29%) developed septic shock. Higher ALBI scores were independently associated with septic shock, with a consistent positive association across several analyses.
The ALBI showed moderate discrimination for septic shock, but performed lower discrimination than SOFA and did not significantly outperform albumin alone.
What do these results mean?
ALBI is a simple laboratory-based marker associated with septic shock in E. coli bloodstream infection. However, these findings do not establish causality or clinical utility. Prospective multicenter studies are needed to determine whether ALBI adds clinically meaningful information beyond established severity measures.
Introduction
Escherichia coli bloodstream infection (E. coli-BSI) is a leading cause of community- and hospital-acquired bacteremia and contributes substantially to morbidity and mortality worldwide.1 Despite advances in antimicrobial therapy and supportive care, some patients with E. coli-BSI deteriorate rapidly and progress to septic shock, which is associated with mortality and healthcare resource use.2 Early recognition of patients at increased risk of septic shock is therefore clinically important for timely assessment and management. While pathogen virulence influences disease severity,3 host-related factors, rather than microbial characteristics alone, may play an important role in the trajectory from bacteremia to septic shock.4,5 Among these host factors, hepatic dysfunction has emerged as an important contributor to adverse outcomes in sepsis.6 Sepsis-associated liver injury7—characterized by cholestasis, impaired synthetic function, and altered bilirubin metabolism—reflects both overwhelming systemic inflammation and microcirculatory failure.
The Albumin–bilirubin (ALBI) score, derived from serum albumin (ALB) and total bilirubin (TBIL), was originally developed to assess hepatic reserve in patients with hepatocellular carcinoma.8 Unlike many composite scores, the ALBI score is based on objective laboratory parameters that are routinely available in clinical practice. Emerging evidence suggests that the ALBI score or grade is associated with disease severity and adverse outcomes in critically ill patients.9–12 Hypoalbuminemia may reflect inflammation-induced redistribution, increased vascular permeability, and impaired hepatic synthesis, whereas hyperbilirubinemia may indicate sepsis-associated cholestatic liver injury and impaired hepatic clearance. Together, these abnormalities may reflect systemic hepatic, inflammatory, and endothelial disturbances associated with severe infection and circulatory dysfunction.
However, the association between ALBI score and septic shock in patients with E. coli-BSI has not been well characterized. We therefore conducted a retrospective cohort study to investigate this association. We hypothesized that a higher ALBI score would be associated with higher odds of septic shock in patients with E. coli-BSI.
Materials and Methods
Study Design
This retrospective cohort study was conducted at Beijing Luhe Hospital, Capital Medical University (Beijing, China). Consecutive adults (≥18 years) with confirmed E. coli-BSI between January 2019 and December 2024 were screened for eligibility. The study protocol adhered to the Declaration of Helsinki and was approved by the institutional review board (No. 2025-LHKY-017-01). Informed consent was waived due to the retrospective design. Reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.13
Study Population
De-identified clinical data were extracted from electronic medical records. Inclusion criteria were: ≥1 positive blood culture for E. coli with compatible clinical features (excluding contaminants) and E. coli as the sole blood isolate. For multiple episodes, only the first was included. Exclusion criteria were age <18 years, missing data of TBIL, or repeat BSI episodes during the same hospitalization.
Microbiological Methods
Blood cultures were processed using the BACT/ALERT 3D automated blood culture system and VITEK 2 Compact systems (bioMérieux, France). Bacterial species identification and antimicrobial susceptibility testing (AST) were performed according to the manufacturer’s instructions. AST results were interpreted based on the Clinical and Laboratory Standards Institute (CLSI) M100 guidelines, and disk diffusion testing followed CLSI M02 standards.14–16 Extended-spectrum β-lactamase (ESBL)-producing and carbapenem-resistant E. coli isolates were identified according to CLSI recommendations and established phenotypic criteria.
Variable Definitions
ALBI Score
Baseline ALB and TBIL measured concurrently with or within 24 hours of blood culture collection were used to calculate the ALBI score and grade. The calculation formula for the ALBI score is (log10 TBIL [μmol/L] × 0.66) + (ALB [g/L] × (−0.085)). The ALBI was analysed as both continuous and categorical variables. Tertiles were defined as T1: ≤ −2.16, T2: −2.16 ~ −1.78, T3: > −1.78. ALBI grades based on established cut-off values (Grade 1: ≤ −2.6, Grade 2: −2.6 ~ −1.39, and Grade 3: > −1.39).8
Covariates
Demographics: age and sex; Comorbidities: diabetes, chronic heart failure, chronic kidney disease, chronic obstructive pulmonary disease, malignant tumor, cerebrovascular disease, and Charlson Comorbidity Index (CCI); bedridden status, hospital-acquired infection (HAI); Primary infection site: urinary tract infection (UTI), intra-abdominal infection, hepatobiliary infection, pneumonia, soft tissue infection, intestinal infection; microbiological resistance: ESBL-producing and carbapenem-resistant; Laboratory indices: white blood cell (WBC) count, hemoglobin, red cell distribution width (RDW), platelet count, C-reactive protein (CRP), TBIL, and ALB; and disease severity indicators: ICU admission, Sequential Organ Failure Assessment (SOFA) score, and sepsis. HAI was defined as an E. coli-BSI occurring ≥48 hours after hospital admission, whereas infections present on admission or occurring within 48 hours of admission were classified as community-acquired. Sepsis was defined conceptually according to the Sepsis-3 criteria as life-threatening organ dysfunction caused by a dysregulated host response to infection, and was operationalized in this study as a total SOFA score ≥2 points. The first diagnostic blood culture was defined as the index blood culture and served as the reference time point. All baseline variables were collected concurrently with or within 24 hours of the index blood culture.
Outcome
The primary outcome was septic shock occurring concurrently with or after the index blood culture collection during the index hospitalization. Septic shock was defined according to the Sepsis-3 consensus as requiring vasopressors to maintain a mean arterial pressure of ≥65 mmHg and having a serum lactate level >2 mmol/L despite adequate fluid resuscitation.
Statistical Analysis
Continuous variables were presented as mean ± standard deviation (SD) for normally distributed data and as median (interquartile range [IQR]) for non-normally distributed data. Categorical variables were presented as frequencies and percentages. Comparisons across ALBI tertile groups were performed using one-way analysis of variance, the Kruskal–Wallis test, the chi-square test, or Fisher’s exact test as appropriate.
Multivariable logistic regression with progressive adjustment was used to evaluate the association between the ALBI and septic shock. Covariates were selected based on clinical relevance, prior literature, and univariable associations. As no missing data were identified for the variables included in the regression models, no imputation procedures were applied. Model 1 was unadjusted. Model 2 was adjusted for age, sex, CCI, and bedridden status. Model 3 was additionally adjusted for ESBL-producing infection, WBC, CRP, UTI, and HAI. Model 4 was further adjusted for sepsis status. Multicollinearity among independent variables was assessed using variance inflation factors (VIFs), with a VIF > 5 considered indicative of substantial multicollinearity and were excluded from the final model.17 Restricted cubic spline (RCS) analysis with four knots was used to assess potential nonlinearity between the ALBI and septic shock. Subgroup analyses were performed according to age (<65 vs ≥65 years), sex, ESBL-producing infection (yes vs no), CCI (<3 vs ≥3), and SOFA score (<6 vs ≥6), using the fully adjusted logistic regression model. Effect modification was assessed using interaction terms and P for interaction.
Several sensitivity analyses were performed to assess the robustness of the findings. The association between ALBI and septic shock was re-evaluated after excluding patients with hepatobiliary infection and, separately, after restricting the analysis to patients with a hospital stay >48 hours, using the fully adjusted model. To address potential reverse causation, patients in whom septic shock was already present at or occurred concurrently with ALBI assessment were excluded. The association between ALBI and sepsis was also evaluated using univariable and multivariable logistic regression, with ALBI modeled continuously and by tertiles. The multivariable model was adjusted for age, sex, CCI, bedridden status, ESBL-producing infection, WBC, CRP, UTI, and HAI. E-values were calculated for the primary fully adjusted association between ALBI and septic shock to assess the potential strength of unmeasured confounding required to explain the observed association. The discriminative performance of ALBI and its individual components, ALB and TBIL, as well as CRP, RDW, and SOFA was assessed using receiver operating characteristic (ROC) curve analysis, with comparisons of the area under the curves (AUCs) using DeLong’s test. The optimal cut-off value was determined using the maximum Youden index. Exploratory variable selection was performed using least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation, with λ.1se selected according to the one-standard-error criterion. Shapley additive explanations (SHAP) analysis based on a naive Bayes model was subsequently used to visualize the relative contributions and directions of selected predictors to the predicted probability of septic shock. Detailed and reproducible methods for the RCS, LASSO, SHAP, and E-value analyses are provided in the Supplementary Methods.
All analyses were performed using R software, version 4.2.2 (The R Foundation for Statistical Computing, Vienna, Austria) and Free Statistics software, version 2.2.4. All statistical tests were two-sided, and a p value < 0.05 was considered statistically significant.
Results
Baseline Characteristics of Participants
A total of 334 patients with E. coli-BSI were included in the final analysis according to the predefined inclusion and exclusion criteria (Supplementary Figure 1). Patients were stratified into tertiles based on the ALBI (T1–T3). Overall, 77 (23.1%) patients had HAI, whereas 257 (76.9%) had community-acquired infections. As shown in Table 1, patients in higher ALBI tertiles were older (mean age: 56.2, 62.5, and 67.2 years for T1–T3, respectively; p < 0.001). The distribution of sex, comorbidities, microbiological resistance, and HAI rate did not differ significantly across tertiles. Hepatobiliary infections were more frequent in higher ALBI tertiles (p = 0.003). Inflammatory markers (eg, RDW and CRP) increased and ALB decreased while bilirubin increased across tertiles (all p < 0.01). Baseline SOFA scores (2.0, 3.0, and 7.0) and the proportion of patients with sepsis (36%, 68.5%, and 91.1%) increased progressively across the tertile groups (all p < 0.001). Importantly, the proportion of patients who developed septic shock increased stepwise from T1 to T3 (5.4%, 27.0%, and 54.5%, respectively; p < 0.001).
Table 1.
Baseline Clinical Characteristics and Outcomes of Patients with E. coli-BSI Patients Stratified by ALBI Score Tertiles
| Variables | Total (n = 334) | ALBI Score Tertile | P value | ||
|---|---|---|---|---|---|
| T1 (n = 111) | T2 (n = 111) | T3 (n = 112) | |||
| Demographics | |||||
| Age, years | 61.9 ± 16.8 | 56.2 ± 17.7 | 62.5 ± 16.7 | 67.2 ± 14.2 | < 0.001 |
| Sex, male, n (%) | 201 (60.2) | 72 (64.9) | 68 (61.3) | 61 (54.5) | 0.273 |
| Comorbidities | |||||
| Diabetes, n (%) | 141 (42.2) | 43 (38.7) | 49 (44.1) | 49 (43.8) | 0.661 |
| Chronic heart failure, n (%) | 41 (12.3) | 12 (10.8) | 13 (11.7) | 16 (14.3) | 0.714 |
| Chronic kidney disease, n (%) | 43 (12.9) | 13 (11.7) | 16 (14.4) | 14 (12.5) | 0.826 |
| Malignant tumor, n (%) | 74 (22.2) | 23 (20.7) | 22 (19.8) | 29 (25.9) | 0.499 |
| Chronic obstructive pulmonary disease, n (%) | 30 (9.0) | 10 (9) | 11 (9.9) | 9 (8) | 0.887 |
| Cerebrovascular disease, n (%) | 76 (22.8) | 18 (16.2) | 25 (22.5) | 33 (29.5) | 0.062 |
| Charlson Comorbidity Index, score | 2.0 (1.0, 3.0) | 2.0 (0.0, 3.0) | 2.0 (1.0, 3.0) | 2.0 (1.0, 3.0) | 0.163 |
| Bedridden status, n (%) | 32 (9.6) | 6 (5.4) | 8 (7.2) | 18 (16.1) | 0.015 |
| Hospital-acquired infection, n (%) | 77 (23.1) | 29 (26.1) | 26 (23.4) | 22 (19.6) | 0.513 |
| Microbiological resistance | |||||
| ESBL-producing, n (%) | 177 (53.0) | 62 (55.9) | 59 (53.2) | 56 (50) | 0.681 |
| Carbapenem-resistant, n (%) | 9 (2.7) | 2 (1.8) | 3 (2.7) | 4 (3.6) | 0.921 |
| Primary Infection Site | |||||
| Urinary tract infection, n (%) | 153 (45.8) | 47 (42.3) | 59 (53.2) | 47 (42) | 0.164 |
| Intra-abdominal infection, n (%) | 62 (18.6) | 20 (18) | 17 (15.3) | 25 (22.3) | 0.398 |
| Hepatobiliary infection, n (%) | 52 (15.6) | 13 (11.7) | 11 (9.9) | 28 (25) | 0.003 |
| Pneumonia, n (%) | 51 (15.3) | 14 (12.6) | 18 (16.2) | 19 (17) | 0.628 |
| Soft tissue infection, n (%) | 24 (7.2) | 5 (4.5) | 7 (6.3) | 12 (10.7) | 0.181 |
| Intestinal infection, n (%) | 17 (5.1) | 7 (6.3) | 5 (4.5) | 5 (4.5) | 0.775 |
| Laboratory indices | |||||
| White blood cell count, 109/L | 10.1 (5.9, 14.1) | 9.3 (6.0, 12.7) | 10.6 (6.6, 15.3) | 10.3 (5.1, 16.3) | 0.067 |
| Hemoglobin, g/L | 103.4 ± 24.5 | 109.7 ± 25.5 | 104.6 ± 23.1 | 96.0 ± 22.9 | < 0.001 |
| Red cell distribution width, % | 14.7 ± 2.6 | 14.3 ± 2.6 | 14.3 ± 2.3 | 15.3 ± 2.9 | 0.009 |
| Platelet count, 109/L | 152.8 ± 94.2 | 185.3 ± 91.4 | 155.8 ± 87.2 | 117.7 ± 92.1 | < 0.001 |
| C-reactive protein, mg/L | 148.2 ± 96.9 | 117.8 ± 79.4 | 161.3 ± 121.1 | 165.3 ± 77.3 | < 0.001 |
| Total bilirubin, μmol/L | 13.8 (8.6, 23.6) | 12.1 (7.4, 17.8) | 12.7 (8.0, 20.6) | 23.5 (11.2, 47.0) | < 0.001 |
| Albumin, g/L | 31.8 ± 5.4 | 37.1 ± 3.4 | 31.8 ± 2.7 | 26.6 ± 3.6 | < 0.001 |
| Disease severity indicators | |||||
| SOFA score, score | 4.0 (1.0, 6.0) | 2.0 (0.0, 3.0) | 3.0 (1.5, 5.0) | 7.0 (4.0, 10.2) | < 0.001 |
| Sepsis, n (%) | 218 (65.3) | 40 (36) | 76 (68.5) | 102 (91.1) | < 0.001 |
| Outcomes | |||||
| Septic shock, n (%) | 97 (29.0) | 6 (5.4) | 30 (27) | 61 (54.5) | < 0.001 |
Notes: Categorical variables are presented as counts and percentages (n, %), normally distributed continuous variables as mean ± standard deviation (SD), and non-normally distributed continuous variables as median with interquartile range (IQR).
Abbreviations: ALBI, Albumin-Bilirubin; BSI, bloodstream infection; E. coli, Escherichia coli; ESBL, Extended-Spectrum Beta-Lactamases; SOFA, Sequential Organ Failure Assessment.
Multivariable Logistic Regression Analysis
The ALBI was consistently associated with septic shock across all models (Table 2). When analyzed as a continuous variable, ALBI demonstrated a significant association, with an unadjusted odds ratio (OR) of 11.91 (95% CI, 6.25–22.69; p < 0.001) and an adjusted OR of 5.12 (95% CI, 2.37–11.05; p < 0.001) in the fully adjusted model (adjusted for age, sex, CCI, bedridden status, ESBL-producing infection, WBC, CRP, UTI, HAI, and sepsis). In tertile-based analyses, patients in the highest tertile (T3) had markedly increased odds of septic shock compared with those in T1 (adjusted OR, 7.40; 95% CI, 2.63–20.83; p < 0.001). When categorized by ALBI grade, grade 3 was associated with a higher estimated odds than grade 1 (adjusted OR, 9.02; 95% CI, 0.92–88.57), although the comparison did not reach statistical significance (p = 0.059); importantly, a significant dose-response trend was observed across grades (p for trend = 0.001).
Table 2.
Association of ALBI Score and ALBI Grade with Septic Shock in Patients with E. coli-BSI
| Variable | Septic shock [n (%)] |
Model 1 | Model 2 | Model 3 | Model 4 | ||||
|---|---|---|---|---|---|---|---|---|---|
| OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | OR (95% CI) | p-value | ||
| ALBI continuous |
97 (29) | 11.91 (6.25~22.69) | <0.001 | 10.12 (5.23~19.57) | <0.001 | 10.23 (4.94~21.16) | <0.001 | 5.12 (2.37~11.05) | <0.001 |
| ALBI tertile | |||||||||
| T1 | 6 (5.4) | 1(Ref) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| T2 | 30 (27) | 6.48 (2.57~16.32) | <0.001 | 6.13 (2.4~15.62) | <0.001 | 4.55 (1.72~12.01) | 0.002 | 2.59 (0.91~7.36) | 0.075 |
| T3 | 61 (54.5) | 20.93 (8.49~51.63) | <0.001 | 17.61 (7.03~44.11) | <0.001 | 15.43 (5.9~40.36) | <0.001 | 7.4 (2.63~20.83) | <0.001 |
| P for trend | <0.001 | <0.001 | <0.001 | <0.001 | |||||
| ALBI grade | |||||||||
| Grade 1 | 4 (11.1) | 1(Ref) | 1(Ref) | 1(Ref) | 1(Ref) | ||||
| Grade 2 | 60 (23.9) | 7.82 (1.04~58.88) | 0.046 | 5.98 (0.79~45.53) | 0.084 | 3.79 (0.49~29.54) | 0.204 | 2.61 (0.28~23.89) | 0.397 |
| Grade 3 | 33 (70.2) | 51.47 (6.43~412.18) | <0.001 | 35.12 (4.31~285.93) | 0.001 | 22.57 (2.66~191.43) | 0.004 | 9.02 (0.92~88.57) | 0.059 |
| P for trend | <0.001 | <0.001 | <0.001 | 0.001 | |||||
Notes: This analysis included 334 patients with E. coli-BSI. Model 1: Unadjusted crude model. Model 2: Adjusted for age, sex, CCI, and bedridden status. Model 3: Further adjusted for ESBL-producing, WBC, CRP, UTI, and HAI. Model 4: Fully adjusted model additionally controlled for sepsis status.
Abbreviations: ALBI; Albumin-Bilirubin; BSI, bloodstream infection; CCI, Charlson Comorbidity Index; CI, confidence interval; CRP, C-reactive protein; E. coli, Escherichia coli; ESBL, Extended-Spectrum Beta-Lactamases; HAI, hospital-acquired infection; OR, Odds ratio; Ref, reference group; UTI, Urinary tract infection; WBC, White blood cell.
RCS Analysis
After adjusting for all covariates in the analytical model, a positive linear association was observed between the ALBI score and septic shock (p for overall < 0.001, p for nonlinear = 0.56; Figure 1). The odds of shock increased progressively with higher ALBI values.
Figure 1.
Restricted cubic spline analysis of the association between ALBI score and septic shock in patients with E. coli-BSI.
Abbreviations: ALBI, Albumin-Bilirubin; BSI, bloodstream infection; E. coli, Escherichia coli; ORs, odds ratios.
Subgroup Analyses
The association between ALBI and septic shock was generally consistent across the examined subgroups, including age (<65 vs ≥65 years), sex (female vs male), ESBL-producing infection (yes vs no), CCI (< 3 vs ≥ 3), and SOFA score (< 6 vs ≥ 6), with no significant interactions (Figure 2).
Figure 2.
Subgroup analyses for the associations between ALBI score and septic shock in patients with E. coli-BSI.
Notes: ORs were adjusted for age, sex, CCI, bedridden status, ESBL-producing, WBC, CRP, UTI, HAI, and sepsis status. P for interaction was presented for each subgroup.
Abbreviations: ALBI, Albumin-Bilirubin; BSI, bloodstream infection; CCI, Charlson Comorbidity Index; CI, Confidence interval; CRP, C-reactive protein; E. coli, Escherichia coli; ESBL, Extended-Spectrum Beta-Lactamases; HAI, hospital-acquired infection; ORs, Odds ratios; SOFA, Sequential Organ Failure Assessment; UTI, Urinary tract infection; WBC, White blood cell.
Sensitivity Analyses
Several sensitivity analyses were performed to assess the robustness of the association between ALBI and septic shock. After excluding patients with hepatobiliary infection, ALBI remained significantly associated with septic shock when analyzed continuously (adjusted OR, 5.21; 95% CI, 2.14–12.7; p < 0.001; Table 3). A similar association was observed among patients with a hospital stay >48 hours (adjusted OR, 3.95; 95% CI, 1.73–9.01; p = 0.001; Table 3). In tertile-based analyses, the highest ALBI tertile (T3) remained significantly associated with septic shock compared with T1 in both cohorts (excluding hepatobiliary infection: adjusted OR, 7.72; 95% CI, 2.54–23.47; p < 0.001; hospital stay > 48 hours: adjusted OR, 5.76; 95% CI, 1.96–16.87; p = 0.001), with significant trends across tertiles (p for trend < 0.001 for both). Although the associations for ALBI grade 3 versus grade 1 were not statistically significant, significant trends across ALBI grades were observed in both analyses (P for trend = 0.029 and 0.017, respectively). E-value analysis suggested that substantial unmeasured confounding would be required to fully explain the observed association (E-value for the point estimate, 3.95; E-value for the confidence limit closest to the null, 2.45).
Table 3.
Sensitivity Analyses After Excluding Patients with Hepatobiliary Infection or Hospital Stay Shorter Than 48 Hours
| Variable | Septic Shock [n (%)] |
Excluding Hepatobiliary Infection | Septic Shock [n (%)] |
Hospital Stay Longer Than 48 Hours | ||
|---|---|---|---|---|---|---|
| Adjusted OR (95% CI) | P-value | Adjusted OR (95% CI) | P-value | |||
| ALBI continuous |
84 (29.8) | 5.21 (2.14~12.7) | <0.001 | 85 (26.7) | 3.95 (1.73~9.01) | 0.001 |
| ALBI tertile | ||||||
| T1 | 6 (6.1) | 1(Ref) | 6 (5.5) | 1(Ref) | ||
| T2 | 29 (29) | 2.41 (0.8~7.21) | 0.117 | 29 (26.4) | 2.16 (0.73~6.4) | 0.166 |
| T3 | 49 (58.3) | 7.72 (2.54~23.47) | <0.001 | 50 (50.5) | 5.76 (1.96~16.87) | 0.001 |
| P for trend | <0.001 | <0.001 | ||||
| ALBI grade | ||||||
| Grade 1 | 1 (4) | 1(Ref) | 1 (4) | 1(Ref) | ||
| Grade 2 | 58 (26.2) | 2.83 (0.3~26.74) | 0.363 | 58 (23.2) | 1.79 (0.18~17.82) | 0.622 |
| Grade 3 | 25 (69.4) | 6.94 (0.65~73.72) | 0.108 | 26 (60.5) | 4.72 (0.44~50.93) | 0.201 |
| P for trend | <0.029 | 0.017 | ||||
Notes: All ORs were adjusted for age, sex, CCI, bedridden status, ESBL-producing, WBC, CRP, UTI, HAI, and sepsis status.
Abbreviations: ALBI, Albumin-Bilirubin; BSI, bloodstream infection; CCI, Charlson Comorbidity Index; CI, confidence interval; CRP, C-reactive protein; ESBL, Extended-Spectrum Beta-Lactamases; HAI, hospital-acquired infection; ORs, Odds ratios; Ref, reference group; UTI, Urinary tract infection; WBC, White blood cell.
To address potential reverse causation, we excluded 15 patients in whom septic shock was already present at or occurred concurrently with ALBI assessment. The association between ALBI and septic shock remained significant, although the effect estimate was attenuated. In the fully adjusted model, each one-unit increase in ALBI was associated with higher odds of septic shock (adjusted OR, 3.41; 95% CI, 1.53–7.62; P = 0.003). Similarly, T3 remained significantly associated with septic shock compared with T1 (adjusted OR, 5.45; 95% CI, 1.89–15.67; P = 0.002; P for trend = 0.001), supporting the robustness of the association when the analysis was restricted to patients in whom ALBI assessment preceded septic shock onset (Supplementary Table 1).
Among the 334 patients with E. coli-BSI, 218 (65.3%) developed sepsis. Higher ALBI was consistently associated with sepsis in both continuous and tertile-based analyses. In the fully adjusted model, each one-unit increase in ALBI was associated with higher odds of sepsis (adjusted OR, 15.23; 95% CI, 6.90–33.64; P < 0.001). Compared with T1, the adjusted ORs were 3.55 (95% CI, 1.90–6.60) for T2 and 15.38 (95% CI, 6.70–35.30) for T3 (P for trend < 0.001; Supplementary Table 2).
ROC Curve Analysis and Comparison of Discriminative Performance
The ALBI showed moderate discriminative ability for septic shock, with an AUC of 0.796 (95% CI, 0.745–0.847). Among its individual components, ALB showed a slightly higher AUC (0.816; 95% CI, 0.766–0.865), but the difference from ALBI was not statistically significant (DeLong test, p = 0.22), whereas TBIL showed poor discrimination (AUC, 0.497; 95% CI, 0.424–0.569). For comparison, CRP and RDW showed AUCs of 0.734 (95% CI, 0.678–0.790) and 0.590 (95% CI, 0.525–0.655), respectively. SOFA demonstrated the highest discriminative performance (AUC, 0.909; 95% CI, 0.877–0.941), which was significantly higher than that of ALBI (DeLong test, p < 0.001). Overall, ALBI showed moderate discrimination for septic shock but was inferior to SOFA (Supplementary Figures 2, 3 and Supplementary Table 3).
Feature Selection and Model Interpretation
Exploratory LASSO selection identified ALBI, sepsis, bedridden status, WBC, CRP, and UTI as selected variables (Figure 3A and B). SHAP analysis ranked sepsis as the most important influential variable, followed by ALBI, bedridden status, WBC, CRP and UTI (Figure 4A and B). The SHAP dependence plot showed an approximately monotonic positive association between ALBI and SHAP values, indicating that higher ALBI scores were associated with higher model-estimated probabilities of septic shock (Figure 4C).
Figure 3.
Variable screening results using LASSO regression. (A) Coefficient variation curves of the 11 included variables with changing regularization parameter λ. (B) Binomial deviance curve for optimal λ value selection in LASSO regression.
Abbreviation: LASSO, least absolute shrinkage and selection operator.
Figure 4.
Model interpretation based on SHAP analysis. (A) Global feature importance ranked by the mean (|SHAP|). (B) SHAP beeswarm summary plot illustrating the magnitude and direction of each feature affecting the risk of septic shock. (C) SHAP dependence plot for ALBI score with LOESS smoothing and rug distribution.
Abbreviations: ALBI, Albumin-Bilirubin; CRP, C-reactive protein; SHAP, Shapley additive explanations; |SHAP|, absolute SHAP value; UTI, Urinary tract infection; WBC, White blood cell.
Discussion
In this retrospective cohort study of adults with E. coli-BSI, we investigated the association between the ALBI and septic shock occurring concurrently with or after the index blood culture during hospitalization. Higher ALBI scores were significantly associated with higher odds of septic shock, and this association remained significant after adjustment for demographic characteristics, comorbidities, infection-related factors, inflammatory markers, and sepsis status. The association was also observed consistently across ALBI tertiles and grades, while RCS analysis demonstrated a positive linear association. Subgroup and sensitivity analyses yielded generally consistent findings. Exploratory SHAP analysis also identified ALBI as an important contributor to the model-estimated probability of septic shock. However, ALBI showed only moderate discrimination and performed significantly worse than SOFA. Moreover, ALB alone had a slightly higher AUC than ALBI, although the difference was not statistically significant (0.816 vs 0.796; p = 0.22). These findings suggest that the observed discriminatory performance of ALBI may be largely attributable to its ALB component. Overall, ALBI was associated with septic shock in patients with E. coli-BSI but should be considered a simple laboratory-based marker rather than a replacement for established severity scores.
The temporal relationship between ALBI and septic shock warrants careful consideration. The index blood culture was used as the reference time point, with ALBI measured concurrently with or within 24 hours of the index blood culture, whereas septic shock was defined as occurring concurrently with or after the index blood culture. Thus, ALBI assessment preceded septic shock in some patients but may have overlapped with early clinical deterioration in others. After excluding 15 patients in whom septic shock was already present at or concurrent with ALBI assessment, the association remained significant, although the effect estimate was attenuated. Nevertheless, ALBI abnormalities may reflect underlying hepatic, inflammatory, and systemic disturbances associated with severe infection, while evolving sepsis-related organ dysfunction may also contribute to hypoalbuminemia and hyperbilirubinemia. Therefore, our findings should be interpreted as an association rather than evidence of causality or temporal precedence.
The ALBI score is a composite index calculated from serum ALB and TBIL levels, reflecting hepatic synthetic function and cholestatic liver injury. Hepatic dysfunction is increasingly recognized as an important component of sepsis-related organ dysfunction.18 The liver plays critical roles in immune regulation, acute-phase protein synthesis, detoxification, metabolic homeostasis, and maintenance of vascular integrity.6,19 During severe infection, systemic inflammation, endotoxemia, and microcirculatory dysfunction can impair hepatic perfusion and function, leading to hypoalbuminemia, hyperbilirubinemia, and dysregulated inflammatory responses.20,21 Previous studies have shown that hypoalbuminemia is associated with mortality and multiple organ dysfunction in septic patients, whereas elevated bilirubin levels are associated with sepsis-related cholestatic injury and poor prognosis.20,22–25 For example, Sun et al reported that lower serum ALB levels were associated with increased hospital mortality and multiple organ dysfunction syndrome in surgical septic patients,25 while Kramer et al found that early hepatic dysfunction characterized by hyperbilirubinemia was independently associated with increased hospital mortality in critically ill patients.24 Collectively, these findings provide biological and clinical context for the observed association between ALBI and septic shock. To our knowledge, few studies have specifically evaluated this association in patients with E. coli-BSI. Our findings extend previous evidence by demonstrating an association between ALBI and septic shock in this population.
Several mechanisms may potentially underlie the observed association. Hypoalbuminemia during severe infection may result from reduced hepatic synthesis, increased vascular permeability, inflammation-driven redistribution, and accelerated catabolism. Reduced ALB levels may decrease plasma oncotic pressure and contribute to tissue edema and microcirculatory dysfunction.26,27 Hyperbilirubinemia may reflect sepsis-associated cholestatic liver injury, impaired hepatic clearance and hepatic microcirculatory dysfunction associated with systemic inflammation,6,24,28 and has also been associated with oxidative stress and mitochondrial dysfunction.21 In addition, pro-inflammatory cytokines, including interleukin-6, tumor necrosis factor-α, and interleukin-1β, and oxidative stress during severe infection may further impair hepatic function, endothelial integrity, and ALB metabolism.29–31 These interconnected processes may partly explain the association between ALBI and septic shock; however, these mechanistic interpretations remain speculative and cannot establish a causal pathway.
This study has several strengths. First, we focused on a microbiologically confirmed and pathogen-specific cohort of patients with E. coli-BSI, reducing the heterogeneity inherent in mixed-pathogen BSI populations. Second, we applied multiple complementary analytical approaches, including multivariable logistic regression, RCS, subgroup and sensitivity analyses, ROC analysis, E-value analysis, exploratory LASSO regression and SHAP analysis, to examine the consistency and potential interpretation of the observed association. Third, the ALBI score is derived from inexpensive and routinely available laboratory parameters, making it readily measurable and potentially useful for further investigation in patients with E. coli-BSI.
Several limitations should be acknowledged. First, the single-center retrospective observational study may introduce selection bias, limit generalizability, and preclude causal inference. Second, although extensive adjustment was performed and E-value analysis suggested that relatively strong unmeasured confounding would be required to fully explain the observed association, residual confounding cannot be excluded. Third, ALBI was assessed only at baseline, precluding evaluation of dynamic changes during hospitalization; detailed causes of in-hospital death were unavailable. Fourth, the temporal relationship between ALBI and septic shock remains uncertain because ALBI assessment may have overlapped with early clinical deterioration. Although the association remained significant after excluding patients with septic shock present at or concurrent with ALBI assessment, reverse causation cannot be completely excluded. Finally, ALBI showed moderate discrimination but did not outperform SOFA, and its clinical value beyond established severity scores remains uncertain. SHAP findings should also be interpreted as model-specific associations rather than causal effects. Prospective multicenter studies with serial ALBI measurements are warranted to validate these findings and clarify the temporal and biological relationships between ALBI and septic shock.
Conclusion
Higher ALBI scores were independently associated with septic shock in adults with E. coli-BSI. ALBI showed moderate discrimination but was inferior to SOFA and not significantly better than ALB alone. These findings indicate an association rather than established predictive or clinical utility. Prospective multicenter studies with serial measurements are warranted to validate these findings and determine whether ALBI provides clinically meaningful information for risk assessment.
Acknowledgments
No acknowledgments are required for this work.
Funding Statement
No funding was received.
Abbreviations
ALB, albumin; ALBI, albumin-bilirubin; AST, antimicrobial susceptibility testing; AUC, area under the curve; BSI, bloodstream infection; CCI, Charlson Comorbidity Index; CI, confidence interval; CLSI, Clinical and Laboratory Standards Institute; CRP, C-reactive protein; E. coli, Escherichia coli; ESBL, extended-spectrum beta-lactamases; HAI, hospital-acquired infection; IQR, interquartile range; LASSO, least absolute shrinkage and selection operator; ORs, odds ratios; RDW, red cell distribution width; ROC, receiver operating characteristic; RCS, restricted cubic splines; SD, standard deviation; SHAP, Shapley additive explanations; SOFA, Sequential Organ Failure Assessment; STROBE, Strengthening the Reporting of Observational Studies in Epidemiology; TBIL, total bilirubin; UTI, urinary tract infection; VIFs, variance inflation factors; WBC, white blood cell.
Data Sharing Statement
All data generated or analysed during this study are included in this published article and its Supplementary Information Files.
Ethical Approval and Informed Consent
The study was approved by the Ethics Committee of Beijing Luhe Hospital, Capital Medical University (No. 2025-LHKY-017-01) and conducted in accordance with the Declaration of Helsinki. Informed consent was waived due to the retrospective design.
Author Contributions
All authors made a significant contribution to the work reported, whether that was in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising, or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.
Disclosure
The authors declare no competing interests.
References
- 1.Kb L. Incidence of bloodstream infection: a review of population-based studies. Clin Microbiol Infect. 2013;19. doi: 10.1111/1469-0691.12144 [DOI] [PubMed] [Google Scholar]
- 2.G M, Mn A-H. Overall burden of bloodstream infection and nosocomial bloodstream infection in North America and Europe. Clin Microbiol Infect. 2013;19. doi: 10.1111/1469-0691.12195 [DOI] [PubMed] [Google Scholar]
- 3.Maldonado N, López-Hernández I, López-Cortés LE, et al. Association of microbiological factors with mortality in Escherichia coli bacteraemia presenting with sepsis/septic shock: a prospective cohort study. Clin Microbiol Infect. 2024;30:1035–14. doi: 10.1016/j.cmi.2024.04.001 [DOI] [PubMed] [Google Scholar]
- 4.Butler JM, Peters-Sengers H, Reijnders TDY, et al. Pathogen-specific host response in critically ill patients with blood stream infections: a nested case-control study. EBioMedicine. 2025;117:105799. doi: 10.1016/j.ebiom.2025.105799 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Fröding I, Hasan B, Sylvin I, Coorens M, Nauclér P, Giske CG. Extended-Spectrum-β-Lactamase- and plasmid AmpC-Producing Escherichia coli causing community-onset bloodstream infection: association of bacterial clones and virulence genes with septic shock, source of infection, and recurrence. Antimicrob Agents Chemother. 2020;64:e02351–19. doi: 10.1128/AAC.02351-19 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Yan J, Li S, Li S. The role of the liver in sepsis. Int Rev Immunol. 2014;33:498–510. doi: 10.3109/08830185.2014.889129 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Guo Y, Guo W, Chen H, Sun J, Yin Y. Mechanisms of sepsis-induced acute liver injury: a comprehensive review. Front Cell Infect Microbiol. 2025;15:1504223. doi: 10.3389/fcimb.2025.1504223 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Johnson PJ, Berhane S, Kagebayashi C, et al. Assessment of liver function in patients with hepatocellular carcinoma: a new evidence-based approach-the ALBI grade. J Clin Oncol. 2015;33:550–558. doi: 10.1200/JCO.2014.57.9151 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Luo Y, Li Z, Liu J, Chong Y, Wu B. Prognostic value of the albumin-bilirubin score in critically ill patients with heart failure. Ann Palliat Med. 2021;10:12727–12741. doi: 10.21037/apm-21-3424 [DOI] [PubMed] [Google Scholar]
- 10.Yang C, Yu Z, Peng B, Mao C, Li J, Cao Y. ALBI grade is associated with clinical outcomes of critically ill patients with AKI: a cohort study with cox regression and propensity score matching. Mediators Inflamm. 2024;2024:1412709. doi: 10.1155/2024/1412709 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Wang J, Yang P, Qin C, et al. Predictive role of the Albumin-Bilirubin score in ICU patients with cirrhosis and sepsis: insights from a large retrospective cohort. BMC Gastroenterol. 2025;25:520. doi: 10.1186/s12876-025-04111-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Chen W, Chen Q, Tang Z, et al. Association of Albumin–Bilirubin (ALBI) grade with 28-day all-cause mortality in patients with acute respiratory distress syndrome: a retrospective analysis of the MIMIC-IV database. Mediators Inflamm. 2025;2025:9930648. doi: 10.1155/mi/9930648 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453–1457. doi: 10.1016/S0140-6736(07)61602-X [DOI] [PubMed] [Google Scholar]
- 14.Dallas SD, McGee L, Limbago B, et al. Development of doxycycline MIC and disk diffusion interpretive breakpoints and revision of tetracycline breakpoints for Streptococcus pneumoniae. J Clin Microbiol. 2013;51:1798–1802. doi: 10.1128/JCM.00125-13 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Hong X, Wang X, Xiang X, et al. Evaluation of sulbactam-durlobactam disk diffusion and broth microdilution methods for acinetobacter baumannii. Int J Antimicrob Agents. 2026;67:107659. doi: 10.1016/j.ijantimicag.2025.107659 [DOI] [PubMed] [Google Scholar]
- 16.Humphries R, Bobenchik AM, Hindler JA, Schuetz AN. Overview of changes to the Clinical and Laboratory Standards Institute performance standards for antimicrobial susceptibility testing, M100, 31st edition. J Clin Microbiol. 2021;59:e0021321. doi: 10.1128/JCM.00213-21 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.O’brien RM. A caution regarding rules of thumb for variance inflation factors. Qual Quant. 2007;41:673–690. doi: 10.1007/s11135-006-9018-6 [DOI] [Google Scholar]
- 18.Nesseler N, Launey Y, Aninat C, et al. Liver dysfunction is associated with long-term mortality in septic shock. Am J Respir Crit Care Med. 2016;193:335–337. doi: 10.1164/rccm.201508-1660LE [DOI] [PubMed] [Google Scholar]
- 19.Ehlting C, Wolf SD, Bode JG. Acute-phase protein synthesis: a key feature of innate immune functions of the liver. Biol Chem. 2021;402:1129–1145. doi: 10.1515/hsz-2021-0209 [DOI] [PubMed] [Google Scholar]
- 20.Beyer D, Hoff J, Sommerfeld O, Zipprich A, Gaßler N, Press AT. The liver in sepsis: molecular mechanism of liver failure and their potential for clinical translation. Mol Med. 2022;28:84. doi: 10.1186/s10020-022-00510-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Joffre J, Hellman J. Oxidative stress and endothelial dysfunction in sepsis and acute inflammation. Antioxidants Redox Signal. 2021;35:1291–1307. doi: 10.1089/ars.2021.0027 [DOI] [PubMed] [Google Scholar]
- 22.Jacobi J. The pathophysiology of sepsis - 2021 update: part 2, organ dysfunction and assessment. Am J Health Syst Pharm. 2022;79:424–436. doi: 10.1093/ajhp/zxab393 [DOI] [PubMed] [Google Scholar]
- 23.Chen J-W, Liu C-Y, Li S, Wu S-W, Cai C, Lu M-Q. Sepsis-associated liver injury: mechanisms and potential therapeutic targets. World J Gastroenterol. 2024;30:4518–4522. doi: 10.3748/wjg.v30.i42.4518 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kramer L, Jordan B, Druml W, Bauer P, Metnitz PGH, Austrian Epidemiologic Study on Intensive Care, ASDI Study Group. Incidence and prognosis of early hepatic dysfunction in critically ill patients--a prospective multicenter study. Crit Care Med. 2007;35:1099–1104. doi: 10.1097/01.CCM.0000259462.97164.A0 [DOI] [PubMed] [Google Scholar]
- 25.Sun J-K, Sun F, Wang X, Yuan S-T, Zheng S-Y, Mu X-W. Risk factors and prognosis of hypoalbuminemia in surgical septic patients. PeerJ. 2015;3:e1267. doi: 10.7717/peerj.1267 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Vincent J-L, Dubois M-J, Navickis RJ, Wilkes MM. Hypoalbuminemia in acute illness: is there a rationale for intervention? Ann Surg. 2003;237:319–334. doi: 10.1097/01.SLA.0000055547.93484.87 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Artigas A, Wernerman J, Arroyo V, Vincent J-L, Levy M. Role of albumin in diseases associated with severe systemic inflammation: pathophysiologic and clinical evidence in sepsis and in decompensated cirrhosis. Journal of Critical Care. 2016;33:62–70. doi: 10.1016/j.jcrc.2015.12.019 [DOI] [PubMed] [Google Scholar]
- 28.Bhogal HK, Sanyal AJ. The molecular pathogenesis of cholestasis in sepsis. Front Biosci. 2013;5:87–96. doi: 10.2741/e598 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wang X, Li W, Lu J, Li N, Li J. Lipopolysaccharide suppresses albumin expression by activating NF-κB in rat hepatocytes. J Surg Res. 2004;122:274–279. doi: 10.1016/j.jss.2004.07.008 [DOI] [PubMed] [Google Scholar]
- 30.Gou E, Yang Q, Chen J, et al. Association between albumin-bilirubin score and in-hospital mortality in patients with sepsis: evidence from two large databases. Heliyon. 2024;10:e34697. doi: 10.1016/j.heliyon.2024.e34697 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Jiang P, Dou R, Cui Z, et al. Value of albumin-bilirubin score combined with neutrophil count and procalcitonin in ascites in predicting spontaneous bacterial peritonitis in patients with cirrhotic ascites. Lcgdbzz. 2021;37:2097–2101. doi: 10.3969/j.issn.1001-5256.2021.09.019 [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
All data generated or analysed during this study are included in this published article and its Supplementary Information Files.