Prediction of carbapenem-resistant gram-negative bacterial bloodstream infection in intensive care unit based on machine learning.

BACKGROUND: Predicting whether Carbapenem-Resistant Gram-Negative Bacterial (CRGNB) cause bloodstream infection when giving advice may guide the use of antibiotics because it takes 2-5 days conventionally to return the results from doctor's order. METHODS: It is a regional multi-center retrospective study in which patients with suspected bloodstream infections were divided into a positive and negative culture group. According to the positive results, patients were divided into the CRGNB group and other groups. We used the machine learning algorithm to predict whether the blood culture was positive and whether the pathogen was CRGNB once giving the order of blood culture. RESULTS: There were 952 patients with positive blood cultures, 418 patients in the CRGNB group, 534 in the non-CRGNB group, and 1422 with negative blood cultures. Mechanical ventilation, invasive catheterization, and carbapenem use history were the main high-risk factors for CRGNB bloodstream infection. The random forest model has the best prediction ability, with AUROC being 0.86, followed by the XGBoost prediction model in bloodstream infection prediction. In the CRGNB prediction model analysis, the SVM and random forest model have higher area under the receiver operating characteristic curves, which are 0.88 and 0.87, respectively. CONCLUSIONS: The machine learning algorithm can accurately predict the occurrence of ICU-acquired bloodstream infection and identify whether CRGNB causes it once giving the order of blood culture.

Prediction of successful weaning from renal replacement therapy in critically ill patients based on machine learning.

BACKGROUND: Predicting the successful weaning of acute kidney injury (AKI) patients from renal replacement therapy (RRT) has emerged as a research focus, and we successfully built predictive models for RRT withdrawal in patients with severe AKI by machine learning. METHODS: This retrospective single-center study utilized data from our general intensive care unit (ICU) Database, focusing on patients diagnosed with severe AKI who underwent RRT. We evaluated RRT weaning success based on patients being free of RRT in the subsequent week and their overall survival. Multiple logistic regression (MLR) and machine learning algorithms were adopted to construct the prediction models. RESULTS: A total of 976 patients were included, with 349 patients successfully weaned off RRT. Longer RRT duration (7.0 vs. 9.6 d, p = 0.002, OR = 0.94), higher serum cystatin C levels (1.2 vs. 3.2 mg/L, p < 0.001, OR = 0.46), and the presence of septic shock (28.1% vs. 41.5%, p < 0.001, OR = 0.63) were associated with reduced likelihood of RRT weaning. Conversely, a positive furosemide stress test (FST) (60.2% vs. 40.7%, p < 0.001, OR = 2.75) and higher total urine volume 3 d before RRT withdrawal (755 vs. 125 mL/d, p < 0.001, OR = 2.12) were associated with an increased likelihood of successful weaning from RRT. Next, we demonstrated that machine learning models, especially Random Forest and XGBoost, achieving an AUROC of 0.95. The XGBoost model exhibited superior accuracy, yielding an AUROC of 0.849. CONCLUSION: High-risk factors for unsuccessful RRT weaning in severe AKI patients include prolonged RRT duration. Machine learning prediction models, when compared to models based on multivariate logistic regression using these indicators, offer distinct advantages in predictive accuracy.

Chasing the landscape for intrahospital transmission and evolution of hypervirulent carbapenem-resistant Klebsiella pneumoniae.

The spread of hypervirulent carbapenem-resistant Klebsiella pneumoniae (Hv-CRKP) is a global health concern. Here, we report the intrahospital colonization and spread of Hv-CRKP isolates in a tertiary hospital from 2017 to 2022. Analyses of 90 nonredundant CRKP isolates from 72 patients indicated that Hv-CRKP transferability relies on the dominant ST11-K64 clone. Whole-genome sequencing of 11 representative isolates gave 31 complete plasmid sequences, including 12 KPC-2 resistance carriers and 10 RmpA virulence vehicles. Apart from the binary vehicles, we detected two types of fusion plasmids, favoring the cotransfer of RmpA virulence and KPC-2 resistance. The detection of ancestry/relic plasmids enabled us to establish genetic mechanisms by which rare fusion plasmids form. Unexpectedly, we found a total of five rmpA promoter variants (P9T-P13T) exhibiting distinct activities and varying markedly in their geographic distributions. CRISPR/Cas9 manipulation confirmed that an active PT11-rmpA regulator is a biomarker for the "high-risk" ST11-K64/CRKP clone. These findings suggest clonal spread and clinical evolution of the prevalent ST11-K64/Hv-CRKP clones. Apart from improved public awareness of Hv-CRKP convergence, our findings might benefit the development of surveillance (and/or intervention) strategies for the dominant ST11-K64 lineage of the Hv-CRKP population in healthcare sectors.

Ceftazidime/Avibactam for the Treatment of Carbapenem-Resistant Pseudomonas aeruginosa Infection in Lung Transplant Recipients.

Background: Experience of ceftazidime-avibactam (CAZ/AVI) for carbapenem-resistant Pseudomonas aeruginosa (CRPA) infection in recipients after lung transplantation (LT) is relatively limited. Methods: A retrospective observational study was conducted on lung transplant recipients receiving CAZ/AVI therapy for CRPA infection. The primary outcomes were the 14-day and 30-day mortality. The secondary outcomes were clinical cure and microbiological cure. Results: Among 183 LT recipients, a total of 15 recipients with CRPA infection who received CAZ/AVI therapy were enrolled in this study. The mean age of recipients was 54 years and 73.3% of recipients were male. The median time from infection onset to initiation of CAZ/AVI treatment was 4 days (IQR, 3-7) and the mean duration of CAZ/AVI therapy was 10 days. CAZ/AVI was mainly administered as monotherapy in LT recipients (80%). Among these eligible recipients, 14-day and 30-day mortality were 6.7% and 13.3%, respectively. The clinical cure and microbiological cure rates of CAZ/AVI therapy were 53.3% and 60%, respectively. Three recipients (20%) experienced recurrent infection. In addition, the mean lengths of ICU stay and hospital stay were 24 days and 35 days, respectively, among LT recipients. Conclusion: CAZ/AVI may be an alternative and promising regimen for CRPA eradiation in lung transplant recipients.

Cross-Hospital Sepsis Early Detection via Semi-Supervised Optimal Transport With Self-Paced Ensemble.

Leveraging machine learning techniques for Sepsis early detection and diagnosis has attracted increasing interest in recent years. However, most existing methods require a large amount of labeled training data, which may not be available for a target hospital that deploys a new Sepsis detection system. More seriously, as treated patients are diversified between hospitals, directly applying a model trained on other hospitals may not achieve good performance for the target hospital. To address this issue, we propose a novel semi-supervised transfer learning framework based on optimal transport theory and self-paced ensemble for Sepsis early detection, called SPSSOT, which can efficiently transfer knowledge from the source hospital (with rich labeled data) to the target hospital (with scarce labeled data). Specifically, SPSSOT incorporates a new optimal transport-based semi-supervised domain adaptation component that can effectively exploit all the unlabeled data in the target hospital. Moreover, self-paced ensemble is adapted in SPSSOT to alleviate the class imbalance issue during transfer learning. In a nutshell, SPSSOT is an end-to-end transfer learning method that automatically selects suitable samples from two domains (hospitals) respectively and aligns their feature spaces. Extensive experiments on two open clinical datasets, MIMIC-III and Challenge, demonstrate that SPSSOT outperforms state-of-the-art transfer learning methods by improving 1-3% of AUC.

Active surveillance of carbapenem-resistant gram-negative bacteria to guide antibiotic therapy: a single-center prospective observational study.

BACKGROUND: Carbapenem-resistant gram-negative bacteria (CRGNB) have become a public health concern worldwide. The risk factors associated with CRGNB infection after colonization are unknown, nor is the optimal timing of antibiotic treatment, warranting further investigation. METHODS: A 4-year single-center prospective observational study was conducted. CRGNB-colonized patients were incorporated on admission into our observation cohort for an active surveillance culture program, and analysis of risk factors associated with infections after CRGNB colonization was performed. We divided patients into empirical antibiotic therapy groups and standard antibiotic therapy groups according to whether antibiotics were used before or after cultures yielded a result to explore the relationship between the timing of antibiotics and clinical efficacy. RESULTS: 152 out of 451 CRGNB-colonized patients in the prospective observational cohort developed CRGNB infection. The risk factors associated with CRGNB infection after colonization included CRKP (P < 0.001, OR = 3.27) and CRPA (P < 0.001, OR = 2.97) colonization, history of carbapenems use (P < 0.001, OR = 5.48), and immunocompromise (P < 0.001, OR = 7.07). There were 88 infected patients in the empirical antibiotic therapy groups and 64 in standard antibiotic therapy groups. The mortality was lower in empirical therapy groups than standard therapy groups (17.0% vs. 37.5%, P = 0.004, OR = 0.32). CONCLUSIONS: CRGNB colonized patients who are prone to infection have some high-risk factors included CRKP and CRPA colonization, immunocompromise, and prior carbapenems use. Once infection occurs in CRGNB-colonized patients, early use of effective antibiotics may be associated with reduced mortality, but more studies are needed.

Severe acute kidney injury predicting model based on transcontinental databases: a single-centre prospective study.

OBJECTIVES: There are many studies of acute kidney injury (AKI) diagnosis models lack of external validation and prospective validation. We constructed the models using three databases to predict severe AKI within 48 hours in intensive care unit (ICU) patients. DESIGN: A retrospective and prospective cohort study. SETTING: We studied critically ill patients in our database (SHZJU-ICU) and two other public databases, the Medical Information Mart for Intensive Care (MIMIC) and AmsterdamUMC databases, including basic demographics, vital signs and laboratory results. We predicted the diagnosis of severe AKI in patients in the next 48 hours using machine-learning algorithms with the three databases. Then, we carried out real-time severe AKI prediction in the prospective validation study at our centre for 1 year. PARTICIPANTS: All patients included in three databases with uniform exclusion criteria. PRIMARY AND SECONDARY OUTCOME MEASURES: Effect evaluation index of prediction models. RESULTS: We included 58 492 patients, and a total of 5257 (9.0%) patients met the definition of severe AKI. In the internal validation of the SHZJU-ICU and MIMIC databases, the best area under the receiver operating characteristic curve (AUROC) of the model was 0.86. The external validation results by AmsterdamUMC database were also satisfactory, with the best AUROC of 0.86. A total of 2532 patients were admitted to the centre for prospective validation; 358 positive results were predicted and 344 patients were diagnosed with severe AKI, with the best sensitivity of 0.72, the specificity of 0.80 and the AUROC of 0.84. CONCLUSION: The prediction model of severe AKI exhibits promises as a clinical application based on dynamic vital signs and laboratory results of multicentre databases with prospective and external validation.

Ceftazidime/Avibactam versus Polymyxin B in the Challenge of Carbapenem-Resistant Pseudomonas aeruginosa Infection.

PURPOSE: Ceftazidime/avibactam (CAZ/AVI) monotherapy and polymyxin B-based combination therapy are currently two treatment options for patients with carbapenem-resistant Pseudomonas aeruginosa (CRPA) infection; however, few studies have contrasted the relative efficacy of the two antibiotic regimens. The purpose of this study was to compare the effectiveness of CAZ/AVI and polymyxin B against CRPA infection and analyze the independent predictors of 30-day mortality or survival. PATIENTS AND METHODS: This single-center retrospective observational study included patients with CRPA infection treated with CAZ/AVI or polymyxin B between January 2018 and December 2020. The primary outcomes were the 14-day and 30-day mortality. The secondary outcomes were in-hospital mortality and bacterial clearance. Baseline characteristics and outcomes were compared between the two groups, and COX regression analysis was used to identify predictors of 30-day mortality. RESULTS: A total of 136 patients with CRPA infection were enrolled, including 51 patients in the CAZ/AVI group and 85 patients in the polymyxin B group. The 14-day mortality (5.9% vs 27.1%, p=0.002), 30-day mortality (13.7% vs 47.1%, p<0.001) and in-hospital mortality (29.4% vs 60.0%, p=0.001) in the CAZ/AVI group were significantly lower than the polymyxin B group. The bacterial clearance rate (45.1% vs 12.9%, p<0.001) in the CAZ/AVI group were higher than in the polymyxin B group. After adjustment by propensity score matching, the CAV/AVI group still had lower 30-day mortality (14.3% vs 42.9%, p=0.018) and higher bacterial clearance rate (42.9% vs 14.3%, p=0.018) than the polymyxin B group. The multivariate COX analysis showed that the age was identified as independent predictor of 30-day mortality while CAZ/AVI therapy and central venous catheterization emerged as independent predictors of 30-day survival. CONCLUSION: CAZ/AVI therapy was superior to polymyxin B therapy for patients with CRPA infection, and provided significant survival benefits, but further larger studies were needed to substantiate our findings.

Early prediction of carbapenem-resistant Gram-negative bacterial carriage in intensive care units using machine learning.

OBJECTIVES: This study constructed a carbapenem-resistant Gram-negative bacteria (CR-GNB) carriage prediction model to predict the CR-GNB incidence in a week. METHODS: We used our database to select patients with complete CR-GNB screening records between the years 2015 and 2019 and constructed the model using multivariable logistic regression and three machine learning algorithms. Then we chose the optimal model and verified the accuracy by daily prediction and recorded the occurrence of CR-GNB in all intensive care unit patients admitted for 4 months. RESULTS: There were 1385 patients with positive CR-GNB cultures and 1535 negative patients in this study. Forty-five variables had statistically significant differences. We included 16 variables in the multivariable logistic regression model and built three machine learning models for all variables. In terms of accuracy and the area under the receiver operating characteristic (AUROC) curve, random forest was better than XGBoost and decision tree and better than a multivariable logistic regression model (accuracy: 84%>82%>81%>72%, AUROC: 0.91>0.89=0.89>0.78). In a 4-month prospective study, 74 cases were predicted to have positive CR-GNB culture within 7 days, 132 cases were predicted to be negative, 86 cases were positive, and 120 cases were negative, with an overall accuracy of 85.92% and AUROC of 92.02%. CONCLUSION: Machine learning prediction models can predict the occurrence of CR-GNB colonisation or infection within a one-week period and can guide medical staff in real time to identify high-risk groups more accurately.

[Analysis of high-risk factors and clinical characteristics of sepsis-related liver injury].

OBJECTIVE: To analyze the risk factors and clinical characteristics of liver injury in patients with sepsis and to provide a reference for early recognition, early diagnosis, early intervention, and improve the survival rate of patients. METHODS: The clinical data of sepsis patients admitted to the department of general intensive care unit (ICU) of the Second Affiliated Hospital of Zhejiang University School of Medicine from July 2014 to October 2020 were retrospectively analyzed. According to the occurrence of acute liver injury, patients with sepsis were divided into the liver injury group and the non-liver injury group, and the differences of demographic data, history, history of primary diseases, laboratory indicators on the first time of admission, treatments, the severity of the disease and other indicators were compared and analyzed. Logistic regression was used to analyze the risk factors for sepsis-related liver injury. RESULTS: A total of 527 patients with sepsis were enrolled, and 129 patients with acute liver injury, accounting for 24.48%. Compared with the non-liver injury group, acute physiology and chronic health evaluation II (APACHE II), sequential organ failure assessment (SOFA), pro-brain natriuretic peptide (pro-BNP), serum MB isoenzyme of creatine kinase (CK-MB), total bile acid (TBA), serum creatinine (SCr), blood urea nitrogen (BUN), lactic acid (Lac), lactate dehydrogenase (LDH), C-reactive protein (CRP), procalcitonin (PCT) in liver injury group were significantly increased [APACHE II score: 23.00±10.40 vs. 16.10±8.10, SOFA score: 9.17±4.29 vs. 5.90±3.12, pro-BNP (ng/L): 5 500.0 (1 166.0, 16 865.0) vs. 1 377.2 (448.8, 6 136.5), CK-MB (U/L): 23.0 (13.0, 55.0) vs. 18.0 (13.0, 31.0), TBA (µmol/L): 5.0 (2.4, 12.9) vs. 2.6 (1.4, 4.9), SCr (µmol/L): 146.0 (75.0, 222.0) vs. 71.0 (52.0, 125.8), BUN (mmol/L): 13.4 (8.8, 20.2) vs. 7.9 (4.9, 11.6), Lac (mmol/L): 2.0 (1.4, 4.4) vs. 1.4 (1.0, 2.2), LDH (µmol×s-1×L-1): 6.43 (3.76, 11.99) vs. 4.55 (3.38, 6.63), CRP (mg/L): 113.0 (61.8, 201.0) vs. 95.0 (37.3, 170.1), PCT (µg/L): 3.8 (1.0, 23.3) vs. 0.8 (0.2, 6.4)], prothrombin time (PT), international standard ratio (INR) and activated partial thrombin time (APTT) were significantly longer [PT (s): 19.4±7.6 vs. 16.0±4.0, INR: 1.7±1.0 vs. 1.3±0.5, APTT (s): 54.0±25.8 vs. 44.1±15.1], plasma fibrinogen (FIB), platelet count (PLT), albumin (ALB), and cholesterol (CHOL) were decreased [FIB (g/L): 4.2±2.3 vs. 4.9±1.8, PLT (×109/L): 116.3±74.3 vs. 182.7±108.6, ALB (g/L): 25.4±5.5 vs. 27.6±5.5, CHOL (mmol/L): 2.5±1.2 vs. 3.2±1.3], the probability of shock was significantly increased (91.47% vs. 59.19%), and the duration of shock was prolonged [days: 5.0 (2.0, 9.0) vs. 1.0 (0.0, 3.0)], positive rate of microbial culture (81.40% vs. 71.11%), probability of occurrence of drug-resistant bacteria (67.44% vs. 47.99%) were significantly higher, mechanical ventilation time [days: 6.0 (2.0, 12.7) vs. 2.4 (0.0, 6.9)], continuous renal replacement therapy (CRRT) time [days: 1.2 (0.0, 5.0) vs. 0.0 (0.0, 0.0)], the length of intensive care unit (ICU) stay [days: 9.0 (5.0, 18.0) vs. 7.0 (3.0, 13.0)] were significantly longer, 28-day mortality was significantly higher (80.62% vs. 28.89%), and the differences were statistically significant (all P < 0.05). Further Logistic regression analysis showed that PLT decline, PT prolongation, CRRT duration, shock duration and 28-day mortality were correlated with sepsis-related liver injury [odds ratios (OR) and 95% confidence interval (95%CI) were 0.992 (0.987-0.998), 3.103 (1.507-6.387), 1.198 (1.074-1.336), 1.196 (1.049-1.362), and 0.213 (0.072-0.633), respectively, all P < 0.05]. CONCLUSIONS: Prolonged PT and decreased PLT are independent risk factors for sepsis complicated with liver injury. The long duration of CRRT, long duration of shock, and high mortality are independent clinical characteristics of patients with sepsis-related liver injury.

Convergence of carbapenem resistance and hypervirulence in a highly-transmissible ST11 clone of K. pneumoniae: An epidemiological, genomic and functional study.

Co-occurrence of hypervirulence and KPC-2 carbapenem resistant phenotypes in a highly-transmissible ST11 clone ofKlebsiella pneumoniae has elicited deep concerns from public health stand point. To address this puzzle, we conducted a large-scale epidemiological, clinical and genomic study of K. pneumonia ST11 clones with both hypervirulence and carbapenem resistance in two tertiary hospitals in Zhejiang province. Most of the patients (15/23) were diagnosed with exclusively carbapenem-resistant K. pneumoniae (CRKP) infections. Ten death cases were reported, some of which are due to the failure of antibiotic therapies. As a result, we identified one new rare sequence types (ST449) to KPC-2-producing CRKP, in addition to the dominant ST11. These clinical isolates of K. pneumoniae are multi-drug resistant and possess a number of virulence factors. Experimental infections of wax moth larvae revealed the presence of hypervirulence at varied level, suggesting the complexity in bacterial virulence factors. However, plasmid curing assays further suggested that the rmpA2-virulence plasmid is associated with, but not sufficient for neither phenotypic hypermucoviscosity nor virulence of K. pneumoniae. Intriguingly, all the rmpA2 genes were found to be inactive due to genetic deletion. In total, we reported 21 complete plasmid sequences comprising 13 rmpA2-positive virulence plasmids and 8 blaKPC-2-harboring resistance plasmids. In addition to the prevalent pLVKP-like virulence plasmid variants (~178kb), we found an unexpected diversity among KPC-2-producing plasmids whose dominant form is IncFII-IncR type (~120kb), rather than the previously anticipated version of ~170kb. These findings provide an updated snapshot of convergence of hypervirulence and carbapenem resistance in ST11 K. pneumoniae.

Feasibility and safety of total percutaneous closure of femoral arterial access sites after veno-arterial extracorporeal membrane oxygenation.

To evaluate the safety and efficacy of total percutaneous closure of the femoral artery access site after veno-arterial extracorporeal membrane oxygenation (VA-ECMO) with the Perclose ProGlide device.This retrospective observational study during an almost 2-year period included 21 patients who underwent VA-ECMO in whom the femoral artery puncture site was closed percutaneously with Perclose ProGlide devices. Technical success was defined as successful arterial closure of the common femoral artery, without the need for additional surgical or endovascular procedures. Access site complications were recorded at 24 hours and 30 days after arterial closure, such as major bleeding requiring transfusion or surgical intervention, minor bleeding, groin infection, pseudoaneurysm, and lymphocele.Technical success was achieved in 20 patients (95.2%). One patient required surgical repair for an access site pseudoaneurysm. Eighteen femoral arteries were closed with 2 devices each, while 3 patients required the use of a third device for femoral artery access site closure to achieve adequate hemostasis. No arterial thrombosis, arterial dissection, arterial stenosis, groin infection, or arteriovenous fistula occurred during the periprocedural period (within 24 hours of arterial closure) or during 30-day follow-up.Percutaneous closure with the Perclose ProGlide device is a feasible procedure for closing femoral arterial access sites after VA-ECMO, with a low incidence of access site complications.

α-Mangostin suppresses NLRP3 inflammasome activation via promoting autophagy in LPS-stimulated murine macrophages and protects against CLP-induced sepsis in mice.

BACKGROUND: The major mechanism of sepsis is immunosuppression caused by host response dysfunction. It has been found that α-Mangostin (α-M) is a potential candidate as a treatment for multiple inflammatory and immune disorders. To date, the role of α-M in host response during sepsis remains unexplored. METHODS AND RESULTS: Herein, we examined the effect of α-M on macrophages-mediated host response in the presence of lipopolysaccharide (LPS), and the vital organ function, inflammatory response, and survival rate in septic mice. In murine peritoneal macrophages, α-M induced autophagy and then inhibited LPS-stimulated NLRP3 inflammasome activation, as well as interleukin-1ß (IL-1ß) production. Moreover, α-M improved phagocytosis and killing of macrophages, and increased M2 macrophages numbers after LPS stimulation. Furthermore, in vivo experiment suggested that α-M reduced serum levels of tumor necrosis factor-α (TNF-α), interferon-γ (IFN-γ), IL-1ß, alanine transaminase (ALT), aspartate transaminase (AST), and creatinine (Cr), whilst increased that of interleukin-10 (IL-10) in caecal ligation and puncture (CLP) mice. CONCLUSION: Taken together, these findings showed that α-M-mediated macrophages autophagy contributed to NLRP3 inflammasome inactivation and α-M exerted organ protection in septic mice.

Early use of polymyxin B reduces the mortality of carbapenem-resistant Klebsiella pneumoniae bloodstream infection.

Polymyxin B is one of the last resort option for carbapenem-resistant Klebsiella pneumoniae (CRKP) bloodstream infection in China. Therefore, the timing of administration of polymyxin is frequently delayed. We collected 40 cases of CRKP bloodstream infections (BSIs) treated with combinations based on polymyxin B over 30 months. The primary outcome, 30-day mortality rate, was 52.5% (21/40). Early administration of polymyxin B is meant to administer the drug within 48h of diagnosing bacteremia. Delayed administration was considered when polymyxin B was administered after 48h of bacteremia onset. Polymyxin B duration and total dosages were similar in the two groups (11.57 days versus 11.76 days, p=0.919; 1306.52mg versus 1247.06mg, p=0.711). Compared with delayed administration, early use of polymyxin B-based combination therapy had a significant increase in the rate of bacterial clearance (65.22% versus 29.41%, p=0.025; OR=0.533) and decreased 30-day mortality (39.13% versus 70.59%, p=0.045; OR=0.461) and overall mortality (43.48% versus 82.35%, p=0.022; OR=0.321).

Early use of polymyxin B reduces the mortality of carbapenem-resistant Klebsiella pneumoniae bloodstream infection

ABSTRACT Polymyxin B is one of the last resort option for carbapenem-resistant Klebsiella pneumoniae (CRKP) bloodstream infection in China. Therefore, the timing of administration of polymyxin is frequently delayed. We collected 40 cases of CRKP bloodstream infections (BSIs) treated with combinations based on polymyxin B over 30 months. The primary outcome, 30-day mortality rate, was 52.5% (21/40). Early administration of polymyxin B is meant to administer the drug within 48 h of diagnosing bacteremia. Delayed administration was considered when polymyxin B was administered after 48 h of bacteremia onset. Polymyxin B duration and total dosages were similar in the two groups (11.57 days versus 11.76 days, p = 0.919; 1306.52 mg versus 1247.06 mg, p = 0.711). Compared with delayed administration, early use of polymyxin B-based combination therapy had a significant increase in the rate of bacterial clearance (65.22% versus 29.41%, p = 0.025; OR = 0.533) and decreased 30-day mortality (39.13% versus 70.59%, p = 0.045; OR = 0.461) and overall mortality (43.48% versus 82.35%, p = 0.022; OR = 0.321).