Clinical characteristics and 28-day outcomes of sepsis associated with klebsiella pneumoniae carbapenemase-producing Enterobacterales: a single-center retrospective cohort study Authors: Sheng Huang, Sun Yu, Bin Pan, Xiaoping Yang, Jun Sha, Yilan Wang, Chunyang Xu Published: 2026 ABSTRACT BackgroundSepsis caused by carbapenemase-producing Enterobacterales, particularly Klebsiella pneumoniae carbapenemase (KPC)-producing strains, is associated with high mortality and limited treatment options. However, data regarding dynamic organ dysfunction and prognostic factors in this population remain limited.ObjectiveTo compare the clinical characteristics and outcomes of sepsis due to KPC-producing versus non-KPC-producing Enterobacterales and to identify independent predictors of 28-day mortality.MethodsIn this single-center retrospective cohort study, 320 intensive care unit (ICU) patients with sepsis and microbiologically confirmed Enterobacterales infection (January-December 2024) were included. Patients were classified as KPC (n = 210) or non-KPC (n = 110) using the NG-Test CARBA 5 assay, with polymerase chain reaction (PCR) confirmation when available. The primary outcome was 28-day all-cause mortality from sepsis onset. Multivariable logistic and Cox regression analyses were performed.ResultsThe KPC group had significantly higher 28-day mortality than the non-KPC group (35.2% vs. 23.6%, odds ratio [OR] = 1.77, 95% confidence interval [CI]: 1.04-3.00, P = 0.030). KPC infection remained independently associated with mortality after adjustment (adjusted OR [aOR] = 1.58, 95% CI: 1.02–2.44, P = 0.04). Additional independent risk factors included higher Day-1 Sequential Organ Failure Assessment (SOFA) score (aOR = 1.08), elevated lactate level (aOR = 1.12), and inadequate early antimicrobial therapy (aOR = 1.69; all P < 0.05). In the KPC subgroup, ΔSOFA, positive fluid balance, and low albumin levels were also associated with mortality.ConclusionSepsis caused by KPC-producing Enterobacterales is associated with more severe organ dysfunction and increased short-term mortality. Early administration of appropriate antimicrobial therapy, along with optimized organ support and fluid management, may improve clinical outcomes.Clinical trial numberNot applicable.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12879-026-13481-6. Introduction Sepsis remains a leading cause of mortality in global critical care medicine [1], with persistently high mortality rates in intensive care units (ICUs) [2, 3]. In recent years, the widespread use of antimicrobial agents and the continuous dissemination of resistant strains have led to a substantial increase in sepsis caused by drug-resistant Gram-negative bacteria, posing significant challenges to infection control and early treatment strategies [4, 5]. Among these pathogens, Enterobacterales are one of the most common causes of severe infections and sepsis, owing to their high pathogenicity and propensity for antimicrobial resistance [6]. Carbapenemase-producing Enterobacterales (CPE), particularly strains producing Klebsiella pneumoniae carbapenemase (KPC), have spread rapidly worldwide [7, 8]. KPC enzymes can be horizontally transferred among different Enterobacterales species via plasmids, resulting in complex resistance patterns and considerable epidemiological heterogeneity [9]. A clinical study has suggested that KPC-related infections are associated with limited therapeutic options, higher rates of treatment failure, and adverse clinical outcomes [10]. However, existing evidence remains incomplete in several important aspects. In sepsis management, dynamic changes in early organ dysfunction, as well as the timeliness and adequacy of antimicrobial therapy, are considered critical determinants of prognosis [11]. However, current evidence regarding sepsis associated with KPC-producing Enterobacterales is limited in the following key areas. First, most previous studies have focused on specific bacterial species (e.g., Klebsiella pneumoniae) or particular infection types such as bloodstream infections, rather than evaluating sepsis as a dynamic and systemic clinical syndrome. Second, there is a paucity of data on the temporal evolution of organ dysfunction and its relationship with clinical outcomes in patients with KPC-producing Enterobacterales infection. Third, it remains unclear whether resistance mechanisms independently influence prognosis or whether observed differences in outcomes are primarily driven by disease severity, treatment adequacy, and host factors. In addition, systematic evaluations integrating microbiological characteristics [12], dynamic Sequential Organ Failure Assessment (SOFA) scores, and time-dependent survival analyses are lacking, particularly in real-world ICU populations. Patients in ICU settings often present with greater illness severity, require organ support therapies, and exhibit complex clinical trajectories, highlighting the need for more comprehensive clinical evidence in this context. Therefore, using a real-world clinical cohort, this study aims to compare the etiological characteristics, resistance phenotypes [13], dynamic organ dysfunction, and clinical outcomes between patients with sepsis caused by KPC-producing versus non-KPC-producing Enterobacterales infection, and to identify independent risk factors associated with 28-day all-cause mortality. Methods Study design and population This single-center retrospective cohort study included consecutive adult patients admitted to the ICU with sepsis between January and December 2024. Sepsis was defined according to the Sepsis-3 criteria [14], based on suspected or confirmed infection accompanied by an acute increase in SOFA score ≥ 2 points. The source population consisted of consecutive adult patients admitted to the ICU during the study period. All included patients met Sepsis-3 criteria at the time of ICU admission, which was defined as the time of sepsis onset for this study. Patients were eligible if they had microbiologically confirmed Enterobacterales infection from clinically relevant specimens and complete clinical data available for analysis. Both community-acquired and hospital-acquired infections were included. Infection onset was classified as community-acquired, ward-acquired, or ICU-acquired according to standard definitions. The final analytic cohort was assembled by applying predefined inclusion and exclusion criteria to all screened ICU patients during the study period, as illustrated in the study flow diagram (Fig. 1). Inclusion and exclusion criteria Patients were included if they met all of the following criteria: age ≥ 18 years; diagnosis of sepsis or septic shock according to Sepsis-3 criteria; and isolation of Enterobacterales from clinically relevant specimens. Infection was defined based on a combination of clinical presentation, laboratory findings, radiological evidence, and microbiological results. Only cases judged as true infection by treating clinicians were included. Standard criteria were applied to differentiate infection from colonization or contamination, particularly for respiratory and urinary tract specimens. For patients with multiple admissions or infection episodes during the study period, only the first episode of sepsis was included. Patients were excluded if they died or were discharged within 24 h of admission; had polymicrobial infection with non-Enterobacterales as the primary pathogen; or had missing key clinical variables (e.g., SOFA score, lactate level, or outcome data). Patients receiving long-term immunosuppressive therapy were excluded only when essential clinical data were incomplete. Determination of KPC-producing strains and group definitions Bacterial identification was performed using the VITEK-2 Compact automated system (bioMérieux, France) and/or matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry according to standard laboratory procedures. Antimicrobial susceptibility testing was conducted using broth microdilution or automated systems. Results were interpreted according to Clinical and Laboratory Standards Institute (CLSI) guidelines. Carbapenemase production was screened using the NG-Test CARBA 5 assay. Polymerase chain reaction (PCR) detection of the bla_KPC gene was performed in a predefined subset of isolates (n = 96/320, 30.0%) for confirmatory purposes [15, 16]. Patients were classified into KPC and non-KPC groups based on carbapenemase detection. The non-KPC group included both carbapenem-susceptible isolates and isolates with non-KPC resistance mechanisms [17, 18]. Biofilm formation ability was assessed in all 320 isolates (100% of total). Microbiological procedures not directly contributing to the present analyses (e.g., multilocus sequence typing and combination susceptibility testing) were not included in the main analysis. Clinical specimen processing and detection All clinical specimens were collected and processed as part of routine clinical care according to institutional standard protocols. No additional interventions were performed for the purpose of this study. Clinical data and variable collection Clinical data were extracted from the electronic medical record system, including demographic characteristics, comorbidities, prior antimicrobial exposure, infection site, and ICU-acquired infection. Disease severity was assessed using SOFA score and blood lactate at admission. Dynamic organ dysfunction was evaluated using SOFA scores on Days 1, 3, and 7, and ΔSOFA (Day 3 − Day 1). SOFA scores were summarized as median (interquartile range, IQR) due to non-normal distribution. For visualization purposes, mean values with 95% confidence intervals (CIs) were used in figures to illustrate temporal trends. Missing SOFA values at each time point were not imputed, and analyses were conducted using available data. Treatment-related variables included mechanical ventilation, vasopressor use, continuous renal replacement therapy (CRRT), and fluid balance during Days 1–3. Antimicrobial therapy variables included empirical therapy within 24 h and appropriate therapy, defined as administration of at least one in vitro active agent within 48 h. Empirical regimens mainly included carbapenem-based or β-lactam/β-lactamase inhibitor therapy. The primary outcome was 28-day all-cause mortality, defined from the time of sepsis onset (ICU admission). Secondary outcomes included ICU length of stay, duration of mechanical ventilation, septic shock, and acute kidney injury (AKI), defined according to the Kidney Disease: Improving Global Outcomes (KDIGO) criteria. Source control was defined as any intervention aimed at eliminating the primary focus of infection, including but not limited to abscess drainage, surgical debridement, removal of infected devices (e.g., catheters), or relief of obstruction. Effective source control was defined as the implementation of an appropriate intervention within 48 h of sepsis onset, accompanied by documented clinical improvement or stabilization, as recorded in the medical records. Assessment of source control was based on predefined criteria and determined retrospectively from clinical documentation. Ethics statement The study protocol was reviewed and approved by the Institutional Review Board of Changshu Hospital Affiliated to Soochow University, First People’s Hospital of Changshu City (Ethics Approval Number: SXCS036). Given the retrospective observational nature of this study and the absence of additional interventions, the requirement for informed consent was waived by the ethics committee. All data were anonymized prior to analysis to protect patient privacy. Sample size and statistical power analysis The sample size was determined based on consecutively available cases during the study period. The number of outcome events was sufficient to support multivariable regression analyses, with an events-per-variable (EPV) ratio exceeding 10, indicating acceptable model stability. Statistical analysis Normality of continuous variables was assessed using the Shapiro-Wilk test. Normally distributed variables are presented as mean ± standard deviation (SD) and compared using the independent-samples t test. Non-normally distributed variables are expressed as median (IQR) and compared using the Mann-Whitney U test, with Hodges-Lehmann median differences and 95% CIs reported. Categorical variables are presented as frequencies and percentages and compared using the chi-square test or Fisher’s exact test, as appropriate. Time-to-event analyses were performed using sepsis onset (ICU admission) as time zero. Kaplan-Meier survival curves were constructed and compared using the log-rank test. Multivariable Cox proportional hazards models were used to identify independent risk factors for 28-day mortality, reporting adjusted hazard ratios (aHRs) with 95% CIs. The proportional hazards assumption was assessed using Schoenfeld residuals. For correlated organ dysfunction measures, ΔSOFA (Day 3 − Day 1) was used instead of serial SOFA scores. Candidate variables for multivariable analyses were selected a priori based on clinical relevance and prior evidence. These included KPC infection status, age, sex, Charlson Comorbidity Index, prior antimicrobial exposure, ICU-acquired infection, infection source, pathogen species, Day 1 SOFA score, blood lactate level, adequacy of early antimicrobial therapy, and source control. All candidate variables considered for model inclusion are explicitly listed above to ensure transparency. To avoid overfitting, variables were entered into multivariable models while maintaining an events-per-variable (EPV) ratio ≥ 10. Multicollinearity among candidate variables was assessed using variance inflation factors (VIF), with VIF < 5 considered acceptable. No significant collinearity was observed among variables included in the final models. Model discrimination was evaluated using the area under the receiver operating characteristic curve (AUC), and calibration was assessed using the Hosmer-Lemeshow test. All statistical analyses were two-sided, with P < 0.05 considered statistically significant. Analyses were performed using GraphPad Prism 10.0, R (version 4.3.2), and SPSS version 26.0. Results Study population screening and overall follow-up In this real-world ICU cohort, a total of 341 patients with sepsis and Enterobacterales infection were screened, of whom 21 excluded according to predefined criteria, yielding a final analytic cohort of 320 patients (Fig. 1). Patients were stratified into a KPC group (n = 210) and a non-KPC group (n = 110). Compared with the non-KPC group, patients with KPC-producing infections exhibited more complex resistance phenotypes and a higher proportion of biofilm-forming isolates. Clinically, they showed more pronounced early and sustained organ dysfunction following sepsis onset, reflected by higher SOFA scores over time, and required more intensive organ support, including mechanical ventilation and vasoactive agents. The KPC group also demonstrated lower 28-day survival. Multivariable analyses identified KPC infection, dynamic organ function deterioration, and early treatment-related factors as independent determinants of 28-day all-cause mortality. Detailed results are presented in the following sections. All patients included in the final analysis were admitted to the ICU and met Sepsis-3 criteria at the time of ICU admission. Fig. 1Overview of study design and key findings. (A) Graphical summary illustrating the study framework and main findings. Patients with KPC-producing Enterobacterales infection exhibited more severe dynamic organ dysfunction (increased ΔSOFA), higher lactate levels, and a lower rate of appropriate early antimicrobial therapy, which were associated with increased risk of 28-day mortality. (B) Study flow diagram. A total of 341 patients were screened, of whom 21 were excluded (non-Enterobacterales infection, n = 7; missing key data, n = 6; ICU stay < 24 h, n = 5). The final cohort included 320 patients, comprising 210 in the KPC group and 110 in the non-KPC group Overview of study design and key findings. (A) Graphical summary illustrating the study framework and main findings. Patients with KPC-producing Enterobacterales infection exhibited more severe dynamic organ dysfunction (increased ΔSOFA), higher lactate levels, and a lower rate of appropriate early antimicrobial therapy, which were associated with increased risk of 28-day mortality. (B) Study flow diagram. A total of 341 patients were screened, of whom 21 were excluded (non-Enterobacterales infection, n = 7; missing key data, n = 6; ICU stay < 24 h, n = 5). The final cohort included 320 patients, comprising 210 in the KPC group and 110 in the non-KPC group Species composition and infection site distribution A total of 210 patients in the KPC group and 110 in the non-KPC group were included. The distribution of causative pathogens differed significantly between the two groups (χ² = 27.52, P < 0.001) (Table 1). Klebsiella pneumoniae was the predominant pathogen in the KPC group, accounting for 81.9% of isolates, compared with 54.5% in the non-KPC group (OR = 3.77, 95% CI: 2.25–6.31, P < 0.001). In contrast, Escherichia coli and other Enterobacterales species were less frequently identified in the KPC group (E. coli: 8.6% vs. 24.5%, OR = 0.29, 95% CI: 0.15–0.55, P < 0.001; other Enterobacterales: 9.5% vs. 20.9%, OR = 0.40, 95% CI: 0.21–0.75, P = 0.004). Regarding infection sites, no significant difference in overall distribution was observed between the two groups (χ² = 3.15, P = 0.370). Table 1Distribution of pathogens and sites of infection in patients with Enterobacterales sepsis according to KPC productionVariableKPC group (n = 210)Non-KPC group (n = 110)OR (95% CI)*χ²P valuePathogen distribution (overall)———27.52< 0.001Klebsiella pneumoniae, n (%)172 (81.9)60 (54.5)3.77 (2.25–6.31)24.1< 0.001Escherichia coli, n (%)18 (8.6)27 (24.5)0.29 (0.15–0.55)16.8< 0.001Other Enterobacterales, n (%)20 (9.5)23 (20.9)0.40 (0.21–0.75)8.40.004Site of infection (overall)———3.150.37Respiratory tract, n (%)47 (22.4)32 (29.1)0.71 (0.42–1.21)1.630.202Bloodstream infection, n (%)52 (24.8)26 (23.6)1.07 (0.61–1.87)0.060.812Urinary tract, n (%)58 (27.6)22 (20.0)1.53 (0.86–2.71)2.330.127Intra-abdominal / biliary tract, n (%)53 (25.2)30 (27.3)0.90 (0.53–1.53)0.140.709Note: KPC, Klebsiella pneumoniae carbapenemase; OR, odds ratio; CI, confidence interval*Odds ratios were calculated using the non-KPC group as the reference categoryCategorical variables were compared using the χ² test or Fisher’s exact test, as appropriate Table 1Distribution of pathogens and sites of infection in patients with Enterobacterales sepsis according to KPC productionVariableKPC group (n = 210)Non-KPC group (n = 110)OR (95% CI)*χ²P valuePathogen distribution (overall)———27.52< 0.001Klebsiella pneumoniae, n (%)172 (81.9)60 (54.5)3.77 (2.25–6.31)24.1< 0.001Escherichia coli, n (%)18 (8.6)27 (24.5)0.29 (0.15–0.55)16.8< 0.001Other Enterobacterales, n (%)20 (9.5)23 (20.9)0.40 (0.21–0.75)8.40.004Site of infection (overall)———3.150.37Respiratory tract, n (%)47 (22.4)32 (29.1)0.71 (0.42–1.21)1.630.202Bloodstream infection, n (%)52 (24.8)26 (23.6)1.07 (0.61–1.87)0.060.812Urinary tract, n (%)58 (27.6)22 (20.0)1.53 (0.86–2.71)2.330.127Intra-abdominal / biliary tract, n (%)53 (25.2)30 (27.3)0.90 (0.53–1.53)0.140.709Note: KPC, Klebsiella pneumoniae carbapenemase; OR, odds ratio; CI, confidence interval*Odds ratios were calculated using the non-KPC group as the reference categoryCategorical variables were compared using the χ² test or Fisher’s exact test, as appropriate Distribution of pathogens and sites of infection in patients with Enterobacterales sepsis according to KPC production Note: KPC, Klebsiella pneumoniae carbapenemase; OR, odds ratio; CI, confidence interval *Odds ratios were calculated using the non-KPC group as the reference category Categorical variables were compared using the χ² test or Fisher’s exact test, as appropriate Antimicrobial susceptibility, resistance phenotypes, and biofilm formation ability The resistance profiles of isolates in the two groups are summarized in Table 2. The susceptibility rate to ceftazidime-avibactam (CAZ-AVI) was lower in the KPC group than in the non-KPC group (88.1% vs. 96.4%, OR = 0.28, 95% CI: 0.09–0.82, P = 0.014). The proportion of multidrug-resistant (MDR) strains was significantly higher in the KPC group compared with the non-KPC group (64.8% vs. 38.2%, OR = 2.97, 95% CI: 1.85–4.79, P < 0.001). Similarly, isolates with moderate or strong biofilm formation were more frequent in the KPC group (52.4% vs. 34.6%, OR = 2.08, 95% CI: 1.29–3.36, P = 0.002, Table S1). These findings indicate a higher burden of antimicrobial resistance and virulence-related characteristics in KPC-producing isolates. Table 2Comparison of antimicrobial susceptibility, resistance phenotypesParametersKPC group (n = 210)Non-KPC group (n = 110)OR (95% CI)*χ²P valueSusceptibility rate to CAZ-AVI, n (%)185 (88.1)106 (96.36)0.28 (0.09–0.82)5.9880.014MDR strains (≥ 3 classes of antimicrobials), n (%)136 (64.76)42 (38.18)2.97 (1.85–4.79)20.662< 0.001Note: KPC, Klebsiella pneumoniae carbapenemase; CAZ-AVI, ceftazidime-avibactam; OR, odds ratio; CI, confidence interval. Categorical variables are presented as n (%). Between-group comparisons were performed using the χ² test or Fisher’s exact test (when expected frequencies were < 5). *ORs were calculated using the non-KPC group as the reference Table 2Comparison of antimicrobial susceptibility, resistance phenotypesParametersKPC group (n = 210)Non-KPC group (n = 110)OR (95% CI)*χ²P valueSusceptibility rate to CAZ-AVI, n (%)185 (88.1)106 (96.36)0.28 (0.09–0.82)5.9880.014MDR strains (≥ 3 classes of antimicrobials), n (%)136 (64.76)42 (38.18)2.97 (1.85–4.79)20.662< 0.001Note: KPC, Klebsiella pneumoniae carbapenemase; CAZ-AVI, ceftazidime-avibactam; OR, odds ratio; CI, confidence interval. Categorical variables are presented as n (%). Between-group comparisons were performed using the χ² test or Fisher’s exact test (when expected frequencies were < 5). *ORs were calculated using the non-KPC group as the reference Comparison of antimicrobial susceptibility, resistance phenotypes Note: KPC, Klebsiella pneumoniae carbapenemase; CAZ-AVI, ceftazidime-avibactam; OR, odds ratio; CI, confidence interval. Categorical variables are presented as n (%). Between-group comparisons were performed using the χ² test or Fisher’s exact test (when expected frequencies were < 5). *ORs were calculated using the non-KPC group as the reference Baseline demographic and clinical characteristics No significant differences were observed between the KPC and non-KPC groups in age, sex distribution, or body mass index (BMI) (all P > 0.05) (Table 3). The Charlson Comorbidity Index was higher in the KPC group compared with the non-KPC group [median (IQR): 4 (3–5) vs. 3 (2–4), P < 0.001]. Prior antimicrobial exposure within 3 months was more frequent in the KPC group (44.8% vs. 30.0%, OR = 1.88, 95% CI: 1.13–3.12, P = 0.010). No significant differences were observed between groups in immunosuppressive status or ICU-acquired infection (all P > 0.05). Table 3Baseline demographic and clinical characteristicsVariablesKPC group (n = 210)Non-KPC group (n = 110)Effect size (95% CI)t/χ²/ZP valueAge, years58.25 ± 7.7756.98 ± 8.74MD = 1.27 (− 0.62, 3.16)−1.320.187Male, n (%)136 (64.76)68 (61.82)OR = 1.13 (0.70–1.83)0.270.603BMI, kg/m²21.43 (19.21, 23.53)21.57 (19.28, 23.86)HL = − 0.14 (− 0.76, 0.48)−0.430.666Charlson Comorbidity Index4 (3, 5)3 (2, 4)HL = 1.00 (0.50, 1.50)−5.08< 0.001Immunosuppressive status, n (%)34 (16.19)17 (15.45)OR = 1.06 (0.56–2.01)0.030.864Antimicrobial use within the preceding 3 months, n (%)94 (44.76)33 (30.00)OR = 1.88 (1.13–3.12)6.570.010ICU-acquired infection, n (%)157 (74.76)72 (65.45)OR = 1.56 (0.92–2.63)3.070.080Note: For continuous variables, normally distributed data are presented as mean ± standard deviation and compared using the independent samples t-test; non-normally distributed data are presented as median (interquartile range) and compared using the Mann-Whitney U test, with the Hodges-Lehmann median difference and its 95% confidence interval reported. Categorical variables are presented as n (%) and compared using the χ² test or Fisher’s exact test. OR, odds ratio (calculated using the non-KPC group as the reference, with 95% confidence interval reported) Table 3Baseline demographic and clinical characteristicsVariablesKPC group (n = 210)Non-KPC group (n = 110)Effect size (95% CI)t/χ²/ZP valueAge, years58.25 ± 7.7756.98 ± 8.74MD = 1.27 (− 0.62, 3.16)−1.320.187Male, n (%)136 (64.76)68 (61.82)OR = 1.13 (0.70–1.83)0.270.603BMI, kg/m²21.43 (19.21, 23.53)21.57 (19.28, 23.86)HL = − 0.14 (− 0.76, 0.48)−0.430.666Charlson Comorbidity Index4 (3, 5)3 (2, 4)HL = 1.00 (0.50, 1.50)−5.08< 0.001Immunosuppressive status, n (%)34 (16.19)17 (15.45)OR = 1.06 (0.56–2.01)0.030.864Antimicrobial use within the preceding 3 months, n (%)94 (44.76)33 (30.00)OR = 1.88 (1.13–3.12)6.570.010ICU-acquired infection, n (%)157 (74.76)72 (65.45)OR = 1.56 (0.92–2.63)3.070.080Note: For continuous variables, normally distributed data are presented as mean ± standard deviation and compared using the independent samples t-test; non-normally distributed data are presented as median (interquartile range) and compared using the Mann-Whitney U test, with the Hodges-Lehmann median difference and its 95% confidence interval reported. Categorical variables are presented as n (%) and compared using the χ² test or Fisher’s exact test. OR, odds ratio (calculated using the non-KPC group as the reference, with 95% confidence interval reported) Baseline demographic and clinical characteristics Note: For continuous variables, normally distributed data are presented as mean ± standard deviation and compared using the independent samples t-test; non-normally distributed data are presented as median (interquartile range) and compared using the Mann-Whitney U test, with the Hodges-Lehmann median difference and its 95% confidence interval reported. Categorical variables are presented as n (%) and compared using the χ² test or Fisher’s exact test. OR, odds ratio (calculated using the non-KPC group as the reference, with 95% confidence interval reported) Antimicrobial therapy characteristics Detailed antimicrobial regimens, including empirical and definitive therapy patterns, are summarized in Table 4. The proportion of patients receiving appropriate antimicrobial therapy within 48 h was lower in the KPC group compared with the non-KPC group (60.5% vs. 73.6%, OR = 0.55, 95% CI: 0.34–0.92, P = 0.019). Ceftazidime-avibactam use also differed between groups (88.1% vs. 96.4%, OR = 0.28, 95% CI: 0.09–0.82, P = 0.014). In addition, a positive fluid balance > 1.5 L during Days 1–3 was more frequent in the KPC group (62.9% vs. 50.0%, OR = 1.69, 95% CI: 1.06–2.70, P = 0.027). Table 4Comparison of antimicrobial therapy and fluid management between groupsVariablesKPC group (n = 210)Non-KPC group (n = 110)OR (95% CI)χ²P valueAppropriate antimicrobial therapy ≤ 48 h, n (%)127 (60.48%)81 (73.64%)0.548 (0.34–0.92)5.500.019Ceftazidime-avibactam use, n (%)185 (88.1%)106 (96.36%)0.279 (0.09–0.82)5.990.014Positive fluid balance > 1.5 L (Day 1–3), n (%)132 (62.86%)55(50.0%)1.692(1.06–2.70)4.910.027Note: Data are presented as n (%). Odds ratios (ORs) were calculated using the non-KPC group as the reference. Comparisons between groups were performed using the chi-square test. Appropriate antimicrobial therapy was defined as administration of at least one in vitro active agent within 48 h of sepsis onset. Positive fluid balance refers to cumulative fluid balance > 1.5 L during Days 1–3 Table 4Comparison of antimicrobial therapy and fluid management between groupsVariablesKPC group (n = 210)Non-KPC group (n = 110)OR (95% CI)χ²P valueAppropriate antimicrobial therapy ≤ 48 h, n (%)127 (60.48%)81 (73.64%)0.548 (0.34–0.92)5.500.019Ceftazidime-avibactam use, n (%)185 (88.1%)106 (96.36%)0.279 (0.09–0.82)5.990.014Positive fluid balance > 1.5 L (Day 1–3), n (%)132 (62.86%)55(50.0%)1.692(1.06–2.70)4.910.027Note: Data are presented as n (%). Odds ratios (ORs) were calculated using the non-KPC group as the reference. Comparisons between groups were performed using the chi-square test. Appropriate antimicrobial therapy was defined as administration of at least one in vitro active agent within 48 h of sepsis onset. Positive fluid balance refers to cumulative fluid balance > 1.5 L during Days 1–3 Comparison of antimicrobial therapy and fluid management between groups Note: Data are presented as n (%). Odds ratios (ORs) were calculated using the non-KPC group as the reference. Comparisons between groups were performed using the chi-square test. Appropriate antimicrobial therapy was defined as administration of at least one in vitro active agent within 48 h of sepsis onset. Positive fluid balance refers to cumulative fluid balance > 1.5 L during Days 1–3 Dynamic changes in organ dysfunction and clinical treatment requirements Dynamic changes in organ dysfunction over the first seven days following sepsis onset are presented in Fig. 2, with corresponding data in Table 5. Overall, SOFA scores were consistently higher in the KPC group than in the non-KPC group throughout the observation period. At individual time points, SOFA scores on Day 1, Day 3, and Day 7 were higher in the KPC group [median (IQR): Day 1: 9 (7–12) vs. 7 (5–10); Day 3: 9 (7–13) vs. 7 (5–10); Day 7: 8 (5–12) vs. 6 (3–9); all P < 0.01]. ΔSOFA was greater in the KPC group [0 (− 2, 2) vs. − 1 (− 3, 1), P = 0.010]. Patients in the KPC group more frequently required organ support therapies. Mechanical ventilation (58.6% vs. 44.5%, OR = 1.77, 95% CI: 1.10–2.86, P = 0.020) and vasoactive agent use (62.9% vs. 46.4%, OR = 1.97, 95% CI: 1.21–3.21, P = 0.005) were more common in the KPC group. There was no significant difference in CRRT use between groups (27.1% vs. 18.2%, OR = 1.67, 95% CI: 0.94–2.97, P = 0.070). SOFA data were available for all patients on Day 1, and for 209 (99.5%) and 206 (98.1%) patients in the KPC group on Day 3 and Day 7, respectively. Missing data were minimal and unlikely to affect the overall trends. Fig. 2Dynamic changes in SOFA score during the first seven days following sepsis onset. The trend of SOFA score changes over the first seven days following sepsis onset is illustrated for patients in the KPC-producing Enterobacterales infection group and the non-KPC group. SOFA scores are presented as means with 95% confidence intervals. Compared to the non-KPC group, patients with KPC-producing Enterobacterales infection exhibited higher overall SOFA scores throughout the follow-up period, with a persistently elevated trend over time. Data are presented as mean values with 95% confidence intervals for visualization purposes Dynamic changes in SOFA score during the first seven days following sepsis onset. The trend of SOFA score changes over the first seven days following sepsis onset is illustrated for patients in the KPC-producing Enterobacterales infection group and the non-KPC group. SOFA scores are presented as means with 95% confidence intervals. Compared to the non-KPC group, patients with KPC-producing Enterobacterales infection exhibited higher overall SOFA scores throughout the follow-up period, with a persistently elevated trend over time. Data are presented as mean values with 95% confidence intervals for visualization purposes Table 5Comparison of dynamic organ dysfunction and clinical therapeutic interventionsVariablesKPC group (n = 210)Non-KPC group (n = 110)Z/χ²Effect size (95% CI)P valueSOFA Day 19 (7–12)7 (5–10)−4.96HL = 2.0 (1.0–3.0)< 0.001SOFA Day 39 (7–13)7 (5–10)−4.21HL = 2.0 (1.0–3.0)< 0.001SOFA Day 78 (5–12)6 (3–9)−3.12HL = 2.0 (1.0–3.0)0.002ΔSOFA (Day 3 - Day 1)0 (− 2, 2)−1 (− 3, 1)−2.58HL = 1.0 (0.0–2.0)0.010Mechanical ventilation, n (%)123 (58.6)49 (44.5)5.53OR = 1.77 (1.10–2.86)0.020CRRT, n (%)57 (27.1)20 (18.2)3.32OR = 1.66 (0.92–2.98)0.070Vasoactive agents, n (%)132 (62.9)51 (46.4)7.84OR = 1.97 (1.21–3.21)0.005Note: Non-normally distributed continuous variables are presented as median (interquartile range) and compared using the Mann-Whitney U test, with the Hodges-Lehmann median difference and its 95% confidence interval reportedCategorical variables are presented as n (%) and compared using the χ² test or Fisher’s exact testOR, odds ratio (calculated using the non-KPC group as the reference) Table 5Comparison of dynamic organ dysfunction and clinical therapeutic interventionsVariablesKPC group (n = 210)Non-KPC group (n = 110)Z/χ²Effect size (95% CI)P valueSOFA Day 19 (7–12)7 (5–10)−4.96HL = 2.0 (1.0–3.0)< 0.001SOFA Day 39 (7–13)7 (5–10)−4.21HL = 2.0 (1.0–3.0)< 0.001SOFA Day 78 (5–12)6 (3–9)−3.12HL = 2.0 (1.0–3.0)0.002ΔSOFA (Day 3 - Day 1)0 (− 2, 2)−1 (− 3, 1)−2.58HL = 1.0 (0.0–2.0)0.010Mechanical ventilation, n (%)123 (58.6)49 (44.5)5.53OR = 1.77 (1.10–2.86)0.020CRRT, n (%)57 (27.1)20 (18.2)3.32OR = 1.66 (0.92–2.98)0.070Vasoactive agents, n (%)132 (62.9)51 (46.4)7.84OR = 1.97 (1.21–3.21)0.005Note: Non-normally distributed continuous variables are presented as median (interquartile range) and compared using the Mann-Whitney U test, with the Hodges-Lehmann median difference and its 95% confidence interval reportedCategorical variables are presented as n (%) and compared using the χ² test or Fisher’s exact testOR, odds ratio (calculated using the non-KPC group as the reference) Comparison of dynamic organ dysfunction and clinical therapeutic interventions Note: Non-normally distributed continuous variables are presented as median (interquartile range) and compared using the Mann-Whitney U test, with the Hodges-Lehmann median difference and its 95% confidence interval reported Categorical variables are presented as n (%) and compared using the χ² test or Fisher’s exact test OR, odds ratio (calculated using the non-KPC group as the reference) Clinical outcomes analysis The 28-day all-cause mortality rate was higher in the KPC group than in the non-KPC group (35.2% vs. 23.6%, OR = 1.77, 95% CI: 1.04-3.00, P = 0.030). Patients in the KPC group had longer ICU length of stay and duration of mechanical ventilation [median (IQR): ICU stay: 12 (7–20) vs. 9 (5–16) days, P = 0.010; mechanical ventilation: 6 (2–11) vs. 4 (1–9) days, P = 0.040]. The incidence of septic shock was higher in the KPC group (52.4% vs. 38.2%, OR = 1.78, 95% CI: 1.10–2.88, P = 0.020) (Table 6). Table 6Specific outcome proportions and hospital burdenOutcomeKPC group (n = 210)Non-KPC group (n = 110)Z/χ²Effect size (95% CI)P value28-day all-cause mortality rate, n (%)74 (35.2)26 (23.6)4.62OR = 1.77 (1.04-3.00)0.030ICU length of stay, days12 (7–20)9 (5–16)−2.56HL = 3.0 (1.0–5.0)0.010Duration of mechanical ventilation, days6 (2–11)4 (1–9)−2.06HL = 2.0 (0.0–4.0)0.040Septic shock, n (%)110 (52.4)42 (38.2)5.56OR = 1.78 (1.10–2.88)0.020AKI, n (%)97 (46.2)40 (36.4)2.87OR = 1.51 (0.93–2.46)0.090Note: ORs were calculated using the non-KPC group as the reference Table 6Specific outcome proportions and hospital burdenOutcomeKPC group (n = 210)Non-KPC group (n = 110)Z/χ²Effect size (95% CI)P value28-day all-cause mortality rate, n (%)74 (35.2)26 (23.6)4.62OR = 1.77 (1.04-3.00)0.030ICU length of stay, days12 (7–20)9 (5–16)−2.56HL = 3.0 (1.0–5.0)0.010Duration of mechanical ventilation, days6 (2–11)4 (1–9)−2.06HL = 2.0 (0.0–4.0)0.040Septic shock, n (%)110 (52.4)42 (38.2)5.56OR = 1.78 (1.10–2.88)0.020AKI, n (%)97 (46.2)40 (36.4)2.87OR = 1.51 (0.93–2.46)0.090Note: ORs were calculated using the non-KPC group as the reference Specific outcome proportions and hospital burden Note: ORs were calculated using the non-KPC group as the reference Analysis of 28-day post-discharge survival and mortality risk factors Univariate logistic regression analyses Survival within 28 days from sepsis onset is shown in Fig. 3. The multivariable models were constructed based on pre-specified clinically relevant variables, including comorbidity burden, prior antimicrobial exposure, infection characteristics, and severity indicators. Kaplan-Meier analysis, using sepsis onset (ICU admission) as time zero, demonstrated a lower cumulative survival probability in the KPC group compared with the non-KPC group (log-rank χ² = 4.030, P = 0.045). In univariable analysis, KPC infection, Day 1 SOFA score, Day 1 blood lactate level, inadequate early antimicrobial therapy, and failure to achieve source control were associated with 28-day mortality (all P < 0.05). Source control was assessed based on predefined clinical criteria as described in the Methods. Multivariable logistic regression including KPC infection status, Day 1 SOFA score, lactate level, antimicrobial therapy adequacy, and source control showed that KPC infection remained independently associated with increased mortality (aOR = 1.58, 95% CI: 1.02–2.44, P = 0.040) (Table 7A). Additional independent risk factors included Day 1 SOFA score (aOR = 1.08, 95% CI: 1.03–1.13, P < 0.001), blood lactate level (aOR = 1.12, 95% CI: 1.05–1.19, P < 0.001), and inadequate early antimicrobial therapy (aOR = 1.69, 95% CI: 1.06–2.70, P = 0.030). The model demonstrated good discrimination and calibration (AUC = 0.78; Hosmer-Lemeshow P = 0.33). Subgroup analysis in the KPC cohort (n = 210) showed that blood lactate level (aOR = 1.14, 95% CI: 1.05–1.25, P = 0.002), ΔSOFA (aOR = 1.11, 95% CI: 1.02–1.22, P = 0.020), inadequate early antimicrobial therapy (aOR = 1.82, 95% CI: 1.05–3.17, P = 0.030), and positive fluid balance > 1.5 L during Days 1–3 (aOR = 1.52, 95% CI: 1.00-2.32, P = 0.049) were independently associated with increased mortality (Table 7B). Serum albumin level was identified as a protective factor (aOR = 0.95, 95% CI: 0.91–0.98, P = 0.004). The model also showed good performance (AUC = 0.76; Hosmer-Lemeshow P = 0.41). Fig. 3Kaplan-Meier 28-day survival curves stratified by KPC enzyme production status. Time zero was defined as the onset of sepsis (ICU admission). The difference between the curves was assessed by the log-rank test Kaplan-Meier 28-day survival curves stratified by KPC enzyme production status. Time zero was defined as the onset of sepsis (ICU admission). The difference between the curves was assessed by the log-rank test Table 7Logistic regression analyses for 28-day all-cause mortalityA. Entire cohort (n = 320)VariablesUnivariable OR (95% CI)P valueMultivariable aOR (95% CI)P valueKPC infection (Yes vs. No)1.76 (1.05–2.95)0.031.58 (1.02–2.44)0.04Day-1 SOFA (per 1 point increase)1.13 (1.07–1.20)< 0.0011.08 (1.03–1.13)< 0.001Day-1 lactate (per 1 mmol/L increase)1.20 (1.11–1.30)< 0.0011.12 (1.05–1.19)< 0.001Inadequate early antimicrobial therapy2.02 (1.23–3.32)0.0061.69 (1.06–2.70)0.03Failure to achieve source control1.58 (1.01–2.48)0.0471.36 (0.90–2.07)0.14Note: Model performance (multivariable): AUC = 0.78; Hosmer-Lemeshow χ² = 6.12, P = 0.33B. KPC subgroup (n = 210)VariablesUnivariable OR (95% CI)P valueMultivariable aOR (95% CI)P valueLactate at Day 1 (per mmol/L)1.19 (1.09–1.31)< 0.0011.14 (1.05–1.25)0.002ΔSOFA (Day 3 − Day 1) (per point)1.16 (1.06–1.27)0.0011.11 (1.02–1.22)0.02Serum albumin (per g/L)0.93 (0.89–0.97)< 0.0010.95 (0.91–0.98)0.004Inappropriate early antimicrobial therapy2.06 (1.20–3.55)0.0091.82 (1.05–3.17)0.03Positive fluid balance > 1.5 L (Day 1–3)1.63 (0.95–2.79)0.081.52 (1.00-2.32)0.049Note: Model performance (multivariable): AUC = 0.76; Hosmer-Lemeshow P = 0.41Univariable and multivariable binary logistic regression were performed with 28-day all-cause mortality as the dependent variable. Multivariable models were pre-specified based on clinical relevance and EPV considerations; ΔSOFA (Day 3 − Day 1) was used to represent dynamic organ dysfunction to reduce collinearity with serial SOFA measurements. OR = odds ratio; aOR = adjusted odds ratio; CI = confidence interval; AUC = area under the ROC curve. All tests were two-sided; P < 0.05 was considered statistically significant Table 7Logistic regression analyses for 28-day all-cause mortalityA. Entire cohort (n = 320)VariablesUnivariable OR (95% CI)P valueMultivariable aOR (95% CI)P valueKPC infection (Yes vs. No)1.76 (1.05–2.95)0.031.58 (1.02–2.44)0.04Day-1 SOFA (per 1 point increase)1.13 (1.07–1.20)< 0.0011.08 (1.03–1.13)< 0.001Day-1 lactate (per 1 mmol/L increase)1.20 (1.11–1.30)< 0.0011.12 (1.05–1.19)< 0.001Inadequate early antimicrobial therapy2.02 (1.23–3.32)0.0061.69 (1.06–2.70)0.03Failure to achieve source control1.58 (1.01–2.48)0.0471.36 (0.90–2.07)0.14Note: Model performance (multivariable): AUC = 0.78; Hosmer-Lemeshow χ² = 6.12, P = 0.33B. KPC subgroup (n = 210)VariablesUnivariable OR (95% CI)P valueMultivariable aOR (95% CI)P valueLactate at Day 1 (per mmol/L)1.19 (1.09–1.31)< 0.0011.14 (1.05–1.25)0.002ΔSOFA (Day 3 − Day 1) (per point)1.16 (1.06–1.27)0.0011.11 (1.02–1.22)0.02Serum albumin (per g/L)0.93 (0.89–0.97)< 0.0010.95 (0.91–0.98)0.004Inappropriate early antimicrobial therapy2.06 (1.20–3.55)0.0091.82 (1.05–3.17)0.03Positive fluid balance > 1.5 L (Day 1–3)1.63 (0.95–2.79)0.081.52 (1.00-2.32)0.049Note: Model performance (multivariable): AUC = 0.76; Hosmer-Lemeshow P = 0.41Univariable and multivariable binary logistic regression were performed with 28-day all-cause mortality as the dependent variable. Multivariable models were pre-specified based on clinical relevance and EPV considerations; ΔSOFA (Day 3 − Day 1) was used to represent dynamic organ dysfunction to reduce collinearity with serial SOFA measurements. OR = odds ratio; aOR = adjusted odds ratio; CI = confidence interval; AUC = area under the ROC curve. All tests were two-sided; P < 0.05 was considered statistically significant Logistic regression analyses for 28-day all-cause mortality Note: Model performance (multivariable): AUC = 0.76; Hosmer-Lemeshow P = 0.41 Univariable and multivariable binary logistic regression were performed with 28-day all-cause mortality as the dependent variable. Multivariable models were pre-specified based on clinical relevance and EPV considerations; ΔSOFA (Day 3 − Day 1) was used to represent dynamic organ dysfunction to reduce collinearity with serial SOFA measurements. OR = odds ratio; aOR = adjusted odds ratio; CI = confidence interval; AUC = area under the ROC curve. All tests were two-sided; P < 0.05 was considered statistically significant Multivariable Cox proportional hazards analysis for 28-day all-cause mortality Multivariable Cox proportional hazards models were used to evaluate independent risk factors for 28-day all-cause mortality (Fig. 4; Table 8). In the overall cohort (n = 320), KPC-producing Enterobacterales infection was independently associated with increased mortality (aHR = 1.44, 95% CI: 1.01–2.06, P = 0.040) (Table 8A). Additional independent risk factors included Day 1 SOFA score (per 1-point increase, aHR = 1.07, 95% CI: 1.03–1.12, P < 0.001), Day 1 blood lactate level (per 1 mmol/L increase, aHR = 1.11, 95% CI: 1.05–1.18, P < 0.001), and inadequate early antimicrobial therapy (aHR = 1.62, 95% CI: 1.09–2.41, P = 0.020). Failure to achieve source control was not significantly associated with mortality in the multivariable model (P = 0.11). In the KPC subgroup (n = 210), blood lactate level (aHR = 1.13, 95% CI: 1.05–1.22, P = 0.001), ΔSOFA (aHR = 1.10, 95% CI: 1.03–1.18, P = 0.005), inadequate early antimicrobial therapy (aHR = 1.78, 95% CI: 1.12–2.84, P = 0.020), and positive fluid balance > 1.5 L during Days 1–3 (aHR = 1.46, 95% CI: 1.00-2.14, P = 0.049) were independently associated with increased mortality (Table 8B). Serum albumin level was inversely associated with mortality (per 1 g/L increase, aHR = 0.95, 95% CI: 0.91–0.98, P = 0.004). Fig. 4Adjusted Cox proportional hazards model for 28-day all-cause mortality. A forest plot derived from the multivariable Cox proportional hazards model, used to evaluate the independent associations between KPC-producing infection, related clinical factors, and the risk of 28-day all-cause mortality following sepsis onset (ICU admission). Hazard ratios (HR) and their 95% confidence intervals (CI) are displayed on a logarithmic scale. The vertical dashed line indicates the null value (HR = 1.0) Adjusted Cox proportional hazards model for 28-day all-cause mortality. A forest plot derived from the multivariable Cox proportional hazards model, used to evaluate the independent associations between KPC-producing infection, related clinical factors, and the risk of 28-day all-cause mortality following sepsis onset (ICU admission). Hazard ratios (HR) and their 95% confidence intervals (CI) are displayed on a logarithmic scale. The vertical dashed line indicates the null value (HR = 1.0) Table 8Multivariable Cox proportional hazards models for 28-day mortalityVariablesaHR95% CI P A Entire cohort ( n A = 320) KPC infection1.441.01–2.060.04Day 1 SOFA (per 1 increase)1.071.03–1.12< 0.001Day 1 lactate (per 1 mmol/L increase)1.111.05–1.18< 0.001Inadequate early antimicrobial therapy1.621.09–2.410.02Failure to achieve source control1.330.93–1.910.11 B. KPC subgroup ( n A = 210) Lactate at Day 1 (per mmol/L)1.131.05–1.220.001ΔSOFA (Day 3 − Day 1) (per point)1.101.03–1.180.005Serum albumin (per g/L)0.950.91–0.980.004Inappropriate early antimicrobial therapy1.781.12–2.840.02Positive fluid balance > 1.5 L (Day 1–3)1.461.00-2.140.049Note: aHR, adjusted hazard ratio; CI, confidence interval. Cox proportional hazards models were constructed using the time of sepsis onset (ICU admission) as the starting point for follow-up and 28-day all-cause mortality as the outcome variable. Multivariable models were pre-specified based on clinical relevance and considered the events-per-variable ratio Table 8Multivariable Cox proportional hazards models for 28-day mortalityVariablesaHR95% CI P A Entire cohort ( n A = 320) KPC infection1.441.01–2.060.04Day 1 SOFA (per 1 increase)1.071.03–1.12< 0.001Day 1 lactate (per 1 mmol/L increase)1.111.05–1.18< 0.001Inadequate early antimicrobial therapy1.621.09–2.410.02Failure to achieve source control1.330.93–1.910.11 B. KPC subgroup ( n A = 210) Lactate at Day 1 (per mmol/L)1.131.05–1.220.001ΔSOFA (Day 3 − Day 1) (per point)1.101.03–1.180.005Serum albumin (per g/L)0.950.91–0.980.004Inappropriate early antimicrobial therapy1.781.12–2.840.02Positive fluid balance > 1.5 L (Day 1–3)1.461.00-2.140.049Note: aHR, adjusted hazard ratio; CI, confidence interval. Cox proportional hazards models were constructed using the time of sepsis onset (ICU admission) as the starting point for follow-up and 28-day all-cause mortality as the outcome variable. Multivariable models were pre-specified based on clinical relevance and considered the events-per-variable ratio Multivariable Cox proportional hazards models for 28-day mortality Note: aHR, adjusted hazard ratio; CI, confidence interval. Cox proportional hazards models were constructed using the time of sepsis onset (ICU admission) as the starting point for follow-up and 28-day all-cause mortality as the outcome variable. Multivariable models were pre-specified based on clinical relevance and considered the events-per-variable ratio Discussion In this real-world retrospective ICU cohort, we comprehensively characterized the clinical features, temporal evolution of organ dysfunction, and short-term outcomes of patients with sepsis associated with KPC-producing Enterobacterales infection. Our findings demonstrated that patients in the KPC group exhibited more severe and persistent organ dysfunction, required more intensive organ support, and had a significantly elevated risk of 28-day mortality. Furthermore, multivariable analyses confirmed that KPC infection remained independently associated with mortality after adjustment for major confounders, suggesting its role as a high-risk clinical phenotype in sepsis. A key finding of this study was the role of dynamic organ dysfunction in determining prognosis. Patients in the KPC group exhibited persistently elevated SOFA scores and greater early deterioration, as reflected by ΔSOFA, highlighting the importance of temporal changes in organ function. This finding is consistent with the emerging concept that dynamic assessment may provide superior prognostic value compared with single time-point evaluation in critically ill patients. In addition, our findings highlighted the importance of early risk stratification using both conventional clinical indicators and emerging biomarkers. Blood lactate and SOFA score were consistently associated with mortality across models, indicating their utility as early prognostic markers. Recent studies have further suggested that novel sepsis biomarkers may improve early diagnostic accuracy and risk stratification, particularly in the context of multidrug-resistant infections, where early identification of high-risk cases is critical [19]. Integration of biomarker-based approaches with dynamic clinical assessment may further enhance early decision-making in this population. Relationship between KPC infection and poor prognosis in sepsis: comparison with previous studies Previous studies have suggested that carbapenemase-producing Enterobacterales, particularly KPC-producing strains, are associated with higher mortality [20]. However, most prior investigations have focused on bloodstream infections [21] or single species (e.g., Klebsiella pneumoniae) [22, 23], with limited consideration of sepsis as a complex and heterogeneous syndrome. The present study, encompassing a broader range of infection sites within a real-world ICU population, confirmed that KPC infection was associated with 28-day mortality. These findings support the concept that antimicrobial resistance mechanisms themselves may serve as clinically relevant prognostic indicators in sepsis [24]. Notably, rather than stratifying patients by bacterial species, this study classified cases according to resistance mechanisms, thereby minimizing potential confounding related to species-specific pathogenicity. This approach aligns with the biological characteristics of KPC, which can be horizontally transferred across Enterobacterales species via plasmids [25], and enhances the generalizability of the findings. Value of dynamic organ dysfunction in risk stratification A core feature of sepsis is the dynamic progression of organ dysfunction. This study not only assessed SOFA scores at admission but also systematically analyzed the trend of SOFA score changes during the first seven days following sepsis onset. Patients in the KPC group exhibited higher SOFA score levels at multiple time points and demonstrated significantly limited early improvement in SOFA scores. Notably, within the KPC subgroup, ΔSOFA was associated with 28-day mortality risk in the multivariable Cox analysis, suggesting that severity assessment at a single time point may not fully capture disease evolution. This finding aligns with the emerging emphasis in critical care medicine on the principle that “dynamic assessment outperforms static assessment,” underscoring the importance of continuous monitoring and early trend evaluation in patients with sepsis associated with resistant organisms. Adequacy of early antimicrobial therapy and mortality risk Another key finding is the impact of early antimicrobial therapy adequacy on clinical outcomes. In our study, patients with KPC-producing infections were less likely to receive appropriate antimicrobial therapy in the early phase, and inadequate therapy was independently associated with increased mortality in both the full cohort and the KPC subgroup. These results are consistent with prior studies on KPC/CPE infections [20], and are further supported by recent evidence demonstrating that delays in effective antimicrobial therapy are strongly associated with adverse outcomes, particularly in resistant infections [26]. Although novel agents such as ceftazidime-avibactam have expanded therapeutic options for KPC infections [27], challenges remain in real-world settings, including delayed initiation, suboptimal empirical coverage, and limited accessibility. Clinical implications of fluid management and supportive therapy Furthermore, this study identified fluid balance and host condition as additional prognostic factors in the KPC subgroup. Positive fluid balance was associated with increased mortality, while higher albumin levels were associated with improved outcomes. These findings may reflect the combined effects of capillary leakage, systemic inflammation, and impaired hemodynamic regulation, suggesting that individualized fluid management and supportive strategies are essential in this high-risk population. Collectively, our findings suggest a potential clinical pathway in KPC-related sepsis, in which antimicrobial resistance contributes to delayed effective therapy, leading to sustained organ dysfunction, fluid imbalance, and ultimately increased mortality. This integrated perspective may help explain the interaction between pathogen-related factors and host response in critically ill patients. Study strengths and limitations Several limitations should be acknowledged. First, as a single-center retrospective study, our findings may be subject to residual confounding and limited generalizability. Second, selection bias may have been introduced through exclusion criteria, particularly the exclusion of patients with missing key clinical data and those with very short ICU stays. Third, the non-KPC comparator group was heterogeneous, including both carbapenem-susceptible isolates and organisms with alternative resistance mechanisms, which may have influenced intergroup comparisons. Fourth, although the primary outcome was defined as 28-day mortality from sepsis onset, the retrospective nature of outcome assessment may still introduce variability Finally, the longitudinal analysis of SOFA scores was based on pairwise comparisons and ΔSOFA rather than formal repeated-measures modeling. Although missing SOFA data at later time points were minimal, the potential for bias in longitudinal analyses cannot be entirely excluded. KPC-producing Enterobacterales infection is associated with more severe organ dysfunction and increased short-term mortality in ICU patients with sepsis. Early identification of high-risk patients, prompt initiation of appropriate antimicrobial therapy, and optimization of supportive management strategies may be critical to improving outcomes in this population. Conclusion In this retrospective ICU cohort, KPC-producing Enterobacterales infection was significantly associated with greater organ dysfunction and increased short-term mortality in patients with sepsis. Dynamic changes in organ dysfunction, early antimicrobial therapy adequacy, and fluid balance were also associated with clinical outcomes. These findings highlight the potential importance of early risk stratification and timely optimization of antimicrobial and supportive management strategies in this high-risk population. However, given the observational design of the study, these results should be interpreted as associations rather than causal relationships. Further prospective, multi-center studies are warranted to validate these findings and to better define optimal management strategies for patients with KPC-related sepsis. Electronic Supplementary Material Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1