Development and validation of a nomogram for predicting sepsis in patients with pyogenic liver abscess

Pyogenic liver abscess (PLA) is a severe condition that significantly increases the risk of sepsis. However, there is a notable dearth of research regarding the prediction of sepsis in PLA patients. The objective of this study was to develop and validate a prognostic nomogram for predicting sepsis in PLA patients. A total of 206 PLA patients were enrolled in our study, out of which 60 individuals (29.1%) met the Sepsis-3 criteria. Independent risk factors for sepsis were identified through univariate and multivariate logistic regression analyses. Subsequently, a nomogram was developed based on age, positive blood culture, procalcitonin, alanine aminotransferase, blood urea nitrogen, and d-dimer. The nomogram demonstrated excellent calibration and discrimination, as evidenced by the area under the receiver operating characteristic curve (AUC) values of 0.946 (95% confidence interval [CI], 0.912–0.979) and 0.980 (95%CI 0.951–1.000) in the derivation and validation cohorts, respectively. Furthermore, decision-curve analysis confirmed the clinical utility of the nomogram. This study provides valuable insights for the prevention of sepsis in PLA patients and underscores the potential application of the prognostic nomogram in clinical practice.

www.nature.com/scientificreports/ platelet, prothrombin time, and d-dimer, as well as radiological findings such as septations within the abscess, rim enhancement, internal gas bubble, and pleural effusion were significantly associated with sepsis in the derivation cohort. Furthermore, patients with sepsis were more likely to suffer from septic shock, require intensive care unit admission, and experience infection-related deaths.

Development of a novel nomogram prediction model. Multivariate logistic regression analysis
revealed age, positive blood culture, procalcitonin, ALT, BUN, and d-dimer were independent risk predictors of sepsis in PLA patients within the derivation cohort (Table 2). A nomogram was then developed to integrate these predictors, and the risk score of sepsis was calculated by summing their respective scores using R software (Fig. 2). The calibration curve of the nomogram demonstrated a high degree of concordance between the predicted and observed sepsis outcomes (Fig. 3A). Furthermore, the Hosmer-Lemeshow test yielded a nonsignificant P value of 0.939, coupled with a brier score of 0.077, indicating excellent calibration accuracy. The nomogram also exhibited strong predictive capability for sepsis, as evidenced by an area under the receiver operating characteristic (ROC) curve (AUC) of 0.946 (95% confidence interval [CI], 0.912-0.979) (Fig. 3B).
Validation of the nomogram prediction model. In the validation cohort, we found that our nomogram model was well calibrated (Fig. 3C), with a brier score of 0.047 and a non-significant P value of 0.953 in the Hosmer-Lemeshow test. Additionally, the nomogram exhibited excellent discrimination for sepsis prediction, with an AUC of 0.980 (95%CI 0.951-1.000) (Fig. 3D).

Decision curve analysis.
To evaluate the clinical utility of our nomogram model, we conducted a decision curve analysis to calculate the net benefits across a range of threshold probabilities (Fig. 4). The results demonstrated that our model was a reliable tool for predicting sepsis in patients with PLA, as it achieved good clinical application across nearly the entire range of threshold probabilities.

Discussion
Patients diagnosed with PLA who subsequently develop sepsis are known to face an elevated risk of encountering clinical complications and unfavorable prognoses 2 . It is widely recognized that early and precise risk assessment, along with the administration of appropriate antibiotics and targeted treatment, are critical for successfully managing sepsis in these individuals 11 . Consequently, the identification of risk factors associated with sepsis in PLA patients assumes paramount importance, as it has the potential to significantly enhance clinical outcomes and reduce the incidence of sepsis.
The primary objective of our study was to develop and validate a nomogram model to predict the probability of sepsis in PLA patients during hospitalization. As total bilirubin, creatinine, and platelet levels are already included in the definition of Sepsis-3, we did not incorporate these parameters into our model. Through our analysis, we identified age, positive blood culture, procalcitonin, ALT, BUN, and d-dimer as the most optimal predictors of sepsis in PLA patients. Importantly, our nomogram model exhibited excellent calibration and discrimination capabilities in both derivation and external validation cohorts. Furthermore, the decision curve analysis provided additional insights into the clinical utility of our nomogram model. The analysis demonstrated its potential to guide disease management and treatment planning by generating favorable net benefits across various threshold probabilities. Overall, our findings suggest that our nomogram model could serve as a valuable  www.nature.com/scientificreports/ tool for clinicians to predict the probability of sepsis in PLA patients and guide early intervention to improve patient outcomes. Advanced age is a crucial factor that significantly affects the prognosis and clinical severity of numerous illnesses 12 . However, its influence on PLA remains largely unknown. Studies investigating the role of age in complications of PLA, such as sepsis and septic shock, in the geriatric population have reported controversial findings [12][13][14][15] . In our study, we found that age was an independent risk factor for sepsis in PLA patients. Elderly  Table 1. Comparison of clinical data between patients with and without sepsis in the derivation cohort. AST, aspartate aminotransferase; ALT, alanine aminotransferase; GGT, gamma-glutamyltransferase; BUN, blood urea nitrogen; NLR, Neutrophil-to-lymphocyte ratio; PLR, Platelet-to-lymphocyte ratio; APTT, activated partial thromboplastin time. www.nature.com/scientificreports/ patients with PLA may present with atypical and misleading symptoms and signs upon admission, which can result in the severity of the illness being underestimated 14 . Moreover, elderly patients with PLA often have multiple chronic comorbidities, which could lead to the progression of the disease and the occurrence of sepsis. Our findings suggest that age should be considered an important risk factor in the early identification and management of sepsis in PLA patients, particularly in the elderly population. Bacterial infection via the biliary tract or organs drained by portal veins is a well-established contributor to the development of PLA, which frequently results in sepsis and organ dysfunction. Currently, there are limited studies comparing clinical outcomes between culture-positive and culture-negative PLA patients. One study reported no difference in the incidence of sepsis between culture-negative and culture-positive liver abscess patients when blood and pus cultures were both considered 16 . In contrast, our study revealed an increased risk of sepsis in PLA patients with a positive blood culture alone. For the treatment of PLA, abscess drainage and administration of appropriate antibiotics are the primary therapeutic modalities 17,18 . The choice of antibiotic should be based on the suspected source of infection and the predominant pathogens in the region. It is worth noting that a positive blood culture in PLA patients indicates not only the presence of bacteria in the bloodstream but also an increased risk of sepsis. Therefore, empirical antibiotic therapy should target all potential pathogens, using potent antibiotics such as carbapenems, and may require an extended duration. Culture and susceptibility testing should guide antibiotic therapy, and patients should be closely monitored to ensure treatment efficacy. Prompt attention should be given to any signs of sepsis, warranting immediate intervention.
Procalcitonin is a widely used biomarker for sepsis, providing a real-time response that increases transiently during sepsis and is long enough to detect 19 . Compared to other biomarkers such as C-reactive protein, necrosis factor-alpha, and interleukins, procalcitonin has been shown to be more sensitive and specific in sepsis diagnosis 20,21 . In particular, procalcitonin has been identified as a marker for sepsis diagnosis in liver abscess caused by transcatheter arterial chemoembolization (TACE) therapy among patients with hepatocellular carcinoma 22 . Our study further demonstrated the strong predictive value of procalcitonin in patients with sepsis due to pyogenic liver abscess. Therefore, we recommend the routine procalcitonin examination as an early screening tool for sepsis in these patients to prompt timely and effective treatment.
d-dimer, a specific degradation product of stabilized fibrin, serves as a valuable biomarker for assessing the hypercoagulable state and secondary hyperfibrinolysis 23 . Within the context of sepsis, a complex interplay of pathophysiological processes occurs, involving the release of inflammatory mediators, endothelial cells, and cytokines. These elements collectively activate and amplify the coagulation cascade, leading to a substantial increase in d-dimer concentrations and other coagulation-related biomarkers. Previous studies have already established a correlation between d-dimer levels and the severity of sepsis, as well as their predictive value in determining in-hospital sepsis mortality 24,25 . In our comprehensive analysis of patients with PLA, we substantiates that the concentration of d-dimer upon admission holds potential as a reliable predictor for sepsis in PLA patients. By recognizing the role of d-dimer as a predictive marker, physicians can enhance their ability to promptly identify sepsis, thereby leading to improved patient outcomes. Further research is warranted in this area to elucidate the precise mechanisms linking d-dimer and sepsis, thereby facilitating the development of targeted interventions to combat this life-threatening condition.
Kidney and liver dysfunction represent significant pathological manifestations induced by infection. BUN, the primary end product of protein metabolism, is predominantly excreted by the kidneys in the human body. BUN levels serve as an indicator of protein catabolism and can also reflect renal impairment 26 . Patients with sepsis often experience a substantial increase in the rate of protein catabolism, which is frequently accompanied by acute renal injury 27 . As a result, BUN levels tend to rise in septic patients. Aminotransferases, such as ALT, are commonly used to assess hepatocellular damage. Elevated ALT levels in clinical practice are indicative of acute hepatocellular injury 24 . Our study demonstrates the predictive value of BUN and ALT in assessing sepsis among  www.nature.com/scientificreports/ patients with PLA. Given that sepsis-induced liver and kidney dysfunction can manifest subtly, it is crucial to consider the possibility of such dysfunction in any patient presenting with an infection, particularly when ALT and BUN levels are elevated. However, our study has certain limitations that should be considered. Firstly, its retrospective nature and reliance on electronic medical record systems may introduce potential biases. Furthermore, certain important parameters such as partial pressure of oxygen and serum lactate level were not included in our study due to their infrequent evaluation at admission in our hospital. Future studies should investigate the potential impact of these variables on the prediction of sepsis. Secondly, the predictive value of blood cultures may be constrained by the time required to confirm a positive result. Thirdly, while we conducted external validation using an additional database from another hospital, both databases were obtained from the same region, which may result in some regional deviations and limit generalizability. Lastly, although our study included a relatively large derivation cohort of 206 patients, and a total of 265 patients which exceeds the sample size of many previous studies, further validation using larger databases from prospective studies conducted in different regions or countries is warranted to strengthen the robustness of our findings.
In conclusion, we have developed a novel nomogram that incorporates age, positive blood culture, procalcitonin, ALT, BUN, and d-dimer to predict the probability of sepsis in patients with PLA. Our nomogram is easily accessible, cost-efficient and has demonstrated good calibration, discrimination, and clinical utility. The results of our study have important significance for clinical practice, as the nomogram may assist clinicians in predicting  Clinical data collection. Clinical data, including demographic characteristics (age and sex), vital signs (pulse rate and respiratory rate), clinical symptoms, comorbidities, treatment methods, laboratory findings, and clinical outcomes were collected from the electronic medical record systems of the hospitals. All patients with PLA underwent CECT before drainage of the liver abscess. CECT scans were reviewed for the purpose by two radiologists, and any disagreement was resolved by a third party. The features were collected as previously described [29][30][31] . The assessment of sepsis was conducted within the first 72 h of hospitalization and continued throughout the entire duration of the patient's hospital stay. Sepsis was defined as life-threatening organ dysfunction caused by infection with Sequential Organ Failure Assessment score (SOFA) ≥ 2, according to the Third International Consensus Definitions for sepsis and septic shock (Sepsis-3) updated in 2016 4 .

Statistical analysis.
The chi-squared test or Fisher's exact test was used to analyze categorical variables, and nonparametric tests were used to analyze continuous variables. Continuous variables were described by median (interquartile range [IQR]), and categorical variables were expressed as frequency rates and percentages.
Univariate analysis was performed to identify potential predictors with a significance level of P < 0.05, which were then included in multivariate logistic regression. Variables with P < 0.05 in the multivariable analysis were used to develop a nomogram in the derivation cohort. Calibration curves with 1000 bootstrap samples were plotted to demonstrate the consistency between the predicted probability and the actual probability. The ROC curve was used to evaluate the diagnostic efficacy of the nomogram. The discriminative ability of the nomograms was further measured by AUC. Additionally, the nomogram was validated in an external cohort, and its clinical utility was assessed by decision-curve analysis through calculating net benefits at different threshold probabilities. R version 4.2.0 (Math Soft, Cambridge, Massachusetts) was used for all statistical analyses, and P < 0.05 (twotailed) was considered statistically significant.