Impact of a nomogram-based antibiotic decision-making pathway on prophylactic antibiotic use in acute pancreatitis patients with systemic inflammatory response syndrome.

This study developed and validated a multivariable predictive model and constructed a nomogram based on APACHE II, NLR, PCT, CRP, and PaO₂ to estimate infection risk in patients with AP complicated by SIRS. The model demonstrated excellent discrimination, satisfactory calibration, and favorable clinical utility. By integrating the nomogram into the antibiotic decision-making pathway, clinicians can reduce unnecessary prophylactic antibiotic use, improve individualized risk assessment, and optimize treatment strategies. Future multicenter studies and the incorporation of emerging biomarkers are warranted to enhance model performance and broaden its clinical applicability.

Acute pancreatitis (AP) is an inflammatory disorder of the pancreas caused by multiple etiologies, and its incidence has been steadily increasing worldwide in recent years. Despite advances in diagnosis and treatment, AP continues to be a major contributor to severe disease progression and multiple organ dysfunction, particularly among critically ill patients. [ 1 ] Systemic inflammatory response syndrome (SIRS) is a common and serious complication of AP, characterized by excessive activation of the systemic inflammatory cascade, which exacerbates pancreatic injury and contributes to organ failure, secondary infection, and increased mortality. [ 2 ] Differentiating SIRS from early infection in AP patients remains a major clinical challenge because of their overlapping clinical and biochemical features. [ 3 ] Although inflammatory biomarkers such as C-reactive protein (CRP), procalcitonin (PCT), and the neutrophil-to-lymphocyte ratio (NLR) are commonly used for risk stratification, these parameters alone lack adequate sensitivity and specificity for distinguishing infection from sterile inflammation. [ 4 – 6 ] Furthermore, the absence of a standardized predictive tool that integrates multiple clinical and laboratory variables often leads to empirical prophylactic antibiotic use, which may promote antimicrobial resistance and negatively impact patient outcomes. [ 7 , 8 ] Accurate early prediction of infectious complications in AP patients with SIRS remains an unmet clinical need. Existing predictive models, often based on single indicators or limited variables, demonstrate poor generalizability and limited practical utility. [ 9 ] Therefore, developing a reliable and user-friendly tool that combines both clinical and laboratory data is essential to facilitate timely recognition of high-risk patients and guide rational antibiotic management. In this study, we retrospectively analyzed clinical data from patients with AP complicated by SIRS to develop a multivariable predictive model for infection risk. Based on this model, we constructed a nomogram designed to provide clinicians with an intuitive and individualized risk assessment tool. Incorporating this nomogram into the antibiotic decision-making pathway may help reduce unnecessary prophylactic antibiotic use and improve clinical outcomes in this population.

Conceptualization: Jing Guo, Wei Xiao, Jinghong Han, Qun Mo. Data curation: Jing Guo, Wei Xiao, Jinghong Han, Qun Mo. Formal analysis: Jing Guo, Jinghong Han, Qun Mo. Investigation: Qun Mo. Methodology: Qun Mo. Software: Qun Mo. Validation: Jing Guo, Qun Mo. Visualization: Jing Guo, Wei Xiao, Jinghong Han, Qun Mo. Writing – original draft: Jing Guo, Wei Xiao, Jinghong Han, Qun Mo. Writing – review & editing: Jing Guo, Wei Xiao, Jinghong Han, Qun Mo.

This study was approved by the Ethics Committee of Yanshan People’s Hospital (Approval No. YSRH2024-021). Written informed consent was obtained from all participants or their legal guardians before enrollment. All procedures were performed in accordance with the Declaration of Helsinki and applicable national ethical guidelines. This single-center retrospective cohort study aimed to develop a multivariable predictive model for assessing infection risk in patients with acute pancreatitis (AP) complicated by systemic inflammatory response syndrome (SIRS). Clinical, laboratory, and imaging data were collected from patients admitted to our hospital between January 2024 and April 2025. Based on these data, a predictive nomogram was subsequently constructed and internally validated. Patients were included if they met the following criteria: age ≥ 18 years; diagnosis of AP according to the revised Atlanta classification; diagnosis of SIRS confirmed on admission; availability of complete clinical and laboratory data within 24 hours of admission, including but not limited to white blood cell count (WBC), neutrophil count (NEU), lymphocyte count (LYM), NLR, CRP, PCT, arterial partial pressure of oxygen (PaO₂), and lactate dehydrogenase (LDH). Patients were excluded if they: had a history of severe immunodeficiency or malignant tumors; had comorbidities such as chronic liver disease, renal failure, or significant cardiovascular disease; lacked complete clinical or laboratory records; had postoperative or traumatic pancreatitis. Demographic and clinical data were extracted from electronic medical records, including: Demographics: age, sex, body mass index (BMI), and medical history. Disease-related parameters: AP clinical subtype (mild vs severe), Acute Physiology and Chronic Health Evaluation II (APACHE II) score, and Computed Tomography Severity Index. Inflammatory biomarkers: WBC, NEU, LYM, NLR, CRP, PCT, and LDH levels measured within 24 hours of admission. Oxygenation status: PaO₂. Infectious complications: including bloodstream infections, intra-abdominal infections, and pulmonary infections. Hospitalization outcomes: length of hospital stay, intensive care unit admission rate, and 28-day mortality. Antibiotic use: whether antibiotics were administered, antibiotic category, and duration of therapy. All statistical analyses were performed using SPSS version 27.0 (IBM Corp., Armonk) and R software version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria). A 2-sided P  < .05 was considered statistically significant. Continuous variables with normal distribution were expressed as mean ± standard deviation (SD) and compared using Welch t test. Non-normally distributed variables were summarized as median and interquartile range and compared using the Mann–Whitney U test. Categorical variables were expressed as counts and percentages [n (%)] and compared using Pearson chi-square test or Fisher exact test as appropriate. Univariate comparisons were conducted to evaluate the associations between clinical variables and infection status. Variables with P  < .05 in univariate analysis were considered for multivariable modeling. Variables that showed statistical significance in univariate analysis ( P  < .05) or were considered clinically relevant based on previous evidence were included in the multivariable logistic regression model using a stepwise selection method to identify independent predictors of infection. Adjusted odds ratios and 95% confidence intervals were calculated for each variable. Discrimination: receiver operating characteristic (ROC) curves were generated, and the area under the curve (AUC) was calculated to evaluate the model’s discriminatory power. Calibration: calibration plots were constructed to compare predicted and observed probabilities, and the Hosmer–Lemeshow goodness-of-fit test was applied. Clinical utility: decision curve analysis (DCA) was performed to assess the net clinical benefit of the predictive model compared with traditional strategies. A predictive nomogram was constructed based on the regression coefficients of the multivariable logistic regression model. Regression coefficients were transformed into point scores for each variable, and the total score corresponded to the predicted probability of infection. This visual tool was designed to facilitate early identification of high-risk patients and support clinical decision-making regarding antibiotic therapy.

A total of 200 patients with AP complicated by SIRS were enrolled, including 80 (40.0%) in the infection group and 120 (60.0%) in the non-infection group. Compared with the non-infection group, patients in the infection group had significantly higher disease severity scores (APACHE II), elevated inflammatory markers (WBC, NEU, NLR, CRP, and PCT), higher LDH and Computed Tomography Severity Index values, and lower LYM counts and PaO₂ levels ( P  < .001 for all). No significant differences were observed in age or sex distribution between the 2 groups. Detailed comparisons of baseline characteristics are summarized in Table 1 , and laboratory findings are illustrated in Figure 1 . Baseline characteristics of patients with and without infection. Notes : Values are mean ± SD or n (%). t -values are from Welch t tests; χ ² values from Pearson chi-square tests. APACHE II = Acute Physiology and Chronic Health Evaluation II, CRP = C-reactive protein, CTSI = CT severity index, LDH = lactate dehydrogenase, LYM = lymphocytes, NEU = neutrophils, NLR = neutrophil-to-lymphocyte ratio, PaO₂ = arterial oxygen partial pressure, PCT = procalcitonin, WBC = white blood cell. Laboratory markers by infection status. Variables with significant differences in univariate analysis and high clinical relevance were included in a multivariable logistic regression model, including APACHE II, NLR, PCT, CRP, and PaO₂. The results revealed that APACHE II score (per 1-point increase) was independently associated with a higher risk of infection ( P  = .047). Similarly, NLR (per 1-unit increase) was positively correlated with infection risk ( P  = .001), and elevated PCT and CRP levels were strongly associated with infection ( P  < .001 for both). In contrast, higher PaO₂ levels were independently associated with a lower risk of infection ( P  = .001). Adjusted odds ratios and 95% confidence intervals are presented in Table 2 , and the results are visualized in Figure 2 . APACHE II, NLR, PCT, and CRP were identified as significant risk factors, whereas PaO₂ served as a protective factor, consistent with known pathophysiological mechanisms. Multivariable logistic regression for predictors of infection. Notes : ORs are adjusted for APACHE II, NLR, PCT, CRP, and PaO₂. AP = acute pancreatitis, APACHE II = Acute Physiology and Chronic Health Evaluation II, AUC = area under the curve, CI = confidence interval, CRP = C-reactive protein, NLR = neutrophil-to-lymphocyte ratio, PaO₂ = arterial partial pressure of oxygen, PCT = procalcitonin. Multivariate logistic regression forest plot. A predictive nomogram for infection probability was developed based on the multivariable logistic regression model (Fig. 3 ). The upper scale of the nomogram represents the point value assigned to each variable, followed by individual predictor axes for APACHE II, NLR, PCT, CRP, and PaO₂. The total point score can be calculated by summing the points corresponding to each patient’s values. The predicted infection probability can then be read from the bottom axis of the nomogram. This tool provides an intuitive, individualized risk assessment and assists clinicians in making timely decisions regarding antibiotic therapy and monitoring strategies during the early stages of hospitalization. Nomogram for predicting infection probability. The ROC analysis demonstrated that the predictive nomogram had superior discriminatory power compared with individual biomarkers (PCT, CRP, and NLR) and the APACHE II score (Fig. 4 ). The AUC of the nomogram was 0.82, indicating excellent discrimination. Sensitivity and specificity were balanced near the optimal Youden index, suggesting that the model is suitable for early risk stratification and clinical decision-making regarding suspected infection. ROC curves: nomogram versus individual markers. ROC = receiver operating characteristic. The calibration curve showed strong agreement between the predicted probabilities and observed outcomes across risk strata (Fig. 5 ). The data points closely followed the ideal reference line, indicating excellent calibration of the model. Furthermore, the Hosmer–Lemeshow goodness-of-fit test was non-significant ( P  > .05), supporting the reliability and robustness of the model within the study population. Calibration curve. DCA demonstrated that the predictive nomogram provided greater net clinical benefit compared with the “treat-all” or “treat-none” strategies, as well as single biomarkers and the APACHE II score (Fig. 6 ). Within a clinically relevant threshold probability range of 0.10 to 0.70, the model supported 2 key decision-making strategies: At lower thresholds (0.10–0.30), the model emphasizes early vigilance and close monitoring, helping to avoid missing high-risk infections. Decision curve analysis (DCA). At moderate-to-high thresholds (0.30–0.70), the model assists in reducing unnecessary antibiotic use and intensive monitoring in low-risk patients, achieving a better balance between benefits and risks. Antibiotic utilization was significantly higher in the infection group than in the non-infection group (100.0% vs 45.8%, P  < .001), indicating substantial unnecessary prophylactic antibiotic use among non-infected patients. Clinical outcomes also differed significantly between the 2 groups. Patients with infection had higher intensive care unit admission rates (28.8% vs 7.5%, P  < .001), increased 28-day mortality (15.0% vs 3.3%, P  = .003), and prolonged hospital stays (18.10 ± 6.63 vs 11.83 ± 4.13 days, P  < .001). These findings are summarized in Table 3 and highlight the increased clinical burden associated with infection. Integrating the predictive nomogram into clinical workflows may help clinicians optimize antibiotic use, reduce unnecessary exposure, and identify high-risk patients requiring early targeted interventions. Clinical outcomes by infection status. Notes : Unnecessary antibiotic use is defined as antibiotic administration among non-infected patients within 7 days of admission. Values are mean ± SD or n (%). t values are from Welch t tests; χ ² values from Pearson chi-square tests. ICU = intensive care unit.

In this study, we developed a multivariable predictive model based on APACHE II, NLR, PCT, CRP, and PaO₂, and constructed a nomogram to estimate the risk of infection in patients with AP complicated by SIRS. The nomogram demonstrated excellent discriminatory power (AUC = 0.82), satisfactory calibration, and favorable clinical utility in DCA. By integrating multiple clinical and laboratory variables, the proposed tool provides clinicians with an accurate and practical approach for individualized infection risk assessment and early antibiotic decision-making. Predicting infectious complications in AP patients with SIRS has long been a clinical challenge due to overlapping manifestations of systemic inflammation and infection. [ 10 , 11 ] Previous studies have primarily relied on single inflammatory biomarkers such as CRP or PCT, or on scoring systems like APACHE II, but these indicators have shown limited sensitivity and specificity. [ 12 ] Our findings suggest that combining multiple predictors significantly improves diagnostic accuracy and better reflects the complex pathophysiological mechanisms underlying infection in this population. The results are consistent with earlier studies emphasizing the value of multivariable models in infection prediction. [ 13 – 15 ] A key advantage of our study lies in the integration of the nomogram into the antibiotic decision-making pathway. Unlike conventional strategies that depend on empirical prophylactic antibiotic administration, this approach enables personalized assessment of infection risk before definitive microbiological confirmation. As shown in our analysis, prophylactic antibiotic use among non-infected patients reached 45.8%, indicating a high rate of unnecessary exposure. Implementing this tool may reduce inappropriate antibiotic prescriptions, lower the risk of antimicrobial resistance, and optimize patient management. Moreover, the nomogram provides a user-friendly and intuitive graphical interface that allows clinicians to quantify individualized infection risk based on easily obtainable parameters. This is especially valuable during the early stages of hospitalization when culture results are unavailable and imaging findings are inconclusive. By supporting timely and evidence-based decision-making, the model facilitates early initiation of appropriate interventions for high-risk patients while avoiding overtreatment in low-risk individuals. Despite its strengths, this study has several limitations. First, the single-center retrospective design and relatively small sample size may introduce selection bias and limit generalizability. Future multicenter prospective studies with larger cohorts are needed to validate the robustness and external applicability of the model. Second, we focused on short-term outcomes during hospitalization and did not assess the predictive value of the model for long-term prognosis, including recurrence, chronic pancreatitis progression, or late complications. Third, emerging biomarkers and omics-based parameters, such as genomic, transcriptomic, and metabolomic data, were not incorporated in this model but may further enhance predictive accuracy in future studies. In summary, the proposed nomogram addresses a critical gap in the early identification of high-risk patients with AP complicated by SIRS and provides a practical tool for optimizing antibiotic decision-making strategies in clinical practice.