Nomogram construction and validation for predicting infection recurrence after necrotic tissue debridement in severe acute pancreatitis

Abstract

BACKGROUND

Postoperative infection recurrence remains a major challenge following necrotic tissue debridement in severe acute pancreatitis (SAP), yet reliable tools for early risk stratification are lacking. Although several clinical and laboratory parameters have been associated with infection outcomes, no simple, practical model exists to predict individualized recurrence risk. This study hypothesized that a nomogram incorporating preoperative clinical scores and inflammatory markers could effectively stratify patients by recurrence risk.

AIM

To construct and validate a nomogram for predicting infection recurrence after necrotic tissue debridement in patients with SAP.

METHODS

This study included 312 SAP patients who underwent necrotic tissue debridement at a single center (2023-2025), divided 7:3 into a modeling cohort (n = 218) and a validation cohort (n = 94). Infection recurrence within 30 days post-surgery was the primary outcome. Multivariate logistic regression identified independent risk factors, which were incorporated into a nomogram. Model performance was evaluated by receiver operating characteristic curves, calibration plots, and decision curve analysis.

RESULTS

Multivariate logistic regression identified Acute Physiology and Chronic Health Evaluation II score [odds ratio (OR) = 1.384, 95% confidence interval (CI): 1.147-1.670], computed tomography severity index score (OR = 4.231, 95%CI: 2.336-7.663), and C-reactive protein (OR = 1.051, 95%CI: 1.023-1.079) as independent risk factors for infection recurrence (all P < 0.05). The nomogram showed good discrimination in both the modeling cohort (area under the curve = 0.863, 95%CI: 0.798-0.927) and the validation cohort (area under the curve = 0.809, 95%CI: 0.751-0.867). Calibration was satisfactory (Hosmer-Lemeshow test: P = 0.849 and P = 0.715, respectively). Decision curve analysis confirmed clinical net benefit across risk thresholds of 0.10-0.75 and 0.10-0.85.

CONCLUSION

The nomogram incorporating Acute Physiology and Chronic Health Evaluation II, computed tomography severity index, and C-reactive protein demonstrates good predictive performance for postoperative infection recurrence in SAP patients and may serve as a practical tool for individualized risk stratification.

Key Words: Severe acute pancreatitis; Necrotic tissue debridement; Infection recurrence; Nomogram; Predictive model

Core Tip: This study developed and validated a nomogram to predict infection recurrence after necrotic tissue debridement in patients with severe acute pancreatitis. The model incorporates three independent risk factors: Acute Physiology and Chronic Health Evaluation II score, computed tomography severity index score, and C-reactive protein level. It demonstrated robust predictive performance and clinical utility, offering a practical tool for the individualized risk assessment of postoperative infection recurrence in severe acute pancreatitis patients.

INTRODUCTION

Acute pancreatitis is an inflammatory disorder of the pancreas triggered by premature activation of digestive enzymes within the organ, leading to self-digestion and local inflammation[1]. Its incidence continues to rise globally[2]. If not diagnosed and treated in time, about 20%-35% of patients with acute pancreatitis will advance to severe acute pancreatitis (SAP), which manifests as symptoms like necrosis of the pancreas, generalized systemic inflammation, and multiple organ failure[3,4]. The high mortality rate of SAP is mainly attributed to secondary infection of necrotic tissue in the pancreas and peripancreatic tissue, i.e., infectious pancreatic necrosis, which accounts for 30% of SAP mortality[5]. Necrotic tissue debridement is one of the means of treating infectious pancreatic necrosis. With the advancement of minimally invasive techniques, endoscopic necrotic tissue debridement and other procedures have significantly reduced the traumatic stress and complication rate of patients, but the problem of postoperative infection recurrence has not yet been solved[6]. Therefore, accurate identification of high-risk individuals for infection recurrence in the early postoperative period is crucial for implementing stratified management, optimizing anti-infection strategies and improving the long-term prognosis of patients.

Nomogram models are advanced tools that integrate multiple predictor variables to determine individual risk or probability by assigning relative weights based on regression coefficients[7]. With the simplification of complicated mathematical formulas through nomograms, clinicians are able to quickly and easily determine which patients face a high risk, thus predicting the outcome of such individual patients[8]. The objective of this study is to utilize nomograms to analyze past clinical data of SAP patients undergoing necrotic tissue debridement in order to identify independent risk factors for postoperative infection recurrence, develop a predictive nomogram model, and perform rigorous validation of the model. Ultimately, it aims to produce an advanced risk assessment tool for use in clinical settings.

MATERIALS AND METHODS

General information

A study gathered clinical data from 312 patients who underwent subcutaneous necrotic tissue debridement (SAP) at our hospital from January 2023 to December 2025 to form the training cohort, which was subdivided into a modeling group (218 cases) and a validation group (94 cases). The modeling group was categorized into relapse subgroup (55 cases) and non-relapse subgroup (163 cases); the validation group was classified into relapse subgroup (23 cases) and non-relapse subgroup (71 cases).

Inclusion criteria: (1) Complying with the diagnostic standards for SAP in the Chinese Guidelines for the Diagnosis and Treatment of Acute Pancreatitis (2021)[9]; (2) Undergoing pancreatic necrotic tissue debridement; (3) The preoperative infection was clinically controlled after treatment and the patient was eligible for observation of postoperative infection recurrence; (4) Having complete clinical data and surgical records; and (5) At least 3 months of postsurgical follow-up.

Exclusion criteria: (1) Diffuse abdominal infection or infectious complications requiring emergency surgery before surgery; (2) Immune deficiency disease or long-term use of immunosuppressants; (3) Malignant tumor; and (4) Death during surgery or within 48 hours after surgery.

Methods

Patient-related records are retrieved via the hospital’s digital medical record system, covering: (1) General information: Gender, age, body mass index, tobacco use history, substance use history, prior history of diabetes, and prior history of elevated blood pressure; (2) Clinical scores and blood laboratory indicators: Acute Physiology and Chronic Health Evaluation II (APACHE II) score, acute pancreatitis computed tomography severity index score (CTSI) score, electrolytes, C-reactive protein (CRP), procalcitonin (PCT), albumin (ALB), amylase, and lipase within 24 hours before surgery; (3) Procedure-related outcomes: Surgical approach for necrotic tissue debridement (categorized as endoscopic debridement, laparoscopic debridement, percutaneous step-up debridement, or open debridement, based on operative records), duration of procedure, blood loss during surgery; and (4) Prognostic indicators: Postoperative infection recurrence, defined as a recurrence of infection within 30 days after surgery, which is confirmed by etiology (positive peritoneal pus/blood culture) or imaging suggests a new infection focus (abdominal abscess, infectious effusion, etc.) after necrotic tissue debridement and standardized anti-infection treatment, with clinical symptoms, inflammatory markers, and imaging examinations indicating that the infection has been controlled. All imaging assessments were independently reviewed by two senior radiologists who were blinded to the patients' clinical outcomes; any disagreement was resolved by consensus discussion with a third radiologist. All computed tomography interpretations followed a standardized institutional protocol.

Statistical analysis

Data was subjected to computational analysis via SPSS 21.0 and R statistical software. The Shapiro-Wilk method was used to test for normality. Metric data with a normal probability distribution are denoted by mean ± SD, and pairwise comparisons adopt the parametric t-test. metric data with a non-normal probability density are reported as median (interquartile range), and pairwise comparisons adopt the non-parametric Mann-Whitney U test. Frequencies are denoted as percentages (%), and χ² tests were implemented for inter-sample comparisons. A prediction model was derived from the results of stepwise multiple regression analysis. Its clinician utility was appraised using receiver operating characteristic (ROC) curve analysis, calibrated ROC curves, the Hosmer-Lemeshow (H-L) goodness-of-fit test, and decision curve analysis. A P < 0.05 was deemed clinically meaningful.

RESULTS

Comparison of clinical data between the two groups of patients

Between the two groups, clinically meaningful differences were found in age, history of diabetes, APACHE II score, CTSI score, CRP, PCT, and ALB levels (P < 0.05); in contrast, there were no clinically meaningful differences across groups in the following parameters: Gender, body mass index, history of smoking, history of alcohol consumption, history of high blood pressure, amylase, lipase levels, and surgery-related indicators (P > 0.05), as shown in Table 1.

Table 1 Comparison of baseline data between the two groups, n (%)/(mean ± SD).

	Relapse group (n = 55)	The non-relapse group (n = 163)	t/χ2	P value
Age (years)	62.15 ± 6.23	59.32 ± 7.15	2.969	0.003
Gender			0.247	0.619
Male	32 (58.18)	101 (61.96)
Female	23 (41.82)	62 (38.04)
BMI (kg/m2)	24.82 ± 3.17	24.95 ± 2.86	0.284	0.777
Smoking history			0.667	0.414
Have	24 (43.64)	61 (37.42)
None	31 (56.36)	102 (62.58)
Drinking history			0.210	0.647
Have	18 (32.73)	48 (29.45)
None	37 (67.27)	115 (70.55)
History of diabetes			7.690	0.006
Have	21 (38.18)	32 (19.63)
None	34 (61.82)	131 (80.37)
History of hypertension			0.685	0.408
Have	23 (41.82)	58 (35.58)
None	32 (58.18)	105 (64.42)
APACHE II score (scores)	15.98 ± 1.46	10.91 ± 1.85	18.466	< 0.001
CTSI score (scores)	8.27 ± 1.18	5.06 ± 0.84	22.034	< 0.001
CRP (mg/L)	116.53 ± 19.03	74.21 ± 13.82	17.751	< 0.001
PCT (μg /L)	3.41 ± 0.81	2.59 ± 0.61	7.900	< 0.001
ALB (g/L)	30.15 ± 3.19	36.13 ± 2.47	14.386	< 0.001
Amylase (U/L)	728.06 ± 106.64	728.38 ± 112.36	0.018	0.985
Lipase (U/L)	128.32 ± 20.21	126.88 ± 20.15	0.458	0.647
Operation duration (minutes)	105.80 ± 12.68	103.48 ± 12.62	1.181	0.239
Intraoperative blood loss (mL)	131.67 ± 14.82	135.63 ± 20.27	1.332	0.184

BMI: Body mass index; APACHE: Acute Physiology and Chronic Health Evaluation; CTSI: Computed tomography severity index; CRP: C-reactive protein; PCT: Procalcitonin; ALB: Albumin.

Multivariate logistic regression analysis

Using postoperative infection recurrence as the dependent variable and age, history of diabetes, Apache II score, CTSI score, CRP, PCT, and ALB as explanatory covariates, a multivariate logistic regression analysis was performed. The coding of individual factors is shown in Table 2, and the results are shown in Table 3. The results showed that Apache II score, CTSI score, and CRP were clinically meaningful independent predictors of postoperative infection relapse (P < 0.05), see Table 3.

Table 2 Values assigned to each factor.

Variable	Assignment
Age	Value
History of diabetes	1 = occurred; 0 = did not occur
APACHE II rating	Value
CTSI score	Value
CRP	Value
PCT	Value
ALB	Value
Grouping	1 = occurred; 0 = did not occur

APACHE: Acute Physiology and Chronic Health Evaluation; CTSI: Computed tomography severity index; CRP: C-reactive protein; PCT: Procalcitonin; ALB: Albumin.

Table 3 Logistic regression analysis of factors influencing postoperative infection recurrence.

Variable	B	SE	Wald	P value	Exp (B)	95% confidence interval of Exp (B)
						Lower limit	Upper limit
APACHE II rating	0.325	0.096	11.512	0.001	1.384	1.147	1.670
CTSI score	1.442	0.303	22.648	< 0.001	4.231	2.336	7.663
CRP	0.049	0.014	12.969	< 0.001	1.051	1.023	1.079
Constant	-16.847	3.077	29.973	< 0.001	< 0.001

APACHE: Acute Physiology and Chronic Health Evaluation; CTSI: Computed tomography severity index; CRP: C-reactive protein.

Construction of a predictive model for postoperative infection recurrence in SAP patients

To incorporate the results of the logistic regression analysis, APACHE II score, CTSI score, and CRP were incorporated into the survivorship predictive model to formulate a nomogram model (Figure 1) for visualizing the reinfection rate across patients.

Figure 1 Nodal plot of the predictive model for postoperative infection recurrence risk. APACHE: Acute Physiology and Chronic Health Evaluation; CTSI: Computed tomography severity index; CRP: C-reactive protein.

Performance analysis of predictive models for postoperative infection recurrence in SAP patients

The nomogram prediction model for surgical site infection relapse in SAP patients was validated. The regression curve showed that the H-L in the modeling group was χ² = 2.743, P = 0.849; and in the validation group, H-L was χ² = 3.403, P = 0.715. The model’s prediction of postoperative infection recurrence in SAP patients in both the modeling and validation groups was largely consistent with the actual situation (Figure 2). ROC curves showed that the area under the curve of the model was 0.863 [95% confidence interval (CI): 0.798-0.927] in the modeling group and 0.809 (95%CI: 0.751-0.867) in the validation group (Figure 3). The decision analysis curves show that when the high-risk threshold for the modeling group is between 0.10 and 0.75, and the high-risk threshold for the experimental group is between 0.10 and 0.85, the model has good clinical net benefits (Figure 4).

Figure 2 Calibration curve of postoperative infection recurrence in severe acute pancreatitis patients.

Figure 3 Receiver operating characteristic curve of postoperative infection recurrence in severe acute pancreatitis patients. AUC: Area under the curve; CI: Confidence interval.

Figure 4 Decision curve analysis curve of postoperative infection recurrence in severe acute pancreatitis patients. APACHE: Acute Physiology and Chronic Health Evaluation; CTSI: Computed tomography severity index; CRP: C-reactive protein.

DISCUSSION

Acute pancreatitis is a complex inflammatory disorder triggered by the premature and aberrant activation of endogenous pancreatic enzymes. This autodigestive process damages the pancreatic parenchyma and surrounding peripancreatic tissues, resulting in edema, tissue necrosis, and severe local inflammation[10]. In the absence of appropriate treatment, such an inflammatory response locally may develop into a chronic, systemic pro-inflammatory process that often ends up in critical illness syndrome[11]. From an epidemiological perspective, the worldwide burden of acute pancreatitis is constantly growing; the standardized rate of incidence increased by 4.7%, rising from 17.3 per 100000 persons in 1990 to 18.1 per 100000 in 2019[12]. Although most of the cases take a mild and uncomplicated course and resolve in one week, roughly 20% of people experience progression towards SAP. Such a form of the disease has poor prognosis due to considerable morbidity and high case-fatality rate in the range between 20% and 30%[13,14]. Pathogenetically, the early stage of pancreatic necrosis as well as peripancreatic fluid collections usually develops without involvement of pathogenic microorganisms. However, secondary infections play an important role and about 40% of patients progress towards development of septic pancreatic necrosis[15].

For moderately severe or SAP patients, infected pancreatic necrosis has been a relatively common yet serious local complication. This entails infection that arises from early acute necrotic collections and also infection that occurs in later walled-off necrosis. After developing, infected pancreatic necrosis frequently results in a variety of systemic complications, such as sepsis and multiple organ dysfunction syndrome, which have remained the leading causes of mortality in such patients[16]. In the past, it had been proposed that the development of infected pancreatic necrosis was due to translocation of bacteria from the intestine as a result of its impaired integrity, and this hypothesis is corroborated by the fact that the microbes isolated from pancreatic necrotic fluids share features with those of the intestinal flora[15]. The process of debridement of necrotic tissue is considered one of the main treatment approaches for septic necrotizing pancreatitis. The data provided by various research studies demonstrates that compared to open surgery procedures, minimally invasive endoscopy and staged debridement have several important advantages, including a milder reaction of inflammation after the procedure, faster recovery rates, shorter hospital stays, reduction of indirect costs, and reduced cases of developing pancreatic fistulas[17]. However, despite the development of such medical advances, septic necrotizing pancreatitis remains associated with considerable short-term morbidity rates[18]. Therefore, the timely identification of individuals who are at an increased risk of post-surgery infection is critical. Prompt treatment of patients with identified risks has an extremely significant clinical value due to its implications on their survival rate.

The nomogram is an important method of visual risk prediction based on which complex multivariate regression models are transformed into scoring systems, which can help make accurate predictions regarding individual risk probability in certain patients[19]. During the past decades, the utilization of nomograms has become increasingly common not only among clinicians but also among researchers working within such disciplines as oncology, cardiovascular medicine, and critical care[20-22]. Such popularity can be explained by the simplicity of the technique used, its visualization, and effectiveness in making decisions. In order to estimate the risk associated with postoperative infection recurrence in patients suffering from SAP, the present study built a special nomogram considering three independent risk factors, namely, the APACHE II score, CTSI, and CRP concentration. The main goal of this nomogram lies in creating a forecasting method that will allow predicting postoperative infection recurrence with adequate validity and reliability based on multidimensional logistic analysis results.

The present study revealed that APACHE II scores, CTSI scores, and CRP levels were significantly higher in the relapse group than in the non-relapse group (P < 0.05). Multivariate logistic regression further confirmed these three factors as independent risk factors for postoperative infection recurrence. The APACHE II score is a classic indicator for evaluating critical illness severity, where higher scores reflect aggravated systemic inflammation and impaired organ reserve[23]. This metric reliably identifies pancreatic necrosis with favorable diagnostic efficiency[24]. Elevated APACHE II scores mirror robust compensatory anti-inflammatory response and immune paralysis, which weakens the clearance of translocated gut pathogens and elevates reinfection risk via disrupted intestinal barrier integrity[25,26]. CTSI is one of the two important radiological methods for assessing the severity of acute neuromotor disease. The CTSI score classifies the severity of pancreatic inflammation and necrosis into mild (0-3), moderate (4-6), or severe (7-10 points)[27]. In this study, for every 1 point increase in CTSI score, the risk of infection recurrence increased by 4.23 times (OR = 4.231, 95%CI: 2.336-7.663), indicating the role of local necrosis burden in infection recurrence. The study pointed out that CTSI score is a key imaging tool for assessing the extent of chronic pancreatitis and predicting the risk of secondary infectious necrosis of the pancreas; given that high scores (indicating extensive pancreatic necrosis) are markedly correlated with elevated infection rates and poor prognosis, it is recommended in clinical practice to implement more rigorous monitoring and individualized drainage management strategies for such high-risk patients[28]. An elevated CTSI implies necrosis of a larger amount of pancreatic tissue and extension of the area of peripancreatic effusion. This necrotic pancreatic tissue not only provides a breeding ground for bacteria to reproduce in large numbers, but also creates a barrier through which penetration by antibiotics becomes difficult[29]. While surgical interventions can help eliminate most of the necrotic tissue, it becomes extremely difficult to ensure there will be no remaining tiny bits of necrotic tissue that may cause chronic infections[30]. Also, the extensive necrosis is frequently associated with ductal damage and pancreatic fistulas, thereby enhancing the probability of infection[31]. CRP is an acute phase reactant that is chiefly produced in the liver. CRP is part of the adaptive immune response[32]. In the present study, with every 1 mg/L increment in the level of CRP, there was a 5.1% increase in the likelihood of recurrence of infections (OR = 1.051, 95%CI: 1.023-1.079), indicating that a continuous rise in the CRP level within the early postoperative period might indicate poor control over infections or infection sources. According to studies, CRP has a high prediction rate of SAP infections three days from the time of admission[33]. The CRP level is positively correlated with inflammation, meaning that it is a direct indicator of the body's inflammatory reaction to infections, while, on the other hand, CRP can contribute to immune anti-infective resistance through activation of the complement system and opsonization processes[34]. Beyond serving as a conventional marker, CRP elevation in SAP should be interpreted within the pathophysiological cascade of persistent inflammation and gut-derived bacterial translocation. Sustained systemic inflammation in SAP activates regulatory T-cells, which impair the duodenal mucosal immune barrier and allow facultative pathogens such as Escherichia/Shigella and Enterococcus to translocate across the intestinal wall into the pancreatic necrotic bed[35]. Thus, persistent postoperative CRP elevation may signal not merely ongoing inflammation but failure of intestinal barrier restoration and continued bacterial translocation - a mechanistic link that explains why CRP predicts infection recurrence. Future studies correlating CRP with gut permeability markers may further validate this hypothesis.

Although PCT showed significant differences between the relapse and non-relapse groups in univariate analysis, it did not emerge as an independent risk factor in the multivariate logistic regression model. This may be attributed to several factors. First, while PCT is a more specific marker of bacterial infection, CRP reflects a broader spectrum of inflammatory responses, including sterile inflammation triggered by pancreatic necrosis itself[33,34]. In the context of postoperative infection recurrence - which may arise from both persistent sterile inflammation and superimposed bacterial infection - the broader inflammatory coverage of CRP may have conferred a statistical advantage in the multivariate analysis. Additionally, the relatively small sample size of the recurrence group (n = 55) limited the number of variables that could be reliably included in the multivariate model, potentially reducing the power to detect PCT as an independent predictor. Future studies with larger cohorts and serial PCT measurements are warranted to further elucidate its role in predicting postoperative infection recurrence.

For the model developed based on the three above-mentioned risks, it showed excellent predictive performance. As it can be seen from the ROC graph, in the modeling group, the area under the curve of the model was 0.863 (95%CI: 0.798-0.927). For the validation group, the corresponding value amounted to 0.809 (95%CI: 0.751-0.867). Thus, the developed model demonstrates sufficient discriminative power for differentiating those patients who will develop postoperative infection recurrence and those patients who will not. Training and validation set calibration plot graphs show that there is an excellent agreement between the predicted and actual outcomes. This hypothesis is confirmed by the H-L test which shows that P > 0.05 for both groups. Besides, we have evaluated the clinical utility of our nomogram through decisional curves analysis. It should be noted that at the risk thresholds ranging from 0.10 to 0.75 in the modeling group and from 0.10 to 0.85 in the validation group, the developed nomogram model offered more net benefit than “full intervention” and “no intervention” strategies. From this observation, it can be said that the nomogram model has both excellent predictive accuracy and is useful from a clinical perspective as well because it helps to identify those patients who are at high risk. Clinicians may implement tiered management guided by nomogram-derived risk scores. Patients at high predicted recurrence risk require intensified serial inflammatory testing, shortened follow-up computed tomography scanning and prompt broad-spectrum antibiotic escalation upon inflammatory surges; low-risk individuals may receive routine surveillance with streamlined anti-infective regimens to prevent overtreatment.

This study has several limitations that warrant careful consideration. First, the single-center prospective design may introduce selection and information biases, as enrollment criteria and reliance on medical records could affect data completeness and accuracy. Therefore, the applicability of our findings to other populations with various demographic backgrounds cannot be guaranteed. Another limitation concerns the sample size; namely, it should be noted that it was rather small. Such an imbalance may result in unstable effect sizes for some variables and reduced statistical power for detecting weak but clinically important relationships. Thirdly, the variables included in the present study do not fully account for the presence of possible confounders, such as particular types of pathogens, antibiotics, and postoperative care techniques. Notably, CRP was measured at a single preoperative timepoint in this study. Using a single baseline CRP value may not fully capture the inflammatory trajectory of patients or the impact of postoperative recovery on infection risk. Serial postoperative CRP measurements or dynamic trend analysis could potentially provide more accurate risk stratification and better reflect the resolution or persistence of inflammation after necrotic tissue debridement. However, due to the retrospective nature of this study and the lack of standardized postoperative CRP monitoring protocols, we were unable to incorporate dynamic CRP trends into our predictive model. Future prospective studies with predefined serial CRP monitoring schedules are warranted to further validate whether dynamic CRP patterns enhance the predictive performance of the nomogram. Additionally, heterogeneity in the specific surgical approach for necrotic tissue debridement - including endoscopic, laparoscopic, percutaneous step-up, or open debridement - was not separately analyzed in our model. Given that different procedures may carry varying risks of postoperative infection recurrence due to differences in trauma extent, drainage adequacy, and cavity clearance, the lack of stratification by surgical approach may have introduced unmeasured confounding that could affect the observed outcomes. Future studies with larger sample sizes should further investigate whether the predictive performance of the nomogram differs across surgical approaches. Notably, the time interval from symptom onset to surgical intervention-a well-established prognostic factor in infected necrotizing pancreatitis-was not included as a variable in our model, as the “delayed intervention” strategy has been shown to significantly improve outcomes compared with early debridement[9,10]. The lack of standardized timing of necrotic tissue debridement across patients may have introduced confounding that could affect the observed recurrence rates. Furthermore, etiological characteristics, including the distribution of Gram-negative bacteria, fungal infections, and the proportion of multidrug-resistant organisms, were not systematically collected or analyzed in this study. These microbiological factors are closely linked to the risk of postoperative infection recurrence and may have contributed to the observed outcomes independently of the inflammatory markers included in our model[5,31]. In addition, the study period spanned from 2023 to 2025, during which clinical practice may have evolved, including changes in antibiotic regimens, refinements in endoscopic necrosectomy techniques, and updates in intensive care unit support strategies for SAP patients. Such temporal variations in treatment protocols could have influenced patient outcomes and introduced unmeasured confounding, potentially affecting the stability and generalizability of our predictive model[28,29]. Future research studies need to focus on including additional variable dimensions to increase the model's validity. Finally, although the model performed very well in internal validation, no proper independent validation was carried out. Therefore, further testing across different multipurpose data sets is necessary to ensure its reliability, stability, and universality.

CONCLUSION

In conclusion, APACHE II score, CTSI score, and CRP are independent prognosticators of recurrence of surgical site infection in SAP patients undergoing necrotic tissue removal. The prediction model developed based on these three independent factors demonstrates excellent discrimination, calibration, and clinical net benefit, and may assist in quantifying the probability of postoperative infection recurrence and guiding personalized treatment strategies.