This study was a retrospective single-center cohort study utilizing data from a prospectively maintained AP inpatient database at the First Affiliated Hospital of Nanchang University. The database was established by trained clinicians through structured electronic medical record (EMR) extraction and independently verified by two researchers to ensure data accuracy and completeness. It encompasses comprehensive information, including demographic characteristics, laboratory findings, radiological reports, AP scoring systems (e.g., APACHE II, BISAP), clinical features, and discharge outcomes. All methods were performed in accordance with the relevant guidelines and regulations, and in compliance with the Declaration of Helsinki. The study was approved by the Medical Ethics Research Committee of the First Affiliated Hospital of Nanchang University.

A total of 16,550 patients discharged with a diagnosis of acute pancreatitis (AP) from the First Affiliated Hospital of Nanchang University between January 2005 and December 2024 were initially identified. Hospital admission data for patients with AP were retrieved using the International Classification of Diseases, 10th Revision (ICD-10) codes K85-K85.92. AP diagnosis and severity classification followed the 2012 revised Atlanta criteria (18). Exclusion criteria included: (1): non-index admission for AP (prior AP hospitalization), (2), suspected AP cases, (3) age <18 years, (4) pregnancy, (5) undetermined urban–rural residency, and (6) missing critical laboratory parameters. After screening, 12,214 eligible AP patients were enrolled and stratified into urban (n = 5,002) and rural (n = 7,212) groups based on residential status. Figure 1 depicts the participant selection procedure.

Patients were stratified into urban or rural groups based on residential duration and geographical criteria: urban residency was defined as permanent residence (≥6 months/year) in prefecture-level cities or municipal districts, while rural residency encompassed townships or administrative villages. Residential data were extracted from hospitalization records, which mandated patients to report detailed addresses and contact information. For migrant populations, classification prioritized actual residential address over household registration (e.g., rural-registered individuals residing long-term in urban areas were categorized as urban). Additionally, health insurance type (urban employee/resident insurance vs. new rural cooperative medical scheme) and occupation (agricultural vs. non-agricultural) were used as secondary reference criteria. Cases with indeterminate residency (e.g., temporary migrant workers lacking fixed addresses) were excluded to minimize misclassification bias. This approach ensured alignment between residential exposure and epidemiological risk profiles.

Data collection encompassed five domains: (1) Demographics—age, gender, residential status, smoking, and alcohol consumption; (2) Baseline disease characteristics—subclassified into etiology (biliary, HTG, alcoholic, idiopathic, drug-induced, post-surgical, traumatic, post-ERCP, autoimmune, parapapillary diverticulum of duodenum, or other), severity (mild, moderately severe, or severe AP per revised Atlanta criteria), comorbidities (hypertension, diabetes mellitus, hyperlipidemia, chronic obstructive pulmonary disease [COPD], coronary heart disease, cirrhosis, cerebral infarction, chronic renal failure), and admission clinical status (fever on admission, APACHE II score within 24 h); (3) Treatment process—including healthcare access (prior hospital transfer) and time metrics (symptom-to-admission interval); (4) Laboratory parameters—white blood cell count (×10⁹/L) and serum amylase (U/L); (5) Clinical outcomes—stratified into short-term outcomes (in-hospital mortality, total hospital stay, ICU stay, hospitalization costs) and complications (local: infected pancreatic necrosis, abdominal compartment syndrome, enteric fistula; systemic: acute respiratory failure, acute kidney injury, circulatory failure, septic shock, heart failure, hepatic dysfunction; thrombotic: portosplenomesenteric venous thrombosis, deep vein thrombosis). This classification aligns with acute pancreatitis research frameworks, ensuring logical separation of intrinsic disease attributes from care-related factors.

All statistical analyses were performed using R software (version 4.4.1). Missing data were handled using multiple imputation with chained equations (MICE), generating 20 imputed datasets based on all analysis variables. Continuous variables were evaluated for normality using Shapiro–Wilk tests and presented as mean ± standard deviation (SD) for normally distributed data or median (interquartile range, IQR) for non-normally distributed data. Categorical variables were expressed as frequencies (percentages). Group comparisons were conducted using independent t-tests (for normally distributed continuous variables), Mann–Whitney U tests (for non-normally distributed continuous variables), or chi-square/Fisher’s exact tests (for categorical variables), as appropriate.

Continuous variables that violated linearity assumptions were categorized into clinically meaningful groups: hospitalization costs (≤11,250, 11,251-26,270, >26,270 Yuan), APACHE II scores (<8, 8–12, ≥12), and white blood cell counts (≤10, 10.1–20, >20 × 10⁹/L). Multivariable logistic regression models were developed to identify risk factors, incorporating variables with p < 0.1 in univariate analysis or clinical significance (Supplementary Table 1). Multicollinearity was assessed using variance inflation factors (VIF < 5 for all retained variables). Model performance was evaluated by calculating area under the receiver operating characteristic curve (AUC). Temporal trends were analyzed using 5-year intervals (2005–2009 to 2020–2024). Results were presented as adjusted odds ratios with 95% confidence intervals and visualized in forest plots. All tests were two-tailed with p < 0.05 considered statistically significant.

This study enrolled 12,214 acute pancreatitis patients, comprising 5,002 urban cases (40.9%) and 7,212 rural cases (59.1%). Comparative analysis revealed urban group had significantly higher proportions of males, younger age, greater alcohol, and higher prevalence of hypertension, diabetes, and hyperlipidemia (all p < 0.05). Rural patients demonstrated significantly higher rates of prior hospital transfer, longer symptom-to-admission intervals, elevated APACHE II scores, and greater severe acute pancreatitis incidence (all p < 0.01). While rural patients incurred higher hospitalization costs, no significant differences existed in length of stay or mortality. Temporal trends showed rural case proportions increased markedly post-2015 (rising from 8.6% in 2005–2009 to 36.5% in 2015–2019), whereas urban cases peaked during 2020–2024. Baseline clinical characteristics of the two groups are detailed in Table 1.

As shown in Table 2, acute respiratory failure and acute kidney injury were the most common systemic complications overall, with no significant urban–rural differences (both p > 0.05). Notably, rural patients exhibited significantly higher rates of circulatory failure (3.7% vs. 2.9%, p = 0.022), abdominal compartment syndrome (1.7% vs. 1.1%, p = 0.011), and infected pancreatic necrosis (5.3% vs. 4.3%, p = 0.014). In contrast, other complications including septic shock, heart failure, and thrombotic events showed no statistically significant intergroup differences.

As shown in Table 3, biliary, HTG, and alcoholic etiologies constituted the top three causes of AP in both urban and rural populations. Biliary AP was the most prevalent (54.5%), with significantly higher incidence in rural patients (56.4% vs. 51.7%, p < 0.001). HTG-AP accounted for 28.1% of cases, demonstrating a higher proportion in urban areas (30.6% vs. 26.3%, p < 0.001). No significant urban–rural difference was observed in alcoholic AP, while idiopathic AP was more common in urban patients (8.2% vs. 6.9%, p = 0.010). Notably, statistically significant disparities were identified in drug-induced AP and post-ERCP AP between the two groups. Other etiologies (e.g., traumatic, autoimmune, and periampullary diverticulum-associated AP showed no significant intergroup differences).

The study period witnessed significant epidemiological shifts in AP etiology (Figure 2). Biliary AP demonstrated a consistent downward trend (51.4% → 41.3%), with rural areas maintaining higher prevalence than urban areas (2024, 43.2% vs. 39.0%) (Figure 3). In contrast, HTG-AP showed remarkable growth (21.6% → 42.4%), surpassing biliary AP as the leading etiology in 2023 (43.0%). The urban group consistently demonstrated higher proportions than the rural group in most years, with both groups exhibiting a steep surge in growth rates since 2020 (Figure 4). Alcoholic AP displayed fluctuating proportions (3.6–11.6%) with diminishing urban–rural disparities (Figure 5). Importantly, the urban–rural gaps for HTG-AP and biliary AP narrowed substantially from 7.9 to 5.3% and 9.9 to 4.2%, respectively, suggesting converging epidemiological patterns.

Multivariable logistic regression analysis identified significant risk factors for moderate-to-severe AP (Table 4). As shown in the fully adjusted model (Model 3), rural group was significantly associated with increased disease severity (adjusted OR = 1.13, 95% CI: 1.04–1.24, p = 0.005) compared with urban residence. Hypertriglyceridemia-induced AP (HTG-AP) demonstrated the strongest association (aOR = 2.06, 95% CI: 1.79–2.38, p < 0.001), followed by alcoholic etiology. Clinical predictors including prior hospital transfer, delayed admission, elevated white blood cell count, and higher APACHE II scores were all significantly associated with increased disease severity (all p < 0.001). Notably, patients with APACHE II scores ≥12 had 2.76-fold higher risk of severe outcomes (p < 0.001). Higher hospitalization costs were strongly correlated with disease severity (aOR = 12.6, p < 0.001), likely reflecting the more intensive treatment required for severe cases. Compared with the 2005–2009 baseline period, subsequent years showed elevated risks of severe AP.

The forest plot (Figure 6) displays adjusted odds ratios with 95% confidence intervals for all variables showing statistically significant associations with disease severity (p < 0.05), organized by clinical domains. The final multivariable model (Model 3) demonstrated good discriminative capacity (AUC = 0.724; Figure 7), representing a 39.8% relative improvement over the crude model (Model 1 AUC = 0.525).