Risk factors and early clinical outcomes of acute pancreatitis: A retrospective case series from Kyrgyzstan

Background: Acute pancreatitis (AP) is a common gastrointestinal emergency characterized by unpredictable severity. Early identification of patients at risk for severe disease is essential for timely intervention and improved outcomes, yet reliable prognostic markers remain limited, particularly in Central Asian clinical settings. Objective: To identify early clinical, laboratory, and demographic predictors of acute pancreatitis (AP) severity using statistical and machine learning approaches, to improve early risk stratification and guide prompt clinical management. Methods: This retrospective case series analysed 40 patients diagnosed with AP between 2022 and 2024 at tertiary hospitals in Bishkek, Kyrgyzstan. Admission data, including demographic characteristics, clinical symptoms, and laboratory values, were collected and evaluated using the revised Atlanta criteria. Statistical analyses included correlation testing, subgroup comparisons, and logistic regression, while a machine–learning–based feature importance analysis was used to identify key predictors of severe AP. Results: Several variables were significantly associated with severe AP. Serum amylase >500 U/L (OR = 5.2, p < 0.001), WBC count >15×10?/L (OR = 4.7, p < 0.001), and BMI ?30 (OR = 3.4, p = 0.003) emerged as strong predictors of severity. A strong correlation was observed between total bilirubin and jaundice (r = 0.62, p < 0.001). Obese patients had longer hospital stays compared with non-obese patients (median 12 vs. 7 days; p = 0.021). Machine learning analysis confirmed serum amylase, WBC count, and BMI as the most influential predictors. Conclusion: Serum amylase, WBC count and BMI are practical, easily accessible markers that can support early prediction of AP severity. Incorporating these indicators into initial assessment protocols may enhance risk stratification and optimize clinical decision-making. Prospective multicentre studies are needed to validate these findings and refine AP severity prediction models.