New Perspectives in the Assessment of Severity Scoring in Acute Pancreatitis: Development and Internal Validation of...

Background and Objectives: Existing severity scores for acute pancreatitis (AP), such as the Ranson, Glasgow-Imrie, APACHE II, BISAP, Balthazar scores, are limited by complexity, delayed availability, or insufficient discriminative accuracy. We aimed to develop a new prognostic score using parameters available at admission in any primary or regional care setting, suitable for early risk stratification and clinical decision-making in AP. Materials and Methods: We conducted a retrospective observational study of 194 patients admitted with acute pancreatitis, classified as mild (MILD, n = 59), moderately severe (n = 82) and severe (n= 53) course, according to the Revised Atlanta Classification. Candidate predictors-including demographic, clinical, inflammatory, nutritional, and radiological variables-were screened using univariable analysis, and independent predictors were identified by backward stepwise logistic regression. Discriminative performance was assessed using receiver operating characteristic (ROC) curve analysis and compared with six established severity scores (Ranson, Glasgow-Imrie, Balthazar, CTSI, BISAP, and HAPS). APACHE II was not included in this comparison because some of its required physiological variables were not consistently available outside the intensive care unit. Results: MODSEV (moderately severe/severe) patients showed significantly higher inflammatory, nutritional, and metabolic derangement at admission than MILD patients. Six variables-admission white blood cell count, blood urea nitrogen (BUN), pleural effusion, admission blood glucose level, platelet-to-lymphocyte ratio (PLR), and serum albumin-were retained in the final logistic regression model (ProbScore2), which demonstrated excellent discriminative ability (AUC = 0.873; 95% CI: 0.819-0.927), showing, within this retrospective cohort, a numerically higher AUC than the Glasgow-Imrie (AUC = 0.832), CTSI (0.775), Balthazar (0.758), BISAP (0.725), Ranson (0.676) and HAPS (0.574). A simplified, unweighted 0-6 point bedside version of the score retained most of this discriminative ability (AUC = 0.852), with an optimal cut-off of ≥3 points yielding a sensitivity of 77.0% and a specificity of 79.7%. Conclusions: ProbScore2, a novel six-variable prognostic score, demonstrated excellent discriminative performance for the early identification of moderately severe/severe acute pancreatitis, with a numerically higher AUC than established severity scoring systems evaluated within this same retrospective, single-center cohort (excluding APACHE II, which could not be calculated for all patients). These comparative findings require confirmation through prospective external validation before any conclusions about generalizable superiority can be drawn. Five of the six components are derived from routine admission blood tests, and the sixth (pleural effusion) requires only a plain chest radiograph rather than cross-sectional imaging. The score and its simplified bedside adaptation can therefore be calculated within hours of admission using resources available even in hospitals without intensive care or advanced, CT-based imaging capabilities. ProbScore2 is not yet ready for routine clinical use; prospective external validation in independent, multicenter cohorts is required before clinical implementation.