Development and temporal validation of SCRAPS scoring tool to predict poor prognosis in gram negative sepsis patients requiring early intensive care or high dependency unit admission

Background: Gram-negative sepsis (GNS) is associated with high global mortality rates, particularly in intensive care unit (ICU) settings. Despite the availability of various sepsis severity scores, there is a lack of predictive tools tailored for GNS. Our study aimed to develop and temporally validate a novel prediction scoring tool for poor prognosis in GNS patients requiring early ICU or High-dependency unit (HDU) admission. Methods: This retrospective observational study included 312 adult patients with GNS admitted to the ICU or HDU of a tertiary care hospital from January 2016 to December 2020. Primary outcome of interest was poor prognosis. Candidate variables were grouped into demographics, comorbidities, clinical presentation at admission, laboratory data, resistance pattern, and early complications. Independent predictors of poor prognosis were identified using multivariate logistic and cox regression analysis. A prediction scoring tool was developed and temporally validated on a prospective cohort (October-December 2024). All the analysis was conducted using SPSS version 29.0. The study was meticulously designed, implemented, and documented in compliance with the TRIPOD checklist. Results: Among 312 patients with GNS in the development cohort, 71.4% patients had poor prognosis. We developed an 8-point predictive score comprising of independent predictors: APACHE II score > 18, SOFA score > 7, CRP > 88 mg/L, carbapenem resistance in the first isolate, septic shock within 24 h and RRT requirement within 48 h of admission. SCRAPS score demonstrated strong predictive performance with an area under the curve of 0.801 in both the development and validation cohorts. SCRAPS score ≥ 5 was associated with increased risk of poor prognosis and significantly lower survival (Kaplan-Meier, log rank test, p < 0.001). Conclusion: The SCRAPS score ≥ 5 is a reliable, practical tool for predicting poor prognosis in GNS patients. Its early application could support timely clinical decision-making, improving patient outcomes in the ICU and HDU settings. Clinical trial number: Not applicable.