Adding a Gastrointestinal Subscore to SOFA2 for Mortality Risk Assessment in the ICU: The SOFA2-GIS Study

Background and Rationale The Sequential Organ Failure Assessment (SOFA) score, originally introduced in 1996, has been widely used to describe and monitor organ dysfunction in critically ill patients across six organ systems: respiratory, cardiovascular, hepatic, renal, neurological, and coagulation. A revised version, SOFA2, was developed in 2024 by an international consortium under the European Society of Intensive Care Medicine (ESICM) to better reflect contemporary intensive care practice and to incorporate evidence-based threshold revisions derived from large modern ICU databases.

Despite this update, SOFA2 still does not include the gastrointestinal (GI) system. Acute gastrointestinal dysfunction is highly prevalent in critically ill patients and has been independently associated with increased mortality, prolonged ICU stay, and higher rates of nosocomial infection. The Acute Gastrointestinal Injury (AGI) classification, proposed by the ESICM Working Group on Abdominal Problems in 2012, provides a structured framework to grade GI dysfunction from Grade I (increased risk of GI dysfunction) to Grade IV (life-threatening GI failure). Despite its conceptual robustness, the AGI classification has not been formally integrated into a unified organ dysfunction score.

This study aims to address this gap by developing and prospectively evaluating SOFA2-GIS - a composite score combining the SOFA2 score with a gastrointestinal subscore derived from the AGI classification - and comparing its prognostic performance against the standard SOFA2 score in adult ICU patients. The study will be reported in accordance with the TRIPOD+AI 2024 reporting standards for prediction model studies.

Study Design and Procedures

This is a multicenter, prospective, observational cohort study conducted in tertiary adult intensive care units. Eligible patients will be evaluated at two predefined time points:

T24 (primary time point): Within the first 24 hours of ICU admission, the worst values of clinical and laboratory parameters will be used to calculate the SOFA2 score and the AGI grade. The two components will then be combined to derive the SOFA2-GIS score.

T72 (secondary time point): At the 72nd hour (±6 hours) of ICU admission, both scores will be recalculated using cross-sectional measurements, allowing the calculation of Δ-SOFA2 and Δ-SOFA2-GIS (T72 - T24).

AGI grading will follow the original ESICM 2012 criteria, based on clinical symptoms, feeding intolerance, intra-abdominal pressure, gastric residual volume, diarrhea, and gastrointestinal bleeding. To ensure consistency across centers, all participating investigators will undergo standardized training on AGI assessment. Inter-rater agreement will be evaluated through a pilot phase using shared cases prior to the main enrollment period.

No additional interventions, laboratory tests, imaging studies, or sampling procedures will be performed for the purpose of this study. All data will be derived from information collected during routine clinical care.

Data Collection

Data will be collected at the participating sites using a standardized case report form. In addition to the SOFA2 and AGI components, the following variables will be recorded:

Demographic characteristics, comorbidities (Charlson Comorbidity Index), admission category, and admission diagnosis Vital signs, laboratory parameters, mechanical ventilation parameters (including plateau and peak airway pressures), vasopressor use, and renal replacement therapy requirement Sepsis status according to the Sepsis-3 definition Lactate values at baseline, 24th hour, and 72nd hour, and the 24-hour lactate clearance 24-hour cumulative fluid balance Daily caloric intake (DCI) averaged over the first 72 hours (kcal/kg/day) Feeding intolerance, gastrointestinal bleeding, intra-abdominal hypertension, abdominal compartment syndrome, and decompressive laparotomy Acute kidney injury staged according to KDIGO criteria Clinical outcomes: ICU and hospital length of stay, ventilator-free days, vasopressor-free days, renal replacement therapy-free days, 28-day mortality, and 90-day mortality Each participating site will record anonymized patient data into a shared spreadsheet template. The data will then be transferred to and consolidated at the coordinating center, which will not have access to patient identifiers.

Statistical Approach The discriminative performance of SOFA2 and SOFA2-GIS for predicting 28-day mortality (primary outcome) will be compared using the area under the receiver operating characteristic curve (AUC-ROC), with statistical comparison performed via the DeLong test. Calibration will be assessed using the Hosmer-Lemeshow test and calibration plots. Reclassification performance will be evaluated using the net reclassification improvement (NRI) and integrated discrimination improvement (IDI) indices. Multicollinearity between score components will be assessed using the variance inflation factor (VIF).

Secondary analyses will include: comparison of Δ-SOFA2 and Δ-SOFA2-GIS (T72 - T24) for outcome prediction, subgroup analyses by admission category (medical, surgical, septic) and AGI severity grade, evaluation of the interaction between lactate clearance and AGI grade, and the development of a clinically applicable nomogram based on the best-performing model.

Study Conduct The study will be conducted in accordance with the principles of the Declaration of Helsinki. The coordinating center is the Internal Medicine Intensive Care Unit of Ondokuz Mayıs University. Participating centers will join through an open call disseminated via Turkish intensive care societies, following completion of their respective institutional approvals. Each center will obtain its own local ethics committee approval prior to enrolling patients. Reporting will follow the TRIPOD+AI 2024 guideline.

• Difference in discriminative performance (ΔAUC) between SOFA2-GIS and SOFA2 for prediction of 90-day all-cause mortality [Срок оценки: 90 days from ICU admission]
• Difference in discriminative performance (ΔAUC) between SOFA2-GIS and SOFA2 for prediction of ICU mortality [Срок оценки: Through ICU discharge, up to 60 days from ICU admission]
• Difference in discriminative performance (ΔAUC) between Δ-SOFA2-GIS and Δ-SOFA2 for prediction of 28-day all-cause mortality [Срок оценки: From ICU admission to 28 days]
• Net Reclassification Improvement (NRI) of SOFA2-GIS over SOFA2 for 28-day all-cause mortality [Срок оценки: 28 days from ICU admission]
• Intensive care unit length of stay [Срок оценки: Up to 60 days from ICU admission]