Predicting multiple organ dysfunction syndrome in trauma-induced sepsis: Nomogram and machine learning approaches

Area of Science:

• Critical care medicine
• Trauma surgery
• Sepsis research

Background:

• Multiple organ dysfunction syndrome (MODS) is a severe complication in trauma patients with sepsis, leading to high mortality rates.
• Early and accurate prediction of MODS is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To develop and validate predictive models for MODS in trauma-induced sepsis patients.
• To compare the performance of nomogram and machine learning models against traditional scoring systems.

Main Methods:

• Retrospective cohort study using the MIMIC-IV database (2008-2019).
• Development and validation of nomogram and machine learning models using initial 24-hour ICU data.
• Feature selection via Boruta algorithm; performance evaluation using discrimination, calibration, and decision curve analysis.

Main Results:

• 349 out of 1295 trauma sepsis patients (26.95%) developed MODS, with higher mortality in MODS patients (23.82% vs 11.21%).
• Key predictors included SAPS II score, mechanical ventilation, and vasopressor use.
• Nomogram (AUC 0.757) and Random Forest (AUC 0.769) models showed superior performance to traditional systems in temporal validation.

Conclusions:

• Nomogram and machine learning models offer enhanced predictive accuracy for MODS in trauma sepsis.
• These models can aid in early risk stratification and clinical decision-making.
• Further external validation is recommended to confirm generalizability.