Clinical Scorecard: An Interpretable Gradient Boosting Machine Approach for Forecasting In-Hospital Mortality in Patients with Sepsis-Induced Myocardial Injury: A Multicenter Development, Validation, and Online Clinical Application

At a Glance

Key Highlights

• The GBM model achieved an internal validation AUC of 0.751.
• External validation yielded AUCs of 0.924 and 0.703 in different cohorts.
• Key predictors included APS III Score, Hypertension, Albumin, Diabetes, SOFA Score, ALT, RBC, and Lactate.
• A simplified model with five variables achieved an AUC of 0.789.
• The model is deployed on a user-friendly platform for real-time risk assessment.

Guideline-Based Recommendations

Diagnosis

• Diagnosis of SIMI requires meeting Sepsis-3.0 criteria and elevated cTnT levels.

Management

• Utilize the developed GBM model for risk stratification in SIMI patients.

Monitoring & Follow-up

• Monitor key predictors such as SOFA score and cardiac biomarkers.

Risks

• In-hospital mortality rates for SIMI patients can reach 35%.

Patient & Prescribing Data

Patients with septic myocardial injury from multiple databases.

The model aids in early risk assessment to potentially improve patient outcomes.

Clinical Best Practices

• Incorporate machine learning models for enhanced risk stratification in critical care.
• Utilize multi-center validation to ensure model generalizability.

Related Resources & Content

• MIMIC-IV Database
• eICU Collaborative Research Database