Machine learning predicting mortality in sarcoidosis patients admitted for acute heart failure

Area of Science:

• Cardiology
• Pulmonology
• Medical Informatics

Background:

• Cardiac involvement in sarcoidosis, though rare, is associated with a poorer prognosis.
• Heart failure (HF) in sarcoidosis patients presents unique challenges for mortality prediction.

Purpose of the Study:

• To develop and validate machine learning models for predicting in-hospital mortality in sarcoidosis patients admitted with heart failure.
• Identify key predictors of mortality within this specific patient cohort.

Main Methods:

• Utilized the National Inpatient Sample (NIS) database, identifying 4659 patients with primary HF and secondary sarcoidosis diagnoses.
• Trained and tested logistic regression, random forest (RF), and XGBoosting models using a 7:3 data split.
• Employed Least Absolute Shrinkage and Selection Operator (LASSO) regression for variable selection and GridSearchCV for parameter tuning.

Main Results:

• The random forest model demonstrated the highest predictive performance with the best Area Under Curve (AUC) score and sensitivity.
• Comorbid arrhythmias and fluid electrolyte disorders were identified as the most significant predictors of in-hospital mortality.
• Overall in-hospital mortality for sarcoidosis patients with HF was 2.3%.

Conclusions:

• Machine learning models effectively identify critical predictors of in-hospital mortality in sarcoidosis patients with heart failure.
• Findings highlight the importance of managing arrhythmias and electrolyte imbalances to improve outcomes.