Machine learning interpretability methods to characterize the importance of hematologic biomarkers in prognosticating patients with suspected infection

• Published in: Computers in biology and medicine 2024-12, Vol.183, p.109251, Article 109251
• Main Authors: Upadhyaya, Dipak P., Tarabichi, Yasir, Prantzalos, Katrina, Ayub, Salman, Kaelber, David C., Sahoo, Satya S.
• Format: Article
• Language: English
• Subjects: Accuracy; Adult; Aged; Algorithms; Antibiotics; Biomarkers; Biomarkers - blood; Blood tests; Clinical outcomes; Data analysis; Electronic health records; Emergency medical care; Emergency medical services; FDA approval; Feature importance; Feature interaction effect; Female; Hematologic parameters and vital signs; Hospitals; Humans; Infections; Intensive care; Learning algorithms; Machine Learning; Machine learning interpretability; Male; MDW biomarker; Medical prognosis; Middle Aged; Monocytes; Mortality; Parameters; Patients; Prediction models; Predictions; Prognosis; Regional development; Retrospective Studies; Sepsis; Sepsis - blood; Sepsis - diagnosis; Sepsis screening; Signs; Vital signs
• ISSN: 0010-4825; 1879-0534; 1879-0534
• DOI: 10.1016/j.compbiomed.2024.109251

To evaluate the effectiveness of Monocyte Distribution Width (MDW) in predicting sepsis outcomes in emergency department (ED) patients compared to other hematologic parameters and vital signs, and to determine whether routine parameters could substitute MDW in machine learning models. We conducted a retrospective analysis of data from 10,229 ED patients admitted to a large regional safety-net hospital in Cleveland, Ohio who had suspected infections and developed sepsis-associated poor outcomes. We developed a new analytical framework consisting of seven data models and an ensemble of high accuracy machine learning (ML) algorithms (accuracy values ranging from 0.83 to 0.90) to predict sepsis-associated poor outcomes (3-day intensive care unit stay or death). Local Interpretable Model-Agnostic Explanation (LIME) and Shapley Additive Value (SHAP) interpretability methods were utilized to assess the contributions of individual hematologic parameters. The ML interpretability analysis indicated that the predictive value of MDW is significantly reduced when other hematological parameters and vital signs are considered. The results suggest that complete blood count with differential (CBD-DIFF) alongside vital signs can effectively replace MDW in high accuracy machine learning algorithms for screening poor outcome associated with sepsis. MDW, although a newly approved biomarker for sepsis, does not significantly enhance prediction models when combined with routinely available parameters and vital signs. Hospitals, especially those with resource constraints, can rely on existing parameters with high accuracy machine learning models to predict sepsis outcomes effectively, thereby reducing the need for specialized tests like MDW. [Display omitted] •Novel ML framework predicts sepsis with severe outcomes with high accuracy.•MDW's prognostic value is less significant with routine hematologic data.•Study supports using CBC data and vital signs over MDW for sepsis prognostication.•Ensemble ML models achieved AUC values from 0.83 to 0.90.•LIME and SHAP methods highlight key features influencing sepsis risk.