Identifying mortality factors from Machine Learning using Shapley values - a case of COVID19

Area of Science:

• Medical Informatics
• Computational Biology
• Epidemiology

Background:

• COVID-19 poses a significant global health challenge, with patient mortality influenced by various clinical factors.
• Understanding predictors of mortality is crucial for effective patient management and resource allocation.

Purpose of the Study:

• To apply and interpret machine learning models for predicting COVID-19 patient mortality.
• To identify robust clinical predictors of mortality using a unique dataset from Wuhan, China.

Main Methods:

• Utilized a dataset of 375 COVID-19 patients (201 survivors, 174 deceased) including blood sample data.
• Applied a series of machine learning models, focusing on model interpretability (global and local).

Main Results:

• Identified age, days in hospital, lymphocyte count, and neutrophil count as significant predictors of mortality.
• Machine learning models provided case-by-case insights into variable impact, aiding in anomaly detection.

Conclusions:

• Machine learning models can effectively predict COVID-19 mortality and offer valuable clinical interpretations.
• Interpretable machine learning can assist healthcare practitioners in identifying at-risk patients and understanding mortality patterns.