Interpretable Machine Learning-Based Risk Scoring with Individual and Ensemble Model Selection for Clinical Decision Making

Han Yuan, Jinwee Lee, Mingxuan Liu, Siqi Li, Chenglin Niu, Jun Wen, Feng Xie

Keywords: clinical score, automated machine learning, interpretable machine learning

TL;DR: In this work, we improved AutoScore with additional variable ranking methods and an automatic model selection and demonstrated that these updates generate clinical scores with fewer variables and higher accuracy.

Abstract: Clinical scores are highly interpretable and widely used in clinical risk stratification. AutoScore was previously developed as a clinical score generator, integrating the interpretability of clinical scores and the discriminability of machine learning. Although a basic framework has been established, AutoScore leaves room for enhancement: variable ranking via the random forest and manual model selection. In this work, we improved them with additional variable ranking methods and an automatic model selection. We demonstrated that these updates generate clinical scores with fewer variables and higher accuracy. The code is available at https://github.com/Han-Yuan-Med/comparison.