A data-driven medical knowledge discovery framework to predict the length of ICU stay for patients undergoing craniotomy based on electronic medical records

Abstract

Craniotomy is an invasive operation with great trauma and many complications, and patients undergoing craniotomy should enter the ICU for monitoring and treatment. Based on electronic medical records (EMR), the discovery of high-risk multi-biomarkers rather than a single biomarker that may affect the length of ICU stay (LoICUS) can provide better decision-making or intervention suggestions for clinicians in ICU to reduce the high medical expenses of these patients and the medical burden as much as possible. The multi-biomarkers or medical decision rules can be discovered according to some interpretable predictive models, such as tree-based methods. Our study aimed to develop an interpretable framework based on real-world EMRs to predict the LoICUS and discover some high-risk medical rules of patients undergoing craniotomy. The EMR datasets of patients undergoing craniotomy in ICU were separated into preoperative and postoperative features. The paper proposes a framework called Rules-TabNet (RTN) based on the datasets. RTN is a rule-based classification model. High-risk medical rules can be discovered from RTN, and a risk analysis process is implemented to validate the rules discovered by RTN. The performance of the postoperative model was considerably better than that of the preoperative model. The postoperative RTN model had a better performance compared with the baseline model and achieved an accuracy of 0.76 and an AUC of 0.85 for the task. Twenty-four key decision rules that may have impact on the LoICUS of patients undergoing craniotomy are discovered and validated by our framework. The proposed postoperative RTN model in our framework can precisely predict whether the patients undergoing craniotomy are hospitalized for too long (more than 15 days) in the ICU. We also discovered and validated some key medical decision rules from our framework.