Predicting Length of Stay in Neurological ICU Patients Using Classical Machine Learning and Neural Network Models: A Benchmark Study on MIMIC-IV
Alexander Gabitashvili, Philipp Kellmeyer
TL;DR
The paper addresses predicting ICU length of stay for neurological patients using the MIMIC-IV dataset, framing LOS as a three-class problem and benchmarking both classical ML on static data and neural models on time-series data. It systematically compares KNN, Random Forest, SVM, XGBoost, and CatBoost against LSTM, BERT, and Temporal Fusion Transformer, with Optuna-driven hyperparameter tuning and SMOTE-based class-balancing. Key findings show Random Forest achieves robust performance on static data (~0.68 accuracy), while BERT on time-series data achieves higher accuracy (~0.80), and LSTM reaches ~0.72 under sliding-window augmentation; TFT underperformed due to computational constraints. The study highlights the value of condition-specific LOS prediction, demonstrates effective use of time-series modeling for ICU data, and provides a benchmark framework to guide future neuro-ICU LOS prediction and ICU resource planning.
Abstract
Intensive care unit (ICU) is a crucial hospital department that handles life-threatening cases. Nowadays machine learning (ML) is being leveraged in healthcare ubiquitously. In recent years, management of ICU became one of the most significant parts of the hospital functionality (largely but not only due to the worldwide COVID-19 pandemic). This study explores multiple ML approaches for predicting LOS in ICU specifically for the patients with neurological diseases based on the MIMIC-IV dataset. The evaluated models include classic ML algorithms (K-Nearest Neighbors, Random Forest, XGBoost and CatBoost) and Neural Networks (LSTM, BERT and Temporal Fusion Transformer). Given that LOS prediction is often framed as a classification task, this study categorizes LOS into three groups: less than two days, less than a week, and a week or more. As the first ML-based approach targeting LOS prediction for neurological disorder patients, this study does not aim to outperform existing methods but rather to assess their effectiveness in this specific context. The findings provide insights into the applicability of ML techniques for improving ICU resource management and patient care. According to the results, Random Forest model proved to outperform others on static, achieving an accuracy of 0.68, a precision of 0.68, a recall of 0.68, and F1-score of 0.67. While BERT model outperformed LSTM model on time-series data with an accuracy of 0.80, a precision of 0.80, a recall of 0.80 and F1-score 0.80.