A Hybrid Model for Predicting ICU Discharge Readiness and Estimating Excess ICU Stay Duration

This paper presented a hybrid framework for predicting ICU discharge readiness and estimating excess ICU stay duration. The proposed architecture combines temporal patient information with structured clinical features and generates two complementary outputs: a probability of near-term discharge readiness and an estimate of additional ICU duration. The central idea is to represent both the current clinical state and the trajectory of patient recovery. Such a representation can provide more useful information than relying on a single severity score or a fixed length-of-stay threshold. The framework has potential applications in clinical de-cision support, personalized discharge planning, and ICU resource management. However, reliable deployment requires careful validation, attention to data quality, protection of patient privacy, and appropriate interpretation by clinical pro-fessionals. Future research will focus on multimodal clinical data, explainable prediction, uncertainty estimation, federated learn-ing, and prospective validation across multiple healthcare institutions.  

[1]     A. E. W. Johnson et al., “MIMIC-III, a freely accessible critical care database,” Scientific Data, vol. 3, 2016. [2]     T. J. Pollard et al., “The eICU Collaborative Research Database, a freely available multi-center database for critical care research,” Scientific Data, vol. 5, 2018. [3]     A. Rajkomar et al., “Scalable and accurate deep learning with electronic health records,” npj Digital Medicine, vol. 1, 2018. [4]     H. Harutyunyan, H. Khachatrian, D. C. Kale, and A. Galstyan, “Multi-task learning and benchmarking with clinical time series data,” Scientific Data, vol. 6, 2019. [5]     Z. Che, S. Purushotham, K. Cho, D. Sontag, and Y. Liu, “Recurrent neural networks for multivariate time series with missing values,” Scientific Reports, vol. 8, 2018. [6]     S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, 1997. [7]     K. Cho et al., “Learning phrase representations using RNN encoder-decoder for statistical machine translation,” in Proc. EMNLP, 2014, pp. 1724–1734. [8]     L. Breiman, “Random forests,” Machine Learning, vol. 45, pp. 5–32, 2001. [9]     J. H. Friedman, “Greedy function approximation: A gradient boosting machine,” Annals of Statistics, vol. 29, no. 5, pp. 1189–1232, 2001. [10]    T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proc. ACM SIGKDD, 2016, pp. 785–794.