加護病房的支出正逐年上漲。為了要減少支出，醫護人員傾向讓病患提早離開加護病房。然而，這樣可能會使病患的狀況惡化，最後導致再住回加護病房。如果醫護人員可以知道再住院風險高的病患有哪些人的話，就可以避免那些人的再住院。另一方面，醫護人員可以讓再住院風險較低的患者離開，以降低醫院的費用。因此，加護病房再住院預測對病患的健康和減少加護病房的支出均有幫助。在這項研究中，我們提出了一個用於預測再住院風險的多模型架構。我們使用三種不同的模型來處理三種不同類型的資料，並結合三種模型的結果來得到每個病患再住院風險的最終預測結果。實驗結果顯示，我們提出的方法可以達到0.755的AUROC，證明它可以有效地預測每位病患的加護病房再住院風險。

The cost of intensive care unit (ICU) is increasing annually. For reducing cost, physicians tend to discharge patients from ICU at an early time. However, it may cause bad result of patient’s outcome, finally lead to readmission. Supposed that physicians could know certain patients that prone to readmission, their readmission could be avoided. In contrast, in order to reduce the cost, patients with low readmission risk could be discharged by physicians. Therefore, ICU readmission prediction is helpful for both the physical condition of patients and ICU cost. In this study, we propose a multimodal architecture to predict readmission risk. There are three different models to utilize three different types of data. We combine results of three models to make final prediction of each patient. Experimental results show that our proposed method can achieve up to 0.755 AUROC, prove that it can effectively predict ICU readmission risk for each patient.

[1] Johnson A., Pollard T., Shen L. MIMIC-III, a freely accessible critical care database. Scientific Data. 2016;3(160035).
[2] A.S. Fialho, F. Cismondi, S.M. Vieira, S.R. Reti, J.M.C. Sousa, S.N. Finkelstein. Data mining using clinical physiology at discharge to predict ICU readmissions. Expert Systems with Applications. 2012;39(18):13159-65.
[3] Marta P. B. Fernandes, Cláudia F. Silva, Susana M. Vieira and João M. C. Sousa. Multimodeling for the prediction of patient readmissions in intensive care units. IEEE International Conference on Fuzzy Systems (FUZZ-IEEE). 2014:1837-42.
[4] Whitney A.W., A direct method of nonparametric measurement selection. IEEE Trans Computer. 1971;20(9):1000-3
[5] Sergio Curto, Joao P. Carvalho, Catia Salgado, Susana M. Vieira, Joao M. C. Sousa. Predicting ICU readmissions based on bedside medical text notes. IEEE International Conference on Fuzzy Systems (FUZZ-IEEE). 2016:2144-51.
[6] Durbin, C.G., Jr, Kopel, R.F. A case-control study of patients readmitted to the intensive care unit. Critical Care Medicine. 1993;21(10):1547-53.
[7] William Caicedo-Torres, Jairo Gutierrez. ISeeU: visually interpretable deep learning for mortality prediction inside the ICU. Journal of Biomedical Informatics. 2019;98.
[8] Chen Y.S., Chou C.Y., Chen A.L.P. Early prediction of acquiring acute kidney injury for older inpatients using most effective laboratory test results. BMC Medical Informatics and Decision Making. 2020;20(1):36.
[9] Chollet F. & others. Keras. https://github.com/fchollet/keras. 2015.
[10] Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O., Blondel M., Prettenhofer P., Weiss R., Dubourg V., Vanderplas J., Passos A., Cournapeau D., Brucher M., Perrot M., Duchesnay E. Scikit-learn: machine learning in python, Journal of Machine Learning Research, 2011;12:2825-30.
[11] Cooper G.S., Sirio C.A., Rotondi A.J., Shepardson L.B., Rosenthal G.E. Are readmissions to the intensive care unit a useful measure of hospital performance? Medical Care. 1999;37(4):399-408.
[12] T. Gentimis, A. J. Alnaser, A. Durante, K. Cook, R. Steele. Predicting hospital length of stay using neural networks on MIMIC III data. IEEE 15th Intl Conf on Dependable, Autonomic and Secure Computing, 15th Intl Conf on Pervasive Intelligence and Computing, 3rd Intl Conf on Big Data Intelligence and Computing and Cyber Science and Technology Congress(DASC/PiCom/DataCom/CyberSciTech). 2017:1194-201.
[13] Tao Gui, Liang Zhu, Qi Zhang, Minlong Peng, Xu Zhou, Keyu Ding, Zhigang Chen. Cooperative multimodal approach to depression detection in Twitter. AAAI Conference on Artificial Intelligence. 2019;33(1).
[14] Jinmiao Huang, Cesar Osorio , Luke Wicent Sy. An empirical evaluation of deep learning for ICD-9 code assignment using MIMIC-III clinical notes. Computer Methods and Programs in Biomedicine. 2019;177:141-53.
[15] I. H. Witten, E. Frank, M. A. Hall, C. J. Pal, Data mining: practical machine learning tools and techniques. Morgan Kaufmann. 2016.
[16] Chung Junyoung, Gulcehre Caglar, Cho KyungHyun, Bengio Y., Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling. 2014.
[17] Feng Liu, Liangji Wang, Xiaofeng Zhu, Dingding Wang. Suggestion mining from online reviews using random multimodel deep learning. IEEE International Conference on Machine Learning and Applications (ICMLA). 2019:667-72.
[18] Joseph L. Nates, Mark Nunnally, Ruth Kleinpell, Sandralee Blosser, Jonathan Goldner, Barbara Birriel, Clara S. Fowler, Diane Byrum, William Scherer Miles, Heatherlee Bailey, Charles L. Sprung. ICU admission, discharge, and triage guidelines: a framework to enhance clinical operations, development of institutional policies, and further research. Critical Care Medicine, 2016:1553-1602.
[19] Melina Loreto, Thiago Lisboa, Viviane P. Moreira. Early prediction of ICU readmissions using classification algorithms. Computers in Biology and Medicine. 2020;118.
[20] Trisha Mittal, Uttaran Bhattacharya, Rohan Chandra, Aniket Bera, Dinesh Manocha. M3ER: multiplicative multimodal emotion recognition using facial, textual, and speech cues. AAAI Conference on Artificial Intelligence. 2020;34(2).
[21] Bergstrom N., Braden B. J., Laguzza A., Holman V. The Braden scale for predicting pressure sore risk. Nursing Research. 1987;36:205–10.
[22] Arash Pakbin, Parvez Rafi, Nate Hurley, Wade Schulz, Harlan M. Krumholz, Bobak J. Mortazavi. Prediction of ICU readmissions using data at patient discharge. Annual International Conference of the IEEE Engineering in Medicine and Biology Society(EMBC). 2018:4932-35.
[23] John P., Probabilistic outputs for support vector machines and comparisons to regularized likelihood methods. Advances in Large Margin Classifiers. 2000;10.
[24] Juan C. Rojas, Kyle A. Carey, Dana P. Edelson, Laura R. Venable, Michael D. Howell, Matthew M. Churpek. Predicting intensive care unit readmission with machine learning using electronic health record data. Annals of the American Thoracic Society. 2018;15(7):846-853.
[25] Matthieu Scherpf, Felix Gräßer, Hagen Malberg, Sebastian Zaunseder. Predicting sepsis with a recurrent neural network using the MIMIC III database, Computers in Biology and Medicine. 2019;113.
[26] Rita Viegas, Cátia M. Salgado, Sérgio Curto, João P. Carvalho, Susana M. Vieira, Stan N. Finkelstein. Daily prediction of ICU readmissions using feature engineering and ensemble fuzzy modeling. Expert Systems With Applications. 2017;79(15):244-53.
[27] Ye Xue, Diego Klabjan, Yuan Luo. Predicting ICU readmission using grouped physiological and medication trends. Artificial Intelligence in Medicine. 2019;95:27-37.
[28] Society of Critical Care Medicine, Critical Care Statistics, https://www.sccm.org/Communications/Critical-Care-Statistics.
[29] Canadian Institute for Health Information. Care in canadian ICUs - data tables. 2016.