Intensive Care Unit (ICU) is a special department of any hospital for the critical patients. Premature discharge from ICU is a common incident due to shortage of clinical resources and economic problems. These discharged patients often return to ICU with more critical situation and risk. Thus the objective of this paper is to propose a model, comprising 18 health related variables to predict readmission after discharge and help doctors in taking decision. This paper also examined the performance of three different data mining techniques: Naïve Bayes (probabilistic), decision tree and neural network. The results showed that the proposed model provide the best result in case of artificial neural network with the prediction accuracy of 95.005, AUC of 0.874 and sensitivity of 0.950. The prediction model is trained and evaluated by using almost 11000 data from the MIMIC-III dataset. The skewness of the dataset is also considered while proposing the model.
Inan, Toki Tahmid; Samia, Mahmuda Binte Rashid; Tulin, Iffat Tamanna; and Islam, Muhammad Nazrul, "A Decision Support Model to Predict ICU Readmission through Data Mining Approach" (2018). PACIS 2018 Proceedings. 218.