| Summary: | 碩士 === 國立雲林科技大學 === 工業工程與管理系 === 107 === The safety of inpatients is an important indicator for hospital healthcare quality, so the in-hospital cardiac arrest(IHCA) has always been the main reason. Because the frequency of occurrence is extremely high, and the survival rate is extremely low after first aid. Although, hospital use clinical alert system(CAS) substitute human monitor in order to increase the number of alarm, but the mortality has not decreased. Thus, in this study will promote sensitivity and specificity for CAS. The medical personnel will find patients with unstable conditions in advance after CAS is improved, then give inpatients appropriately medical treatment to effectively reduce the events occurrence and mortality.
Therefore, in this study will collect the data of the clinically abnormal events with the CAS standards and the adult inpatients who over twenty years old from in case hospital database during 2016 to 2018.Using decision tree C5.0 and Logistic regression to find the correlation between independent variables and dependent variables, and then we use the decision tree C5.0 , Logistic regression and support vector machine(SVM) to conduct data mining and analysis to find out the impact. The important factors of in-hospital cardiac arrest events and building a predictive models. Finally, we use the accuracy and ROC curve as the model performance evaluation criteria. Based on this result , the best predicted model is the decision tree C5.0 , the accuracy is 83.93%. This result can be as a reference for physician in clinical medicine.
|
|---|
