Admission prediction model in the adult medical emergency patients

• Main Authors: Lin, Chi-Min, 林啟民
• Other Authors: Chen, Juei-Chao
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/12115243252860856715

碩士 === 輔仁大學 === 統計資訊學系應用統計碩士班 === 100 === Object: Overcrowding at the emergency department continues to be an important issue. Early prediction of hospital admission may reduce waiting time and also provide the valuable information to the clinical doctor. The purpose of the study is to develop a model predicting patient’s final outcome of the adult medical emergency department at the time of ED triage, using routine hospital administrative data. Method: This is a retrospective study, using the data collected by the nursing at the time of triage from Jan. 2011 to Dec. 2011. The variable includes age, sex, past history, chief complaint, biological profile (such as blood pressure, pulse rate, etc.), and the final outcome. Chi-square tests are used to study the association between nominal or ordinal data, and the student T test analyzes continuous data. CART (Classification and Regression Tree) is applied to develop the prediction model. Result: Of 36287 patients, 5602 patients (15.4%) were admitted for further treatment. Variables like Age, respiratory rate, respiratory pattern, oxygen saturation, body temperature, conscious level, diastolic blood pressure , and chief complaint of blood in stool are included in our predictive model. The sensitivity of the model is 36.0% and the specificity is 95.7%. The c-statists of ROC curve is 73.516%. Conclusion: By CART prediction model, we can identify the high-risk groups of admission, providing useful information to clinicians.