| 摘要: | 研究目的：本研究旨在利用機器學習技術，分析慢性腎病患者出院後30天內再次住院的危險因子，以期能幫助臨床醫生及早偵測具有高風險再住院的患者。
研究方法：本研究為回溯性世代研究，蒐集來自臺北醫學大學三院臨床研究資料庫中，共計9,747位慢性腎病患者的數據，研究資?區間為2012年01月01日至2021年12月31日。使用變項包括人口學特徵、疾病因子及出入院狀況。並使用六種機器學習演算法包括羅吉斯回歸、支持向量機、決策樹、隨機森林、梯度提升機及人工神經網路。算法的性能透過比較六種模型的操作者曲線下面積(AUROC)來衡量。最終選擇擁有最佳表現的模型，進行再住院危險因子的排序與分析。
研究結果：結果顯示，在比較六種機器學習模組在測試組的表現後，羅吉斯回歸模型擁有最佳分類能力，其AUROC為0.5810，準確性為0.6283、敏感度為0.4693、特異度為0.6620、精確率為0.2275、F1-score為0.3064。透過最佳模型進行變數重要性分析，結果指出高血脂、自體免疫疾病、痛風和糖尿病為慢性腎病再住院中最重要的危險因子。這也提醒了醫療單位在預防慢性腎病再住院時，應特別關注這些因素，以期減少再住院的機率，改善病患及家屬的生活品質，降低健保醫療的負擔。
Objective: The aim of this study is to utilize machine learning techniques to analyze the risk factors for readmission within 30 days after discharge for patients with chronic kidney disease. The goal is to assist clinical physicians in early detection of patients at high risk of readmission.
Methods: This study is a retrospective cohort study that collected data from the clinical research database of Taipei Medical University Hospital, encompassing a total of 9,747 patients with chronic kidney disease. The study period ranged from January 1, 2012, to December 31, 2021.
The variables used include demographic characteristics, disease factors, and admission/discharge status. Six machine learning algorithms, including logistic regression, support vector machine, decision tree, random forest, gradient boosting machine, and artificial neural network, were utilized. The performance of these algorithms was evaluated by comparing the area under the receiver operating characteristic curve (AUROC) of each model. The model with the best performance was selected for ranking and analysis of readmission risk factors.
Results: The results show that among the six machine learning models compared in the test set, the logistic regression model exhibited the best classification ability. It achieved an AUROC of 0.5810, an accuracy of 0.6283, a sensitivity of 0.4693, a specificity of 0.6620, a precision of 0.2275, and an F1-score of 0.3064. Through the analysis of variable importance using the best model, the results indicate that hyperlipidemia, autoimmune diseases, gout, and diabetes are the most important risk factors for readmission in chronic kidney disease. This also serves as a reminder for healthcare institutions to pay special attention to these factors in order to prevent readmissions for chronic kidney disease. The aim is to reduce the probability of readmission, improve the quality of life for patients and their families, and alleviate the burden on the healthcare system. |
