| 摘要: | 本研究利用Cox比例風險模型存活分析方法，對台北榮民總醫院腹主動脈瘤患者臨床資料進行分析與說明。該資料包括病患基本資料、病歷史資料、臨床檢驗資料與手術資料等。文中將變數分為三類：慢性疾病變數、心血管疾病常規檢驗變數〈術前、術後第一天〉及一般疾病變數。分析的過程中以住院天數當做反應變數，以慢性疾病變數、心血管疾病常規檢驗變數及一般疾病變數當作共變數，進行初期探勘分析並得到各變數的統計摘要。再進行逐步迴歸分析發現與住院天數有顯著相關的變數。因此，資料分析以三個步驟進行分析，首先將原始資料進行分析。接著，因該筆臨床資料中有約百分之五十的遺失值，為了提高資料的使用度，故初步先以各項變數的平均值代入其遺失值，再將處理遺失值後的資料進行初步分析。最後藉由前二次分析所得到的模型，再代入原始資料中做進一步分析，得到最後結果。
經由分析結果得到，與住院天數有顯著相關的變數有：周邊動脈阻塞性疾病、慢性腎臟病、血紅素〈術前〉、白蛋白〈術前〉、尿素氮〈術後第一天〉、血比容〈術後第一天〉、麻醉風險等級、手術時間、加護病房停留時間及出院條件等。其危險比依序為：0.705、0.689、1.091、1.601、0.982、1.053、0.639、0.888、0.968及1.887。
In this study, we use Cox proportional hazards model to analyze the abdominal aortic aneurysms clinical data of Veterans General Hospital in Taipei, Taiwan. The data includes patient’s demographic and characteristic variables, disease history, clinical examination and treatment data. The response variable in this study is hospitalization days. The rest of variables are prognostic factors or covariates that are divided into three categories: chronic disease variables, cardiovascular disease routine tests of pre-surgery and 1-day after surgery, as well as general illness variables. The summary statistics of each covariates and the correlation with the response variable are given. The analysis was carried out in three steps. We first obtain the preliminary fitting of the raw data by using the Cox PH model. Because of the large proportion of missing values in the dataset, the missing values are estimated and then the data were analyzed. Finally, by employing the findings of the previous analysis, we obtain the result of fitting the raw dataset into the Cox PH regression model.
The stepwise Cox PH regression analysis shows that the significant covariates associated with the hospitalization days (and the hazard ratio) are: PAOD ( 0.705 ), Chronic renal disease ( 0.689 ), Hgb_Pre_op ( 1.091 ), Albumin_Pre_op ( 1.601 ), BUN_Day1 ( 0.982 ) ,Hct_Day1 ( 1.053 ), ASA class ( 0.639 ) , Operative time ( 0.888 ), ICU stay ( 0.968 )and Discharge condition ( 1.887 ). |
