查詢結果分析
來源資料
相關文獻
- 白內障術後眼內炎預測模型之研究
- 全民健保資料庫的抽樣方法
- Endophthalmitis as a Complication of Meningococcal Meningitis: Report of One Case
- 角膜潰瘍併發眼內炎
- Enucleation with Primary Hydroxyapatite Implant in Treatment of Recalcitrant Endophthalmitis
- 從醫院管理角度論全民健保資料庫
- 外傷性蠟狀芽胞桿菌全眼炎--病例報告
- Overview of Extremity Replantation in Taiwan by the National Health Insurance Database from 1996 to 2000
- 內因性克雷伯氏肺炎桿菌眼內炎之病例討論
- Endogenous Klebsiella Pneumoniae Endophthalmitis Associated with Prostate Abscess: Case Report
頁籤選單縮合
題名 | 白內障術後眼內炎預測模型之研究=A Study on a Logistic Model for Predicting the Incidence of Endophthalmitis after Cataract Surgery |
---|---|
作者 | 吳建和; 張財源; 王淑蕙; Wu, Chien-ho; Chang, Tsair-yuan; Wang, Shu-hui; |
期刊 | 長榮大學學報 |
出版日期 | 20111200 |
卷期 | 15:2 2011.12[民100.12] |
頁次 | 頁99-115 |
分類號 | 416.744 |
語文 | chi |
關鍵詞 | 眼內炎; Logistic迴歸分析; 類別不平衡; 健保資料庫; Endophthalmitis; Logistic regression; Imbalanced data; NHI claim data; |
中文摘要 | 每種手術或醫療處置都會面臨醫療風險,為了降低醫療風險,做法之一是找出關鍵性指標及建立預測模型,以做為風險管理的決策依據。目前白內障手術非常普遍,而眼內炎為白內障術後最嚴重的併發症,可能有失明的危險。本研究以全民健保資料庫2008至2010年實施白內障手術的申報資料(424,773筆)為對象,運用logistic迴歸建立術後發生眼內炎的預測模型。由於資料有類別不平衡的問題(class imbalance problems),因此在建模過程中,本研究將有發生(positive)及未發生(negative)眼內炎的資料分開並分別以隨機抽樣方式進行檔案切割,產生十組訓練集,同時以此十組訓練集建立十個初次模型。在參考十個初次模型之結果後,選取重要解釋變數,並建立最終預測模型;最終預測模型之平均預測正確率為71.5%。研究結果顯示,與術後發生眼內炎有關的重要因素分別為青光眼疾病史、門住診別、醫院地區別、醫院年手術量、醫師年齡別及醫師年手術量等六項。本研究之分析工具為SAS及SPSS,研究結果可做為醫療服務者及管理者在規劃如何提升白內障手術醫療品質時之參考。 |
英文摘要 | Each type of surgery inherently involves some degree of medical risk. To help decrease the risk of suffering endophthalmitis for patients receiving cataract surgery, it is necessary to find out critical factors related to the event of endophthalmitis and use these factors to build a forecasting model for interpretation and administrative purposes. To this end, we collected 424,773 cases, who received cataract surgery between the year 2008 and 2010 time frame, from Taiwan National Health Insurance Database. To address the problem that the cases collected for this research are imbalanced in terms of the frequency of the outcomes of the response variable, we first use 10 training sets to build 10 preliminary models and then select key model variables from the preliminary models to build the final logistic model. The forecasting model thus built can help interpret the relationships between and odds of model variables. The results, generated by SAS and SPSS, show that the average correct rate of the forecasting model is 71.5%. Important factors related to the event of endophthalmitis after receiving cataract surgery include (1) patients’ medical history on glaucoma illness, (2) ambulatory or hospital surgery, (3) geographic location of a hospital, (4) hospital surgery volume per year, (5) doctors’ age, and (6) doctors’ surgery volume per year. This research is a good reference for medical service providers and managers who intend to improve the quality of cataract surgery. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。