由於醫學資料型態具有多樣且複雜的特性,經由主觀的判讀可能造成結果的偏差,則對醫療決策具有一定的影響力,因此選擇適當的資料分析方法有其必要性,在策略選擇的過程中,以類別化的資料型態較能幫助決策之決定。病患透析前的臨床檢驗值是腎臟科醫師常用來判斷病人腎臟功能的主要因子之一,作為病患是否應進入透析治療階段的重要參考指標。故本研究目的為探討適當的資料轉化方法運用在透析前檢驗資料對醫療決策品質的評估。 此研究前期,我們設計一個以透析病患為中心的透析醫療照顧資料庫系統(Dialysis Medical Care System; HD-PD-DBS)提供臨床紙張病歷建檔的資料庫。此資料庫可以完整的收集透析相關資料內容。本研究資料收集來源與內容為某地區醫院所有血液透析病人共212人之所有透析前(Before-dialysis)、透析期間(包括HD與PD)、透析後(After-dialysis)紙張病歷之就醫記錄與透析期間臨床透析醫療作業記錄,其內容包括病患基本資料、疾病史、門診、急診、住院就醫記錄、檢驗與檢查資料、處方用藥記錄、營養衛教、血管通路評估、一般人工量測、設備材料維護、等全部登錄於所設計的資料庫內。經資料庫之檢驗資料截取,篩選出87位病患之首次透析前一年就醫檢驗資料有763筆、23個檢驗變項,做為研究目的之資料轉化內容。 檢驗資料分析結果,以χ2,篩選出透析前臨床檢驗值有BUN、CR、NA、K、CA、WBC、RBC、MCV、MCHC、PLT共10項可供醫療決策之參考變項。經判別分析方法驗證資料轉化,發現原始檢驗值比轉化後的檢驗值好,其可能原因是,人體之生理值雖在正常範圍內,但檢驗值偏高或偏低之趨勢,仍對醫療診斷之判斷有助益。
Medical data has varied format because of its complex nature. Subjective selection and application could be biased, and then affect medical decision. Therefore, appropriate data format selection is necessary for better application. Categorical data can be helpful while making decision. Patients’ clinic examinations data before dialysis is one of main factors and important criteria for kidney disease treatment and prevention. The purpose of this study is to search for an appropriate data transformation method that can be used to analysis clinic examination data before dialysis for better medical decision-making quality. Before starting data transformation research, a patient centered Electronic Dialysis Medical Care System (HD-PD-DBS) was designed and constructed. This database can transfer current paper based medical record and collect all patients’ related data completely. Currently collected data contains 212 patients’ 1 year before dialysis, during dialysis and 3 year after dialysis medical care data, which includes personal information, medical history, outpatient medical records, emergency medical records, inpatient medical records, examination data, medications, nutrition and education, etc. For the proposed research, 87 patients’ 1 year before dialysis examinations data with 763 records and 23 variables were abstracted from HD-PD-DBS. In results, 10 selected variables (BUN, Cr, Na, K, Ca, WBC, RBC, MCV, MCHC, PLT) using Chi-square criteria can be defined as important examination item for dialysis treatment decision-making. Multi-data analysis methods (eq. Canonical correlation coefficient; Discriminant analysis correct classification rate) find that the original examination data better than transformed data in this study. It may because physiology examination data tells only less information when only normality and abnormality are marked. The data trend maybe important and contains more information about the disease, therefore is useful to help to make better medical decision.