應用資料探勘技術對肩關節攝影檢查鑑別診斷差異性之探討――以某區域教學醫院為例

A Study on the Differential Diagnosis of Shoulder Arthrogram with the Application of Data Mining - Taking a Regional Hospital for an Example

在2002開始實施總額的制度之下能夠使用的額度受到限制，在臨床執行醫療業務上也產生排擠效應，如何讓檢查更有效的運用，使得有限的額度能照顧到更多需要的病人將是各個醫院努力的目標。本研究統計2008~2011年，個案醫院在骨科門診的病人量及放射線科進行肩關節攝影之病患，研究病患數約在1,000人左右。在資料分析方面分為二部份，第一部分依醫院評鑑檢查之陰性率將閾值設定在50%，在鑑別診斷準確性的研究得到結果一致性為57.7%，達到所設定之閾值，而伴隨出現次數最多的則是與脊椎相關疾病之診斷碼；在第二部份邏輯迴歸分析，經由分析結果找出關聯性最大的診斷碼分別是715.9和727.61與715.9和840.4二組，研究分析所得到的結果與臨床醫師在臨床的診斷的邏輯是一致的。

After the implementation of the so-called total amount policy in 2002, the access of all available funds is further tightened up. This makes various clinical medical practices feel the squeeze effect from each other. How to make more effective use of medical checks and to employ limited amount of allowance to take care of more patients in need is now a major goal of efforts at each and every hospital. In this study we looked into the statistical data from 2008 to 2011 of the case hospital, of the total number of orthopedic outpatients, and approximately 1,000 of those having had their shoulder scanned by arthrography radiology. The data analysis is divided into two parts. The first part was based on the negative rate of hospital accreditation to have the threshold set at 50%. Then in a differential diagnostic accuracy study we got a consistent result of 57.7%, surpassing the threshold set, which was accompanied most frequently by diagnostic codes associated with spinal related diseases. In the second part of the logistic regression analysis, we identified the most relevant results analyzed by the diagnostic code to be two groups, 727.61 with 715.9 and 715.9 with 840.4. By the way, the results obtained in our analysis turned out to be consistent with the clinical diagnostic logic by the clinicians.