全民健保的財務規劃原本是以自給自足為原則,而在1995年全民健保實施以來,卻屢傳財務失衡的危機,隨著健保費用不斷的高漲下,為了避免醫療資源的浪費,有效的評估醫療資源分配成為一項重要的評估性指標。 肝癌是全世界五大常見的惡性腫瘤之ㄧ,台灣亦是肝癌高盛行區,近幾年的研究顯示,全球各地區肝癌發生率皆有上升的趨勢,其治療所花費的資源與費用比例是相當可觀的,因此,如何有效評估住院天數對於醫療資源分配與醫院管理是一個重要的議題且具挑戰性的工作。 本研究主要挖掘影響肝癌手術的相關因子來做探討,以健保資料庫中肝癌之患者為研究對象,運用人工智慧法中的粒子群最佳化演算法、倒傳遞類神經網路、支援向量機、分類迴歸樹、案例式推理等有效的評估肝癌手術患者的住院天數與醫療費用,以提供相關醫療人員參考依據。研究結果顯示,住院天數整體而言,除了單分類迴歸樹,其他準確度皆高於85%以上,其中以粒子群最佳化演算法結合倒傳遞類神經網路最高,為96.78%,在醫療費用準確度也達87.86%,結果顯示系統準確度之可行性,以提供醫師作為臨床診斷輔助參考,對於醫院醫療資源分配與病患家屬後續照顧上更能有效的控管。
The originally enactment planning for the national health insurance was based on the income-costs self-sufficiency principle. Since the promulgation of the national health insurance act in 1995, it has always fallen into the crisis and financial imbalance and deficits. Subsequently, the avoidance of medical resource wastes integrated with the effective estimation of the medical resource apportionment has thus become one of important evaluation indicators. Especially both the domestic and foreign researches all show that the length of stay for the surgery will directly influence the total medical costs, meanwhile. The hepatocellular Carcinoma ranks the top 5th common malignant tumor globally amid Taiwan as a high-prevalent area unexceptionally. The researches in the past few years indicate that the global occurrence hepatocellular Carcinoma rates all remain in an ever-increasing tendency. The medical resources and medical treatment cost spent for such the hepatocellular Carcinoma are significantly tremendous. Such a situation, therefore, constitutes an important issue and a challengeable mission for how to evaluate effectively the length of stay for the surgery and medical resource appropriation and medical institution management. With targeting on exploring all the relevant variables influencing the liver surgery, it has been adopted in this research some effective evaluation algorithm belonging to the Artificial Intelligence Method including the Particle Swarm Optimizer, Back-Propagation Neural Network, Support Vector Machine, Classification and Regression Tree, and Case-Based Reasoning for trying the overall average hepatocellular Carcinoma surgery patients’ length of stay (number of days), and medical treatment costs for providing relevant medical professionalism for evaluation basis and reference. The research results indicate that except the Classification and Regression Tree Algorithm, all other evaluation methods demonstrate precision rates more than 85% among which the portfolio of Classification and Regression Tree and Back-Propagation Neural Network achieves as high as 97.86%, ranking the top 1, and also a precision rate of 87.86% for analyzing medical treatment cost. Having reaching such so high feasibility, all those research results validate the adopted systematic precision rates can provided physicians auxiliary reference data for clinical diagnosis and assist the medical institutions in applying more effective control measures in medical resources apportionment and for help patients’ family conducting follow-up caring.