全民健康保險支付制度其主要目的是提供人民獲得公平的醫療服務,提升全體國民健康品質。由於健保給付制度不健全,造成醫療資源過度浪費,財務的失衡嚴重虧損。隨著健保費用不斷攀升,避免不必要的醫療資源浪費,有效的評估醫療資源分配為一項重要的評估性指標。胃癌是我國常見的惡性腫瘤之一,它在消化器官癌症中排名第四,是僅次於肝癌、結腸癌及口腔癌的惡性腫瘤。透過統計數據顯示,胃癌死亡率雖然有下降趨勢,然而胃癌的可怕性仍是不可輕忽的,其診斷與治療所花費的資源費用比例是相當可觀的,如何有效評估住院日與醫療費用將是一個重要的議題。本研究主要探討影響胃癌手術之相關因子,透過參考醫療相關文獻與醫師諮詢,篩選出相關影響變數,以健保資料庫中胃癌病患為研究對象,並且運用決策樹、倒傳遞神經網路、支持向量機、粒子群最佳演算法、案例式推理於評估術後住院日與醫療費用,以提供相關醫療人員作為參考之依據。研究結果顯示,針對住院日方面,決策樹結合支持向量機在測試準確度及ROC曲線面積下皆有不錯的表現,分別為81.48%、54%;費用部分,模型準確度為83.33%,本研究利用資料庫系統並設計使用者介面,以提供醫師臨床評估之輔助工具,減少不必要的資源浪費,提高醫療服務品質與效率。
The national healthcare insurance was initially intended for providing the people equal access to medical treatment and enhancing the health quality of the commonwealth. Improper disbursement of the national healthcare insurance has already resulted in excessively waste of medical resource and deteriorated the financial revenue. The increasingly soaring national healthcare expenses synchronize with unnecessary medical resource waste. The medical resource utilization would be an important index of evaluation. The gastric cancer is one of the most frequently-incident malicious tumors and ranks the top 4 among all the digestive-organ cancer, next to only the liver cancer, colorectal cancer and oral cancer. The domestic and foreign medical research indicate that the hospitalization duration varies directly with the medical expenses. It become an important issue as how to effectively evaluate the length of stay and medical expenditure. This study investigates the factors related to gastric cancer surgery, by reference to the medical literature and consult with physician, screening the relevant impact variables of Gastric cancer patients in the health care database for research, The decision tree, artificial neural network, support vector machine, particle swarm optimization and case-based reasoning have been applied in this research to evaluate the length of stay and medical expenditure. It is expected that the results could provide the relevant medical staff as a basis for reference. The results show that the terms of length of stay, the decision tree combined with support vector machine offers better performance in accuracy and areas under the ROC, which is 81.48% and 0.54 respectively, for the medical expenditure, the system accuracy is 83.33%. In this study, the database system was applied and the user interface was designed to provide the physician an assistant tool for clinical evalution. It is expected that the results could reduce the unnecessary waste of resources and enhance the quality and efficiency of medical services.