應用人工智慧於急重症病患之轉院評估研究

急重症病患指的是急診檢傷分類中，嚴重度達到一至二級的傷重患。病況急需加護病房照顧或須立即實施緊急手術之傷患，是屬於緊急救治中最為急迫，最為需要醫療照顧的病患。當加護病房滿床或無法立即接受手術治療時，急重症病患必須進行轉院治療。轉院治療是為了病患的生命著想，但是往往忽略到患者的安全考量。對於急重症病患的轉院規範與效率，關係到醫院的醫療品質、院方及醫師的效率、病患及家屬的滿意度、甚至於關乎到急重症病患的生命安全。因此，如何有效的提升急重症病患轉院之品質，將是非常重要的議題。本研究乃是以急診急重症病患作為研究對象，與雲林地區某個案醫院進行合作，運用人工智慧的技術，將急重症病患進行分析，從流程架構、參數選取、資料處理，建構出一個「急診急重症病患轉院之評估系統」。本研究藉由案例式推理建構系統之介面，同時運用類神經網路決定案例式推理中各因子之最佳權重。研究結果顯示，以案例轉入醫院名稱顯示，於五筆案例內之準確率為82.35%；而以醫院權屬別區分之下，其正確評估之準確率則為82.94 %。期望此系統可以提供醫師做為臨床之輔助評估的參考依據，對於醫院、醫師、重症病患及家屬將有實質的助益。

關鍵字

急重症病患；病患轉院；類神經網路；案例式推理

並列摘要

Critical-Emergency patient is indicated that the patient maimed degree of aggravate that calibration first to second. The patient needs to take care in intensive care unit and administration of emergency surgery. Above-mentioned patient was the most imminently and attention in cure. The patient had to referral when the intensive care unit was overflowing or could not have a surgery immediately. Referral the hospital treatment is considers for sickness life, but often neglects to patient's safe consideration. Regarding the Critical-Emergency patient extension standard and the efficiency of the hospital, relate the medical quality of the hospital, and the hospital efficiency, patient and family member's degree of the satisfactions, and even closes to the Critical-Emergency patient’s safety. Therefore, how to promote the effective to the Critical-Emergency patient does referral research to the hospital, will be the count for much subject. This study uses the Critical-Emergency patient at a hospital in the Yunlin County as its case research subject. It puts the Critical-Emergency patient into categories from the most severely ill to the least urgent using artificial intelligence technology. From the procedure structure, parameter selection, and data processing, it establishes a model ‘Emergency patient in acute disease to evaluate system of the referral’. This research uses the construction system based on CBR interface and applies Neural Network to decide the optimal-weighted for CBR factors. The result showed that the accuracy of transferring hospital display names within the five cases rate is 82.35%; the accuracy of the evaluation based on the discrimination of hospital ownership rate is 82.94%. This research expects the system could provide a reference point to evaluate clinical for doctors, and it would be practically beneficial to hospitals, physicians, Critical-Emergency patient and their family.