經由天然或人為災難所引發的大型傷患事件回應為現今緊急醫療回應系統之重大挑戰之一。在進行大型傷患事件回應決策時,必須即時地決定如何將傷患由災點救出、運送至醫院,以最快獲得醫療救援。針對此問題本研究提出一隨機動態模式,其同時可以考慮傷患嚴重性等級隨機性、傷患治療隨機時間,以決定一組較佳的救護任務派遣計畫。此模式為一辭書式多目標模式,傷患嚴重性等級越高其獲得緊急醫療救護的優先性等級就越高,尋求每一傷患等級的總傷患回應時間最小化並按照傷患嚴重性等級分級進行之。所謂的傷患回應時間包括傷患於災點等待救援時間、傷患運輸時間、傷患於醫院等待救護時間以及傷患治療時間。假設傷患嚴重性等級隨機性具有馬可夫鏈特性,故可根據這一期的傷患嚴重性等級預測下一期的傷患嚴重性等級,將此資訊導入求解大型傷患事件回應決策的啟發式演算法中,可以避免產生短視的決策。此外,受限於大型傷患緊急救護回應真實問題的狀態空間及決策空間維度過高,故本研究發展一模擬基礎式近似動態規劃演算法求解此大型傷患事件緊急醫療回應隨機動態模式。最後,藉由數個數值測試例評估與驗證此近似演算法的有效性。
Response to mass casualty incidents (MCI) caused by natural or man-made disasters is one of the greatest challenges to medical emergency response systems (MERS). During the emergency response to mass casualty incidents decisions relating to the extrication, transporting and treatment of casualties are made in a real-time, sequential manner. In this thesis, a novel stochastic dynamic programming (SDP) model of this problem is proposed. The stochastic nature of casualty health and treatment time are considered to determine ambulance dispatches assignment. The model is of a multi-objective nature, utilizing a lexicographic view to combine objectives in a manner which capitalizes on their ordering of priority. That is, injuries of higher level of severity have higher priority. Each objective is to minimize the total response time of casualties at the specific level of severity, including waiting times at emergency sites, transportation times, waiting times at hospitals, and treatment times. The uncertainty follows “Markov chain” properties in which the correlations of the variations in the consecutive periods are high and the severity status of casualties in next period is stochastically determined by the present one. These decision results can lead the course of the response operation, thus avoid the myopic decision making which could result from the use of a sequential, heuristic decision making process. Because of the size of the state and action spaces for realistic problems. A simulation-based approximate dynamic programming algorithm is developed to solve the proposed SDP model. The model is evaluated over several potential problems, with results confirming its effective nature.