空間分析輔助緊急醫療資源派遣決策之研究

The Study for Dispatch Decision of Medical Emergency Resources with Real-Time Spatial Analysis

高瑞鴻(Jui-Hung Kao)

指導教授 ： 歐陽彥正

共同指導教授 ： 孫維仁(Wei-Zen Sun)

國立臺灣大學/電機資訊學院/生醫電子與資訊學研究所/博士(2018年)

https://doi.org/10.6342/NTU201800739

摘要

過去研究指出，需要急救患者發生到院前呼吸、心跳停止之狀況至抵達醫療院所過程中，所接受救護車上緊急處置與最救護資源分派之有效性為病患最主要的存活因子。
目前救護車標準分派程序為，救護指揮中心( Emergency Command Center)之緊急救護派遣員 （Emergency Medical Dispatch, EMD）依據當下有限的資訊與經驗直覺進行分派。然而，過去研究顯示到院前心跳停止(Out of Hospital Cardiac Arrest, OHCA)病病患的年齡、種族、發生時間、救護人員的救護能力和送醫過程…等，都影響病患的預後與存活率。以此思之，目前極少關於利用大數據與機器學習為基礎之智慧型即時性空間分析輔助緊急醫療資源派遣決策系統研究。
本研究計劃結合地理區域化緊急醫療資源資訊與救護車地理資訊收集資料庫，評估OHCA病患最有利之緊急醫療處置模式，提供例如:最近醫療院所、急救責任醫院（First-Aid Capabilities Hospital）等級、最適處理科別與分科、醫院床位資訊、特別分科需求…等建議選項以供緊急救護派遣員決策之參考。
此外，在緊急救護客觀策略提供上，本研究認為可透過大數據分析找出最佳化的人口數、區域發展程度，對緊急醫療資源分配，將結合RVKDE其換算空間涵蓋區對OHCA病人存活率之影響，或者透過增加EMT-P的空間配置與機動性，都可反映空間因素對於緊急醫療資源分配問題中的重要性，其研究結果可提供醫療資源分配之政策決議參考。
我們回顧新北市「救護指揮中心」於2010年至2011年收集發生到院前心跳停止之病患的收案案例，研究結果說明患者到院前所發生的各種狀況。本研究在研究方法上分為兩個研究進行，在研究一部份，首先採用國際通用「烏特斯坦格式」標準（Utstein Style)進行資料萃取，接著利用迴歸分析]瞭解這些變因與OHCA兩小存活率之間的關係並萃取OHCA建模的風險因子(Risk Factor)，找出OHCA風險因子係數，在研究二部份，將「OHCA空間風險因子」分析結果與「緊急醫療資源區域化數量」因子，形成空間衡量指標，做為測量不同地區人口需求，獲得醫療資源服務的數量，進而產生「緊急醫療資源區域化評估與設置最佳模式」之決策建議。
本研究將以OHCA病患為研究對象，同時將「Relaxed Variable Kernel Density Estimator」、「緊急醫療資源區域化評估與設置最佳模式」結合，此分析方式，將有助於「緊急救護派遣員」在OHCA病患到院前提供最佳醫療資源分配決策擬定，進而根據不同OHCA病患擬定最佳化救護派遣策略，本研究模型可提供醫療急救領域，實用的且有效率之分析工具，其分析結果將有助於救護派遣時參考依據。

關鍵字

緊急醫療 ； 資源管理 ； 最佳化 ； 決策支援 ； 資訊系統

並列摘要

According to previous research, the most important factor for a patient’s survival is the emergency treatment in the ambulance and the effective allocation of emergency resources when emergency patients experience out-of-hospital cardiac arrest (OHCA) before arriving at the hospital.
The current procedure for deciding which hospital a patient is sent to is followed by the emergency medical dispatch (EMD) sent by the emergency command centre, and they make decisions based on limited information and previous experience. Previous research indicated that the age and race of OHCA patients and the timing and rescue capacity of EMD in addition to the process of being taken to the hospital would impact a patient’s recovery and survival rate. However, very few researchers conducted studies on real-time spatial analysis of emergency medical resource allocation and dispatch based on big data and machine learning.
This study intended to combine the regional information of emergency medical resources and the data base of geographic information for ambulances to assess the best method of medical treatment for OHCA patients, such as the closest medical centre, the level of first-aid capabilities, the proper medical department, the information for hospital beds, and the special needs of the medical department so that the EMD can make decisions based on these suggestions.
To provide a strategy for an emergency ambulance service, this research proposed using big data analysis to examine the impact on the survival rate of OHCA patients based on the following features: the optimized population, level of regional development, allocation of emergency medical resources and the optimal execution time for the chain of survival and advanced cardiac life support. Additionally, through increasing the spatial allocation and flexibility of emergency medical technician-paramedic (EMT-P), it can reflect the importance of spatial factors in the challenge of allocating emergency medical resources. The results of this study can provide recommendations for a policy of medical resource allocation.
We evaluated the cases of OHCA patients between 2010 and 2011 collected by the ambulance commend centre in New Taipei City, and the research results showed patients’ different medical conditions before arriving at the hospital. This study utilized two types of research methods. First, the study used the Utstein style to conduct data extraction and adopted a regression analysis to explore the relationship between these factors and on the OHCA patient’s survival rate, extracting an OHCA patient’s risk factor and calculating the coefficient of risk. Then, we utilized the method of decision trees to establish the OHCA’s risk decision model, adding risk factors with coordinates, and formed the OHCA spatial risk factors. In the second part of the research methods, we combined OHCA spatial risk factors with the regionalized quantity of emergency medical resources and spatial distribution of first-aid capability hospitals as well as the demand for medical resources and formed the spatial matrix to evaluate the demand for medical services. Finally, a policy recommendation for the optimized region and an allocation strategy for emergency medical resources were made based on the matrix.
The object of this research is to study OHCA patients in this context. The research method combined a relaxed variable kernel density estimator with the optimized region and allocation for emergency medical resources so that the results can assist EMD in making the proper medical decision before arrival at the hospital. In addition, EMD can formulate the optimal strategy for ambulance dispatch. The model built in this research can provide a practical and efficient analytic tool for the medical emergency field and its results will provide reliable references for emergency dispatch.

並列關鍵字

Emergency Medical ； Resource Management ； Optimization ； Decision Support ； Information System

參考文獻

REFERENCE

[1] Y.-J. Syue, Y.-L. Yen, S.-Y. Cheng, C.-W. Hung, K.-H. Wu, and Y.-R. Lin, "Characteristics and Risk Factors of Out-of-Hospital Cardiac Arrest Within 72 Hours After Discharge," The American journal of the medical sciences, vol. 350, no. 4, pp. 272-278, 2015.

[2] E. Markusohn et al., "Primary percutaneous coronary intervention after out-of-hospital cardiac arrest: patients and outcomes," Hypertension, vol. 12, p. 48, 2007.

[3] D. A. v. Hoeijen, M. T. Blom, A. Bardai, P. C. Souverein, A. d. Boer, and H. L. Tan, "Reduced Prehospital Survival Rate After Out-of-Hospital Cardiac Arrest in Patients with Diabetes Mellitus Type 2: A Prospective Community-Based Study," Circulation, vol. 130, pp. A246-A246, 2014.

[4] H. Davies, A. Loosely, S. Dolling, and R. Eve, "Predicting survival in patients admitted to intensive care following out-of-hospital cardiac arrest using the Prognosis After Resuscitation score," Critical Care, vol. 18, no. Suppl 1, p. 491, 2014.