【作者】 张雪婷；

【导师】 于国栋；

【作者基本信息】 山东大学 ， 管理科学与工程， 2023， 硕士

【摘要】 近年来,以“上车即入院”为特征的新型5G急救模式对于打造端到端的应急救援体系,缩短急救需求响应和入院交接时间具有重要价值和意义。它通过5G通讯技术手段与传统救护车融合,搭载新型移动式医疗检查和监测设备,可以实现病人病情的快速诊断和信息实时传输。以实现最小化病人等待时间和病人生命损失为目标,学术界关于院前急救网络设计和院内医疗资源规划配置的研究成果丰硕。但现有研究大都针对院前急救与院内急诊治疗两个孤立的系统展开,很少考虑救护车的规划部署以及院内专科医疗能力的协同规划问题,这可能制约急救需求与应急响应能力的匹配效率的提升,进而造成较长的需求响应和交接等待时间,延误急救病人救治。鉴于此,考虑到越发普及的5G救护车部署,以及新型急救模式带来的良好效果,本文将对5G联网救护车条件下的院前院内急救资源协同规划问题展开深入研究。现实中,考虑到院前急救调度过程中存在两个比较严重的问题:一方面,近年来全国各地救护车空跑现象严重,救护车资源存在严重的浪费,严重威胁到急症病人的生命安全。目前提出的“5G急救分级调度系统”在呼叫达到时口头判别危急重症病情的方法主要依靠调度指挥人员的主观判断,无法从根本上解决救护车资源浪费的问题。另一方面,最新的急救《条例》明确指出了病人具有自行选择医院的权力,在实际救援过程中病人上了救护车后既可以服从医护人员的指示到达指定的医院,也可以提出要求去自身偏好的医院。因此,如何在减少空车浪费和满足病人的选择行为下联合优化院前和院内的急救资源是一个具有挑战性的问题。为此,本文基于排队理论和多项Logit用户选择理论,基于院前封闭排队网络流和院内医疗能力单排队服务模型分析急救网络流量,然后构建考虑不确定需求发生场景下的两阶段随机规划院前院内资源联合规划模型,实现在未知的需求发病情况和空车场景发生时,规划的院前院内资源能够满足院前的总等待延迟时间和响应损失成本以及入院的排队等待延迟时间和排队失效损失之和最小。具体地,本文首先描述院前封闭排队网络流和院内排队视图,针对需求发生的不确定性构建三个两阶段随机规划模型:第一个是基本的端到端的两阶段随机规划的系统优化模型,目标是最小化资源部署成本和相关的系统运营成本,并实现事前资源规划和事后稳定状态下救护车响应和二次响应以及病人入院分配的决策;然后考虑跟旅行时间、排队等待延迟和排队失效概率有关的病人的择院选择行为并构建了纳入多项Logit选择模型的端到端急救资源用户均衡选择双层规划模型;之后考虑5G医联体情况下医疗资源在医院之间存在共享的院内资源协同规划模型。由于本文构建的多阶段非线性模型属于NP-Hard问题,直接求解非常困难,而现有的研究大多采用启发式算法或者近似方法来获得近似解。因此本文一共运用了三种算法求解:首先使用分段逼近算法对非线性模型进行近似求解并分析近似误差;然后本文根据模型特征使用分段线性松弛子问题并嵌套经典的Benders分解算法框架加速求解;接着结合上述两种算法特点通过构造增强的线性松弛子问题设计了一种非凸-嵌套广义Benders分解算法用于多阶段非线性非凸模型的高效精确求解,能够有效提升院前院内资源联合规划的决策效率。最后,本文通过济南市五区的实际需求进行案例分析,对所提模型和算法进行有效性验证,通过算法性能分析发现广义嵌套Benders分解算法与另外两种算法相比在求解速度和求解质量上都有明显的优势,再通过模型效果分析和灵敏度分析说明了设计的三个模型的有效性和可靠性。因此证实了本文所提模型和算法对于5G救护车联网使能背景下的院前院内资源协同规划研究具有重要意义。更多还原

【Abstract】 In recent years,the new 5G emergency mode characterized by "admission on board" is of great value and significance for building an end-to-end emergency rescue system and shortening the emergency demand response time and admission handover time.It is integrated with traditional ambulances through 5G communication technology,and is equipped with new mobile medical examination and monitoring equipment,which can realize the rapid diagnosis of patients’ conditions and real-time transmission of information.With the goal of minimizing patients’ waiting delay and life loss,the academic community has achieved fruitful results on the design of pre-hospital emergency network and the planning and allocation of in-hospital medical resources.However,the existing research mostly for two isolated system of pre-hospital emergency and in-hospital emergency treatment,rarely consider the planning of ambulance and hospital specialist medical ability of collaborative planning,which may restrict the emergency demand and emergency response ability matching efficiency,thus cause longer demand response and handover time,and emergency patient treatment delay.In view of this,considering the increasingly popular deployment of 5G ambulances and the good effects of the new emergency mode,this paper will carry out an in-depth study on the collaborative planning of pre-and-in hospital emergency resources under the condition of 5G connected ambulance.In reality,there are two serious problems in the process of pre-hospital emergency dispatch.On the one hand,in recent years,the phenomenon of empty ambulance running is serious,which seriously threatens the life safety of acute patients.At present,the proposed "5G emergency hierarchical dispatching system"method of verbally judging the critical condition when the call is reached mainly depends on the subjective judgment of the command personnel,and cannot fundamentally solve the problem of waste of ambulance resources.On the other hand,the latest emergency regulations clearly point out that patients have the right to choose the hospital by themselves.In the actual rescue process,the patient can follow the instructions of the medical staff to arrive at the designated hospital,or request to go to the hospital with their own preferences.Therefore,how to jointly optimize pre-and-in hospital emergency resources while reducing empty vehicle waste and satisfying patient choice behavior is a challenging problem.To this end,based on the queuing theory and the multinomial logit user-choice theory,this paper constructs a network traffic model based on the pre-hospital closed queuing network flow and the in-hospital medical capability single-server queuing model,and then build a two-stage stochastic programing model for joint planning of pre-and-in hospital resources under the demand occurrence uncertainty.In the event of unknown demand onset and empty ambulamce scenario,the planned pre-hospital resources can minimize the sum of the total pre-hospital waiting delay time and the queuing delay time and the in-hospital queuing failure loss.Specifically,this paper first describes the pre-hospital closed queuing network flow and the in-hospital queuing view,and constructs three two-stage stochastic planning models for the uncertainty of demand occurrence.The first model is the basic end-to-end stochastic system-optimal model,which aims to minimize the resource deployment costs and related system operation costs and realize the here-and-now decisions of resource planning and wait-and-see decisions of ambulance initial dispatch and secondary dispatch and patients admission allocation under steady state;The second model is to consider the hospital choice behavior of patients related to travel time,queuing delay and the probability of balking and construct an end-to-end user equilibrium choice bi-level planning model incorporated with the multinomial logit choice model.The third is to consider the pre-hospital and in-hospital resource collaborative planning model in the case of inter-hospital sharing of medical resources in 5G medical alliance.Since the multi-stage nonlinear model constructed in this paper belongs to NP-Hard problem,which is very difficult to solve it directly,and most of the existing studies use heuristic algorithm or approximate method to obtain approximate solutions,so the following three algorithms are proposed in this paper.Firstly we use piecewise linear approximation algorithm to solve these models approximatively and analyze the approximation error.Then designs the classical benders decomposition algorithm using the linear relaxation subproblem,and then further designs a non-convex nested generalized benders decomposition algorithmuse using the enhanced linear relaxation subproblem method according to the model characteristics to solve the model efficiently and accurately,which can effectively improve the decision-making efficiency of the joint planning of pre-and-in hospital resources.Finally,this paper makes a case analysis of the actual demand in Jinan city,applies the above proposed models and algorithms and verifies their effectiveness.It can be found that the nested generalized benders decomposition algorithm has obvious advantages in the solution speed and the solution quality compared with the piecewise linear approximation and the relaxed benders decomposition algorithms through the algorithm performance analysis,and also the validity and reliability of the three models are illustrated through model effect analysis and sensitivity analysis.Therefore,the proposed models and algorithms in this paper are of great significance for the collaborative planning of pre-and-in hospital resources in the context of 5G ambulance networking enabling.更多还原