| With an aging population,the number of outpatient visits is increasing year by year,and the shortage and uneven distribution of medical resources in China has put tremendous pressure on our healthcare system.As the first step in providing medical services to patients,the resource utilization and service efficiency of outpatient services have a great impact on the patient’s medical experience.With the promotion of the "Health China 2030",medical institutions at all levels are trying to optimize the operation of the outpatient service information platform,trying to schedule patients in advance through online appointments to achieve reasonable allocation of limited medical resources,improve the efficiency of outpatient services,and thus improve the patient’s medical experience.However,in the current outpatient service practice,appointment scheduling is still manually operated mainly by empiricism and lacks effective automated and dynamic decision-making methods,resulting in a large amount of actual service process data accumulated in outpatient clinics not being scientifically utilized.Therefore,the optimal decision support system for outpatient appointment scheduling service still has much room for improvement.At the present stage,because the medical habits of outpatients in China have not yet been completely changed,multiple types of patients such as advance appointment,on-site appointment and same-day patients coexist,and the outpatient clinics of large hospitals in China are generally composed of multi-level doctor resources such as senior doctors and junior doctors,taking into account the random behavioral factors existing in the arrival process,registration process and waiting process of various types of patients,as well as the random behavioral patterns of doctors in the process of receiving consultations etc.,it is crucial to accurately capture these key behavioral mechanisms that have a significant impact on outpatient service efficiency and incorporate them into the outpatient appointment scheduling decision optimization system.Under the State Council’s proposal to promote the development of medical big data and the new requirements of the 14th Five-Year Plan of the national medical and health service system,how to mine and analyze the accumulated data information of outpatient clinics,while taking into account the complex random behaviors(and adaptive behaviors)of multiple levels of doctors and patients,and propose a medical outpatient appointment scheduling system that is applicable to the current stage of medical consultation in China?It is a hot issue of national importance and people’s livelihood that needs to be solved,and it has attracted a lot of attention from all walks of life.This dissertation takes the outpatient appointment system of several large tertiary general hospitals in China as the research object,investigates the realistic medical outpatient scenario in China,collects a large amount of historical data of outpatient service process at the same time,and systematically summarizes and composes the frontier research literature in this field,focusing on the following problems in the optimization research of the current medical outpatient appointment scheduling system:1)In terms of optimization scenarios,the existing studies have not taken into account the queuing network of outpatient services composed of multi-level doctors,the uneven utilization of doctors’ resources among different levels,the limited waiting patience of patients leading to the abandonment,change of numbers or even departure of patients in the middle of the waiting process,and the resulting waste of medical resources and patient dissatisfaction.2)In terms of utilization of data resources:the existing studies have not considered the behavioral patterns of doctors and patients,nor have they verified,interpreted and revealed their behavior through data to verify,explain and reveal their behavioral mechanisms;the outpatient appointment scheduling optimization model thus established cannot reflect the impact of dynamic random behaviors of doctors and patients(such as patients’ preferences,patients’ patience and how doctors’ treatment speed is adjusted,etc.)on system performance and optimization decision results;3)In terms of decision methods,most of the existing studies adopt traditional queuing theory and mathematical planning decision methods to solve the problems of long lead time,For dynamic appointment scheduling optimization problems with complex scenarios(multi-stage,multi-doctor,and multiple non-deterministic coexistence),doctor-patient behavior adaptation,and time-varying consultation status,more flexible modeling and optimization algorithms that can cope with the explosion of data dimensions are needed.To address these shortcomings,this study focuses on the complex stochastic behaviors of both doctors and patients in the outpatient service process.By mining and analyzing the data from the Hospital Information System,we verify the behavioral patterns of patients’preference for high-level doctors and their patience in response to the system state,and reveal and explain the behavioral mechanisms of doctors’ service speed in response to workload and the behavioral mechanisms of interactions between doctors’ work efficiency in a complex queuing network composed of multiple levels of doctors.The outpatient appointment scheduling service system is modeled by a discrete-event simulation method on the basis of fully considering the factors of stochastic behavior of multiple types of patients in the outpatient system of two levels of doctors,and a multi-objective simulation optimization algorithm combining simulation budget allocation and genetic algorithm is used to obtain an approximate optimal scheduling scheme;on the basis of fully considering the factors of doctor-patient adaptive behavior within the multi-level doctor service system,a Markov decision process is modeled for the outpatient appointment scheduling service system,and a policy iteration algorithm with a deep neural network approximating the value function of dynamic programming is used to obtain the approximate optimal scheduling strategy,and finally,the policy comparison and numerical experimental analysis are used to provide decision support and policy recommendations for decision makers.The main work of this dissertation includes.(1)Data modeling on patient choice preferences and patience-limited behavior patterns in medical outpatient clinicsCluster analysis,logistic regression analysis and survival analysis were used to explore the behavioral characteristics of patients’ choice preferences and the expression and measurement of patients’ waiting patience limits in two types of levels of medical resources;a fitting formula for the probability of patients’ survival(whether to abandon from the system)during the visit was established to reveal the behavioral mechanisms of patients’choice,waiting and abandonment in outpatient clinics.It was found that patients’ choice of different types of outpatient services was uneven,and patients with different demographic characteristics had relatively clear preferences for visit duration and doctor title;patients’patience limits for waiting before exiting midway generally fluctuated with the average congestion of the system.By comprehensively analyzing and summarizing the behavioral factors and behavioral characteristics of patients at each stage of receiving outpatient services,the research findings provide ideas for improving patient satisfaction in outpatient appointment scheduling services,and also provide strong data support for scenario modeling of subsequent outpatient appointment scheduling optimization problems,especially considering the preferences,patience limits and abandonment behaviors of multiple types of patients,which helps optimize resource allocation more effectively.(2)Analysis of the mechanism of adjusting outpatient doctor service speedIn the classical queuing theory literature,a server is usually assumed to work at an exogenous and constant speed.The adaptiveness of doctors’ working speed is hardly considered in the traditional outpatient appointment scheduling service system,but outpatient practice and literature research show that doctors’ service speed is largely affected by workload and scheduling situation.In this dissertation,we investigate the adaptive mechanism of workload on outpatient multilevel doctor service speed and the impact on system performance metrics in the context of workload and scheduling adjustments through empirical tests and queueing theory analysis,respectively.Based on the summary of the literature,an empirical analysis framework that considers the mechanism of interaction between doctors of different levels and familiarity levels is proposed,and the hypothesis that outpatient doctors rely on workload and team familiarity to adjust service speed is validated and analyzed by actual data from outpatient clinics in partner hospitals.Based on the results of the empirical analysis,an analytic queuing model close to the realistic scenario in China is considered;the model is a multi-server Erlang queuing system with heterogeneous servers,non-homogeneous customers,and limited memory space,and its basic characteristic values and performance metrics of the queuing system when the number of doctors is 2 and 3 are derived from the queuing theory perspective.It is found that the steady-state patient number distribution in the system depends not only on the average of the service time but also on the joint distribution function of the number of patients and the service time;the doctors adjust the service rate by estimating the workload;and the overall efficiency of the system is higher when the team size is large and the familiarity among doctors is high when the team doctors collaborate with each other.The results not only reveal the workload-dependent service time regulation mechanism and demonstrate the importance of more accurately capturing doctor behavior mechanisms in a multi-level resource clinic environment,but also,more importantly,evaluate and quantify the adaptive behavior of multi-level doctors in medical outpatient clinics,provide a reference theoretical basis for the management and standardization of the outpatient service process,and provide reliable data for the subsequent dynamic planning of outpatient appointment scheduling.(3)Multi-objective simulation optimization of medical outpatient appointment scheduling considering multiple types of patients’ selection preferences and abandonment behaviorsIn medical outpatient services,uneven utilization of medical resources at different levels,inefficiency and patient dissatisfaction have become major problems for hospital managers due to patients’ selection behaviors for different levels of doctors and different consultation times.To address this problem,for the first time,patients’ choice preference behavior for junior doctors and senior doctors is considered.Through the data analysis modeling,patient selection preferences and abandonment behaviors that seriously affect outpatient service efficiency and patient satisfaction are considered within the appointment scheduling optimization problem,and a data-driven Discrete Event Simulation(DES)model is established by using outpatient process data fitting as the actual data support through the Discrete Event Simulation method model.The model takes into account the time preference,no-show and cancellation behaviors of patients with appointments,and considers the complex patient flow caused by the imbalance in the selection of doctor resources and the patience limit of waiting time,including patients choosing to concentrate on senior doctors resulting in untimely visits,patients choosing to quit because their patience limit is reached,and patients choosing other doctors to re-queue or leave directly after dropping out,etc.We establish a multi-server parallel outpatient complex queueing network considering both appointment and same-day patients,and by jointly allocating and scheduling the capacity of the appointment patients who choose both senior and junior doctor resources,thus maximizing both patient satisfaction and hospital revenue.The multi-objective computing of budget allocation and genetic algorithm are combined to obtain an approximate multi-server Pareto joint capacity plan and patient scheduling scheme.The simulation model is validated by a case study based on real data from year-round outpatient clinics,and the proposed optimization method can improve the performance of the outpatient scheduling system in general.The simulation optimization framework can provide an effective scheduling scheme for all multi-server service systems involving consumer choice and impatience behaviors.It provides ideas for multi-objective optimization problems with multiple behavioral factors for multiple types of patients in outpatient clinics,and provides practical management insights to managers of outpatient appointment scheduling service systems.(4)Approximate dynamic planning of outpatient appointments considering the multiple levels of doctors and the adaptive service speedWe further consider delayed adaptive patient abandonments and load-adaptive doctor service speed,two common phenomena in outpatient practice.We formulate this dynamic appointment scheduling problem in terms of a Markov Decision Process(MDP)model.Through the data modeling analysis,we develop an analytical framework based on the Markov Decision Process for dynamic appointment scheduling in a multi-class queuing system with parallel servers in an outpatient service context.First,a near-optimal tradeoff between consultation revenue and patient satisfaction,as well as doctor workload at each level,is made in the model by sequentially assigning each appointment request to an available time slot for a particular level of doctor.Second,we derive the properties of the model and propose a heuristic policy based on the tradeoff between appointment times and doctor seniority,and employ Approximate Dynamic Programming(ADP)to design a feasible approximate policy iteration algorithm for the intractable stochastic dynamic optimization problem.The value function approximation method based on the neural network is used,and a set of queueing-based features is selected as the basic function according to the characteristics of the outpatient system to ensure the stability and convergence of the algorithm and produce a good trade-off between the accuracy of the solution and the computational efficiency.Finally,it is also verified through practical cases that the constructed scheduling strategy based on approximate dynamic programming outperforms current systems in practice in various performance metrics such as the difference in average daily waiting queue length between two types of doctors,the average number of patients served per day,the number of patients evaluated exiting per day,and the average patient waiting time per day,giving practical operational management of medical outpatient clinics in China from process data.It provides an integrated and fully automated optimization algorithm framework from process data to dynamic modeling to optimal strategy for the practical management of Chinese medical outpatient practice,and provides practical management insights for managers of the complex outpatient appointment scheduling problem unique to China. |
PDF Full Text Request