| Healthcare service is one of the most important sectors for human health.Recent years,with the improvement of health requirements,healthcare service also faces an increasing demand,so many countries have made massive investment to promote the development of healthcare service.Although the quality of healthcare service has been greatly improved with the progress of economy and technology,the medical resources waste is quite common due to lack of effective operations management.Relevant researches show that effective strategy policy and operations management methods have played an important role in the increase of healthcare service,consequently give rise to relief the contradiction of healthcare supply and demand.Therefore,the operations management of healthcare service has been one of the research hotspots.There are many uncertainty factors in the process of healthcare service,such as daily number of patients,patients’ service duration,and patients’ behaviors which include unpunctuality,no-show,cancellation and so on.Those uncertainty factors not only lead to the low effective of healthcare service,but also increase the cost and decrease the revenue of hospital.Due to the above reasons,it is significant to take the uncertainty factors into consideration in healthcare service operations management.Meanwhile,for the sake of the public benefit and equity of healthcare service,the optimizing strategy should be suitable and robust under different situations.Distributionally robust optimization method is an effective tool for stochastic optimization to make decision by minimizing the worst objective with partial information of random variables,that’s why it has been paid close attention recently.In this thesis,we assume that we only know partial moment information about the probability distribution and adopt distributionally robust optimization methods to study some problems in healthcare service operations management.The structure can be summarized as following:Firstly,as the first step of healthcare service,the quality of outpatient service scheduling has great impact on the efficiency of successive services and even the overall hospital operation.In the first model,we study a single service platform appointment scheduling problem with the number of patients and their sequence of appointments are fixed during a single outpatient service period.Take the mean and covariance of random service durations into account,we establish the distributionally robust appointment scheduling model which has min-max form,and we solve this distributionally robust schedules by minimizing the expectation of the weighted sum of patients’ waiting time and the doctor’s overtime.We show that the original model can be reformulated as a two-stage optimization problem which can be solved by decomposition algorithm.Numerical result shows that patient’s service duration increases with sequence of arrival when patient’s waiting cost equals to doctor’s overtime cost,it is an appropriate strategy for general clinics.And patient’s service duration presents a ”dome” shape which means the service time increases at first then decreases when patient’s waiting cost less than doctor’s overtime cost,it is an good strategy for scarce clinics.Secondly,in reality,the moments of random variables usually be derived from statistical sample which is often difficult to completely characterize the random variables due to little historical data or inaccuracy of forecast methods.Based on the first research content,we investigate a stochastic appointment scheduling problem in an outpatient clinic with a single doctor considering the uncertainty moments of random variables.In this section,we propose a distributionally robust appointment scheduling model that describes uncertainty in the form of confidence region for the mean and the covariance matrix of a random vector.Furthermore,we also reformulate the model in terms of a two-stage optimization problem and solve it by a decomposition algorithm.Result also shows that patient’s service duration should increase with sequence of arrival in general clinics,and patient’s time arrangement should increase at first then decreases in scarce clinics.The third model is carried out at the background that current availability of inpatient beds is under strain in Chinese hospitals.And the availability of beds should not only meet the demand of regular patients,but also satisfy the demand of emergency patients.So we consider the partitioning of inpatient beds between regular patients and emergency patients,assume that the number of beds is fixed and the objective is to minimize patients’ total weighted waiting time.Using the knowledge of queuing theory to describe patients’ waiting time,we provide an a distributionally robust optimization model based on the moments information of uncertain inpatient duration.Similar to prior the solving method,we reformate the model to a simplified form that facilitate decomposition algorithms.Result shows that the bed arrangement of emergency patients should increase with their urgency degree,and emergency patients’ bed arrangement should satisfy their expected demand when the urgency degree is big enough.Finally,we deploy a distributionally robust model for ordering quantities of radioactive tracer when considering the high cost and demand of nuclear medical diagnosis as well as the half-life characteristic of radioactive tracer which is widely used in nuclear medical diagnosis.Taking the moments of random patients’ demand for nuclear medical diagnosis into account,the objective of this model is to maximize the worst case total expected revenue.In particular,the problem can be reformulated as a second-order cone program which can be solved easily by optimization toolbox.Result shows that the demand fluctuation has greater impact on optimal revenue compared with the mean of patients’ demand,and the cost of radioactive tracer is also an important influence factors for total expected revenue.Result shows that rational measures should be adopted to control volatility of patient’s demand,such as effective examine schedule and pre-examine check.This thesis proposes some distributionally robust optimization models for healthcare service operations management by considering some uncertainties in the process of healthcare service.In particular,the methods and models are expansions for healthcare service operations management theory and can be used to improve the management level of hospitals.At the meanwhile,this thesis can provide decision support to improve the effective and service quality of hospital operations,and have some practical guiding significance to increase healthcare resource utilization and enhance hospital’s competitiveness. |
