OR in healthcare: A European perspective

The European Working Group “Operational Research Applied to Health Services” (ORAHS) is one of the domain specific EURO Working Groups organized by EURO - the European Association of Operational Research Societies. In this paper we report on the development of ORAHS as a platform for OR in health, and analyse the papers presented at meetings over the 35 years of its existence. We propose a two-way framework for analysis, where one dimension is the nine stages of the product life cycle: identifying consumer requirements, designing a new service to meet these requirements, forecasting demand for such a service, securing resources for it, allocating these resources, developing Programs & Plans to use these resources for delivering the service, establishing criteria for service delivery, managing the performance of the service, and finally, evaluating its performance. The other dimension is a three-level classification into broad application areas referring to processes at different levels in healthcare: Patients & Providers, Units & Hospitals, and Regional & National. We use this framework to carry out a quantitative analysis of all the papers presented during the meetings of ORAHS since its inception in 1975. We then describe developments over this period in applying OR approaches and techniques to healthcare, and present an overview of the main application areas and challenges.     

SimLean: Utilising simulation in the implementation of lean in healthcare

Discrete-event simulation (DES) and lean are approaches that have a similar motivation: improvement of processes and service delivery. Both are being used to help improve the delivery of healthcare, but rarely are they used together. This paper explores from a theoretical and an empirical perspective the potential complementary roles of DES and lean in healthcare. The aim is to increase the impact of both approaches in the improvement of healthcare systems. Out of this exploration, the ‘SimLean’ approach is developed in which three roles for DES with lean are identified: education, facilitation and evaluation. These roles are demonstrated through three examples of DES in action with lean. The work demonstrates how the fusion of DES with lean can improve both stakeholder engagement with DES and the impact of lean.     

A shift scheduling model for employees with different seniority levels and an application in healthcare

This paper addresses the problem of scheduling medical residents that arises in different clinical settings of a hospital. The residents are grouped according to different seniority levels that are specified by the number of years spent in residency training. It is required from the residents to participate in the delivery of patient care services directly by working weekday and weekend day shifts in addition to their regular daytime work. A monthly shift schedule is prepared to determine the shift duties of each resident considering shift coverage requirements, seniority-based workload rules, and resident work preferences. Due to the large number of constraints often conflicting, a multi-objective programming model has been proposed to automate the schedule generation process. The model is implemented on a real case in the pulmonary unit of a local hospital for a 6-month period using sequential and weighted methods. The results indicate that high quality solutions can be obtained within a few seconds compared to the manually prepared schedules expending considerable effort and time. It is also shown that the employed weighting procedure based on seniority levels performs much better compared to the preemptive method in terms of computational burden.     

A multi-criteria model for auditing a Predictive Maintenance Programme

Auditing tools can play a key role in the continuous improvement of maintenance policies, in particular to enhance predictive maintenance (PM). This paper proposes a multi-criteria model for auditing a Predictive Maintenance Programme (PMP) developed and implemented in the General Hospital of Ciudad Real (GHCR) in Spain. The model has a two-level structure, with top level auditing areas specified by second level auditing criteria on which the performance of the PMP should be appraised. This structure resulted from the analysis and discussion of an internal questionnaire to the management, technical and consulting staff of GHCR. This also guided the association of a performance scale with each criterion, describing several reference levels of accomplishment. Using the MACBETH (Measuring Attractiveness by a Categorical Based Evaluation Technique) approach, a hierarchical additive value model was constructed, with criteria weights and value scales derived from staff judgments of comparison of different reference levels and profiles of performance. This model enables managers to measure the performance of the PMP and its added value for the hospital, not only against each audit criterion individually, but also on each area and in overall terms. Integrated in a management “tableau de bord”, the model outputs permit the identification of PMP deficiencies requiring urgent intervention and corrective measures for its continuous improvement.     

Model development in discrete-event simulation and system dynamics: An empirical study of expert modellers

An empirical study comparing the model development process followed by experts in discrete-event simulation (DES) and system dynamics (SD) modelling is undertaken. verbal protocol analysis (VPA) is used to study the modelling process followed by ten expert modellers (5 SD and 5 DES). Participants are asked to build simulation models based on a case study and to think aloud while modelling. The generated verbal protocols are divided into seven modelling topics: problem structuring, conceptual modelling, data inputs, model coding, verification & validation, results & experimentation and implementation and then analyzed. Our results suggest that all modellers switch between modelling topics, however DES modellers follow a more linear progression. DES modellers focus significantly more on model coding and verification & validation, whereas SD modellers on conceptual modelling. Observations are made revealing some interesting differences in the way the two groups of modellers tackle the case. This paper contributes towards the comparison of DES and SD.     

Simulation modelling for contracting hospital emergency services at the regional level

Hospital emergency services are closely connected to demographic issues and population changes. The methodology presented here helps to assess the effects of the forecasted demand changes on the next-year emergency unit workloads. The objective of the study is to estimate the expected volume of emergency hospital services, as measured by the number and costs of medical procedures provided to patients, to be contracted by the Polish National Health Fund (NFZ) branch at the regional level to cover the forecasted demand. A discrete-event simulation model was developed to elaborate the credible forecasts of the function components, the fundamental elements of the contract values granted by the NFZ for emergency departments for the following year. Emergency department-level data were drawn from the NFZ regional branch registry to perform a statistical analysis of emergency services provided to patients in 17 admission units and emergency wards in 2010. The model results indicate that the predicted increase in two age groups, i.e., the youngest children and the older population, will have different effects on the number and value of hospital emergency services to be considered in the contracting policy. There is potential for a discrete-event simulation to support strategic health policy decision making at the regional level. The value of this approach lies in providing estimates for the what-if scenarios related to the prognosis of changing acute demand.     

Combining simulation and goal programming for healthcare planning in a medical assessment unit

This paper describes a detailed simulation model for healthcare planning in a medical assessment unit (MAU) of a general hospital belonging to the national health service (NHS), UK. The MAU is established to improve the quality of care given to acute medical patients on admission, and to provide the organisational means of rapid assessment and investigation in order to avoid unnecessary admissions. The simulation model enables different scenarios to be tested to eliminate bottlenecks in order to achieve optimal clinical workflow. The link between goal programming (GP) and simulation for efficient resource planning is explored. A GP model is developed for trade-off analysis of the results obtained from the simulation. The implications of MAU management preferences to various objectives are presented.     

Uncertainty and value of information when allocating resources within and between healthcare programmes

A two-stage stochastic mathematical programming formulation has been developed to optimally allocate resources within and between healthcare programmes when there is an exogenous budget and the parameters of the healthcare models are variable and uncertain. This formulation solves the optimal resource allocation problem and calculates the expected value of acquiring additional information to resolve the uncertainties within the allocation. It is shown that the proposed formulation has several advantages over the chance constrained and robust mathematical programming methods.     

A branch-and-price approach for integrating nurse and surgery scheduling

A common problem at hospitals is the extreme variation in daily (even hourly) workload pressure for nurses. The operating room is considered to be the main engine and hence the main generator of variance in the hospital. The purpose of this paper is threefold. First of all, we present a concrete model that integrates both the nurse and the operating room scheduling process. Second, we show how the column generation technique approach, one of the most employed exact methods for solving nurse scheduling problems, can easily cope with this model extension. Third, by means of a large number of computational experiments we provide an idea of the cost saving opportunities and required solution times.     

Dynamic job assignment: A column generation approach with an application to surgery allocation

We consider the assignment of jobs to heterogeneous agents in a dynamic system with a rolling time horizon. An example is a hospital operating theatre where the jobs are surgeries and the agents are the surgeons. The paper is presented in the context of surgery allocation and the system is characterized as follows: Patients are grouped into categories and they arrive continually following a stochastic process. Patients in each group have specific time limits within which they need treatment and if it cannot be accommodated then the patients are outsourced. The service level is the percentage of patients in each group treated within the time limit. Surgery durations are stochastic and depend on the surgeon conducting the surgeries. Each surgeon has limited time available and expected overtime is penalized by a non-decreasing convex function. We develop a column generation approach for the assignment of already arrived patients and tentative future patients to surgeons on specific days. It balances the conflicting objectives of including as many arrived patients as possible within their time limits, maximizing the service level of future patients, and minimizing the expected overtime of surgeons. A computational study is conducted with the model embedded in a rolling time horizon frame. The study indicates that the assignment of patients based on our model increases system performance in terms of service level and reduced overtime compared to a First-Come-First-Served (FCFS) policy when the arrival rates of patients are medium to high compared to the capacity of the system.     

A statistical process control approach to selecting a warm-up period for a discrete-event simulation

The selection of a warm-up period for a discrete-event simulation continues to be problematic. A variety of selection methods have been devised, and are briefly reviewed. It is apparent that no one method can be recommended above any other. A new approach, based upon the principles of statistical process control, is described (SPC method). Because simulation output data are often highly autocorrelated and potentially non-normal, the batch means method is employed in constructing the control chart. The SPC method is tested on seven data sets and encouraging results are obtained concerning its accuracy. The approach is also discussed with respect to its ease of implementation, simplicity, generality of use and requirements for parameter estimation.     

A stochastic appointment scheduling system on multiple resources with dynamic call-in sequence and patient no-shows for an outpatient clinic

This research focuses on the stochastic assignment system motivated by outpatient clinics, especially the physical therapy in rehabilitation service. The aim of this research is to develop a stochastic overbooking model to enhance the service quality as well as to increase the utilization of multiple resources, like therapy equipment in a physical therapy room, with the consideration of patients’ call-in sequence. The schedule for a single-service period includes a fixed number of blocks of equal length. When patients call, they are assigned to an appointment time for that block, and an existing appointment is not allowed to be changed. In each visit, a patient might require more than one resource and a probability of no-show. Two estimation methods were proposed for the expected waiting and overtime cost with multiple resources: Convolution Estimation Method and Joint Cumulative Estimation Method for the upper and lower bound value; respectively. A numerical example based on a physical therapy room was used to show that this stochastic model was able to schedule patients for better profitability compared with traditional appointment systems based on four prioritization rules. The workload in each appointment slot was more balanced albeit more patients were assigned to the first slot to fill up the empty room.     

Combining approaches for evaluating auditing populations: A simulation study

This research explores ways of combining four distinct bounds for the mean error in an auditing population. Two competing objectives for a bound are to be close to the true mean being estimated and to be reliable: not less than the true mean in more than 5% of estimations. The optimal combination should provide the best balance of these competing objectives. Estimating the mean error by a single approach is difficult because typically most accounts have no error and the distribution of the errors among those that do is discontinuous and highly skewed. This study reveals that the weights in the optimal combination are not constant but depend on the characteristic of the population being estimated. The optimally combined bound is only 7% smaller overall than the best of the constituents. However, while the best of the constituents fails in 50% of most challenging populations, the optimal combination never fails.     

Development and implementation of a real time statistical control method to identify the start and end of the winter surge in demand for paediatric intensive care

Winter surge management in intensive care is hampered by the annual variability in the winter surge. We aimed to develop a real-time monitoring system that could promptly identify the start, and accurately predict the end, of the winter surge in a paediatric intensive care (PIC) setting. We adapted a statistical process control method from the stock market called “Bollinger bands” that compares current levels of demand for PIC services to thresholds based on the medium term average demand. Algorithms to identify the start and end of the surge were developed for a specific PIC service: the North Thames Children's Acute Transport Service (CATS) using eight winters of data (2005–12) to tune the algorithms and one winter to test the final method (2013/14). The optimal Bollinger band thresholds were 1.2 and 1 standard deviations above and below a 41-day moving average of demand respectively. A simple linear model was found to predict the end of the surge and overall demand volume as soon as the start had been identified. Applying the method to the validation winter of 2013/14 showed excellent performance, with the surge identified from 18th November 2013 to 4th January 2014.An Excel tool running the algorithms has been in use within CATS since September 2014. There were three factors which facilitated the successful implementation of this tool: the perceived problem was pressing and identified by the clinical team; there was close clinical engagement throughout and substantial effort was made to develop an easy-to-use Excel tool for sustainable use.     

Disseminating a warning message to evacuate: A simulation study of the behaviour of neighbours

Large-scale evacuations are a recurring theme on news channels, whether in response to major natural or manmade disasters. The role of warning dissemination is a key part in the success of such large-scale evacuations and its inadequacy in certain cases has been a ‘primary contribution to deaths and injuries’ (Hayden et al., 2007). Along with technology-driven ‘official warning channels’ (e.g. sirens, mass media), the role of unofficial channel (e.g. neighbours, personal contacts, volunteer wardens) has proven to be significant in warning the public of the need to evacuate. Although post-evacuation studies identify the behaviours of evacuees as disseminators of the warning message, there has not been a detailed study that quantifies the effects of such behaviour on the warning message dissemination. This paper develops an Agent-Based Simulation (ABS) model of multiple agents (evacuee households) in a hypothetical community to investigate the impact of behaviour as an unofficial channel on the overall warning dissemination. Parameters studied include the percentage of people who warn their neighbours, the efficiency of different official warning channels, and delay time to warn neighbours. Even with a low proportion of people willing to warn their neighbour, the results showed considerable impact on the overall warning dissemination.     

Locating specialized service capacity in a multi-hospital network

Multi-hospital systems have become very common in today’s healthcare environment. However, there has been limited published research examining the opportunities and challenges of pooling specialized services to a subset of hospitals in the network. Therefore, this paper considers how hospital networks with multiple locations can leverage pooling benefits when deciding where to position specialized services, such as magnetic resonance imaging (MRI), transplants, or neonatal intensive care. Specifically, we develop an optimization model to determine how many and which of a hospital network’s hospitals should be set up to deliver a specialized service. Importantly, this model takes into account both financial considerations and patient service levels. Computational results illustrate the value of optimally pooling resources across a subset of hospitals in the network versus two alternate approaches: (1) delivering the service at all locations and requiring each site to handle its own demand, or (2) locating the service at one hospital that handles all network demand.     

A branch and bound based heuristic for makespan minimization of washing operations in hospital sterilization services

In this paper, we address the problem of parallel batching of jobs on identical machines to minimize makespan. The problem is motivated from the washing step of hospital sterilization services where jobs have different sizes, different release dates and equal processing times. Machines can process more than one job at the same time as long as the total size of jobs in a batch does not exceed the machine capacity. We present a branch and bound based heuristic method and compare it to a linear model and two other heuristics from the literature. Computational experiments show that our method can find high quality solutions within short computation time.     

Optimizing specimen collection for processing in clinical testing laboratories

We study the logistics of specimen collection for a clinical testing laboratory that serves sites dispersed in an urban area. The specimens that accumulate at the customer sites throughout the working day are transported to the laboratory for processing. The problem is to construct and schedule a series of tours to collect the accumulated specimens from the sites throughout the day. Two hierarchical objectives are considered: (i) maximizing the amount of specimens processed by the next morning, and (ii) minimizing the daily transportation cost. We show that the problem is NP-hard and formulate a linear Mixed Integer Programming (MIP) model to solve the bicriteria problem in two levels. We characterize properties of optimal solutions and develop a heuristic approach based on solving the MIP model with additional constraints that seeks for feasible solutions with specific characteristics. To evaluate the performance of this approach, we provide an upper bounding scheme on the daily processed amount, and develop two relaxed MIP models to generate lower bounds on the daily transportation cost. The effectiveness of the proposed solution approach is evaluated using realistic problem instances. Insights on key problem parameters and their effects on the solutions are extracted by further experiments.     

Discrete Conditional Phase-type models utilising classification trees: Application to modelling health service capacities

Discrete Conditional Phase-type models (DC-Ph) consist of a process component (survival distribution) preceded by a set of related conditional discrete variables. This paper introduces a DC-Ph model where the conditional component is a classification tree. The approach is utilised for modelling health service capacities by better predicting service times, as captured by Coxian phase-type distributions, interfaced with results from a classification tree algorithm. To illustrate the approach, a case-study within the healthcare delivery domain is given, namely that of maternity services. The classification analysis is shown to give good predictors for complications during childbirth. Based on the classification tree predictions, the duration of childbirth on the labour ward is then modelled as either a two or three-phase Coxian distribution. The resulting DC-Ph model is used to calculate the number of patients and associated bed occupancies, patient turnover, and to model the consequences of changes to risk status.