Editorial

We were delighted to be offered the opportunity to edit another Special Issue of JORS on OR in health relatively soon after the previous one, in February 2005 (JORS 2005), which was edited by Ruth Davies and David Bensley. This was a tough act to follow, but the high quality of the submitted papers shows that health OR is a very active field where a great deal of exciting and innovative research is being carried out worldwide. The papers in this Special Issue include several which were given at the 31st annual meeting of the EURO Working Group on OR Applied to Health Services (ORAHS), held at the University of Southampton in August 2005, but they are by no means limited to this conference. These 14 papers serve to illustrate the breadth and depth of health OR, both within the UK and worldwide. We have papers from Australia, the US, Sweden and the Republic of Ireland, and three from Canada. However, seven are from the UK, bearing out the findings of EPSRC in their 2004 Review (EPSRC, 2004) that health OR research is an area in which the UK is a world leader. However, this research strength is not always translated into practical implementations, and this is a theme which is explored in several of the papers in this Special Issue. In contrast with the 2005 Special Issue, which focussed on successful implementations of OR models, the current issue contains several papers describing excellent collaborative research with healthcare providers with potentially wide application but (as yet) little evidence of take-up beyond the original partner organization. We also have several review and discussion papers that we hope will stimulate a response from the health modelling community. The papers cover the spectrum of OR approaches, from ‘soft’ (political and cognitive mapping) to ‘hard’ (scheduling and optimization) and a similar spectrum of application areas, from population environmental health to hospital-acquired infections, and from Paediatric Intensive Care to acute care for elderly patients; truly from cradle to grave.

The UK National Health Service (NHS) has witnessed significant change in recent years with a greater emphasis placed on re-design of processes and modernization. In particular, much attention has been focussed on improving inpatient flows to improve system efficiency. In the discussion paper by Proudlove et al, the authors suggest that one might expect OR to play a major contribution in this field. As well as reviewing the literature, the authors evaluate two projects commissioned by the Department of Health for forecasting emergency admissions and patient flow modelling. The paper concludes that regrettably the contribution by OR has been limited, despite the fact that there is a large number of OR people publishing in this field. The authors suggest that there tends to be a bias towards large complex models that may often be unnecessary and perhaps even be a hindrance for providing clear insight and guidance for problem owners. Instead, more sophistication in understanding the requirements of the environment rather than ever-more complex models is required. The paper concludes with a helpful series of recommendations for more effective OR engagements, such as being open to insights from other disciplines, and focussing more on the needs of NHS personnel who are tasked with implementing change to make a difference.

The paper by Sachdeva et al provides a good illustration of precisely this point. The authors discuss some of the possible reasons for lack of take-up of OR models, and describe a study set in the Paediatric Intensive Care Unit in Wisconsin Children's Hospital. A very simple simulation model is used to investigate patient flows and bottlenecks in the system, but this is combined with a cognitive mapping approach in order to gain acceptance from hospital staff. In attempting to combine both hard and soft OR approaches, Sachdeva et al show that both have a role to play in gaining model acceptance by users.

Decision trees, Markov processes and discrete event simulation are three commonly used techniques for evaluating healthcare interventions. With the aid of examples from coronary heart disease, the paper by Cooper et al is a highly formative discussion on the relative strengths and weaknesses of each of these three techniques. The authors develop guidelines for the choice of modelling technique according to the characteristics of the healthcare intervention. Generally, decision trees are suitable for acute interventions but cannot capture disease recurrence, whereas Markov models are suitable for simple chronic interventions. Furthermore, it is recommended that population-based models be used to provide health outcomes and cost-effectiveness analyses, but the likely modelling complications are better resolved through the use of discrete event simulation. This paper will surely be of assistance to anyone required to model healthcare interventions.

Radiation therapy for cancer is a dangerous treatment which can have serious or even fatal consequences if administered inappropriately, and the optimal dose depends on correctly determining the stage of disease. It is crucial to minimize the risks of over-treating as well as under-treating patients. The paper by Ekaette et al, the first of the three Canadian papers, describes joint work with the Radiology Department at the University of Calgary, the Tom Baker Cancer Centre in Calgary and the Alberta Cancer Board. A simulation model is developed which uses clinical data to estimate the probability of incorrect staging of breast cancer, based on several of the diagnostic tests commonly in use in Alberta, and quantifies the consequences of misdiagnoses using various combinations of these tests. Ekaette et al make recommendations to the Alberta Cancer Board about the optimal combination of tests that should be performed. This work clearly has general implications that are not restricted to Alberta or even to Canada, since the tests used at the Tom Baker Centre are standard ones.

It is interesting to reflect that when people talk about healthcare, they are generally referring to bad health rather than to good health. The paper by Thunhurst makes this distinction and focuses on preventing the ‘upstream’ causes of ill health, rather than on dealing with the ‘downstream’ problems of ill health once it has arisen. Thunhurst observes that OR has historically been applied more frequently to the downstream aspects of health service delivery, and attempts to redress the balance by discussing the potential contribution of OR to the upstream aspects. The paper describes the application of OR approaches in the area of Health Impact Assessment (HIA). HIA is a process for evaluating the impact of policies, programmes and projects on population health. Thunhurst was part of a team commissioned by the Irish Health Research Board to study the Irish government's waste management strategy, and to investigate the health and environmental impact of different landfill and incineration policies. Two visual techniques, political mapping and cognitive mapping, were used to represent the perspectives of the different players in this problem. These greatly aided mutual understanding, and Thunhurst argues eloquently for increased use of problem structuring methods in dealing with these complex societal issues.

Andersson and Varbrand present a dynamic algorithm for the automated allocation of ambulances. The research was carried out in collaboration with the Swedish ambulance service SOS Alarm AB. In the dynamic ambulance dispatching problem, the aim is to assign an ambulance when a 999 call comes in so that two simultaneous (and possibly conflicting) objectives are met: firstly, the time to reach the patient should be as small as possible, but secondly, the potential of the system to respond to future calls (the ‘preparedness’) should be compromised as little as possible. Therefore it is not always optimal to send the nearest ambulance, if this would adversely affect the preparedness. There is a considerable OR literature on the static ambulance location problem—where to site ambulance stations—but the dynamic allocation problem has been relatively little studied. The DYNAROC algorithm uses a new definition of preparedness, based on the geographical distribution of ambulances in zones, and also allows for three call priority levels. An ambulance already assigned to a lower priority call can be diverted to a high-priority call and can still contribute to the preparedness. The system has been implemented with a geographical information system interface, so that different zones are colour-coded by preparedness, and has been successfully tested by simulation based on data for Stockholm. The paper is a good example of OR health modelling work—a nice combination of a theoretical development of a smart idea and a practical implementation in a real setting.

A classical application of simulation in healthcare is described in the second paper from Canada, by Vasilakis et al. The problem here was to evaluate two different policies for managing surgical outpatient waiting lists. In Canada, like the UK, patients requiring surgery are referred to a hospital specialist by their general practitioner, and have to attend outpatient clinics both before and after the actual operation. Under the first policy, which is currently more common, individual surgeons maintain their own outpatient waiting lists: patients are referred to a named surgeon and attend outpatient clinics for that surgeon only. Under the second policy, all patients are placed in a single pooled list for outpatient appointments, so that a patient will be seen by the first available surgeon, although the named referral surgeon will still carry out the actual operation. It is not intuitively obvious whether this will improve the overall waiting time, since patients may just end up waiting longer between the first clinic visit and admission for the operation itself. The model uses the Statecharts formalism to represent the different tasks that surgeons have to perform, and incorporates different urgency levels of patients, who have different acceptable waiting times. It was implemented using cardiac surgery data from a major hospital in British Columbia.

Shaw and Marshall examine the length of stay of heart failure patients admitted to hospital. Heart failure patients require significant healthcare resources with the number of hospital admissions predicted to significantly rise over the coming years. The authors suggest that length of stay is a reliable indicator for measuring the quantity of such resource consumption and use a special type of Markov model, a Coxian phase-type distribution, which is fitted to real data from Belfast hospital, Northern Ireland. The research demonstrates that this methodology accurately captures length of stay, whereby the patients can be thought of as passing through one, two or three phases of care. By more closely predicting individual heart patient hospital stay, the authors are able to predict various hospital measures such as occupancy and apply renewal theory to calculate the mean number of admissions required to maintain a ward at a constant size over a defined planning horizon. This paper demonstrates the value of capturing the stochastic nature and distribution of length of stay as opposed to using simple average values for planning hospital resources.

The supply and demand of blood products is of concern in many countries worldwide. UK demand forecasts show an increase within 20 years of 20% relative to supply. It is therefore important to examine the blood supply chain and improve the management of blood inventories by determining policies that lead to reductions in shortages, wastage and overall costs, while increasing service levels and improving safety procedures. This challenge is explored by Katsaliaki and Brailsford who use discrete event simulation to model a blood supply chain for a UK hospital and regional blood centre. The uniqueness and value of this research lies in the ability to model at a detailed holistic perspective of the blood supply chain and hence results in a better understanding of the entire system and the ability to model system coordination. The authors evaluate different inventory policies and demonstrate how the supply chain can be improved through a number of measures such as changes to holding stock and blood delivery schedules. Although this is still very much research in progress, the approach and insights gained can be taken forward and applied to more complex and realistic systems with a larger network of hospitals and regional centres.

There has been much recent media coverage concerning hospital-acquired infections, such as methicillin-resistant Staphylococcus aureus (MRSA). These so-called ‘superbugs’ have heightened public anxiety and are the cause of increased mortality and morbidity, as well as placing a burden on hospital resources and patient care. Previous studies have shown that an infection increases length of stay and overall cost of care between two and three times. The development and use of tools for monitoring surgical wound infections is reported in the paper by Sherlaw-Johnson. This paper describes work undertaken in collaboration with the University College London Hospitals (UCLH) to develop appropriate real-time outcome monitoring tools that are easy for hospital staff to use and interpret. Although the application of monitoring techniques for surgical wound infections is not new, the contribution of this paper to the literature is the use of the variable life-adjusted display (VLAD) method that can adjust for case-mix. This paper also demonstrates the benefits of combining VLAD and Cumulative Sum charts in order to assess the randomness of observed deviations. The developed monitoring, which has been validated at UCLH and has the potential for more widespread use, comprises of a number of easy to interpret outcome measures which can allow hospital managers to more quickly respond to unusually high levels of infection.

Long waiting lists are a universal problem, and Canada is no exception. The third Canadian paper, by Patrick and Puterman, describes a method to increase the utilization of and reduce the waiting times for a scarce resource—in this case, computed tomography (CT) scanning—for which there is unpredictable demand. The CT department at the Vancouver General Hospital, with whom this research was carried out, serves both high-priority inpatients and lower priority outpatients. Currently, all inpatient demand must be met on the day of the request. Patrick and Puterman develop a method that allows a percentage of non-emergency inpatient demand to be carried over to the next day, by utilizing a pool of on-call outpatients who can respond quickly to available capacity. The policy is tested by simulation and shows a significant reduction in the growth rate of outpatient waiting times.

The paper by Ceglowski et al presents a new methodology to extract patient treatment groups from existing data using non-parametric methods, which are then combined with discrete event simulation to provide an insight into the complex relationship between patient urgency, treatment and queues for treatment in an emergency department (ED). The methodological framework is then applied to the ED in one of the Melbourne's teaching metropolitan hospitals. By identifying core patient treatments provided by the ED, and hence capturing homogeneous patient clusters, it is subsequently relatively straightforward to analyse resource needs and to provide a utility function of the demand placed on the system by patient type. The resulting work has proved valuable within the hospital to reinforce the linkage between hospital policy and access to care, and has identified which patient groupings to target for reducing ward transfer times in order to free up ED beds.

A key issue for all industrialized countries is the ageing population and the increasing cost of healthcare for the elderly. McClean et al consider the problem of acute hospital treatment for older patients, and use a Markov reward model to study the cost-effectiveness of different discharge policies for these patients. McClean et al derive the interesting result that it may be more cost-effective to keep acute elderly patients in hospital longer, to ensure they are really fit for discharge—counter to the current trend for early discharge back into the community with the support of such schemes as Intermediate Care. This approach enables a broader, systems view to be taken, ensuring that all the true social costs are taken into account as well as the hospital costs.

Finally, Eldabi et al propose considered futures for the use of simulation as a problem-solving technique within healthcare settings. This discussion paper is sure to start a welcome debate on the future of simulation that will offer the opportunity to speed up the increased effectiveness of the technique. Critical analysis has been applied to survey data based on trends identified by a selection of experts in the field, academics and industrialists. This reveals that whole system approaches with more joined up modelling or mixed methods are likely to be high on the agenda for future research. The paper then presents options for how simulation could be used within the healthcare domain and identifies common themes, such as persuading service providers and clinicians that simulation as a system level tool can make a critical contribution, that represent challenges for the OR healthcare modelling community. The authors suggest that more methodological research in this area, such as the creation of a widely accepted frame of reference, is needed.

We hope you enjoy reading these papers, which in our opinion convey the strengths and diversity of the health OR modelling community, and demonstrate the wide applicability and range of innovative approaches and techniques in this field.

Dr Sally Brailsford is a senior lecturer in Management Science in the School of Management, University of Southampton. Her research interests include simulation modelling methodologies, system dynamics, health service research and disease modelling, and the modelling of human behaviour in healthcare systems. She won the UK OR Society's 2004 Goodeve Medal for her JORS paper on system dynamics modelling of emergency healthcare (Brailsford et al., 2004), and is on the editorial board of Health Care Management Science and the Journal of Simulation. She is the secretary of the EURO Working Group on OR Applied to Health Services, and is currently Vice President of the UK OR Society.

Dr Paul Harper is a senior lecturer in Operational Research within the School of Mathematics, University of Southampton. His research interests are in simulation, healthcare modelling, OR for developing countries and data mining. Within the healthcare domain, he has studied early detection, screening and treatment issues for various diseases and medical conditions, including diabetic retinopathy, HIV/AIDS and cancer. Furthermore, Dr Harper has worked with a number of healthcare organizations to help form policy on resource capacities and simulation modelling of patient pathways. Application areas have included intensive care and hospital inpatients, outpatient departments and current work on stochastic location–allocation models for geographical provision of healthcare services.