USING SIMPLIFIED DISCRETE-EVENT SIMULATION MODELS FOR HEALTH CARE APPLICATIONS

Transcription

1 Proceeings of the 2011 Winter Simulation Conference S. Jain, R.R. Creasey, J. Himmelspach, K.P. White, an M. Fu, es. USING SIMPLIFIED DISCRETE-EVENT SIMULATION MODELS FOR HEALTH CARE APPLICATIONS Anthony Virtue John Kelly EC Harris LLP ECHQ, 34 York Way Lonon, N1 9AB, UK Thierry Chaussalet University of Westminster 115 Cavenish Street Lonon, W1W 6UW, UK ABSTRACT Simulation moeling has been aroun for many years an prouce many papers. Arguably, there has been a lack of impact in the health arena, some may say ue to moeling such a large, complex, iverse an often interconnecte inustry. Other observations suggests that acaemics get reware for publishing large complicate moels with etaile analysis rather than focusing on the requirements of the environment or the nees of implementation. This paper attempts to a to the moeling ebate by suggesting that average simulation process times can act as estimators for real length of stay. This paper will also illustrate how average process time moels coul be use to help reconfigure emergency care services moels. Average time simulation moels have the potential to make a valuable contribution to moeling an they support simplifie, transparent moels with shortene evelopment time. 1 INTRODUCTION This paper will show that average simulation process times can act as estimators for real length of stay in health care environments. Using average time can help to simplify moels, make them more transparent an reuce moeling evelopment time. This paper will use real, reaily available A&E ata to evelop these simplifie moels. Using a steppe approache, this paper will escribe how using real local A&E ata can be use to evelop simplifie moels in a shortene evelopment time. These simplifie moels, clearly linke to real local ata, can help stakeholers to open up black boxes an get a better unerstaning of high level interactions of their health care system. The ata will also be use to illustrate how average time simulation moels might be use to help reconfigure emergency care services an provie moeling insights into those services. These simplifie moels coul be of great benefit as the author s experience is that stakeholers often esire moels with a short turnaroun time. These moels can be use to provie quick illustrations of new scenarios, if require leaing to more intensive investigation/moeling. Simulation moeling has been aroun for many years an many papers have been prouce covering a range of applications in health care. Health relate simulation moeling papers inclue moeling geriatric length of stay by El-Darzi et al. (1998) an Vasilakis an Marshall (2005). Discrete-event simulation (DES) moeling irectly relate to health care inclues: moeling of nurses an nursing cost in an intensive care unit (Griffiths et al. 2005); evaluation of strategies for prevention of mother-to-chil HIV transmission (Rauner et al. 2005); stuy of an NHS walk-in centre (Ashton et al. 2005); an use of simulation to improve the bloo supply chain (Katsaliaki an Brailsfor 2007). More specifically, work linking DES to emergency epartments inclues a stuy combining ata mining an DES for a value-ae view of a hospital emergency epartment (Ceglowski et al. 2007) an unerstaning emergency epartment per /11/$ IEEE 1154

2 formance using simulation (Günal an Pi 2006). Whilst Fletcher et al. (2007) evelope a national generic A&E moel an iscusse the application of its use at a local level. Despite the number of simulations papers prouce there has been a relative lack of real worl involvement an an even greater lack of evience for real worl benefit (Taylor et al. 2009). One reason may be that health care elivery is a complex process an employs a huge number of employees elivering health care over a wie range of services. Harper an Pitt (2004) highlighte a number of issues an challenges particular to health care moeling incluing: scale, complexity an changes (emographic change, social an behavioral change, organizational change, political change, strategic change, technological an clinical change); iversity; buy-in an creibility; conflicting obectives; an ata issues. Another observation note by Taylor et al. was that acaemics are reware for publishing in high-quality ournals an that real worl links are either implicitly unerstoo or not require as research built on previous work. In oing so, Taylor et al. suggeste that researchers became isengage from real worl issues. Similarly Proulove et al. (2007) observe that much of the publishe work was conucte by acaemics an their performance measure value large an complicate moels with etaile statistical analysis of results rather focusing on the requirements of the environment an the nees of the people who nee to implement changes. Günal an Pi (2010) highlighte the extent to which DES moels are use for real ecisions is rarely iscusse an that stake-holers nee to be convince of the benefits. Specifically with regars to poor moeling aoption in hospitals, Sinreich an Marmor (2004) suggeste the reluctance of hospital management to accept change, especially if suggestions come from a black box. To help improve acceptance of moeling in health care Sanchez et al. (2004) suggeste that simulation professionals in health care neee to improve their personal capabilities to: make vali verifie moels; better unerstan their customer s business nees; an to provie customers with answers an insights to their business. Barnes et al. (1997) suggeste three key elements to successful simulation in health care were: Communication an participation; User-frienly simulation software; an using simulation as a ecision making tool. This paper will use conventional A&E ata an illustrate how this service coul be reconfigure into an Urgent Care Centre (UCC). The evelope moel will show examples how patients might be groupe to moel patient pathways. The ability to moel specific patient pathways, such as ault emergency treatment, ault minor treatment, elerly treatment an paeiatric treatment coul have irect inputs into health care planning processes an provie valuable real worl information to health care planners an their clients. These moels will also be able to provie valuable insights into utilization of treatment area, queues an length of stay. Although the moels escribe in this paper are focuse on emergency care, opportunities present themselves to evelop similar moels in other health care environments. The moels evelope are moular by esign, supporting rapi evelopment for new applications. These moels woul have relatively low levels of abstraction as efine by Sinreich an Marmor (2004). Sinreich an Marmor escribe low levels of abstraction as moels esigne for specific systems an moels that are simple an easy to use after a quick explanation. Their simplicity supports rapi feeback, early process reviews an iscussions - supporting Bowers et al. (2009) observation that the evelopment phase is the biggest benefit of simulation moeling. Lower levels of abstraction marks a move away from generic moels as escribe by Corington-Virtue et al. (2005b), Corington-Virtue et al. (2005a), Günal an Pi an Fletcher et al. (2007). Lower levels of abstraction were also applie to moel patient groupings in contrast to clustere patients by iagnosis groups by Corington-Virtue et al. (2006). Low levels of abstraction an the application of reaily available ata to rive the moel helpe to make these evelope moels easier to unerstan for the health community, minimizing obections to black box moels as reference by Sinreich an Marmor (2004). The remainer of this paper is outline as follows. Section 2 provies an overview of the health care planning process, A&Es an Urgent Care Centres. Section 3 will outline the stage approach use to evelop the moel where stage 1 showe how real ata was use to group patients (by acuity an type) an efine their arrival patterns. Stage 2 etermine resources using patient group activity an stage 3 escribes running an calibration of the moel by comparing errors in length of stay between real an mo- 1155

3 ele patient groups with auste process times if applicable whilst stage 4 escribes moel valiation an verification. Sections 4 follows with a iscussion on the finings an Section 5 outlines limitations an further work in context of this paper. 2 HEALTH CARE PLANNING AND EMERGENCY CARE PROCESSES 2.1 Health care planning Health care planning is often use to etermine the size an layout of health care facility. The health care planning process function might inclue eman an capacity planning to: etermine the correct number an size of rooms or treatment areas; assess amissions avoiance, length of stay reuctions, emographic an planne services changes; an esign moels of care, scheules of accommoation, room layout an esign to optimize patient flow an movement through clinical areas. Health care planning often work closely with estate planners as well as other managers to help facilitate user groups an stakeholer meetings. Health care planners also often consier patient types an patient groupings to esign appropriate pathways. In Englan, health care planners often work in accorance with builing guiance ocuments known as Health Builing Notes (HBNs). Often accumulate ata is use to etermine capacity an eman, for example, average process time, patients per week etc. This information often provies useful information with regars to the health planning process (for instance the average number of rooms require over a perio of time) however a weakness is that it is often not sensitive to variation uring the time perio. DES is an ieal tool to complement the health care planning process in that it can provie information on variance throughout the moeling process. DES can also be aapte to moel specific patient routes (pathways) aing valuable insight to the health care planning process. 2.2 Accient an Emergency (A&E) an Urgent Care Centre (UCC) processes A&Es (also known as emergency epartments) are an area of interest to health care planners as they receive a range of patients requiring ifferent services. Better analysis of patient groups an their pathways allows focusing of care towars particular patients. This focus can help to improve both health outcomes for the patient as well as reuce their length of stay in the facility. A&E epartments in the UK treat a wie range of patients, ranging from minor inuries to life threatening conitions an it forms a significant source of hospital amissions. Within a traitional A&E moel, patients often arrive into A&E via two generic routes: Ambulance arrivals - where patients arrive via an ambulance; or Walk-in arrivals - nonambulance arrivals. On arrival into A&E, patients are often triage into categories for treatment by clinical staff. Table 1 illustrates triage coe categories an their escriptions for the A&E ata source use for this stuy. The triage coes in Table 1 illustrates the wie range of patients that arrive in A&E. Triage types range from immeiate treatment (coe 1) to non-urgent treatment (coes 5 an 6). After treatment in A&E, patient are either ischarge home or amitte into hospital. Treatment within A&E is often supporte by a range of activities incluing imaging an pathology. Table 1: A&E Triage coes by escription, number of attenance an % of attenance Triage coes Description Number % 1 Immeiate Very urgent 11, Urgent 38, Stanar 69, Non-urgent Non-urgent 1, Total 122,

4 Table 1 shows that 58% of patients require treatment for minor illness an inuries (coes 4 to 6) an 42% (coe 1 to 3) of patients require care that is more urgent. Arguably, patients with minor inuries an illnesses might receive more effective treatment by General Practitioners (GPs) or Nurse Practitioners. New moels of care suggest that conventional A&E s may not be the most effective metho of treating such a wie range of patients. Alternative moels of care suggest that emergency care might be elivere more efficiently by segregating maor an minor patients. One alternative care moel is to supplement an A&E provision with an Urgent Care Centre (UCC). UCCs (supporte by GPs an specialist nurses) are becoming increasingly common in the UK an they focus their care elivery towars attenees with minor inuries an illnesses leaving conventional A&Es to focus their care elivery towars more seriously ill patients. Emergency Nurse Practitioners or Emergency Care Practitioners sometimes run UCCs. A multi-professional team incluing GPs, nurses an pharmacist usually supports Emergency Nurse Practitioner s. Often, especially in urban areas, UCCs are locate aacent to A&E epartments an act as the point of entry into the emergency epartment. There is evience to show that when compare to A&E clinicians, primary care physicians eliver efficient quality care. To further support the clinical case a stuy at the City Hospital, Birmingham, UK (Ansari et al. 2008) showe Primary care physicians saw more patients at lower cost compare to senior house officers with no ifference recognize in the number of investigations. Reviewing Table 1, this stuy assume that triage coes 4, 5 an 6 met the criteria of UCCs minor illness an minor inuries an triage coe 1, 2 an 3 met the criteria for A&E. As such, Table 1 was upate to show A&E an UCC criteria an this is illustrate in Table 2. Table 2: Table 1 A&E triage coes re-assigne to A&E an UCC coes Triage Description New moel 1 Immeiate A&E 2 Very urgent A&E 3 Urgent A&E 4 Stanar UCC 5 Non-urgent UCC 6 Non-urgent UCC 3 MODEL DEVELOPMENT USING REAL DATA Increasingly, it is consiere goo practice to separate treatment pathways for aults an paeiatrics to focus treatment an care to their requirements. As such, elerly an paeiatric patients are sometimes consiere special groups. Elerly patients are sometimes in a confuse state an their range of illnesses often results in extene stays in emergency epartments. In contrast, paeiatric patients are often prioritize for treatment. Within this moel, paeiatric an elerly patients were efine as special cases an as such efine their own iniviual pathways. In the moel, paeiatric an elerly patients were efine by age: paeiatric 16 years of age or uner an elerly 75 years of age or oler. The remaining ault group was ivie into two istinct pathways epenent on acuity: Ault - A&E; an Ault - UCC. Although resuscitation patients often require the most urgent attention, they forme a relatively low number of the overall attenances an consequently were not moele in this stuy. A summary of the UCC patient groups (Pathways) erive from real ata (1 year s activity) is shown in Table 3. Moels were evelope using the simulation software Simul8 working in conunction with interactive input an output Excel worksheets. 1157

5 Table 3: UCC pathways (erive from A&E ata) showing number of patients an percentages. Resuscitation ata inclue for completeness but was not moele as a pathway. Pathway Number of patients Percentage (%) Ault-A&E 31, Ault-UCC 47, Elerly 12, Paeiatric 30, Resuscitation Total 122, An overview of the moel schematic is shown in Figure 1. The first step in the moel (Pathway Arrivals) generate simulation icons in accorance with their pathway arrival pattern. Once generate, simulation icons passe though a queue bin before entry in to a simulation work centre to be processe. Queue bins acte as waiting areas (queuing area) if resources were unavailable at the work centre. In Figure 1, queue bins are shown as Pathway Queues. At the simulation work centres (shown as Pathway Process in Figure 1), patient icons were treate in simulation time using average process times generate by the moel. The average process time in this moel only refers to irect octor/nurse clinical treatment time. The treatment area might be a room or a cubicle an term room in this stuy relates to either treatment area type. This moel also assume that resources (rooms) were fully staffe to treat simulation patients uring the process time. Resources (shown as Pathway Resources in Figure 1) were assigne (if available) at the start of the simulation work centre process an release on completion. Exit from the simulation work centre complete the simulation process. At each step in the simulation, time labels were attache to the simulation icons. These labels were use to calculate an store arrival times, queuing times an treatment times for all the simulation icons traveling on their moele pathways. The same sub-routines were use by all the work centre process to log an store the time labels. This technique both optimize moular esign (using verifie coe) an supporte rapi evelopment of new pathways. Figure 1: Schematic showing simulation moel process. Simulation ata was recore for 24 hours commencing at minight after a 24-hour warm up perio. A typical trial run was 50 simulations. The remainer of this section will show the stage moel evelopment an illustrate how real ata was use to efine arrival patterns, set room resources an provie evience of moel valiation an verification between real an simulate ata. 1158

6 3.1 Extracte ata to group patients by acuity an type an efinition of their arrival patterns - Stage 1 This sub-section will escribe how real ata was extracte to group patients into pathways an use to efine their arrival patterns. Patient categories in the moel were assigne pathways as shown in Table 3. Using the efine patient pathways, each pathway arrival profile was extracte from real A&E ata an this use to generate arrival profiles (Pathways Arrivals) in the simulation moel. An example of the arrivals profile for ault-ucc is illustrate in Figure 2. Figure 2 shows the average arrival by hour over the ata collection perio of a year. In aition, Figure 2 shows the calculate percentage proportions split over 24 hours. The hourly percentage uses the arrivals per ay (129.14) to calculate the patients per hour, which is use to calculate the inter arrival time. Using the ault-ucc example, 0 ArrivalHr (00:00 hrs to 00:59 hrs) saw 2.29% of the activity over an average 24 hours an this equate to 2.96 patients per hour ( patients per ay x 2.29%). The inter arrival time (20.30 = 60/2.96) was importe at the start of the moe run an upate hourly uring the run cycle. Figure 2: Ault - UCC arrival input example. 3.2 Determine pathway room resources an process times using pathway activity - Stage 2 This sub-section escribes how room resources were etermine in the moel. The number of room resources moele for each pathway was efine by calculations efine in Health Builing Note (HBN) (Facilities for primary an community care services). HBN specifically relate to primary an community services an as such is particularly suite to moel UCCs. In contrast, HBN although not specifically esigne for emergency epartments, it was use to moel A&E room (cubicle) requirements for this stuy. Using HBN (with an occupancy of 80% over 24 hours, 7 ays a week) the initial number of ault-a&e rooms were calculate as follows. Using 31,641 ault-a&e attenances an assuming 100 minutes average process time: patient/weeks per year = 31,641/52 = 608; appointment 1159

7 time = 608*(100/60) = 1,014 hrs/week; room availability at occupancy = 24*7*0.8 = 134 hrs/week; rooms = 1,014/134 = 8 (roune up). The other pathway rooms were calculate using the metho escribe above an are shown in column Rooms in Table 4. Table 4: Pathway initial assume process times an number of rooms. Note the process time assumes irect octor/nurse clinical treatment. Pathway Process time (mins) Rooms Ault-A&E Ault-UCC 40 5 Elerly Paeiatric 40 3 The assume process times (also shown in Table 4) reflecte approximate times base on health planning experience with a number of Trust s. The process times shown in Table 4 represente a starting point in the evelopment in the moels. Process times were calibrate to best fit the actual ata an this process is escribe in the next stage. 3.3 Run an calibrate the moel by comparing errors in length of stay between real an moele pathways an austing process times if applicable - stage 3 The average process (treatment) times shown in Table 4 represente a starting point to calibrate the moel s length of stay to the actual length of stay. To calibrate the moel, each pathway the moel ran a range of process times to compare the moele length of stay against the actual length of stay. Specifically, for each pathway, the cumulative istribution of the actual length of stay was compare with the cumulative istribution of the moele length of stay for a range of process times. The cumulative istribution is shown by: F 0.95 x1, x2 f t t x1, x2, where F x1 is the cumulative istribution for actual length of stay, x2 t F is the cumulative istribution for the moele length of stay an f in steps in 15 minutes up to 95% of the cumulative istribution of the actual length of stay. The process times (roune to the nearest whole number) with the minimum errors between actual an moele length of stay ( F x1 F x2 ) is shown in the auste column Table 5. These auste process times were in effect use to calibrate the moel an use to valiate an verify the moel. By observation, Table 5 suggeste the similarity between the initial an the auste clinical treatment time supporte the estimates of clinical process time for the chosen pathways. Table 5: Pathway initial an auste pathway process times. Note the process time assumes irect octor/nurse clinical treatment. Pathway Process time (minutes) Initial Auste Ault-A&E Ault-UCC Elerly Paeiatric

8 3.4 Moel valiation an verification - stage 4 Input-output valiation techniques were applie to both pathway arrivals an pathway lengths of stays to valiate an verify the moel. Using pathway arrivals an the methoology escribe by Banks et al. (2000), systems outputs (real ata) Z were compare with moele outputs W where = the hour of the ay 1 to K (1 to 24), Z W an; S K 1 K K K 1 Using the paire t -test, the null hypothesis was represente by Z W 0 an teste against the alternative hypothesis of Z W 0, where: t 0 S K Arrivals Testing the t-statistic for pathway arrivals conclue acceptance (at a 95% confience level) of the null hypothesis for all pathways. As such, this test provie evience that moele ata effectively matche real ata an coul act as an effective arrival generator for pathways in the moel. As an example, real an moele arrival profiles for ault - UCC (average over 24-hour perio) is illustrate in Figure 3. Figure 3: Average actual (real) arrivals versus average simulation arrivals for ault-ucc. 1161

9 Figure 4: Actual (real) length of stay actual versus simulate length of stay for ault-ucc Length of stay Paire Z tests were performe on pathway lengths of stay where as above, real ata Z was compare with moele ata W, where = 1 to K was 95% of the real length of stay in steps of 15 minutes. Testing the Z-statistic for pathway lengths of stay conclue acceptance (at a 95% confience level) of the null hypothesis ( Z W 0) for all pathways. As an example, real an moele arrival profiles for ault - UCC is illustrate in Figure 4. Figure 5 illustrates average an meian lengths of stay for pathways an overall (all pathways). Paire t tests were also performe on average an meian ata for pathways an all (the combine ata) to compare real an moele ata as shown Figure 5. Test results here also conclue (at a 95% confience level) acceptance of the null hypothesis for all pathways an therefore reinforce evience above of goo correlation between the real lengths of stay an the moele length of stay for pathways. As such, the paire test provie evience that moele outputs reflecte real outputs an as such valiate the moel. This valiation also suggeste an aequate level of moel coing (verification) within the parameters of the evelope moel. 4 FINDINGS Figure 5: Length of stay comparisons between actual (real) an simulate. This paper has shown how average simulation process times can act as estimators for real length of stay an provie pathway information in a health care setting. This paper has shown too that a simplifie moel coul be use to help reconfigure emergency care services an provie moeling insights into those services. The moel showe pathways lengths of stay, moele using average process (treatment) times, 1162

10 riven by real arrival ata with HBN calculate resources, matche real pathway lengths of stay within a 95% confience level. Simplifie moels irectly linke to real ata an with reuce moel evelopment time, allow stakeholers to open up the black box an better appreciate high-level interactions in their systems. This opens up the moel to a number of arenas incluing; testing parameters such as queues uring run time to etermine when might they appear in the cycle, how big those queues might get an how long might they last?; are queues pathway specific; what impact oes changing resources have; an what impact oes changing process time have? For illustration, the ault-ucc queue output is shown in Figure 6. The moel also suggests similarities in the irect clinical treatment time use an the pre moel assumptions. The finings however o appear to show a ifference between the average clinical time (process time) use to rive the work centre an the overall length of stay. For example, ault-ucc use an average process time of 42 minutes an this resulte in a moele average time of 115 minutes (or meian time of 102 minutes) against a real worl average of 113 minutes (or real worl meian of 101 minutes). One area to investigate woul be the resources allocate. In this moel (an often in the real worl) resources are calculate using activity average over a perio. We can see from the arrival graphs that arrivals (reference Figure 3) increases mi morning, stays quite high in the ay before ropping in the evening. Similarly, we can see queues starting to increase at aroun mi morning - see Figure 6 before ropping of in the late evening. In this example, variable resources (high in the ay, low at night) coul be an area for further investigation. The ability to moel patients pathways in a shorten evelopment time using simple, transparent moels attempts to overcome obections to black box moels as escribe by Sinreich an Marmor. Moeling assumptions coul be quickly moifie an the moel re-run. For example, in the light of current NHS fiscal constraints, simplifie moels coul be use to explore the impact of fewer resources on pathways an the impact on their length of stay. The moels evelope here were particularly focuse towars aressing real worl links, the nees of stake-holers an to be use as a ecision making tool as iscusse by Taylor et al., Günal an Pi an Sanchez et al. The finings from these simplifie moels coul well be a pre cursor to more etaile pathway stuies an moels that are more complicate 5 LIMITATIONS AND FURTHER WORK Figure 6: Ault-UCC Moele Queues This stuy represents an early stage of evelopment an work continues to further evelop the moel. Work is continuing to better unerstan the ifferences in the real an moele worls between the average treatment time an the average length of stay. Other areas for evelopment inclue moels to assess the impact of other variables such as ay of week, month of year an the impact of patient mix. The moels iscusse in this paper were evelope in a moular format with a min to support rapi moel evel- 1163

11 opment to moel patient pathways in other health care settings. This too is an area of future work. The example above iscusse fixe resources with changing input. Changing resources throughout the ay is another area of moel evelopment. REFERENCES Ansari, Z., A. Saha, H. Azra, P. John, an P. Ahee Urgent care Centres: A new evelopment in the Unite Kingom. Annals of Emergency Meicine 51:550. Banks, J., J. S. Carson, B. L. Nelson, an D. M. Nicol Discrete-Event System Simulation. 3r e. Upper Sale River, New Jersey: Prentice-Hall, Inc. Barnes, C., J. Quiason, C. Benson, an D. McGuiness Success stories in simulation in health care. In Proceeings of the 1997 Winter Simulation Conference, eite by S. Anraóttir, K. J. Healy, D. H. Withers, an B. L. Nelson, Piscataway, New Jersey: Institute of Electrical an Electronics Engineers, Inc. Bowers, J., M. Ghattas, an G. Moul Success an failure in the simulation of an Accient an Emergency epartment. Journal of Simulation 3: Ceglowski, R., L. Churilov, an J.Wasserthiel Combining ata mining an iscrete event simulation for a value-ae view of a hospital emergency epartment. Journal of the Operational Research Society 58: Corington-Virtue, A., T. Chaussalet, P. Millar, P. Whittlestone, an J. Kelly A system for patient management base iscrete-event simulation an hierarchical clustering In Proceeings of the 19th IEEE International Symposium on Computer-Base Meical Systems. Corington-Virtue, A., P. Whittlestone, J. Kelly, an T. Chaussalet. 2005a. Developing an application of an accient an emergency patient Simulation moeling using an interactive framework. In Proceeings of the 31st Annual Meeting of the EURO Working Group on OR Applie to Health Services. Corington-Virtue, A., P. Whittlestone, J. Kelly, an T. Chaussalet. 2005b. An interactive framework for eveloping simulation moels of hospital accient an emergency services. 114: Meical an Care Compunetics 2. El-Darzi, E., C. Vasilakis, T. Chaussalet, an P. Millar A simulation moelling approach to evaluating length of stay, occupancy, emptiness an be blocking in a hospital geriatric epartment. Health Care Management Science 1: Fletcher, A., D. Halsall, S. Huxham, an D. Worthington The DH accient an emergency epartment moel: a national generic moel use locally. Journal of the Operational Research Society 58: Griffiths, J., N. Price-Lloy, M. Smithies, an J. Williams Moelling the requirement for supplementary nurses in an intensive care unit. Journal of the Operational Research Society 56: Günal, M., an M. Pi Unerstaning accient an emergency epartment performance using simulation. In Proceeings of the 2006 Winter Simulation Conference, eite by L. R. Perrone, F. P. Wielan, J. Liu, B. G. Lawson, D. M. Nicol, an R. M. Fuimoto, Piscataway, New Jersey: Institute of Electrical an Electronics Engineers, Inc. Günal, M., an M. Pi Discrete event simulation for performance moeling in health care. Journal of Simulation 4: Harper, P., an M. Pitt On the challenges of healthcare moeling an a propose proect life cycle for successful implementation. Journal of the Operational Research Society 55: Katsaliaki, K., an S. Brailsfor Using simulation to improve the bloo supply chain. Journal of the Operational Research Society 58: Proulove, N., S. Black, an A. Fletcher OR an the challenge to improve the NHS: moelling for insight an improvement in in-patient flows. Journal of the Operational Research Society 58:

12 Rauner, M., S. Brailsfor, an S. Flessa Use of iscrete-event simulation to evaluate strategies for the prevention of mother-to-chil transmission of HIV in eveloping countries. Journal of the Operational Research Society 56: Sanchez, S., D. Ferrin, T. Ogazon, J. Sepulrea, an T. War Emerging issues in healthcare simulation. In Proceeings of the 2000 Winter Simulation Conference, eite by J. A. Joines, R. R. Barton, K. Kang, an P. A. Fishwick, Piscataway, New Jersey: Institute of Electrical an Electronics Engineers, Inc. Sinreich, D., an Y. Marmor A simple an intuitive simulation tool for analyzing the performance of emergency epartments. In Proceeings of the 2004 Winter Simulation Conference, eite by R. G. Ingalls, M. D. Rossetti, J. S. Smith, an B. A. Peters, Piscataway, New Jersey: Institute of Electrical an Electronics Engineers, Inc. Taylor, S., T. Elabi, G. Riley, R. Paul, an M. Pi Simulation moelling is 50! Do we nee a reality check? Journal of the Operational Research Society 60: Vasilakis, C., an A. Marshall Moelling nationwie hospital length of stay: opening the black box. Journal of the Operational Research Society 56: AUTHOR BIOGRAPHIES ANTHONY VIRTUE is a part-time PhD stuent within the Health an Social Care Moeling Group (HSCMG) at the University of Westminster an his primary interest is the application of iscrete-event simulation moeling techniques to healthcare services an systems. Anthony works full-time an is employe as a Senior Consultant within EC Harris LLP applying healthcare moeling techniques within their Strategy an Transformation Group. Anthony obtaine a BEng in Mechanical Engineering from Brunel University, an MSc in Engineering Management from Bristol University an an MSc in Decision Sciences from the University of Westminster. His aress is anthony.virtue@echarris.com JOHN KELLY was a founing Director of RKW Healthcare Consultants an is now a Partner with EC Harris LLP within its Strategy an Transformation Group. His backgroun is in sociology an business strategy an he has 30 years experience in the health sector. John has a special interest in healthcare futures an has presente at maor conferences in the UK an overseas. He also has extensive experience in eveloping innovative Moels of Care for acute/community an mental health services an in securing the support of key staff to strategic change through workshops an seminars. John has been involve in a wie variety of strategic/service an estate reviews an briefing/evaluation stuies covering maor acute an mental health Trusts. He is also the author of key guiance ocuments for the planning of health facilities base on DoH research proects an consultancy work for the NHS an private sector. His aress is ohn.kelly@echarris.com THIERRY CHAUSSALET is currently Professor of Healthcare Moelling at the University of Westminster, which he oine after stuying in France an the USA. His research interests lie in the evelopment an use of informatics an moelling techniques to improve healthcare management. He serves currently on the Eitorial Boar of various healthcare moelling an informatics ournals. He is member of the EPSRC Peer Review College, Chair of the Health an Social Services Stuy Group an member of the Council of the Operational Research Society. Active promoter of the use of moelling in the health services, he has been member of the Steering Group of MASHnet, the UK healthcare moeling an simulation network, since 2005, an of the recently establishe Cumberlan Initiative, a consortia of over 15 UK universities eveloping a centre for system moelling an service science in health care. His aress is chausst@westminster.ac.uk 1165