Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
Using Computer Simulation in Operating Room Management: Impacts on Process Engineering and Performance Andre Baumgart1, Anja Zoeller2, Christof Denz3, Hans-Joachim Bender3, Armin Heinzl2, Essameddin Badreddin1 1
Automation Laboratory, University Mannheim, Mannheim, Germany {andre.baumgart, badreddin} @ti.uni-mannheim.de
2
3 Department of Business Department of Anaesthesiology Administration and and Critical Care Medicine, Information Systems, University Hospital Mannheim, University of Mannheim, Mannheim, Germany, Mannheim, Germany, {christof.denz,
[email protected] joachim.bender}
[email protected] @anaes.ma.uni-heidelberg.de
Abstract Operating rooms are regarded as the most costly hospital facilities. Due to rising costs and decreasing reimbursements, it is necessary to optimize the efficiency of the operating room suite. In this context several strategies have been proposed that optimize patient throughput by redesigning perioperative processes. The successful deployment of effective practices for continuous process improvements in operating rooms will require that operating room management sets targets and monitors improvements throughout all phases of process engineering. Simulation can be used to study the effects of process improvements through novel facilities, technologies and/or strategies. In this paper, we propose a conceptual framework to use computer simulations in different stages of business process management (BPM) lifecycle for operating room management. Additionally, we conduct simulation studies in different stages of the BPM lifecycle. The results of our studies provide evidence that simulation can provide effective decision support to drive performance in operating rooms in several phases of the BPM lifecycle.
1. Background and Motivation Today’s hospitals face an ever changing economic and political environment, where they must continuously adapt their structure, processes, and technologies to new conditions [9], [11], [18]. Operating rooms (ORs) are the high-cost and highrevenue service centers of hospitals. Therefore, it is essential to optimize the effectiveness and efficiency of the OR suite [24]. Various new management methods and techniques have been developed over the last years to approach the need for constant change and
improvement in OR management [13], [14], [20], [35], [37], [42]. Additionally, several operational research methods have been presented recently that cover master surgery schedules and robust surgery planning through the introduction of slack time [4], [5], [6], [7], [17], [19], [26], [31], [41], [43]. Most strategies focus on the improvement of perioperative efficiency, such as reducing waste time or overtime of cases, adding more cases during regular operating hours and increasing profitability of the operating suite. This requires an effective OR management during the day for operational and real-time control of processes and workflows. The proposed strategies predominantly depend on their assumptions about the mean throughput time of cases and, thus, are not applicable to existing perioperative processes in general. The studies provide techniques that investigate special types or classes of surgeries, where the OR setup costs and/or mean durations of activities are similar, such as in [20], [37]. Thus, variations in the planning and scheduling or portfolio effects of the surgical program comprise less uncertainty. Additionally, the applicability of more advanced planning techniques [17], [19] or master surgery schedules [4], [5], [31] to real-time operational control is restricted because of their focus on a mid-term planning horizon [6], [38]. The interactions and practical implications of the proposed techniques with long term strategic goals of the hospital or operational (real-time) controls of the process have been analyzed scarcely [26], [42]. Moreover, the methods mainly consider parallel processing and resource constraints for individual surgeries or classes of surgeries. Various real world resource conditions and constraints are neglected – see Section 5 for a detailed discussion. Still, OR management shows serious potential in improving
1530-1605/07 $20.00 © 2007 IEEE
1
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
processes and the effectiveness of operational management [24], [44]. Generally, simulation is mentioned in the literature as one mean to improve the effectiveness of change management methods, such as Total Quality Management, Just in Time or Business Process Reengineering [11]. Simulation also provides a structural environment to understand, analyze, design, (experimentally) execute and evaluate business processes and workflows for different process stakeholders. It is typically used to quantitatively evaluate different configurations and alternatives of processes. Computer simulations are not new in health care management [22], [40]. We propose a framework to use computer simulation to support process engineering in perioperative processes and to show its efficacy and potentials for process improvement [11], [34]. Several challenges have to be addressed and managed for introducing simulation in process (re-) engineering [8]. The following challenges for effective OR modeling, simulation and management summarize the experiences found in literature [29], [34] and/or reflect our findings in our field study (see Section 4): - The decision maker has to be supported at each stage in the BP lifecycle. Thus, the integration of simulation into the decision making process must be guaranteed during process analysis, design, enactment and evaluation. - The decision maker needs a tool for comparing various improvement strategies (e.g. by introducing additional resources in the patient process and applying overlapping induction or parallel processing to increase patient throughput) for OR management. These strategies depend on several factors (e.g. operation types) and imply different assumptions (e.g. similar quality of resources) for their evaluation. Consequently, their use in a broader context is often perilous. - The various operating room improvement strategies rely on different information and data needs. Previous studies could not or seldom provide the requisite detail in patient processes and resource utilization for an enhanced process analysis. The main contribution of the paper is the conceptual model introduced in Section 3. It proposes the integration and application of computer simulation studies for process reengineering and improvement in operating theatres in order to enhance decision support capabilities for the process stakeholders during the entire BPM lifecycle. Consequently, simulation-based decision making is not only used to respond ‘a posteriori’ to what has happened, but also ‘a priori’ in order to better streamline the current practices or new
operational research strategies. In addition to other complementary frameworks [6], [38] the model can be applied to continuous process improvement and the use of large scale scenario-based simulations. The remainder of the paper is structured as follows: In Section 2, we provide an overview of the extensive literature of modeling and simulation of operating room processes, decision support and computer simulation in perioperative processes in hospitals. The conceptual framework is introduced in Section 3. Section 4 presents results and experiences from experimental simulations and field study based simulations. In Section 5, we discuss the validity of the proposed framework and what needs to be done to effectively apply it. Additionally, we present several limitations of process oriented analysis in operating theatre management. We conclude our paper in Section 6 with some remarks on future work.
2. Simulation in OR Management Computer simulation is a technique to study changes over time of a dynamic model of a real-world system. The full power of simulation is realized when it is used to analyze dynamical systems with complex interactions among various components and processes [23], [27]. The simulation model allows safely, rapidly and efficiently studying and assessing the impact of new methods, strategies and/or technologies in the perioperative and operative process [11], [39]. OR suites and other health care processes have been studies extensively in literature. This section summarizes the main recently conducted research in health care and OR management to show the qualitative and quantitative advantages of simulation [27]. Qualitative advantages can be found in the analysis stage of modeling the system under research. Regularly, the modeling process raises important additional research questions about the system and its behavior [2]. Process participants, like doctors or nurses, use process models as a focus of discussion for process improvement and create a common understanding about the problem and explore alternative solutions in an interactive way [10]. Additionally, simulation is seen as a tool for communication and an element of the planning process. It is used to enable an informed debate between different process stakeholders, e.g. medical staff, hospital administration and computer departments, for searching for optimal solutions [15]. Concerning quantitative process orientation, simulation studies are applied in [3], [22], [25], [29] to examine one-time changes in OR processes. For
2
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
Goal Specification, Evironmental Analysis Organizational Analysis
Monitoring Results
Implemented Processes Me & R asur es es ult s
Process Enactment Workflow Execution/ Run Time Measure -ments
Metrics/ Audit Events
Me a sure Resu s & lts
Process Implementation Workflow Modeling and Implementation
TargetMetrics
Process Evaluation Warehousing / Controlling/ Process Mining
Export for other reporting purpose
Process Models
Specification and Modeling
Measures for Improvement Ex post
Measures & Results
Simulation
Ex ante
s u re as lts Me esu &R
Process Design
Simulation
Process Monitoring Business Activity Monitoring
Simulation In-process (online)
Figure 2. Simulation-oriented Business Process Management Lifecycle [45]. certain surgery unit layouts they give advice on scheduling policies, predictions on demand for personnel and rooms, or optimize certain perioperative performance measures such as waiting times or time between surgeries. Moreover promising results for process reengineering in medical care processes are presented, where process changes where analyzed with statistical and simulation techniques and where simulation provided the best quality solution [30]. [22] reviews the use of discrete event simulation in healthcare systems, listing over one hundred studies. [40] examines health-care simulation studies published in the Winter Simulation Conference Proceedings from 1997 through 2004. They summarize background research on the sources of data available for emergency departments and hospitals. Most studies neglect real-time control approaches supporting daily OR management and do not discuss the respective information need on critical process interfaces and influencing variables [4], [42], [44]. Accordingly, a simulation model and tool can be useful in monitoring the performance of the system [15] and if integrated into an extensible and well-interfaced simulation environment real-time data is used for process analysis and improvement. We use the Arena® simulation environment for the discrete event simulation models in Section 4.
3. A Conceptual Framework In general, processes transform input into output along a path of activities, which may utilize resources such as personnel or devices [36]. A process is a sequence of activities that is necessary to manipulate
an object of economic interest to the hospital, and that achieves a specific goal [1], [45]. Business Process Management (BPM) creates alignment among the individual process components input, output, resources, process structure, and process goals through a lifecycle of planning, organizing, scheduling, directing and controlling. Alignment is achieved, if the overall process performance increases in process quality (e.g. long-term patient health) and quantity (e.g. faster operations, lower costs). It is an iterative approach in form of a continuous process management lifecycle that assists organizations to achieve, maintain, and improve their processes [45]. Figure 1 presents the main stages and components of the BPM lifecycle. In addition, it shows the integration of simulation support in different stages of the lifecycle for OR management. The lifecycle contains the following main phases: The process lifecycle starts with a definition of organizational and process goals, and an assessment of environmental factors and constraints. The completeness of the goal specification and the organizational analysis defines the parameters and thus the constraints for process design. The process design phase is the identification of those processes an organization wishes to analyze, redesign, and/or automate. These are specified and mapped using (semi-)formal modeling methods. In the implementation phase process models are compiled into the operational environments which can either be manual or automated, e.g. via BPM software. Individual process instances are executed and their performance is monitored and controlled in real time. Ex post process evaluation and improvement, is provided by audit trails which were produced during the process enactment and monitoring stages. During this stage data from multiple process instances is aggregated to discover potential for process improvement and flaws of process design and implementation. Feedback strategies for process redesign and performance improvement can be formulated based on the results and timing of process measurement and evaluation. Thereby, in (perioperative) process management of OR suites, we can integrate simulation into the stages of the BPM lifecycle. If operational perioperative process management is modeled as a transformation process, as proposed here, any discussion of control needs to consider ex ante, inprocess (online), and ex post integration of simulation [28]. Ex ante simulation is a preventive action in the design phase of process management. These ex ante controls are used to perceive analysis and design flaws.
3
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
Simulation can be used for inspection as an ex ante control for process planning and scheduling of doctors and nurse, or medical devices [14], [21], [32], [33]. Additionally, new perioperative or operative strategies could be tested if they are applicable to the specific hospital environment or fit into the design goals of the hospital or OR management [4] . In-process or online simulation is applied during the enactment phase of the process. Because not all possible events can be foreseen or if detailed planning is not feasible or desirable, controls can seek to mitigate negative consequences of wrong design and implementation decisions of processes at runtime. Caused by the high uncertainty in the medical and surgical process, processing times can be seldom predicted accurately. As a result, predefined schedules will not hold for the entire planning period. To prevent the need for rescheduling this problem is usually dealt with by using planned slack [31]. Nevertheless, overlapping cases and interdependence of resource activities in the perioperative process must incorporate real-time controls to ensure the efficient (re-) scheduling of resources. This may also include the activation of additional resources in case of emergencies or the cancellation of cases, when operating hours of the OR suites are not fulfilled during the day. Obviously, there are several restriction when integrating simulation studies in the decision making process during process runtime. This requires above all a timely response of the hard- and software systems and sufficiently fast simulation studies and actions of the decision maker. Ex post simulations are traditionally used to control process performance and the management of negative consequences after process execution. Data collection during process enactment may be integrated into the simulation study in order to identify inefficiencies. This may either lead to a better management or control of similar process instances in the future or the redesign of perioperative processes depending on the impact of process or service quality failures. Ex post simulations have been constantly reported and successfully applied for many years now [40]. Nowadays, many business process vendors implement simulation toolkits into their process engineering software as well, such as Metastrom, IDS Scheer etc. Of course, a comprehensive survey of potential for computer simulation support in the lifecycle would include an evaluation mechanism for comparing simulation studies and real processes before and after the introduction of new perioperative strategies and technologies as proposed by [33]. This would require a detailed documentation of the redesign efforts and
additional data collection, which can imply an enormous effort if this is not automated [44]. In addition, the acceptance of computer simulations has to meet several challenges including the acceptance of staff and decision makers [8]. Both issues are the focus of future work and will not be discussed in this paper in detail. Typically, each of the above introduced simulation concepts face these problems individually and explain some of the problems why simulation is hardly integrated into decision making process [34]. Nevertheless the conceptual model can be seen as a first step into this direction to extend the purpose of simulation for BPM.
4. Experimental and field study simulation Operating room or perioperative processes contain a certain structure and consist of mainly five subprocesses. These subprocesses are: - Transport and Admission to OR suite (TA): This includes the general preparation of the patient, transportation to the operating theatre and the patients admission to the OR suite. - Induction or induction room (IR): Usually anesthesia doctors and nurses prepare the patient for the surgery through induction and attach monitoring devices. - Surgery or operating room (OR): Operating nurses and doctors prepare the patient, material and device for the surgery. After successful preparation the surgical procedure is performed. - Anesthetic emergence (EMR): This includes the detachment of devices and wake up. - Transport to post-anesthesia care units (PACUs): The patient is returned to the ward area for post surgery monitoring and care. The personnel resources – typically anesthesia doctors (ANDs), anesthesia nurses (ANNs), operating nurse (OPNs), surgeons (SD) and non-medical personnel (NMP) - involved in the different sub processes depend on the organizational design of operating suites, perioperative process strategies, medical cases and more. Typically, operating suites in hospitals operate several streams of the perioperative process in parallel. Figure 2 shows the configuration analyzed in our field study and used in our simulation studies with four IRs and ORs. Thus, the resource interactions and interdependencies of processes inhibit complex structure and behavior. The ultimate goal of perioperative redesign is to reduce or optimize nonoperative time in the OR. In consequence, the simulation model must specify the following main elements [33]:
4
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
-
Process elements contain tasks and routing alternatives, and the business logic of perioperative processes. Resources are characterized by their types, work assignment policies, number and availability. Performance measures or objectives can include service times, lead times, arrival rate of new cases, work-in-progress and/or resource utilization.
Figure 2. Generic process structure of parallel subprocesses in OR suite. The next subsections will outline our field study based simulation. Additionally, experimental simulations will briefly be outlined to indicate effective simulation integration for other stages of the process lifecycle proposed in Section 3.
4.1 Ex post simulation 4.1.1 Problem Formulation The overall objective of the simulation study was the analysis, modeling and simulation of processes and workflows of an OR suite of the University Hospital Mannheim. Therefore, a detailed process analysis was
Figure 3 a. OR suite process.
conducted by interviewing experts of all process participant groups and by documenting ten working days in a field study, where all important patientoriented process activities and corresponding time durations were logged. The relevant information of the field study was gathered in a manual data collection, in a general medical surgery unit organized as service centre with central surgery unit management, inpatient ORs and working with overlapping anesthetizing [12], [20]. The resulting system model of the OR suite shows the main process activities according to Figure 2. Starting from this rather high-level model, we have built a higher resolution model after the process analysis to identify the main bottlenecks of the four parallel processes. The main scope of the model is to identify how different numbers of anesthesia doctors and how different release order times for the next patient (this could be considered as varying slack for all process activities before surgery) effect the overall ANDs resource utilization. In our study we use the Arena® simulation software package. 4.1.2 Model Design and Data Collection The model layout for the main process activities is exemplified in Figure 3a and 3b. The figures show the basic process steps in the simulation environment. The data collection of the field study revealed 92 patient records with a detailed summary of all process activities, resources and activity durations. The field study was necessary because the standard documentation of the processes were not detailed enough to analyze the interactions of resource utilization and overlapping anesthetization within the four parallel processes. The patient records serve as historical system input data for the simulation. The model validation was done twofold. The model was verified and conceptually validated by a research team of computer engineers, information systems researchers and technically skilled doctors and nurses of the hospital. The senior medical staff and operating theatre management also approved the key model assumptions and simulation results by performing a structured walk-through of the model. [27] Moreover, we performed a validation of the output of the
Figure 3 b. Detailed OR 1 subprocess.
5
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
Table 1. Input parameters: Patient process activities, resources and time durations. Process Step
Activity
Resources
Duration [min]
TA
Transport to OR
1 NMP
1.5+ GAMMA(7.43,2.41)
Admission
2 OPNs
1.5+LOGN (4.14,3,93)
Narcosis
1 AN, 1ANN
RV
Special 1 AND, Instrumentation 1 OPN Special 1 SD, 1 OPN bedding Surgery 2 OPNs, 3 SDs After surgery 2 SDs care Wake-up 1 OPN OR cleaning 1 NMP
RV
IR
OR
EMR
PACU
Transport to PACU
1 ANN, 1 OPN
Change bedding Post Surgery Care
1 OPN
RV RV RV RV -0.5+ LOGN(15.8,23.6) -0.5+ LOGN (2.12,2.52) -0.5+ LOGN(2.25,1.27) -0.5+ LOGN(4.18,5.26)
1 ANN
simulation model and a trace-driven validation through a basic inspection approach. We have evaluated whether the mapping of activities and subtasks of a surgery process is adequately comprehensive and whether the interplay and coordination of the four ORs (including the concept of overlapping anaesthetization) is represented close to reality. Therefore, we validate the model according to process patient throughput time and OR running times. The trace-driven validation of the simulation showed a shorter patient throughput time than reality, with a median of 13 minutes. With an average of daily 2.5 surgeries per OR the OR running time consequently differed with a median of 34.5 minutes from the simulation model [44]. Although the basic inspection approach has its disadvantages [27]
the overall credibility of the model was approved by the operating theatre management [44]. 4.1.3. Experimental Design and Simulation The simulation was conducted under several assumptions that will be summarized here shortly. The simulation is initialized with the 92 patient records. This may be considered as a basic surgery schedule for the ten working days because all patient where executed according to the basic surgery schedule of the OR management. Each patient must be ordered or release by the surgeon for the operation. Thus, release order times can significantly influence the utilization of resources in the overlapping procedure process, e.g. waiting surgeons in the OR or higher utilization of anesthesia doctors in the IR. In this study we only consider different numbers of ANDs and varying order release times for the next patient because we focus on perioperative activities for timing overlapping anesthetization. The simulation is executed as a trace driven simulation [27] for 10 working days. For the most important activities influencing the perioperative performance real historic data values have been used. All other process activities have been approximated by theoretical distributions. The main processes and activities, resources and durations are summarized in Table 1. The distributions have been calculated by the Arena Input Analyzer, where GAMMA is gamma(alpha, beta), WEIB is Weibull(beta, alpha), and LOGN is lognormal(logMean, logStd). RV indicates that for this activity real data values are used. We obtained stable result for using 30 replications of simulation runs presented in the next subsection. 4.1.4 Results and Interpretation The data of the field study disclosed that the optimal overall (from transportation until end of anesthesia preparation) preparation time for the patient
Patient Ordering Times [min]
100 90 80 70 60 50 40 30 20 10 0
95 94 96 96
41
51
Patient Ordering Times [min]
61
21 100
87 88 88 89
78 80 81 82 67
72 74 75 59
63 65 66
Average Percentage [%]
Average Percentage [%]
21
90 80
41
51
61
95 94 94 93
96 94 96 95
96 99 95 96
4
5
6
80 79 81 81
70 60 50 40 30 20 10 0
3
4
5
6
7
Number of ANDs
Figure 4. Average utilization of ANDs for different numbers of ANDs and different patient ordering times.
3
Number of ANDs
Figure 5. Average utilization of OR 1 for different numbers of ANDs and different patient ordering times.
6
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
is 41 minutes in average. In contrast, the empirical duration of an ordered patient until the patient is in the OR is in average 74 minutes (with standard deviation of 35 minutes) although the operating room is released after 54 minutes in average. As a result, the process theoretically reveals potential improvements. In Figure 4 and Figure 5, we study the effect of different time periods for ordering the next patient before the end of the operation. Figure 4 shows the average utilization (in %) of an anesthesia doctors for different number of doctors (3-7 for all four streams) and different patient ordering times (21, 41, 51, 61 minutes, which are based on data analysis of the process for key process steps before surgery) before the end of operation. Figure 5 shows the average utilization of operating room one with respect to different patient ordering times. The simulation results in Figure 4 show a clear trend of decreasing anesthesia doctor utilization when increasing the available doctors from 3 to 7. The result reflects the intuition that more resources produce the same service output. Nevertheless, the introduction of another doctor from 6 to 7 results in an even bigger decrease in utilization of doctors than before. This could reflect a critical point in the perioperative process where the introduction of an additional doctor has a negative impact on performance. An earlier ordering time of patient increases the utilization of anesthesia doctors, which is also straightforward because the earlier the patient arrives in the preparation room the longer is the time of anesthesia presence. Thus, overlapping anesthetizing requires the introduction of additional doctors with the hope for higher patient throughput. Figure 5 shows the effect of variation of anesthesia doctors and patient ordering times on operating room utilization; we use OR 1 as an example. Additional doctors have only strong positive effect on utilization of the operating room from 3 to 4 doctors. Later, the effect is rather weak in our simulation model. The effect of different release order times on operating room utilization is not observable because there is no significant change in utilization. The OR model implies inefficiencies at the transition from preparation room to the OR because of rather small improvements of earlier ordering times of patients. The results of the effects studied in our Table 2. Simulation model design parameters per stream. Activity IR OR surgery OR suture EMR
Resources 1 ANN, 1 AND, 1 IR 2 SD, 1 AND, 1 OPN, 1 OR 2 SD, 1 AND, 1 OPN, 1 OR 1 OR, 1 AND, 1 OPN
Duration [h] Expo (.5) Expo (1) Expo(.33) Expo(.25)
simulation reflect and confirm the findings in recent studies and their ongoing discussions in literature [12], [24]. Despite the small number of patient records used in the simulation study, which is one of the major shortcomings, the simulation revealed several problems in the OR. Hence the model serves as the starting point for the optimization of processes in the University Hospital of Mannheim. In further studies we will integrate a larger amount of data records and the variation of different resource types.
4.2 Ex ante simulation: experimental study The investigation of new perioperative strategies for process design or redesign is essential for a reliable and profound decision making which strategies or innovations should be introduced prior to final implementation into the real process. The use of simulation at this stage of the BPM lifecycle can provide many insights into the effectiveness of perioperative strategies. We try to clear this proposition by briefly reporting some results of an experimental study. The goal of this experimental study was the evaluation of different operating hours of the OR suite and the variation of surgical durations in the OR subprocess. As data input (= time duration of activities) serves the estimation of doctors and nurses of our field study. The simulation model is a high-level model of the process identified and analyzed in the previous subsection. Table 2 presents the main activities for one of the four streams and provides the corresponding process activities, the involved resources and the time durations used for simulation (Expo(beta) is exponential(beta)). The experimental simulation studies the utilization of resources and the output of patients or work in-progress (patients in the system at the end of operation time) for the four streams OR suite. Thereby, we vary the operating hours per day of the service center from 6 to 16 hours. Figure 6 and Figure 7 present results for one hour surgery cases and operating hours from 6 to 16 hours per day. The results in Figure 6 suggest that short surgeries with fixed resource allocation slightly improve the utilization of resources when operating hours are extended. Moreover, Figure 7 implies the introduction of specific operating hours for cases with certain time distribution of surgical time in the OR. In the case presented here, 10 hours of operation seem to be the ideal duration for optimally timed patient throughput under the given process parameters. The gained results can be related to the findings in the literature where optimal OR suite runtimes depend on
7
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
ANNs
SDs
OPNs
Patient Output
Number of Patients
Average Utilization [%]
ANDs 800% 700% 600% 500% 400% 300% 200% 100% 0% 6h
8h
10h
12h
14h
16h
OR Suite Operating Time [h]
Figure 6. Average resource utilization with a constant number of resources and an increase in operation time of OR suite. similar characteristics of the operation types and variance of activity durations in the OR process. In [31] the planned slack is minimized by using portfolio effects of planned surgeries to build master surgery schedules and [24] compare different perioperative strategies with regard to their impact on economics and show that the surgery types investigated reveal similar variances in procedure durations. Although experimental, our studies can be used to further investigate the parallel processing, rapid turnover times and extra throughput of patients. Accordingly, cases with shorter operative times benefit most from the extra resources through the enhancement in perioperative efficiency which is also confirmed in literature [24].
4.3 In-process simulation support In addition to the traditional ex post analysis of patient queues and resource allocation policies, we are more and more interested in modeling, simulation and analysis of perioperative processes in real-time. Thereby, the computer simulation environment is part of the information and decision support system in a distributed software architecture, where real world system data is provided to the simulation environment. This can be used to provide the decision makers with up-to-date information about the system state. The decision maker can then use the data for simulation during operation hours. For example, the appearance of an emergency case can necessitate the re-scheduling of planned cases in the operating room suite. The injection of the emergency case into the normal operation of perioperative processes can be reported instantly to the simulation system. The results of the online simulation can offer valuable management alternatives for readjusting the operative program during the day. We realized a prototype (as part of a simulation environment) that integrates dynamically external data into the simulation that is provided by external webservices. The work is
Work In Progress
25 20 15 10 5 0 6h
8h 10h 12h 14h OR Suite Operating Time [h]
16h
Figure 7. Patient throughput and work in progress with different operation times. currently under research, but first studies show promising results for online simulation. Though online simulation can provide a powerful addition to operational control of perioperative processes, several limitations exist in practice for its application. Most simulation tools assume steady-state processes instead of transient processes. Thus, adequate tool support must be guaranteed to use online simulation. Additionally, online data collection in perioperative processes is often done manually and time dependent information may not be delivered on time. This could be supported by recent technological developments such as automated data collection for hospital information systems by medical device manufacturers, and/or ubiquitous and mobile systems. [42]
5. Discussion and Limitations Simulation proved to be an effective tool for validation revealing many insights into the perioperative process. Additionally, the studies in Section 4 indicate opportunities that simulation can be effectively integrated into the BPM lifecycle. Recent studies in perioperative process redesign reported patient throughput improvements in various field studies [14], [16], [20], [24], [35], [37]. The studies suggested different strategies to improve the perioperative process. They could be used to further study the benefits and drawbacks of simulation studies in operating room management. Moreover, although the studies do not report the effort and costs for investigating and implementing their strategies, simulation studies could have provided a safe and economic means to provide decision support in applying these methods. Although simulation has been used for many years our simulation studies revealed several problems and limitations in transferring the perioperative processes onto a simulation model. Several artifact and process entities in operating theatres can not easily be transferred and represented in simulation model. This is not caused by a specific tool but by the
8
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
characteristics of the processes and resources. Other studies using simulation mention similar generic problems [33]. The first problem is related to the resources in the perioperative process. Doctors and nurses are normally involved in more than one process at a time. Simulation models do not reflect the sharing of resources over different processes. In the perioperative process we studied anesthesia nurses switch between processes almost randomly depending on the workload or other conditions in that process. This can only be incorporated by sophisticated modeling techniques. The second problem is similar to the first one in the sense that resources can incorporate skills of different professions. For example, an anesthesia nurse and an operating nurse can be used interchangeably for several tasks in the process. These overlapping skill levels could be represented by a specific skill set for specific task in the process. The third problem is that the perioperative processes face different workloads in case of complications in the surgery or in case of emergencies. The introduction of new process strategies in some parts of the process can also increase the workload in another part. Simulation tools could incorporate elasticity models for different workload to integrate this phenomenon [33]. The studies presented in the literature assume that the quality of treatment remains constant. None of the recent studies makes notes on the potential risks for the patients. Simulation concerning quality models could raise attention to this problem, but of course have to be validated by empirical studies. Other problems for modeling are for example the quality of the surgical process and the steadily increasing performance maximization of perioperative activities, the fact that people often work with different speed or the surgical performance provided in the operating theatre. The above mentioned problems reveal several issues in mapping perioperative artifact onto simulation models. These are not so much of technical nature but of ontological nature of the problem domain under research. Although several simulation tools might solve some of the problems, it is better to find new ways to model the specific nature of the OR suite.
6. Conclusion and Future Work In this paper, we proposed a conceptual framework to use computer simulations in different phases of the BPM lifecycle for OR management. Furthermore, we conducted (experimental) simulation studies to evaluate the impact on decision making before and during process execution. Thereby, we conducted a
field study and analyzed the behavior of the system in detail with regard to different number of resources. The challenges identified in Section 1 have been addressed in various aspects. The process stakeholders are supported by simulation at each stage in the BP lifecycle. As a result, the decision making process guarantees ex ante, ex interim and ex post support. The improvement strategies proposed in literature and the various operational research techniques for OR management can be integrated into the simulation environment. Thus, analytical results enhance simulation-based decision making by transparent illustrations [4]. The simulation studies in Section 4 show that high level analysis may not reveal the veritable bottleneck in the perioperative process. Therefore, simulation models must be designed with requisite detail in the patient processes and resource utilization for an enhanced analysis. In the future we will work on several problems mentioned in Section 5 and the successful deployment of high-throughput practices throughout operating rooms.
7. References [1] W. v. d. Aalst and K. v. Hee, Workflow Management: Models, Methods, and Systems. The MIT Press Cambridge, Massachusetts London, England, 2002. [2] J. G. Anderson, "Evaluation in health informatics: computer simulation," 2002. pp. 151-164. [3] K. Barkaoui, P. Dechambre and R. Hachicha, "Verification and optimization of an operating room workflow," in Proceedings of HICSS 2002. [4] J. Belien, "Exact and Heuristic Methodologies for Scheduling in Hospitals: Problems, Formulations and Algorithms," Katholieke Universiteit Leuven, 2006. [5] J. Belien and E. Demeulemeester, "Building cyclic master surgery schedules with leveled resulting bed occupancy," European Journal of Operational Research, 2005. In Press. [6] J. T. Blake and M. W. Carter, "Surgical Process Management: A Conceptual Framework," Surgical Services Management, vol. 3, September 1997. pp. 31-37. [7] J. T. Blake, M. W. Carter and S. Richardson, "An analysis of emergency room wait time issues via computer simulation," INFOR, vol. 34, 1996. pp. 263-272. [8] M. A. Centeno and M. Carrillo, "Simulation in practice: Challenges of introducing simulation as a decision making tool," in Proceedings of WSC 2001. [9] T. H. Davenport, Process Innovation: Reengineering Work through Information Technology. Harvard Business School Press, Boston, 1993. [10] R. Davies and H. T. O. Davies, "Modeling patient flows and resources provision in health systems," International Journal of Management Science, vol. 19, 1994. pp. 173-182. [11] M. den Hengst, V. Hlupic and W. L. Currie, "The increasing need for integrating simulation and collaboration
9
Proceedings of the 40th Hawaii International Conference on System Sciences - 2007
to support change management programs," in Proceedings of HICSS 2004. [12] F. Dexter, "Deciding whether your hospital can apply clinical trial results of strategies to increase productivity by reducing anesthesia and turnover times," Anesthesiology, vol. 103, 2005. pp. 225-228. [13] F. Dexter and R. H. Epstein, "Operating room efficiency and scheduling," Current Opinion in Anesthesiology, vol. 18, 2005. pp. 195-198. [14] F. Dexter, A. Macario, D. A. Lubarsky and D. D. Burns, "Statistical method to evaluate management strategies to decrease variability in operating room utilization: Application of linear statistical modeling and Monte Carlo simulation to operating room management," Anesthesiology, vol. 91, 1999. pp. 262-274. [15] J. E. Everett, "A decision support simulation model for the management of an elective surgery waiting system," Health Care Management Science, vol. 5, 2002. pp. 89-95. [16] S. Freytag, F. Dexter, R. H. Epstein, C. Kugler and R. Schnettler, "Allocating and scheduling operating room time," Chirurg, vol. 76, Jan. 2005. pp. 71-79. [17] A. Guinet and S. Chaabane, "Operating theatre planning," International Journal of Production Economics, vol. 85, 2003. pp. 69-81. [18] M. Hammer and J. Champy, Reengineering the Corporation: A Manifesto for Business Revolution. Harper Business, 1993. [19] E. W. Hans, G. Wullink, M. v. Houdenhoven and G. Kazemier, "Robust surgery loading," Submitted to European Journal of Operational Research, 2005. [20] R. Hanss, B. Buttgereit, P. H. Tonner, B. Bein, A. Schleppers, M. Steinfath and Bauer,Jens Scholz and Martin, "Overlapping induction of anesthesia: an analysis of benefits and costs." Anesthesiology, vol. 103, 2005. pp. 391-400. [21] A. Jebali, Hadj Alouane, Atidel B. and P. Ladet, "Operating rooms scheduling," International Journal of Production Economics, vol. 99, 2006. pp. 52-62. [22] J. B. Jun and S. H. Jacobson, "Application of discreteevent simulation in health care clinics: A survey," J. Oper. Res. Soc., vol. 50, Palgrave Macmillan. 1999. pp. 109. [23] D. W. Kelton, R. R. Sadowski and A. D. Sadowski, Simulation with Arena. New York: McGraw-Hill, 1998. [24] D. C. Krupka and W. S. Sandberg, "Operating room design and its impact on operating room economics," Anesthesiology, vol. 19, 2006. pp. 185-191. [25] N. K. Kwak, P. J. Kuzdrall and H. H. Schmitz, "The GPSS Simulation of Scheduling Policies for Surgical Patients," Management Science, vol. 22, INFORMS: Institute for Operations Research. 1976. pp. 982-989. [26] M. Lamiri, F. Grimaud and X. Xie, "Optimization methods for surgery planning under uncertain demand for emergency surgery," in Proceeding of INCOM 2006. [27] A. M. Law and W. D. Kelton, Simulation Modeling and Analysis, 3rd ed. Boston: McGraw-Hill, 2000. [28] M. A. Lewis, "Cause, consequence and control: towards a theoretical and practical model of operational risk," J. of Operations Management, vol. 21, 2003. pp. 205-224. [29] J. C. Lowery and J. A. Davis, "Determination of operating room requirements using simulation," in Proceeding of WSC 1999.
[30] M. J. Miller and D. M. Ferrin, "The application of simulation methodology in a hospital's six sigma project," in Proceedings of WSC 2005. [31] J. M. v. Oostrum, M. v. Houdenhoven, J. L. Hurink, E. W. Hans, G. Wullink and G. Kazemier, "A model for generating master surgical schedules to allow cycles scheduling in operating room departements," Departement of Applied Mathematics, University of Twente, Tech. Rep. 1789, December 2005. [32] M. L. Pinedo, Planning and Scheduling in Manufacturing and Services. Berlin, Springer, 2005. [33] H. A. Reijers and van der Aalst,Wil M.P., "The effectiveness of workflow management systems: Predictions and lessons learned," International Journal of Information Management, vol. 25, 2005. pp. 458-472. [34] S. Robinson, "Discrete-event simulation: from the pioneers to the present, what next?" J. Oper. Res. Soc., vol. 56, Palgrave Macmillan. 2005. pp. 619-629. [35] W. S. Sandberg, B. Daily, M. Egan, Goldman,James E Stahl and Julian M., R. A. Wiklund and D. Rattner, "Deliberate perioperative systems design improves operating room throughput." Anesthesiology, vol. 103, 2005. pp. 406418. [36] H. Smith and P. Fingar, Business Process Management the Third Wave. Meghan Kiffer Press, 2003. [37] P. M. Torkki, R. A. Marjamaa, M. I. Torkki, P. E. Kallio and O. A. Kirvelä, "Use of anesthesia induction rooms can increase the number of urgent orthopedic cases completed within 7 hours." Anesthesiology, vol. 103, Aug. 2005. pp. 401-405. [38] J. M. H. Vissers, J. W. M. Bertrand and G. De Vries, "A framework for production control in health care organizations," Production Planning & Control, vol. 12, Taylor & Francis 2001. pp. 591-604. [39] P. Volkner and B. Werners, "A simulation-based decision support system for business process planning," Fuzzy Sets and Systems, vol. 125, 2002. pp. 275-287. [40] K. P. White Jr, "A survey of data resources for simulating patient flows in healthcare delivery systems," in Proceeding of WSC 2005. [41] G. Wullink, Gademann,A J R M., E. W. Hans and A. van Harten, "Scenario-based approach for flexible resource loading under uncertainty," International Journal of Production Research, vol. 42, 2004. pp. 5079. [42] Y. Xiao, M. Wasei, P. Hu, P. Wieringa and F. Dexter, "Dynamic management in perioperative processes: A modeling and visualization paradigm," in Proceeding of INCOM 2006. [43] T. Young, S. Brailsford, C. Connell, R. Davies, P. Harper and J. H. Klein., “Using industrial processes to improve patient care,” BMJ 328pp. 162-164. Avail.: bmj.com. [44] A. Zoeller, D. Christof, A. Baumgart, H. J. Bender, E. Badreddin and A. Heinzl, "Information nedd and IT demands for business process reengineering in operation room management," in Proceeding of ECEH, 2006. [45] M. zur Muehlen, Workflow-Based Process Controlling. Foundation, Design, and Application of Workflow-Driven Process Information Systems. Logos, Berlin, 2004.
10
