Вы находитесь на странице: 1из 8

A simulation study to analyze the impact of different emergency physician shift structures in an emergency department

A SIMULATION STUDY TO ANALYZE THE IMPACT OF DIFFERENT EMERGENCY PHYSICIAN SHIFT STRUCTURES IN AN EMERGENCY DEPARTMENT
Rene Alvarez, IE.1,2 Guillermo A. Sandoval, MBA.3,4 Sergio Quijada, Colonel, PhD Modeling and Simulation.5 Adalsteinn D. Brown, PhD.3,4
2

Saint Michaels Hospital, Toronto, Ontario, Canada Centre for Research in Healthcare Engineering, Department of Mechanical and Industrial Engineering, University of Toronto, Canada 3 Health System Strategy Division, Ministry of Health and Long-Term Care, Government of Ontario, Canada 4 Department of Health Policy, Management and Evaluation, University of Toronto, Canada 5 Army Simulation Center, Chilean Army, Chile

Corresponding author: Rene Alvarez, Centre for Research in Healthcare Engineering, Department of Mechanical and Industrial Engineering at the University of Toronto. 5 King's College Road, Toronto, Ontario Canada M5S 3G8; telephone: 416-864-6060 ext 7110; fax: 416-846-5779; e-mail: rene.alvarez@utoronto.ca. Individuals or organizations whose assistance is acknowledged: Hospital Padre Hurtado, Santiago de Chile.

ABSTRACT
Background: Hospitals need to provide timely access to emergency department services in an effective and efficient way. Computer simulation has become a widely accepted tool for evaluating different operational schemes to support managerial decision making processes. We report a simulation study to analyze the impact of different emergency physician shift structures on two emergency department (ED) performance indicators. Objectives: i) To improve emergency physician utilization rates, measured as the total time spent by the emergency physician attending patients divided by the duration of the shift, and ii) To improve average patient turnaround times, measured as the time elapsed between the instant the patient arrives to the ED and the instant the patient leaves the ED. Methods: We constructed a discrete event simulation model that generates patients that follow different paths throughout the ED. The model could simulate the emergency department behavior when changing the emergency physicians schedules and the number of physicians attending patients simultaneously. We evaluated 6 different scenarios including the current operation. Results: By changing the emergency physician shift structure and the number of emergency physicians attending patients simultaneously, the model showed a 49.5% reduction on patient turnaround times, and a 71.8% increase in physician utilization rates. At the same time, the ED requirement of emergency physician hours per day decreased by 38.1%. Conclusions: Computer simulation showed that it is possible to decrease the number of emergency physician hours per day and, at the same time, decrease patient turnaround times. Key words: Computer simulation; Emergency department; Emergency physician shift structure; Patient turnaround times; Physician utilization rates.

Proceedings of the 35th International Conference on Operational Research Applied to Health Services (ORAHS) July 12-17, 2009, Leuven, Belgium ISBN: 9789081409902

A simulation study to analyze the impact of different emergency physician shift structures in an emergency department

INTRODUCTION
Healthcare expenditures have persistently grown in the last several decades worldwide.1 At the same time, hospital care continues to absorb a substantial portion of total health expenditures despite efforts to shift care to other settings.2-3 This pressures hospital management to reduce costs and increase productivity. On the other hand, emergency department (ED) overcrowding has become a global concern since it threatens the capacity of the hospitals to provide quality and timely emergency care.4 Reducing costs while improving turnaround times is a challenging management objective. However, the use of simulation tools to plan and design new operational schemes might be one potential approach to address these two objectives simultaneously. In the past several years, computer simulations have proven to be a powerful tool in assisting the managerial decision making process in this complex, stochastic health care environment.5-8 It has been suggested that simulation modeling is the best suited tool to re-balance capacity variables associated with process improvements and in turn realize the desired cost savings.9 There exists a common understanding in the health industry that although computer simulations do not provide the optimal solution and do not predict the future, they do provide answers to what if questions.10 The main goal of this study was to develop a simulation model to understand how different operational schemes could impact two important ED performance indicators: emergency physician utilization rates (proportion of available time attending patients), and patient turnaround times (total time spent in the system).

METHODS AND MATERIALS


Data used in this study were obtained from a publicly funded teaching hospital located in a low income neighborhood in Santiago de Chile. The hospital has 351 beds, and provides emergency, ambulatory, inpatient, and palliative care services for certain conditions. Patients include children, adults, and newborns. More specialized patients, such as those requiring oncology, neurosurgery, and dermatology services, are transferred to a nearby major hospital. The hospital has an annual budget of close to US$32 million (2006) and its catchment population borders 450 thousand. The hospital has a pediatric emergency department (ED) that attends between 200 and 300 patients daily. Traditionally during the month of July, the pediatric ED experiences the highest demand of the year due to upper respiratory tract infections (URTI) caused by low temperatures and the presence of smog in the air during south-hemisphere wintertime. The ED is divided in two main care services areas: i) pediatric care, and ii) surgical care. During 2005, 89% of patients sought pediatric care and 11% surgical care. Figure 1 describes the patient flow inside the ED. Currently, the total number of physician hours in the ED is 126 per day. In the pediatric area of the ED, three pediatricians work in 24 hour shifts simultaneously, starting at 8:00 am. At any given hour, however, only one of the three pediatricians attends patients. The other two do administrative work, rest, or occasionally attend patients. In other words, the pediatric area of the ED operates as a system with only one server (one pediatrician) most of the year. The three pediatricians rotate systematically (every hour) during the day to attend patients. During high demand winter days there is an additional pediatrician who supports the team (from 3:00 pm to 9:00 pm). During this time, the pediatric area of the ED operates as a system with two servers in parallel (two pediatricians). On the other hand, in the surgical area of the ED, 2 surgeons work in 24 hour shifts simultaneously, starting at 8 AM. These surgeons, however, provide services to other units within the hospital (e.g. clinics, operating rooms) and they are not exclusively dedicated to the ED. From a simulation perspective, these surgeons were considered two servers in parallel but with limited availability during the day.

Proceedings of the 35th International Conference on Operational Research Applied to Health Services (ORAHS) July 12-17, 2009, Leuven, Belgium ISBN: 9789081409902

A simulation study to analyze the impact of different emergency physician shift structures in an emergency department

Figure 1: ED patient flow diagram for both pediatric and surgical patients Similar to Dawson and OConnor (1994), we defined the following simulation objectives: To explore different emergency physician shift structures to: i) improve physician utilization rates, measured as the total time spent by the physician attending patients divided by the duration of the shift, and ii) improve average patient turnaround time, measured as the time elapsed between the instant the patient arrives to the ED and the instant the patient leaves the ED. 11 The simulation scenarios were defined based on the following criteria: a. Physician shift structure following the demand pattern: Figure 2 shows that at night demand decreases dramatically affecting overall physician utilization rates. This suggested that it could be more efficient to consider two 12-hours shifts, instead of the actual 24-hour shifts. The night shift could consider a reduced number of physicians on duty in comparison to the day shift. In addition, Figure 2 also shows that it might prove useful to delay the shift starting time from 8 am to 10 am, following the demand pattern. Physicians on duty attending patients simultaneously: Currently, physicians attend patients continuously in 24-hours shifts rotating every hour (i.e. one server). However, having 12-hours shifts (including one hour for lunch or rest) physicians could work simultaneously and continuously assisting patients (i.e. parallel servers). This could decrease patient turnaround time in comparison to the current situation. Full time ED dedicated surgeons: As mentioned before, surgeons are not exclusively dedicated to the ED. Having full time ED surgeons could reduce patient turnaround time.

b.

c.

Proceedings of the 35th International Conference on Operational Research Applied to Health Services (ORAHS) July 12-17, 2009, Leuven, Belgium ISBN: 9789081409902

A simulation study to analyze the impact of different emergency physician shift structures in an emergency department

Figure 2. Patient arrival rate (every 15 minutes) for both pediatric and surgical patients, used to replicate a typical high demand day. Each 15-minute rate was obtained by averaging daily patients arrivals during July 2005 In order to achieve the simulation objectives, five simulation scenarios were agreed upon with the management of the hospital in addition to the base scenario (Table 1).

Type of emergency physician

Pediatrician Supporting Pediatrician Surgeon Pediatrician Pediatrician Supporting Pediatrician Surgeon Pediatrician Pediatrician Supporting Pediatrician Surgeon

Number of emergency Duration physicians per [hours] shift Base Scenario 3 24 1 6 2 24 Scenario A 2 12 1 12 0 --2 24 Scenario B 2 12 1 12 1 6 2 24

Shift Starts 8:00 AM 3:00 PM 8:00 AM 10:00 AM 10:00 PM --8:00 AM 10:00 AM 10:00 PM 3:00 PM 8:00 AM Ends 8:00 AM 9:00 PM 8:00 AM 10:00 PM 10:00 AM --8:00 AM 10:00 PM 10:00 AM 9:00 PM 8:00 AM

Proceedings of the 35th International Conference on Operational Research Applied to Health Services (ORAHS) July 12-17, 2009, Leuven, Belgium ISBN: 9789081409902

A simulation study to analyze the impact of different emergency physician shift structures in an emergency department

Scenario C Pediatrician Pediatrician Supporting Pediatrician Surgeon Surgeon Pediatrician Pediatrician Supporting Pediatrician Surgeon Surgeon Pediatrician Pediatrician Supporting Pediatrician Surgeon Surgeon Table 1: Simulation Scenarios A discrete event simulation model of the ED was implemented in ARENA (Rockwell Automation, Inc. Warrendale, PA, USA). The model replicates the arrival of patients at the ED in a typical high demand day using a non-homogeneous Poisson process. This process is widely used in discrete event simulation when simulating arrivals of entities to a system. It has the following characteristics: events occur continuously and independently of one another (the number of occurrences counted in disjoint intervals are independent from each other), more than one event cannot occur at the same instant, stationary increments (the probability distribution of the number of occurrences counted in any time interval only depends on the length of the interval), and arrival rates change over time. 2 1 0 1 1 Scenario D 2 1 1 1 1 Scenario E 3 1 1 1 1 12 12 6 12 12 10:00 AM 10:00 PM 3:00 PM 10:00 PM 10:00 AM 10:00 PM 10:00 AM 9:00 PM 10:00 AM 10:00 PM 12 12 6 12 12 10:00 AM 10:00 PM 3:00 PM 10:00 AM 10:00 PM 10:00 PM 10:00 AM 9:00 PM 10:00 PM 10:00 AM 12 12 --12 12 10:00 AM 10:00 PM --10:00 AM 10:00 PM 10:00 PM 10:00 AM --10:00 PM 10:00 AM

The arrival of patients to the ED could be represented by a non-homogeneous Poisson process because it follows all of the above characteristics. In our simulation model, the Poisson arrival rate was changed every 15 minutes to accurately represent the current demand pattern (Figure 2). Rates were calculated as the average rate of the arrivals registered during July 2005 for each 15 minutes gap. In 2005, there were not extraordinary circumstances that affected the demand pattern, so it is possible to consider this year as a typical year. Data collected during 2005 by the hospital was analyzed to determine: i) the proportion of patients requiring pediatric care and surgical care (type of patient), and ii) the proportion of patients that follow each of the multiple routes inside the ED (Figure 3). In the first stage of the simulation model, a decision making module splits the arrival of patients, assigning to each of them the attribute type of patient (89% for pediatric patients and 11% for surgical patients). Then a predetermined route is assigned to each patient (e.g. reception triage physician x rays physician exit). This technique (sequencing) allows dealing with multiple paths and several kinds of entities without reducing the accuracy of the model. In this way we avoid using multiple decision making modules in each station of a route.

Proceedings of the 35th International Conference on Operational Research Applied to Health Services (ORAHS) July 12-17, 2009, Leuven, Belgium ISBN: 9789081409902

A simulation study to analyze the impact of different emergency physician shift structures in an emergency department

Figure : Venn diagram showing routes followed by patients using a base of 1,000 patients. 475 patients were attended by the emergency physician and did not require any further treatment, imaging, or examinations. The hospital has a comprehensive historical database with records of time spent by patients in each stage of the ED. Using information collected in 2005 we fitted probability distributions for processing times to be used in the simulation model (i.e. reception, triage, x-rays, treatment, and examinations). Log-normal probability distributions were used to represent most of these processing times. In a few cases, however, processing times were simulated using empirical distributions. As an example, emergency physician attention times were modeled using these empirical distributions since the log-normal assumption was probed not to be the most adequate. The model design and its assumptions were approved by the hospital management and ED physicians. As a way to demonstrate the capability of the model to capture the complexity and stochastic nature of the ED, the current situation was replicated as the base scenario. ED management issues a monthly comprehensive report that includes ED performance indicators, including waiting times. These indicators were used to compare the base scenario simulation results with the actual behavior, and thus to validate the model. The use of model animation and a comprehensive visual control panel was particularly useful in this stage to improve communication between simulation specialists and physicians. The validation process included: (i) probability distribution calibration, (ii) patients flow corroboration (e.g. waiting time after triage), and (iii) input/output verifications (e.g. patient arrivals per hour). The process ended when the model replicated the current situation within a 5% difference in: i) average patient turnaround time, and ii) the total daily patient arrivals. Once the hospital management and the physicians approved the model, we evaluated the different scenarios presented in Table 1. For this study we chose a typical high demand day (i.e. during July), and a total of 48 hours of ED operation were simulated. When simulation starts the system is empty, therefore it was necessary to consider the first 24 hours of each replica as a warm-up period in order to achieve more realistic results. After 100 replicas the average simulation results became fairly stable, therefore we decided to perform this number of replications of each scenario. By using paired t-test, we compared the results of the different scenarios.

Proceedings of the 35th International Conference on Operational Research Applied to Health Services (ORAHS) July 12-17, 2009, Leuven, Belgium ISBN: 9789081409902

A simulation study to analyze the impact of different emergency physician shift structures in an emergency department

RESULTS
Statistic tests confirmed that results were statistically different, and scenario E presented the best combined indicators. In comparison to the base scenario, scenario E showed a 49.5% reduction in average patient turnaround time (from 1.75 hours to 0.88 hours), and a 71.8% increase in pediatrician utilization rates (from 44.7% to 76.8%). Daily requirement of physician hours in the ED decreased considerably as well. Scenario E requires only 78 hours per day (6 physicians working 12-hour shifts, and one pediatrician working from 3:00 pm to 9:00 pm). This change in the shift structure represents a 38.1% improvement in comparison to the current 126 hours (base scenario). Results for all scenarios are presented in Table 2.

Indicator Pediatrician utilization rates Average [%] Maximum [%] Surgeon utilization rates Average [%] Maximum [%] Average patient turnaround time Average [hours]
*

Base* 44.70% 53.70% * * 1.75

A 100.00% 100.00% 9.00% 14.50% 5.21

B 91.30% 91.3 9.50% 15.50% 1.98

C 99.70% 100 19.50% 36.00% 5.13

D 93.30% 100 20.00% 35.50% 2.05

E 76.80% 92.80% 20.00% 37.50% 0.88

Surgeons utilization in the base scenario cannot be calculated because they do not work exclusively for the ED

Table 2: Simulation Results after 100 replications

DISCUSION
Although these results seem both more efficient and effective than the base situation, they could have been influenced by a particularly inefficient base situation. The 24-hour shifts were initially established because physicians had difficulties accessing the hospital due to poor infrastructure of the city where the hospital is located. This situation, however, has changed. Investments in modern highways have been made lately and accessing the hospital is now easier. Similarly, having one pediatrician attending patients (one server) seemed to be inefficient when, in fact, there were three physicians present at the ED. This operational scheme was initially implemented by the ED management to balance the work load of the pediatricians. Although scenario E in our simulation seems to be a better configuration from different perspectives, it might not be seen as such by emergency physicians. It implies less time working more intensively, which can have negative impacts on physicians income, satisfaction, and administrative work load (e.g. checking laboratory results, speaking with other physicians, overseeing technicians and nurses). From the modeling perspective, when the combinatory of routes and types of entities (patients) is large (as in this case) the complexity of the simulation model could rise dramatically. To deal with this problem we assigned stochastic routes as a sequence attribute to each simulated patient. Results show that the use of Venn diagrams and the sequence technique seems to be an effective solution to simplify the abstraction of the real world. One limitation of this study is that results are not transferable to other situations. Simulation models are constructed and calibrated for a specific situation and a specific set of objectives, and therefore it is not possible to use them universally. Even a slight modification in the ED layout requires important changes in the model and a new calibration process has to be conducted. Another limitation of this study relates to the level of detail included in our model. To reduce the complexity of the model and in line with the objectives of the simulation, we did not consider all the staff working in the ED (e.g. nurses who provide a number care services). However, results showed that

Proceedings of the 35th International Conference on Operational Research Applied to Health Services (ORAHS) July 12-17, 2009, Leuven, Belgium ISBN: 9789081409902

A simulation study to analyze the impact of different emergency physician shift structures in an emergency department

simulation models can be effectively used to evaluate different physician allocation scenarios and thus, could be used to analyze different staff structures as well. Finally, the model was not intended to test the behavior of the ED during unexpected outbreaks, natural disasters, or other unanticipated external phenomena. Future research could use computer simulation to test the ED behavior when the normal demand pattern dramatically changes as in those situations.

CONCLUSION
This computer simulation project enabled the hospital to test new emergency physicians shift configurations to understand their effect on physician utilization rates and patient turnaround times. The model was constructed to answer what if questions and to serve as a managerial tool to assist decisions. The simulation shows that it is possible to increase emergency physician utilization rates, decrease the number of physician hours per day and, at the same time, decrease patient turnaround times. During each stage of the project, the hospital management and ED physicians were involved. They collaborated in defining the problem, establishing the project goals, discussing the model logic, defining the scenarios, and discussing the results. Thus, since the beginning of the process, the use of computer simulation served as a communication tool to achieve a common understanding of the system and find ways to improve it.

REFERENCES
1. Chandrakant, PS. (2003) Public Health and Preventive Medicine in Canada. Elsevier, Toronto, Canada, Fifth edition. Pages 200-201. 2. Aiken, LH; Sochalski, J; Lake, ET. (1997) Studying outcomes of organizational change in health services. Med Care; 35(11): NS6NS18. 3. Brown, AD; Alikhan, LM; Sandoval, GA; Seeman, N; Baker, GR; Pink, GH. (2005) Acute care hospital strategic priorities: perceptions of challenges, control, competition and collaboration in Ontario's evolving healthcare system. Health Q 2005; 8: 36-47. 4. Schull, MJ; Slaughter, PM; Redelmeier, DA. (2002) Urban emergency department overcrowding: defining the problem and eliminating misconceptions. CJEM: Journal of the Canadian Association of Emergency Physicians, Mar 2002; 4, 2 5. Baesler, F; DaCosta, M. (2003) The use of simulation and design of experiments for estimating maximum capacity in an emergency room. Proceedings of the 2003 Winter Simulation Conference. 6. Samaha, S; Armel, W. (2003) The use of simulation to reduce the length of stay in an emergency department. Proceedings of the 2003 Winter Simulation Conference. 7. Blasak, R; Armel, W; Starks D; Hayduk, M. (2003) The use of simulation to evaluate hospital operations between the emergency department and a medical telemetry unit. Proceedings of the 2003 Winter Simulation Conference. 8. Hung, H; Whitehouse, S; ONeill C; Grey, A P; Kissoon, N. (2007) Computer Modeling of Patient Flow in a Pediatric Emergency Department Using Discrete Event Simulation. Pediatric Emergency Care, Volume 23, Number 1, January 2007, pages 5-10. 9. Sanchez, SM; Ferrin, DM; Orgazon, T; Sepulveda, JA; and Ward, TJ. (2000) Emerging issues in health care simulation. Proceedings of the 2000 Winter Simulation Conference. 10. Lowery, JC. (1996) Introduction to simulation in health care. Proceedings of the 1996 Winter Simulation Conference. 11. Dawson, KA; OConnor, K. (1994) How to conduct a successful emergency center staffing simulation study. Proceedings of the 1994 Annual HIMSS Conference 3: 273-89.

Proceedings of the 35th International Conference on Operational Research Applied to Health Services (ORAHS) July 12-17, 2009, Leuven, Belgium ISBN: 9789081409902

Вам также может понравиться