1

An Approach for Scheduling the Operating Rooms of a Hospital

Islam Ali, Prashanth Natarajan, Harshit Sodhani and Amina Lyazidi

IE 632 Scheduling Models Dr. Seokcheon Lee

April 20, 2012

Table of Contents
ABSTRACT INTRODUCTION PROBLEM DEFINITION BACKGROUND LITERATURE REVIEW SIGNIFICANCE SOLUTION APPROACH MODEL CHARACTERISTICS MAIN ARENA MODEL PROCESS DESCRIPTION ARENA ARRIVAL PROCESS ARENA QUEUING AND SCHEDULING PROCESS ARENA OR PROCESS ARENA RECORDING AND DISPOSAL PROCESS EVALUATION BENCHMARK CONCLUSION REFERENCES 3 4 5 7 7 9 10 10 10 12 12 13 14 14 16 ERROR! BOOKMARK NOT DEFINED. 19 20

Abstract
We hear about health care issues in the US a lot in the news, and as many issues there are, there are a tenfold more solution proposals that would or would not solve these issues or at least alleviate their impact. One central issue of health care is its exceedingly high costs that limit the access of healthcare services to the population that can afford these costs or have a proper insurance coverage. One way to address these issues is to effectively manage the Operating Rooms (OR) as they are the most costly and yet the most lucrative units in a hospital. We chose the scheduling model of Arnaout and Kulbashian 2008, which configured a parallel machine environment with stochastic processing times and sequence dependent setup times. The objective was to minimize the makespan, which infers maximizing the utilization of the operating rooms. Using Arena as scheduling and benchmarking tool, LEPST (Longest Expected Processing with Setup Time) was shown to yield the optimal scheduling alternative. As an expansion to their work, we developed the same model taking into consideration dynamic scheduling, that accounts for operations arriving over time as it reorder the scheduled operations whenever a new one arrives. We also used Arena simulation tool to compare between the heuristics and to get the heuristic that yields the best results.

Introduction
Operating rooms are known to be the sink of many expenses. Indeed, operating rooms are the most costly units in hospitals as they require for their maintenance and operation many resources from equipment, time and personnel. They also represent the highest revenue pool of hospitals due to the high costs of surgeries and service fees that US health care is known for. The operating rooms in fact represent more than 40% of a hospitals total revenues and the same amount in terms of the total expenses (Peltokorpi, 2011, p. 370). These characteristics render the operating rooms a strong driver of the performance of any hospital. Thus, it is of primal importance to well manage the OR, however, this task is not without challenges that stem from conflicting priorities of different staff (physicians, nurses, and the hospital management staff) and also from the lack of the resources the OR needs. The challenge is increased if we took into consideration the aging of the population that makes this latter more needful of OR services. (Cardoen, B., Demeulemeester, E., Belin). In order to account for all these, a rigorous scheduling should be put into practice that is tailored to the specific hospital at hand and that would take care of all the constraints and challenges. Such schedule would be compared to other alternatives in order to ascertain the use of an optimal schedule, insuring the highest possible productivity of the operating rooms with the least consumption of resources. This would tremendously increase the revenues of the hospital and decrease its expenses.

Problem Definition
We are concerned with finding an optimal schedule for elective surgery cases for a hospital. As the term elective surgery may induce to confusion, we give hereby a short definition of an elective surgery:

An elective surgery is a planned, non-emergency surgical procedure. It may be either medically required (e.g., cataract surgery), or optional (e.g., breast augmentation or implant) surgery. (In Encyclopedia of Surgery, 2012)

We are considering a hospital with 4 operating rooms and 5 different major types of operations are allotted to these operating rooms. Each of these major surgery types has a different stochastic duration, along with a preparation phase before the actual surgery is conducted. The length of this preparation phase changes according to which type of operation is scheduled before the currently scheduled surgery. Therefore, this problem, converted into job scheduling, is equivalent to a parallel machine model with stochastic processing times, with as constraints, sequence dependent setup times, and with as objective, minimizing the total completion time. We will be also taking into consideration that surgeries will be arriving over time, having a release date, a feature in the model that could also be used for dynamic scheduling and which is an expansion of the work done by Arnaout, J-P. M., and Kulbashian S., 2008.

6 Thus, our OR scheduling problem is as follows: The operating rooms are the machines. The selective surgeries are the jobs. The surgeries can be done in any OR (machine). Operations have a release date. The objective is to find a schedule for the surgeries by minimizing:

This problem is known to be an NP-hard problem, and thus, an appropriate heuristic will be devised for solving it.

Background
Literature Review
Arnaout and Kulbashian 2008, addressed the problem of scheduling selective surgeries in the operating rooms of a Lebanese hospital. They configured the problem as a parallel machine scheduling problem with stochastic processing times for the jobs and with sequence dependent setup times. They chose as an objective the minimization of the makespan, which translates to maximizing the utilization of the operating rooms. Since this problem is NP-hard, the authors proposed three main heuristic s: LPT (Longest Processing Time First) and SPT (Shortest Processing Time First), LEPT (Longest Expected Processing First) and SEPT (Shortest Expected Processing Time First), and LEPST (Longest Expected Processing with Setup Time). The authors compared and verified the last two main heuristics using ARENA simulation tool, which showed that the most effective heuristic algorithm is LEPST. The authors ended up by suggesting considering dynamic scheduling as an extension to their work. Cardoen, Demeulemeester, and Belin have surveyed and collected papers that presented the latest breakthroughs on OR planning and scheduling and classified them according to the specific field from which they tackle the scheduling of OR. The authors identified seven fields in which they organized their taxonomy. These fields are as follows: Patient Characteristics, performance measures, decision level, type of analysis, solution technique, uncertainty, and applicability of research. The authors suggested as future work, to reduce the number of fields and to compile the data without causing the loss of any of it during this process of reduction. The reason behind these potential reduction efforts is to generate a more transparent and simple classification scheme.

Dexter, Macario, and Traub, 1999 had as an objective to find an optimal algorithm for scheduling add-on elective cases that would maximize the utilization of OR. They used computer simulation to evaluate 10 scheduling algorithms, such as on-line and off-line algorithms in terms of their performance at scheduling on on-going number of add-on elective surgeries. They collected data about 2 OR suites about their relative hours of open OR time reserved daily for add-on cases and the duration of each surgical case. The optimal algorithm found was the Best Fit Descending with fuzzy constraints. Serhat, Denton , Fowler, and Huschka, 2011 were comparing a number of 12 heuristic algorithms that would schedule the operating rooms in an Outpatient Procedure Center. Those algorithms would put an end to the issues of the uncertainty of the duration of surgeries that has a bad effect on the patients waiting time, the utilization of resources and the overtime costs. The researchers based their study in an OPC at Mayo Clinic, in Rochester, Minnesota, where an OPC being a complex system with various surgical groups. They first built a simulation model that evaluated the different heuristics. Next they embedded this simulation model to a hybrid solution method that included a genetic algorithm and appointment heuristics. This hybrid method was used to construct a near Pareto optimal set of schedules. They propose as a future research to investigate other more complicated scheduling techniques that would take into consideration additional resource types that have not been considered in this paper and evaluate their effect.

Significance
The surgical operating rooms are, as previously stated, the most resource and money-draining units in hospitals, and even in the whole health service production (Peltokorpi, 2011, p. 370). Peltokorpi, 2011, has devoted a research study on the effect that operative management decisions have on the productivity of operating rooms. The researcher identified 11 hypotheses that identify a sort of link between the use of analytical decisions in regards to operating rooms and the productivity of these latter. One of these hypotheses was: Operating units with a narrow service scope have higher productivity than units providing a wide range of services (Peltokorpi, 2011, p. 370). This research was conducted in a 26 multi-hospital units in 15 hospitals. 24 units were located in Finland, one in Germany and one in the US. The results of this study showed that the most influential decisions of productivity are personal management, case scheduling and performance measurement, whereas strategic decisions that relate, for instance, to the size of the surgical units, the scope or academic status have a more marginal influence on the productivity of operating rooms. Consequently, we see that the practice of organizing schedules to operating rooms is highly important as it significantly influences their productivity, and so, significantly influences the productivity of the whole hospital. This study also showed that the sort of operative practices that surgical units should use depend on the strategic positions of these units. In other words, the hospital units that are focused need sophisticated case scheduling, but flexible hours, incentives and multi-skilled hospital staff are more adequate to hospital units that are central and ambulatory. Thus, not only scheduling ORs is important, but also tailoring the scheduling to the type of these units is important for a better productivity of the OR.

10

Solution Approach
Model Characteristics
In order to study the operating of the hospital, a model was created on Arena to get a clear picture of the allocation of the patients to the operating rooms that were free at the time of arrival. Main Arena Model OR Process
Pr o c e s s 1 Cr e a t e 2 t y p e As s i n g 1 O R

Arrival Process Arrival Pr

Queuing and Scheduling Process

Recording and Disposal Process

F a

s l

Re l a s e e

De l y a

As s i n g

Di p o s e s

0
De c i e d
Cr e a t e 3 t y p e As s i n g 2
E l s e

1
P r P r P r P r o o o o c c c c e e e e s s s s s s s s O O O O R R R R 1 2 3 4 . . . . W W W W I I I I P = = 0 P = = 0 P = = 0 P = = 0

Pr o c e s s 3

De l y a

Pr o c e s s 4

0
Cr e a t e 4 t y p e As s i n g 3 Di p o s e s 2

0
Re l a s e e 2

Cr e a t e 5

t y p e

As s i n g

As s i n g

As s i n g

1 6

De c i e d

3
E n E n E n E n E n t t t t t t i t i t i t i t i y y y y y . . . . . T y T y T y T y T y p p p p p e e e e e = = t = = t = = t = = t = = t y y y y y p p p p p e e e e e 1 2 3 4 5

E l s

As s i n g

1 7

As s i n g

1 8

Di p o s e s

As s i n g

1 9

The model is described as below: The model considers patients who come to the hospital, must undergo one of five types of operating procedures. All the operating rooms in the hospital are equipped to carry out any of the five operations (with a setup time in between). The time gap between the arrivals of patients is considered to be exponentially distributed.

11 Each operation has its own processing time assigned as given below:

Operation Type 1 2 3 4 5

Processing Time (In Hours) UNIF(1.2, 1.5) UNIF (2.2, 2,3) UNIF (6.5, 6.9) UNIF (6.8,7.2) UNIF(7.5,8.1)

Also, there is a setup time between operations in the operating room. This is the time required to prepare the room for the operation. This might involve cleaning the room, arranging the operating instruments and so on. The setup times are not constant for the different pair types of operations, i.e. the setup time between operations depends on the operation that has just been completed and the operation that is to be done.

12

Process Description
Arena Arrival Process As the patients enter the hospital, they must enter a queue until they are allocated a free operating room. The patient released is the one with the lowest attribute value (which will be described shortly). The queue is achieved with the help of a Seize module.

Create type 1

Assign 15

Create type 2

Assign 1

Create type 3

Assign 2

Create type 4

Assign 3

Create type 5

Assign 8

13

Arena Queuing and Scheduling Process

S eize 1 Updating entity attribute bypass

T ru e

Fa ls e

Decide 1
P ro P ro P ro P ro c c c c e e e e s s s s s s s s OR OR OR OR 1 2 3 4 .W .W .W .W IP= IP= IP= IP= = = = = 0 0 0 0

Els e

Delay 2

D ispose 2

0
Release 2

A ssign 7

A ssign 16

Decide 3
En En En En En tity tity tity tity tity .T y .T y .T y .T y .T y p p p p p e e e e e == == == == == ty ty ty ty ty p p p p p e e e e e 1 2 3 4 5

Els e

A ssign 17

A ssign 18

D ispose 3

A ssign 19

When a patient enters the Seize module, he/she is sent to the Operating Room which is free. Whether an operating room is free or not is communicated to the Seize module with the help of an allocated resource carrying a value of either 1 (Busy) or 0 (Free).

14 Arena OR Process

Process OR 1

Process OR 2

Process OR 3

Process OR 4

In these four blocks, the selected cases/patients undergo their operations. Arena Recording and Disposal Process

Assign 5

Record 1

Release 1

Delay 1

Assign 6

Dispose 1

When a patient leaves the operating room, the count is recorded and the type of patient leaving is recorded. This is used to update the setup time as follows:

15 If patient of type 1 exits the operating room, an additional amount of set-up time is added to the processing time of the patients in the queue. There is a setup time defined for each pair of patient types. Continuing with the example, when patient 1 exits the operating room, a setup time S12 is added to patient 2, S13 is added to patient type 3 and so on until patient of type 5. If patient of type 2 was to exit from the operating room then a setup time of S23 would be added to patient type 3 and so on. This is important as the additional setup time is used to record the attribute which is used to decide which patient has to go into the next free operating room. Patients are disposed of after their operation is completed. As the patient leaves, a Record module registers his/her operations setup time to be used for the ORs setup time of the next patient. The existent patient triggers the release module which lets one patient to enter the now free operating room following the queuing principle as explained above. Also a variable called Bypass Variable is set to one. This variable is used to decide whether the entity (patient) must enter the operating room or must go through the updating of the processing times and wait back in the queue module.

16

Evaluation
After implementing our suggested solution procedure for the studied scheduling problem using Arena simulation software, we experimented with different approaches for ordering schedulable jobs (patients) waiting to be processed (operated). Three queue disciplines were tested in this stage; FIFO (First in first out) LEPST (Longest Expected Processing with Setup Time) SEPST (Shortest Expected Processing with Setup Time)

In order to assess the performance of the three approaches, we implemented all of them as different scenarios in our simulation model and compared the expected waiting time for the different types of patients by conducting ANOVA analysis.

Figure 1: Total time for type 1 patients across all three scenarios

17

Figure 2 : Total time for type 3 patients across all three scenarios

18

Figure 3: Total time for type 5 patients across all three scenarios

19

Conclusion
We have identified the operating rooms unit as field of study and research that would address many issues that health care is facing, since it is one of the important unit in a hospital, cost and resource-wise. We also showed in the significance part of this paper that tackling the productivity of operating rooms and controlling it by an appropriate scheduling model has a high importance and return to the hospital. We chose to expand on the research done by Arnaout and Kulbashian 2008, by taking the same parallel machine environment with the same constraints of stochastic processing times and case-related set-up times and expanding them into an on-line heuristic scheduling algorithm that updates its scheduling plan whenever a new operation comes up. Our own study would be further developed and expanded if it takes into consideration the nonterminating simulation that characterizes emergency rooms and including within the scheduling of normal operating room units.

20

References
Arnaout, J-P. M., and Kulbashian S. (2008). Proceedings from Winter Simulation Conference 08: Maximizing the Utilization of Operating Rooms with Stochastic Times Using Simulation. S. J. Mason, R. R. Hill, L. Mnch, O. Rose, T. Jefferson, J. W. Fowler eds. Cardoen, B., Demeulemeester, E., Belin, J. Operating room planning and scheduling: A literature review. Retrived from https://lirias.kuleuven.be/bitstream/123456789/165923/1/KBI_0807.pdf Dexter, F., Macario, A., Traub, R. D. (1999). Which Algorithm for Scheduling Add-on Elective Cases Maximizes Operating Room Utilization? Anesthesiology, 91, 1491-1500. Elective surgery. (2012). In Encyclopedia of Surgery. Retrieved from
http://www.surgeryencyclopedia.com/Ce-Fi/Elective-Surgery.html

Peltokorpi, A. (2011). How do strategic decisions and operative practices affect operating room productivity. Health Care Management Science, 14, 370-382. Serhat, G., Denton, B. T., Fowler, J. W., Huschka, T. (2011). Bi-criteria Scheduling of Surgical Services for an Outpatient Procedure Center. Production and Operations Management, 20, 406-417.