A method for generating an Operating Room Masterplan

(1)

A method for generating an

Operating Room Masterplan

Forming an operating room session planning for Nij Smellinghe

with a levelled clinical flow

Johan Bos

(2)

Master thesis Econometrics, Operations Research and Actuarial Studies

(3)

A method for generating an

Operating Room Masterplan

Forming an operating room session planning for Nij Smellinghe

with a levelled clinical flow

Johan Bos

∗

Abstract

In this thesis the focus is on the Operating Room (OR) planning in hospitals. In particular on the session planning, which is the distribution of OR time to different patient groups (e.g. to specialisms). In the thesis a new planning method is introduced, in which not

only production agreements are taken into account but also the flow of patients from the OR to the clinic. The goal of the model is to level the daily bed occupation in the clinical departments as much as possible. The new planning method is tested in a case study done

at Nij Smellinghe hospital in Drachten.

The planning method is translated into a Mixed Integer Linear Programming model, by which an ‘optimal as possible’ session planning can be generated. The data used in this model is all based on the provided dataset from Nij Smellinghe. By solving this model for

several scenarios and parameter values, a number of alternative session plannings are obtained. These are compared with each other and the current planning situation by means of a simulation study. From this it is seen that the new method indeed takes care of a better levelling of patient flows. Disadvantage however is that the utilisation rate and therefore the number of treated patients decrease somewhat in the new model, which can cause increase of waiting times. In the scenarios it is also investigated what the hospital can do to level the

clinical flow further. From this it turns out that permitting planning of sessions in the weekends additionally lowers peaks and variation of clinical occupation a lot.

Groningen, 2nd March 2009 Master thesis for the MSc in Operations Research Faculty of Economics and Business University of Groningen Supervisor: prof. dr. M.H. van der Vlerk Co-assessor: dr. J.W. Nieuwenhuis ORTEC bv: Bern`ad Willems Nij Smellinghe: Johan Beugeling

∗_{Master student in Econometrics, Operations Research and Actuarial Studies University Groningen; Student}

(4)

Preface

You are reading the report I wrote as my master thesis in Operations Research at the university of Groningen. For this thesis I did a graduation research at ORTEC in Groningen in cooperation with Nij Smellinghe hospital in Drachten. ORTEC is one of the largest providers of advanced planning and optimisation software solutions and consulting services. Because Operations Research tech-niques play an important role in their work, I decided to contact ORTEC for doing an internship for my master thesis. In consultation with ORTEC, I formed a research design in which I use Operations Research techniques to relate operating room planning to the patient flow in the clinic of a hospital.

First of all I want to thank Bern`ad Willems for his help, supervision and feedback during my time at ORTEC. The useful comments, ideas and help he gave me during our contacts greatly helped me to write my thesis. From the first moments to the last steps of my research, he helped me in doing my research at ORTEC.

I would also like to thank Johan Beugeling from Nij Smellinghe. He guided me in and through the hospital and made sure I was able to get the data and information from the hospital that was vital for my research. Furthermore his useful feedback helped me to stay on the right track. I would also like to thank Pieter Buwalda, John van Arnhem, Sietze Anema, Jan Terpstra and Gerrit van der Velde with whom I had interesting contacts to get me somewhat familiar in the hospital world.

At the university I was guided by Maarten van der Vlerk. I would like to thank him for the sub-stantial comments during the progress of my thesis and the interesting and useful consultations we had together. Furthermore I want to thank Hans Nieuwenhuis as my co-assessor. I would also like to thank Erik Hoving for reading my first thesis version. During my study time we worked together many times and once again he gave me a lot of useful comments and help.

From ORTEC I would further like to thank all my colleagues for the nice and instructive time I had at ORTEC Groningen. Especially I want to mention Eric Zeegers for getting in contact with Nij Smellinghe and for the further help he gave me. And of course I want to thank my roommates at ORTEC Richard, Gerolf, Rob and Andr´e for their hospitality, interest and for the good time I had at ORTEC.

Finally I would like to thank my girlfriend, family and friends. Who encouraged me in times every-thing did not go as planned and for the believe in me they always have.

I hope you enjoy reading my thesis.

(5)

1 Introduction

The Dutch health care sector has been developing rapidly in recent years. Recent developments result in a growing need for efficiency. One of these changes is the introduction of a kind of free-market principle in the health care sector. In 2005 the Dutch government introduced the so-called Diagnosis Treatment Combinations (Diagnose Behandel Combinaties, DBC). A DBC contains the problem a patient has, the diagnosis the hospital poses and the needed treatment. For this whole combination, one price must be charged to health insurers. The idea is that the price of these DBC’s can be freely determined by the hospital, so that hospitals can compete with each other in price and quality. At this moment this principle only holds for a fraction of all the DBC’s (20% in 2008 and up to 70% planned in 2011). These developments make it more and more important for health care organisations to grant care in an efficient way, resulting in lower costs.

Talking about cost reduction in hospitals or health care institutions in general, remains a tricky business. Health care providers often argue that improving efficiency in this sector has a negative effect on quality of care. It is desirable that a good quality is and remains the main objective to hospitals, but (in)efficiency should not be lost out of sight. Litvak and Long (2000) address this problem, they state that variability is the key element in causing inefficiency. They state that a hospital faces two types of variability: natural and artificial variability. Natural variability is random by its nature, examples are uncertain lengths of stay, unplanned arrivals of emergency patients, etcetera. Decreasing this variability is often not directly possible. On the other hand the artificial variability originates from poor scheduling policies and therefore unnecessarily in-creases costs and inefficiency. An example is, an Operating Room (OR) schedule that results in a bed-shortage in the beginning of each week and overcapacity at the end of the week (note that OR stands for operating room and not for operations research). By defining ‘smart’ scheduling procedures, this artificial variability can be reduced and hence efficiency can be improved.

In this thesis, the possibilities to lower this artificial variability will be investigated in a case study for Nij Smellinghe hospital in Drachten. To accomplish this, a model is defined which can generate a medium-term OR planning. In this model the patient mix flowing from the OR to the clinic is taken into account. The OR department of a hospital is a large cost-driver and plays a central role in the hospital. The planning of the OR has a substantial influence on other departments of the hospital, such as the Intensive Care and nursing wards. Hence the OR planning also has a major influence on resource requirements in the clinic, such as beds and personnel. So overall efficiency could be improved if this relation is incorporated in the generation of an OR planning. If one could optimally level the patient mix coming to the clinic, this can help in planning your resources. The goal of this research is not to generate one planning that can directly be implemented in the hospital. The research should give insights in the possibilities for improving clinical flow. To be able to investigate these possibilities to form an OR planning with a balanced flow, the model is used to generate different alternative OR plannings. These different plannings are evaluated and compared to the current situation. The idea is that, if one is able to level the flow to the clinic, this would result in having less over- and under capacity of both beds and personnel. Resulting in less working pressure, patients lying on a ‘wrong’ department, extra personnel and costs. But this could well have other (negative or positive) consequences, therefore this research aims to discuss improvements and drawbacks the alternative plannings have.

(8)

(9)

2 Context of the research

2.1 Nij Smellinghe hospital Drachten

The foundation of Nij Smellinghe hospital in Drachten dates back to 1945, when several churches in Drachten set-up a health care institution. The hospital was then called Protestants Ziekenhuis Drachten and was located in a former furniture factory. In 1971 the hospital moved to its current location and was renamed Nij Smellinghe. Around 1990 Nij Smellinghe rebuild and extended much of the hospital. Nowadays Nij Smellinghe is a modern hospital with many modern technologies and facilities. The mission of the hospital is that ‘the patient is leading for our actions’. The hospital gives a lot of attention to the quality of their care and patient satisfaction. This has resulted in a sixth place in the Dutch hospital top 100 list, provided in 2008 by the Algemeen Dagblad. It wants to accomplish these goals with a controlled cost development, one of their policy objectives is that the organisation should characterise itself by quality, effectiveness and efficiency.

To get some insights in the size of the hospital, some key numbers for Nij Smellinghe coming from the 2006 year account are presented. Nij Smellinghe had a total of 339 clinical beds, distributed over ten clinical departments. The number of treated patients in the hospital is increasing over the past few years. In 2007 over 10,000 patients had surgery in the OR and in the first two months of 2008 the hospital admitted a total of 2238 patients. Because of these increasing numbers, the occupation of both beds and personnel in the clinic reaches its capacity limits more and more. Therefore it can be advantageous for Nij Smellinghe to look for ways by which they could lower this pressure on their clinical capacity.

2.2 The OR complex

2.2.1 General information

In Nij Smellinghe there are seven operating rooms which together with the recovery room (con-sisting of ten beds) form the OR complex. In the OR complex the specialisms are active that perform non-polyclinic surgery. In Nij Smellinghe there are eleven such specialisms which are as-signed OR time, namely: general surgery (CHI), orthopaedics (ORT), urology (URO), gynaecology (GYN), plastic surgery (PLA), oral surgery (MHK), ear/nose/throat surgery (ENT), ophthalmol-ogy (EYE), cardiac surgery (CAR), neurosurgery (NCH) and pain suppression (ANA). A typical OR team consists of an anaesthesiologist, a specialist and a number of assistants. The number of assistants needed, depends on the type of surgery carried out.

The different specialisms operate by means of a predetermined OR planning. This planning is made in different stages. On the long run, production agreements are made for each specialism (case-mix planning). These production agreements consist of the yearly number of wished surgeries to be done by each specialism. The agreements result in a schedule in which blocks of OR time are assigned to the different specialisms (session planning). Finally these blocks are filled with actual surgeries (operational planning). More about the OR planning and the current planning situation in Nij Smellinghe can be found in Chapter4. In the planning of actual patients a distinction must be made between so-called elective and emergency patients. An elective operation is an operation that can be planned by the hospital days in advance. In general there is no very short time frame in which an elective patient must be operated. Emergency patients on the other hand, arrive at the hospital unexpected and depending on their degree of acuteness need to be operated within a few minutes or hours. Nij Smellinghe distinguishes four different priority types, namely: elective (surgery can wait >8 days), semi-elective (surgery in 1-8 days), emergency (surgery within 24 hours) and acute (surgery as soon as possible).

2.2.2 The path of an OR patient

(10)

surgery, he or she is put on a waiting list by the doctor. A patient then first gets a preoperative screening by an anaesthesiologist to discuss the procedure and the used anaesthesia. Now the patient needs to wait until he or she is scheduled in the OR schedule by the secretariat of the polyclinic. An elective patient hears at least eight days before surgery when he or she is expected (see also Chapter 4). A patient possibly needs to be admitted at a nursing department one or more days before surgery. These preoperative clinical days (nearly always zero or one day) are not included in this research, focus will only be on the trajectory of a patient after surgery. Now the patient undergoes actual surgery in one of the operating rooms. The different steps in actual surgery are not of particular importance in this research. After surgery the patient generally goes to the recovery room, where he or she recovers from the operation and anaesthesia. Some patients can go directly to the clinic after surgery (e.g. in the case of some local anaesthesia) and some patients need to go to the Intensive or Medium Care Units (ICU/MCU). If recovered, the patient goes to a nursing department in the clinic in general. It is also possible that the patient still has to go to ICU or MCU after recovery. Finally the patient is discharged, meaning that a patient goes home, to another hospital or to another care institution. The typical flow through the hospital of a patient who needs surgery is given in Figure1.

Figure 1: Flow diagram of an elective OR patient through the hospital.

2.3 Variability in hospitals

Variability is an aspect that causes inefficiency in organisations. The health care sector in particular is subject to all sorts of variability. To get some insight in these variability and efficiency problems, consider the following. Variability in the number of patients lying on a certain nursing department makes it hard to optimally allocate the needed resources. A possibility is to calculate average patient numbers and base the resource allocation on this. But due to the variability, this results in undercapacity and hence diminishing quality during peak demands. This fact is known as the flaw of averages, see Figure2.

(11)

Figure 2: Cartoon representing the flaw of averages. Source: http://www.stanford.edu/.

patients and this can be eliminated without influencing quality of care. Hence the challenge for hospitals is to eliminate this last type of variability and find a good balance between costs and needed resource capacity to cope with the remaining variability.

A methodology for coping with this variability is introduced by Litvak and Long (2000). To illustrate this further, consider a hospital in which patients arrive at predetermined times, all with the same disease, length of stay (LOS) and probability of treatment success. This hospital can work 100% efficient (Litvak,2005). It is clear that reality is quite different. Not all patients act the same on treatments, emergency patients arrive at random times and not all surgeons can treat every patient. These examples of variability are inherent to the hospital world and are simply occurring because we live in an uncertain world. This type of variability is called natural variability. This variability cannot be decreased without influencing the quality of care. Hospitals should instead try to manage this natural variability. So that they can act on fluctuations in a good way and allocate the available resources as well as possible.

But more often than not, this is not the only type of variability occurring. If one looks at the census of elective and emergency patients admitted in a hospital’s clinic on each day during a year, it could very well be possible that the variability for elective admissions is equal or even larger than the variability stemming from emergency admissions (McManus et al., 2003). Elective patients are by definition patients that can be planned and hence the elective patient flow can in theory be optimally regulated over the different clinical departments. Hence variability resulting from the planning of these patients can be diminished without diminishing quality of care, this type of variability is an example of so-called artificial variability, as it is stated inLitvak and Long(2000). To see why hospitals take advantage in reducing this variability, consider a schedule in which the day care department has (on average) ten elective admissions on Monday and twenty on Thursday. If the department has fifteen beds, this schedule causes capacity problems on this department. On Mondays the capacity is not optimally used and on Thursdays the working pressure is high, patients need to be diverted to other departments where personnel is less suited for these patients, et cetera.

(12)

2.4 Literature research

In this section the literature is discussed that is published and is related to the subject of OR planning. These publications serve as references to the choices made and the methods used. We also address to what extent this research distinguishes from and contributes to the existing liter-ature.

General discussion and examples of applications of operations research in health care are found in

Brailsford and Harper (2007);Van Houdenhoven et al.(2006);Van Houdenhoven(2007);Vissers and de Vries(2005). These sources give a good overview of the planning problems which hospitals face and what operations research can do to help in these situations.

In this research the focus is in particular on applying operations research to the problem of OR planning. The problem of forming an OR planning has been subject of a lot of research done in the area of operations research in health care. The literature can, just as the OR planning prob-lem itself, be divided into several stages. In the short run, hospitals face the planning probprob-lem of scheduling actual surgeries into the available operating time, this phase is on the operational plan-ning level. Distinction is made between operational planplan-ning of elective patients and scheduling add-on or emergency patients. This research is focussing however on the tactical planning phase. This stage is called the session planning phase. In this stage blocks of OR time are distributed among several session groups. The distribution of total available OR time among the specialisms is done in the strategic level, called the case-mix planning. More about these OR planning phases is discussed in Chapter4.

A number of interesting articles have been written on the operational planning of add-on patients.

Dexter et al.(1999) look for an optimal algorithm to schedule elective patients that arrive after the planning is made. By means of simulation they look for a heuristic that gives the highest utilisation of OR time and conclude that a best fit descending algorithm with fuzzy constraints yields the best results. Other articles on online planning consider dealing with emergency patients. Based on data from the Erasmus Medical Center, Van Houdenhoven (2007) concludes that it is better to reserve time for emergency patients in each OR rather than having one particular emer-gency room. In De Jong (2007) a similar research is found, that is done at the Medical Centre Leeuwarden.

The operational planning of elective patients is also widely researched. Guinet and Chaabane

(2003) propose a two phase scheduling approach for scheduling patients. In the first phase, avail-able ORs are filled with patients and in the second phase each filled OR is rescheduled such that the several needed resources are synchronised. The paper focusses on the first step, an assignment problem is defined to represent this. This problem consists of assigning N patients to the available OR capacity. They use a variant of the Hungarian algorithm to solve this planning problem.Jebali et al.(2006) consider a similar approach.Ozkaraham(2000) defines an integer goal programming model with several goals, to schedule patients into available blocks of OR time.Ogulata and Erol

(2003) use a hierarchical approach for planning patients with efficient resource usage. First, pa-tients are selected who are then assigned to different surgery groups and as a last step the time and place of operation are determined. Sier et al. (1997) use the method of simulated annealing to schedule operations and provide a list of factors that affect the operating schedule.Hans et al.

(2008) consider the problem of forming a robust surgery planning by taking into account sto-chastic operation durations. They aim to minimise total planned slack time on the one hand and probability of overtime on the other. This goal is achieved by forming a base solution and several heuristics are compared for rescheduling planned operations in the base solution.

(13)

still restrictions due to limited capacity of the resources needed. They solve the problem of mak-ing a session plan in their paper by formmak-ing a mixed integer programmmak-ing model that minimises total undersupply with regard to the case mix. Restrictions are added to model the limited re-source capacity.Beli¨en and Demeulemeester(2007) also consider the problem of forming a session planning, but the objective of their model is to minimise total expected bed shortage. By adding restrictions they ensure that each specialism is assigned a predetermined amount of OR time. They impose a dependency between the type of surgery and the number of occupied beds during each time-step. A binomial random variable is used to describe the number of occupied beds in each step. They also incorporate a stochastic number of patients treated each session, this is done by using a multinomial distribution. The resulting model is a non-linear integer programming model, which is linearised and several heuristics are compared to solve the problem.Oostrum et al.(2008) more or less combine the previous two methods in their approach to form a session plan. They look at a model with two objectives, namely a high OR utilisation and a levelled bed occupancy. Operations are divided per specific type and they only include operation types that occur quite frequently. Probabilistic constraints are included to restrict the probability of overtime, these are later discretised to end up with an integer programming model. Solving their model is done in two phases. Firstly, bed requirements are ignored and column generation is used to select a set of possible OR day schedules that satisfy all restrictions and minimise utilisation. In the second phase the selected OR schedules are assigned to particular weekdays, this is done in such a way that bed requirements are levelled. FinallyVissers et al.(2005) introduce a model that takes into account various resources to be levelled. Their research focusses on one particular (cardiothoracic) department. For several resources they define a desired target level and use a mixed integer goal programming model to satisfy these targets as well as possible.

This research is different in that a combined method of linear programming and simulation is used. The linear programming model is a tool to rather easily generate several alternative session plannings. Next the simulation model enables to evaluate and compare these different generated OR plannings. The simulation model enables us to also incorporate stochasticity. Furthermore in most researches forming an OR planning restricts to certain patient types or specialisms, while in this research this will be modelled for the planning of all operating specialisms.

(14)

3 Research description

As mentioned, this research is focussing on getting insights in levelling clinical flow from the OR, by revising the OR planning method. A mathematical planning model will be set up to accomplish this goal. This model enables to generate several alternative OR plannings. These resulting alternatives will be evaluated and compared to each other and to the current situation. This research is done by a case study for Nij Smellinghe hospital in Drachten. Focus will be on generating alternative OR session plannings (called masterplan OR in Nij Smellinghe). Making such a planning is a medium to long-term planning problem which hospitals face. In the session planning, the hospital want to satisfy the production agreements made as well as possible. The problem is of particular interest because the operating theatre is first of all a costly part of the hospital and secondly it is a shared resource for many specialisms. As a consequence the OR department cannot be seen as an independent part of the hospital. The planning made here, influences other resources in the hospital, such as need for personnel and beds. Not only in the OR department, but also for example at the nursing wards and Intensive Care.

If this relation to the clinic is not taken into account in the planning, this could cause artificial variability. As discussed in Section2.3, this type of variability can be reduced to improve efficiency. Consider for example an OR planning in which 20% of the surgeries requires a stay in the ICU for a couple of days. Suppose now that all these surgeries are planned at the end of the week. Then there will probably be a shortage of beds and personnel at the ICU in the weekends, while in the middle of the week there are many empty beds. These fluctuations not only lead to unequal bed occupancies, but also to periods of overstaffing and at other times a very high working pressure. This can influence employee satisfaction and even quality of care. Hence it is wise to try to balance the patient flow from the OR to different parts in the clinic. By doing this, one also balances the needs for scarce resources. A way in which this can be done, is by trying to define an appropriate session planning which takes into account a balanced clinical flow.

Of course this balancing should be seen in a certain perspective. A session planning cannot have as its only objective to balance clinical occupation. Production agreements should also be taken into account. An example is a planning in which the needed clinical beds are effectively balanced, but needs specialisms to change OR every 30 minutes. In these situations the session planning is not workable for the OR department. Bottom-line is that to form an efficient OR session planning, that is not only efficient to the OR department but also to the rest of the hospital, several criteria need to be taken into account.

3.1 Objective and research question

(15)

This research is of particular interest to Nij Smellinghe because the clinical occupation reaches the maximum clinical capacity more and more. Hence the capacity in the clinic can be a bottleneck in further growth. From the data we also see that the pressure on the clinic is high and capacity limits are more often reached in Nij Smellinghe. This research should give insights in the improvements that are possible in levelling the flow coming from the OR, to relieve the pressure on the clinic. Possible other (negative and positive) consequences that changing the OR session planning has, on for example waiting times and flexibility, should be taken into account as well.

To see why there possibly exists artificial variability in the clinic, we briefly describe how the planning is currently made in Nij Smellinghe (more about this in Chapter4). Nij Smellinghe has a session planning (called masterplan OR) in which OR time is given to each specialism which has elective sessions in the OR. Then the available blocks of OR time are filled with actual surg-eries, in a non-central way by the polyclinics. But every polyclinic is more or less free to fill their sessions. Because there is little to no communication between the polyclinics and with the clinic, this can cause artificial variability in the planning. If for example on a certain day all specialisms plan their sessions with ‘big’ surgeries, this results in a tremendous pressure on resources on the long-stay department and possibly intensive care. It must be said that the polyclinics are not entirely free to fill the sessions, since to each session a number of OR assistants is attached (see Chapter 4). But we expect that it is possible to improve the levelling of flow from OR to clinic further, the possibility to achieve this is investigated here. This leads to a more balanced working pressure in the clinic and fewer patients that need to be diverted to other departments, because the most suitable is full. Hence by doing this research, Nij Smellinghe hopefully gets insight in the possibilities to form an efficient OR session planning, both by means of satisfying production agreements and on balancing the resulting patient flow to the clinic.

To accomplish this, first of all the way the OR session planning is currently made is looked at. Then a mathematical programming model is presented which enables us to generate several session plan-nings. How we find several alternative masterplans from this model, is discussed in Section3.3.2. The goal of the model is to form a planning with regulated clinical flow, this model is discussed in Chapter6. Note however that the masterplans resulting from this model serve more as a sort of base model. The resulting plannings cannot be directly implemented as a new masterplan in the hospital. The research aims more at getting insight in the possible improvements in levelling clinical flow, by changing the masterplan OR.

Finally the current situation must be compared to the generated alternatives. To do this, a simula-tion is formed which is used to calculate a number of performance measures. It is straightforward that alternative masterplans cannot be tested by applying them in practice first. In that case sim-ulation is a good tool to imitate the implementation of the masterplan in practice. Furthermore an advantage of simulation is that this is a good way to incorporate the stochasticity involved in the problem.

But a simulation model has as disadvantage that it only evaluates a planning that is put in, hence it does not provide us with alternatives. A possible solution to this problem could be to use meta-modelling. This means that based on several results of the simulation model, the performance measures are tried to be described in terms of the input parameters (Barton,1998). In this way the optimal output is approached. This approach also does not guarantee optimality, but could drive the solution toward optimality. Another solution to this problem could be to define the problem as a stochastic programming problem. Hence as a linear programming model in which stochasticity is incorporated, for more about this subject see Klein Haneveld and van der Vlerk

(16)

Summarising the research described above, boils down to formulating a research question and several sub questions. The research question this research tries to answer is:

• Can (artificial) variability be reduced in day-to-day patient flow to the clinic by forming a mathematical model to generate realistic alternative OR masterplans in which this is taken into account?

To answer this research question, a number of sub questions that first need to be looked at, are identified.

• How is the masterplan OR currently formed and what are the most important factors/ restrictions to take into account?

• What issues come forward in analysing the current situation in Nij Smellinghe?

• How can a mathematical model be formed that can generate OR masterplans, in which the flow to the clinic is taken into account?

In the next section, the assumptions will be described that are made to restrict the size of the research and keep it tractable. After that the methods used in answering the research question and sub questions posed, will be discussed in more detail.

3.2 Restrictions

In answering the research question, we only consider elective patients. This restriction is imposed because by definition, elective patients can be planned some time in advance. Hence in their plan-ning, a good balancing of clinical flow can be incorporated. In this research, emergency patients are defined as patients that need to have surgery within 24 hours after hospital admission. Because these patients need to be operated upon in such a short time, it is not possible to take into account the levelling of needed resources in the planning of these patients. Therefore in this research the focus is on (semi-)elective OR patients.

Furthermore we assume that the planned elective program does not change during a day. This has several implications, it means that no surgeries are cancelled by the hospital, that patients always show-up for surgery and that emergency patients can always be operated after the planned program at the end of the day. This assumption is justified by the policy Nij Smellinghe has, that in general planned operations are never cancelled. In practice it appears that only a small proportion of emergency patients need to be operated right away, most emergency patients can be treated after the elective program is finished. If a patient does need immediate surgery, the elective program is always continued after this operation. All in all the comparison of different session plannings remains valid, if we assume in each model that the elective program remains solid over the day.

The research specifically aims at the OR to clinic relation. This implicates in fact two restrictions. First of all, focus will only be on the influence the OR planning has on the clinic. The planning probably has influence on more parts of the hospital, but this research restricts to the clinic. This is done because this is where Nij Smellinghe thinks that improvements can be made. By focussing on the clinic, the model stays tractable and it enables to give a quantitative insight in the possible improvements. Second this also implies that focus is only on patients admitted in the clinic from the OR. This is done because the research is focussing on the effect the OR planning has on the flow to the clinic. Of course the OR is not the only place from which patients come to the clinic (other examples are the emergency department, treatment house or directly from the polyclinic). But the stream stemming from the OR consists of a large part of the clinical admissions (about 50 to 60% in Nij Smellinghe).

(17)

17:00pm. This is because for this period, sessions are planned in the hospital. The need for re-sources is of course not bound to these hours, but the evening and night hours are not taken into account here. Firstly, because a relation between the needed resources and session plans that are defined only in these hours, is looked at. Secondly for the evening and night hours the resource requirements per patient are different than during daytime.

During the weekends there are also no sessions planned. Hence the performance measures are only calculated during weekdays. For example for the day treatment department, incorporating week-ends in measuring the bed occupancy, would result in a lower average and higher variance because in the weekend there are no patients present here in this model. Weekend days do get incorporated in the model, because for the LOS of a patient these days should be taken into account. If a patient lies on a long-stay department on Monday for a stay of one week he has to leave again on Monday one week later. We assume that patients can be discharged from the hospital at every moment. Hence a patient who can leave the hospital on Sunday, will also leave on Sunday.

3.3 Methodology

In Section 3.1 various sub questions are defined. According to these, the methods used in this research will be described. First of all the methods used in analysing the current situation are described, by means of looking at the available data and current masterplan. Next the method for making a mathematical model to generate alternative masterplans is discussed. Finally the method for evaluating and comparing the alternative masterplans is discussed.

3.3.1 Analysing the current situation

The current planning situation at Nij Smellinghe will be looked to in more detail, to see if im-provements can be made with respect to regulating the flow from the OR. To be able to do this, first we describe and discussed how the masterplan OR is currently made in Nij Smellinghe in Chapter 4. The general known theory for making an OR planning will be described, this frame-work also applies for the situation at Nij Smellinghe. This planning approach consists of different stages, along this framework it will be discussed in detail how the OR planning is made in Nij Smellinghe. The most important restrictions holding for the formation of a session planning will also be discussed. Once it is clear how the masterplan is currently made in Nij Smellinghe, the production agreements holding for Nij Smellinghe in 2007 are presented and the masterplan used in that year will be presented. The information from 2007 is used because that is also the period over which the dataset provided by the hospital runs. Further analysis on the provided data is found in Chapter5.

The numerical values of the performance measures are analysed by means of a simulation model, for more about this model see Section 3.3.3. The current session planning is used as input in this simulation model and several performance measures are given as output. The generation of these numbers for the current situation firstly gives some intuition for the improvements that are possible. If for example a large variation is seen in day-to-day flow to one or several departments, then this is where improvements can possibly be made. But more important, the simulation model enables to compare the performance of the generated session plannings with the current situation. This enables to look at the advantages and disadvantages of the current masterplan and the al-ternatives.

The performance measures on which the different session plannings will be judged in this research are listed here:

• Variation in daily number of patients lying on the different clinical departments; • Total number of patients treated (in total and per department);

(18)

• Average daily bed occupation in different departments; • Average waiting time per patient group before operation; • Total deviation of the production agreements made.

It will be briefly discussed why these performance measures are interesting in this research and how they are found. More about these measures can be found in Section7.3.2. There the short discussion started here, will be continued and the mathematical definition of the calculated performance measures will be given.

The generated alternative masterplans have two main objectives, it is therefore at least necessary to say something about how well these goals are met. The satisfaction of production agreements can just be measured in total hours of deviations to the preferred hours. It could be interesting to split this total deviation in over- and undersupply of allocated hours. The performance measures of the second objective should give insight in the day-to-day variation in bed occupation on the different clinical departments. A first indication of this variation is found by calculating average and variance of the daily number of patients on each department. However the variance is not enough to give a complete view of the variation. For example the variation can be caused by having many small deviations below and a few very large deviations above the average, the other way around is also possible. Therefore some more describing statistics for the daily number of patients will be given. This is further discussed in Section 7.3.2. Finally the alternative masterplans could have other positive or negative consequences, compared to the current situation. We would also like to get insight in these consequences, to investigate the performance of the alternatives. In this research we choose to look at the average waiting times of patients in each group and to the utilisation rate and the total numbers of patients treated. Using a different planning method can cause the utilisation of the sessions to change. If the same amount of sessions are used as in the current situation and the utilisation rate decreases, this will also cause the total number of patients to decrease. Furthermore it could be possible that an alternative masterplan is able to level clinical flow better, but that it causes some patient groups to wait longer for surgery. Furthermore the total number of treated patients to each department is calculated. To level clinical flow better, the mix of treated patients to each department could be different than in the current situation.

3.3.2 Generating session plans

For a workable planning, levelling the clinical flow cannot be the only objective in making an efficient session planning. Hence the model for generating session plans should incorporate mul-tiple objectives. To be able to simultaneously minimise deviations to the desired levels of these objectives, we choose to define a Mixed Integer Linear Programming (MILP) model to generate session plannings with minimal deviations to the multiple objectives. Hence on the one hand the model aims to level the patient mix going from OR to different clinical departments. The second objective to be achieved is that the production agreements made for each specialism are satis-fied as well as possible. The model is build in such a way that it tries to satisfy these objectives as much as possible, but not at all costs. To accomplish this, the model is specified as a goal programming model. Meaning that weights are attached to the several objective functions in the model. These weights give the relative importance of the objectives to be satisfied. These weights do not come up straightforward, but by doing sensitivity analysis on these weights, one can find which assessment gives a good session planning. In addition the model should also incorporate the most important restrictions holding for the OR planning in Nij Smellinghe. The actual model is presented in Chapter6.

(19)

integer decision variables correspond to the number of sessions assigned to a group in each time-step. As opposed to a MILP model, another solution could be to solve the linear programming (LP) relaxation and round the assigned sessions to integer values. However, because most values for the assigned sessions will be zero or one in the solution, this makes the rounding approach not very suitable in this research. Hence, the LP relaxation approach is expected to result in a solution that is quite far from the optimal solution. Therefore a MILP model seems suitable for generating the session plannings. There is an important drawback in this approach, the arrival process and length of stay and surgery for the patient groups are stochastic parameters. A way to incorporate uncertainty is to model the problem by means of a simulation model instead of a linear programming model, in this simulation model the patient data could be randomly generated from given distributions. Disadvantage of such a simulation model is that in this way, one is able to evaluate a session planning but it is not possible to generate (optimal) alternative plannings. For these reasons the choice is made to use a combination of these two options. To generate alternative session plannings, a MILP model is used in which deterministic parameters are used. However the deterministic values of these parameters are based on the data of (uncertain) surgical duration and LOS, for more about this see Chapter 6. Next the generated plannings following from this model will be put in a simulation model in which they are evaluated and compared. More about this simulation part can be found in the next section.

In the generated alternatives each patient group is assigned blocks of available OR time. In these blocks, patients are operated who all belong to a certain patient group. These patient groups are divided both by the operating specialism as well as the type of department on which a patient is admitted. The resulting planning tries to regulate the flow to these clinical departments and to minimise deviations from the case mix.

We are talking about generating several alternatives all the time, this is because the MILP model does not generate one optimal masterplan. In the model several assumptions must be made about input parameters and the modelling in general. As mentioned before, the weights in satisfying the several objectives have to be chosen. These are not defined unambiguously however. Therefore several alternatives need to be generated, where we use different combinations of weights and look what the impact is on the performance of the resulting plannings. Other examples of input pa-rameters that need some further investigation are the papa-rameters for modelling LOS and surgery duration in the MILP model. Changing these parameters is certainly expected to influence the resulting masterplan and hence also the performance of the generated planning. In addition to generating alternative plannings based on the input parameters, also some interesting alternative scenarios are investigated. For example, what happens to the performance of the planning if we specify that sessions can also be done in the weekends? Or can we incorporate the expected care demand of patients in the model and what influence does this have? For which scenarios and input parameters we generate alternative plannings and how this is done, will be further discussed in Chapter8.

3.3.3 Evaluation and comparison by simulation

(20)

A general description of simulation is stated as follows:1

simulation is an attempt to represent a real-life system with a model to determine how a change in one or more variables affects the rest of the system. Simulation will not provide optimization except by trial and error. It will provide comparisons of alternative systems or how a particular system works under specified conditions. When a simulation model has been constructed, it may be used over and over to analyze different kinds of situations.

This definition is used here because it gives a good description of why simulation is chosen in this particular research, for a more formal definition see e.g.Law and Kelton (2000). Simulation is a method that is often used when a problem is too difficult to solve analytically. In a simulation model it is possible to imitate the true process quite precisely. Decisions that are made for the process, can therefore be evaluated in a simulation model, before applying them in practice. In this context for example, it would obviously not be possible to try some alternative session plannings in the hospital to see how they perform. As mentioned already in the previous section, an advantage of using simulation here is that it enables us to deal with stochastic parameters (such as LOS of a patient). An alternative session planning is generated by means of a simplified MILP model. Because we are not sure about some of the input in the MILP model and how some modelling choices affect the resulting masterplan, several alternatives with different parameters values are generated. Subsequently the simulation model is used to evaluate how these plannings perform in a situation which is more similar to practice. Stochasticity is incorporated by analysing and fitting distributions on the provided data, see Chapter5. Hence by means of the simulation model it is also possible to evaluate how an alternative session planning performs in comparison with the current situation in the hospital. A second advantage of evaluating the session planning by means of a simulation model is that the needed resources can be monitored in relative small time-steps. Incorporating such small time-steps in the MILP model, would make this too large for a good evaluation.

Summarising, a MILP model is developed to generate several alternative session plannings, in which a balanced flow into the clinic is taken into account. In this model the most important restrictions in forming a session planning in Nij Smellinghe are incorporated. In this model no stochasticity is assumed, but in practice some parameters used are of an uncertain nature. To cope with this, a simulation model is defined to evaluate the alternative session plannings generated by the MILP model. This simulation model does take into account the stochastic nature of the patient characteristics. In this way different alternative session plannings are evaluated and compared. The set-up of the research is schematically depicted in Figure3.

Figure 3: Global set-up of the research.

(21)

4 The operating room planning

In this section we discuss how an OR planning is made in general and how it is currently formed in Nij Smellinghe in particular. As a start two methods of OR planning can globally be distinguished. The first is the block planning approach, as is more or less the standard nowadays. In this method blocks of OR time are assigned to each specialism in which they can operate their patients. The other possibility is to make the planning on a ‘first come, first serve’ basis. This method can lead to a very unequal distribution of OR time among specialisms and only works if personnel is highly flexible. In this research the block planning approach will be considered, which is currently used in Nij Smellinghe. In making an OR planning in this way, a number of different stages are identified. These stages are represented in Figure 4 and discussed further in the remainder of this section. After describing each stage we discuss how these are currently implemented in Nij Smellinghe.

Figure 4: Different stages in the OR planning.

4.1 Case mix planning

In the strategic planning the available OR time is divided among the operating specialisms. This division is also referred to as the case mix planning (seeBeli¨en and Demeulemeester, 2007). The case mix is mainly based on the realised production in the previous years. But also things like preferred production, available capacity of the OR and possible production agreements made with health insurers should be taken into account. Based on these production agreements, a translation is made to give the total hours of OR time preferred for a certain specialism in a year. This distribution is made at the strategic planning level which means that it is fixed for a long term and only changes if the patient supply changes or if production agreements cannot be met. In this research this strategic planning step is considered fixed and the case mix is based on the current production agreements used in the hospital. In Nij Smellinghe these production agreements are determined based on realised production in previous years and preferred production for the coming year. If there are trends in these data, this is also taken into account. Nij Smellinghe uses a model to translate the production agreements to a total number of hours that need to be allocated to each specialism. The global idea is that based on surgery length, the number of emergency operations and some more retrospective information the production agreements are translated into a number of hours of OR time each specialism must have. The case mix per specialism in total hours that is used in Nij Smellinghe is given in AppendixA.1. Since this model is confidential we do not discuss it in more detail in this research. The division of hours made in the case mix planning is not binding, but the tactical planning is made such that it satisfies the case mix as well as possible.

4.2 Session planning

(22)

Blake et al.(2002) give a definition of a MSS.

The master surgery schedule is a cyclic timetable that defines the number and type of operating rooms available, the hours that rooms will be open, and the surgical groups who are to be given priority for the operating room time.

Such an MSS is made at the tactical planning level, meaning that it is an medium to long term planning problem. A MSS can be applied until the case mix planning or the available OR capacity changes. Often a hospital has two master surgery schedules, one for ‘normal’ capacity weeks and one for so-called reduction periods. In reduction periods, where hospitals have lower resource capacities (due to holidays and vacations) the second MSS is applied in which OR utilisation is lower than in normal weeks.

In Nij Smellinghe this master surgery schedule is called the masterplan OR, so in this report the terms masterplan, masterplan OR and session planning will be used as synonyms. The generation of alternative masterplans for Nij Smellinghe is the focus of this research. At Nij Smellinghe the masterplan is also defined in a cyclic way covering a period of four weeks. Globally the masterplan is similar for each of these four weeks. But to satisfy production agreements better, there are a few sessions that are only done once every two or four weeks. Furthermore there is also a distinction between normal periods (41 weeks in 2007) and reduction periods (11 weeks). The masterplan covers all weekdays (Monday to Friday) and an operating day in general ranges from 8:00am to 17:00pm. At the moment there are three types of possible OR session days in Nij Smellinghe, see Figure5. Specialisms can be assigned to morning and afternoon sessions (ranging from 8:00am to 12:15pm and 13:15pm to 17:00pm). In between there is a lunch break and 15 minutes of set-up and break-down time are needed before and after each session. It is also possible to define a long session in which one specialty operates in an OR from 8:00am to 17:00pm. This means that the total OR time is 30 minutes longer, because before and after the lunch break no set-up or break-down time is needed. A third type of OR day that is used in Nij Smellinghe are days with so-called sluder sessions. These are sessions of one hour, from 8:00am to 9:00am in which children needing short and common surgeries are treated. These are in general short surgeries such that up to four children can be operated in such a sluder session. These sessions are planned to regulate the flow to the childrens department somewhat.

Figure 5: Different OR session days in Nij Smellinghe. Source: Nij Smellinghe

(23)

time that is reserved for emergency patients per specialism is also determined based on historical data, for more about this see Section4.3. Given these hours a masterplan is made. The masterplan describes a four week period in which specialisms are assigned OR time by taking into account the number of hours a specialism needs. The masterplan that was used by Nij Smellinghe in 2007 is given in AppendixA.2.

As is seen in the masterplan, the blocks of OR time are currently divided by surgical specialty. The patient mix that should be operated in the available blocks is not specified explicitly. A patient mix can consist of different types of patients, which can vary in for example type of surgery, postoperative admission department, LOS, amount of needed care after surgery, et cetera. But Nij Smellinghe does have two characteristics they can assign to an OR block to regulate this to a certain degree. Namely an anaesthesia technique and a number of operating assistants. A patient that has local anaesthesia and few assistants will more often be a patient for day treatment, than one having total anaesthesia and four assistants. This is not always true but it gives an indication of the type of patients operated in a session.

4.3 Operational planning

The scheduling of the actual surgeries happens in the week planning. This planning is made on the operational level, meaning that it is made in the short term, for example such a planning can be made once every week. For this short-term planning two phases are distinguished, namely operational online and offline planning. In the operational offline planning, elective patients are scheduled in advance into the available sessions for each specialty. This is often done at a local level and each specialty has its own way of filling its sessions. The operational online planning considers dealing with any unpredictable and unanticipated events. In view of the OR planning one can think of the arriving of emergency patients, delay of an operation, no-show of patients, et cetera.

The filling of the masterplan with actual surgeries is done as follows in Nij Smellinghe. The filling of the available blocks happens in a non-central way, by the polyclinics of the operating specialisms. Each morning the planning committee (consisting of someone who is occupied with planning of beds in all clinical departments, the head of the OR department and one day treatment nurse) discusses the schedule used eight days later. They look if the planning that is non-centrally filled ‘fits’ in the clinic. This means that there should not be too many big surgeries, resulting in a stay on the Intensive Care Unit but also for example not too many day treatments, resulting in a capacity shortage in this department. If they foresee problems, some surgeries are interchanged and then the planning is approved and fixed. The planning is thus made in a rolling way, meaning that this Wednesday the schedule for next Thursday is approved, the following day the one for next Friday and so on. If the planning for next week is approved it is in theory known which patients will be operated, when these surgeries will take place and in which OR.

It was already seen that to cope with emergency patients, a certain amount of time is reserved at the end of each OR day. This reserved time is based on the average amount of emergency arrivals a specialty has. For example, consider that in the past it appeared that 10% of the urology operations were emergency operations. Urology should only fill their allocated blocks for 90%, such that they can operate emergency patients in the remaining 10%. This is of course on average, so on some days there are no emergencies and the session is finished early and on other days they have to work overtime because a number of emergency patients arrived.

(24)

time at the end of each day. Of course there are also acute patients that need surgery right away, in that case the elective program is just continued after this surgery. In that case all planned surgeries start somewhat later, but they are never cancelled.

(25)

5 Data analysis

The data used in this research is provided by Nij Smellinghe. The data basically consists of two datasets. The first contains the case mix and masterplan OR used in the hospital. The numbers presented in this research consist of the production agreements and masterplan over 2007. This is done because the patient data used in this research is also over this period. The hospital uses straightforward calculations to translate the yearly production agreements for each specialism into a number of preferred hours per week. Here a distinction is made between normal and reduction weeks. This first dataset provided by Nij Smellinghe is presented in AppendicesA.1andA.2. The second dataset contains a range of patient specific parameters, for all patients that got surgery in Nij Smellinghe in 2007. The patient data is generated by Nij Smellinghe directly from the Hospital Information System2_{. All patients operated between 1 January and 31 December are}

included in the dataset. The dataset contains a total of 10,246 patients. For each operated patient the following parameters are given:

• Date of operation; • Description of operation; • Operating room; • Operating specialism; • Specialist; • Priority class; • Anaesthesia technique;

• Gross starting time of operation; • (Net) starting time of surgeon; • (Net) end time of surgeon; • Gross ending time of operation; • Admission department;

• Postoperative trajectory score; • Admission date and time; • Discharge date and time.

Some of these parameters need some further explanation. First the priority class indicates if the operation was elective, semi-elective, emergency or acute. There is a distinction between the gross starting time and the time at which a surgeon actually starts with incision. Some time is needed to prepare the patient for actual surgery. At the end of the surgery, again it takes some time before the patient is ready to leave the OR. The admission department variable specifies the predefined department where a patient is expected to be admitted. Possibilities are: day treatment, short-stay, childrens department, ICU, Coronairy Care Unit (for cardiac patients, at the same department as the ICU) or one of the five other long-stay departments (each with their own specialisms). Finally the postoperative trajectory score is a score used in Nij Smellinghe, indicating the postoperative class to which a patient belongs. For example there are patients going to the recovery after surgery before going to their department, other patients need to go to the ICU directly after surgery. More about this score is discussed at the end of Section5.2.

To arrive at the final dataset used in this research, some filtering must take place. Because the research only considers elective surgeries in weeks with normal capacity, only these operating days are taken into account. Hence patients operated during holidays and vacations (reduction weeks) are filtered out. Furthermore the research is focussing on (semi-)elective patients, so the data must be restricted to these priority classes only, hence acute and emergency patients are also filtered out. Eventually a dataset containing data on 8806 treated patients remains.

Further analysis on this dataset is discussed in four parts. First of all to be able to generate

(26)

a masterplan with levelled clinical flow, a number of session groups are defined based on the dataset. This means that patients are divided into one of a number of possible groups, where each group is assigned blocks of OR time in a masterplan. The division is made because in this case a masterplan can be generated in which clinical flow is taken into account. Next, frequencies of different patient groups will be looked at. For example number of patients per specialism, frequencies per postoperative department and number of surgeries per day. Next, a distribution to model the surgical duration of a group of patients is fit. Standard statistical techniques are used to fit a proper distribution for the surgical procedure times. Finally the same is tried for fitting a distribution for the length of stay of a patient group.

5.1 Specifying patient groups

First of all the groups are specified to which blocks of OR time are distributed in the masterplan. These session groups must be chosen such that in the model, both production agreements per specialism, as well as levelling the flow to the clinical departments can be satisfied. Therefore the patients in these groups should have similar properties both in the OR as in the clinic. Examples of such patient characteristics are operation duration, LOS and postoperative trajectory. On the other hand, the specification of these session groups cannot be too specific, because a group should contain enough patients to assign blocks of OR time to in the cyclic schedule. As seen in AppendixA.2the sessions are currently only divided between the different operating specialisms in Nij Smellinghe. Because groups should be formed that have similar resource requirements, it is reasonable to also start with this division in the model specified here. Furthermore this first division makes sure that we can evaluate how good production agreements per specialism are satisfied in a masterplan.

For surgical patients there are roughly three possible postoperative trajectories (after eventual recovery and ICU stay). First of all a patient can go to the day treatment and short-stay department (ds). Here patients go who can get discharged from the hospital within a day. In addition, patients going to the short-stay department in general need to stay one or two nights. Patients going to these departments need relatively light clinical care. The second possibility is patients who go to the childrens department (c). There are a few specialisms, such as ENT and MHK for which a relative large part of the treated patients are children. The final possibility is that patients go to the long-stay department (ls). In Nij Smellinghe there are several different long-stay departments and each specialism sends patients to their own department. Patients lying on these departments often undergo surgeries which results in a longer recovery and observation in the clinic. Patients are staying on these departments, varying from one or two nights till up to over a month.

To be able to take into account the flow to each of these departments, in this model the sessions are not only divided by operating specialism but also by the postoperative department where its patients go. For example a resulting masterplan could specify that on Monday morning urology can treat patients going to day treatment and on Wednesday afternoon ENT can operate patients going to the childrens department. However for some groups there are too few patients to assign separate blocks of OR time to, for example ophthalmology only has a few patients a year that need to go to a long-stay department. In these cases we specify that several groups are taken together, such that in those sessions for example both ophthalmology patients going to long-stay and childrens department are treated. On the other hand, for example all the patients operated by neurosurgery and cardiac surgery go to the long-stay department, so for these specialisms there is only one type of session.

(27)

have the preference of being combined. When applying all these considerations to the patient data, this results in 19 different groups. Hence there are 19 different types of sessions that need to be scheduled by the model presented in Chapter 6. See Figure 6 for an overview. In this figure the lowest level represents the eventual session groups. The higher bars represent the groups that are aggregated, as described above. More about how these session groups are used in the MILP model and how this model is build is found in Chapter6.

Figure 6: Specification of patient groups in the model.

These patient groups are referred to in the rest of the thesis by attaching an L (for long-stay), D (for day treatment / short-stay) or C (for childrens department) behind the specialism code and combinations of those for combined groups. For example, CHID represents the group with general surgery patients going to day treatment and MHKLC patients from oral surgery going to long-stay and childrens department.

5.2 Frequency distributions

In this section the provided dataset is analysed, to present frequency distributions for a number of parameters. First, the daily number of patients treated in Nij Smellinghe will be looked at. This gives some insight in the variability resulting from the operating schedule. A large variation in day to day number of treated patients can result in a variation of patients in the clinic. Note that this does not have to be the case, because there is of course a difference between number of patients arriving in the clinic on a day and the number of patients lying in the clinic on a certain day. But it gives some intuition for the possible improvements. The total number of surgeries done each day is derived from the dataset. The average for each weekday is calculated and depicted in Figure7. Furthermore we find from the data that there are only small variations in the number of treated patients per weekday over a year. But once in a while there are some outliers both above and below the mean. This could be caused by the fact that once in a while a session is called off or an extra sessions could be planned, resulting in more extreme variations.

(28)

It is seen that, with the current masterplan considered, there is some variation in surgeries over the week. Especially on Friday significantly less patients are treated than during the other weekdays. This difference can be explained by looking at the current masterplan (see AppendixA.2). Here we see that especially on Thursday and Friday the OR capacity used, is smaller than on Monday and Wednesday. A phenomenon that is quite often encountered in hospitals can be a possible explanation for this: the available personnel (surgeon) capacity is lower at the end of the week. Another explanation could be that this is caused by the lower clinical capacity in the weekends. If most patients with a LOS of more than one and less than three days are treated in the end of the week, they would still be in the hospital in the weekend.

Figure 8: Daily number of elective surgeries.

To get some more insight in the daily variation of surgical patients, Figure8 shows the number of elective patients treated in all considered days in 2007 (hence excluding weekends and reduction weeks). From this we conclude that as stated before, there are some outliers in the number of treated patients, but the variability is mainly caused by the weekly pattern. Meaning that the number of patients treated each Monday, Tuesday, et cetera, does not vary much in general. This could mean that in each repeated session, a specialism operates approximately the same number of patients.

Table 1: Number of elective surgeries per specialism.

Specialism Frequency Percentage

ANA 765 8.7 CAR 71 0.8 CHI 2058 23.4 GYN 671 7.6 ENT 926 10.5 MHK 210 2.4 NCH 86 1.0 EYE 1484 16.9 ORT 1908 21.7 PLA 158 1.8 URO 469 5.3 TOTAL 8806 100,0%

Next the total number of elective surgery patients per specialism is specified in Table 1. We see that general surgery treats most patients and cardiac and neurosurgery are the smallest specialisms with only around 80 elective patients a year. These frequencies should bear some similarity with the production agreements holding for Nij Smellinghe. This cannot be precisely compared because the frequencies are based on patient numbers rather than allocated OR time, and furthermore because in the dataset reduction weeks and emergency surgeries are not taken into account. Nev-ertheless, we see that this similarity indeed holds true.

A method for generating an Operating Room Masterplan

A method for generating an

Operating Room Masterplan

Forming an operating room session planning for Nij Smellinghe

with a levelled clinical flow

Johan Bos

A method for generating an

Operating Room Masterplan

Forming an operating room session planning for Nij Smellinghe

with a levelled clinical flow

Johan Bos

Preface

Contents

1

Introduction

2

Context of the research

2.1

Nij Smellinghe hospital Drachten

2.2

The OR complex

2.3

Variability in hospitals

2.4

Literature research

3

Research description

3.1

Objective and research question

3.2

Restrictions

3.3

Methodology

4

The operating room planning

4.1

Case mix planning

4.2

Session planning

4.3

Operational planning

5

Data analysis

5.1

Specifying patient groups

5.2

Frequency distributions