Analyzing room usage on the Amsterdam UMC, location AMC outpatient department using historical appointment data

(1)

Department of Medical Informatics

Master Thesis

Analyzing room usage on the Amsterdam

UMC, location AMC outpatient department

using historical appointment data

K.M. de Lange

Mentor

dr. Edith van de Vijver

Afdeling Zorgsupport

Amsterdam UMC, location AMC

Tutor

Marieke Sijm

(2)

Student K.M. de Lange

Student number: 12020915 E-mail: k.m.delange@amc.uva.nl

Mentor

dr. Edith van de Vijver Afdeling Zorgsupport

Amsterdam UMC, location AMC

Tutor Marieke Sijm

Department of Clinical Informatics University of Amsterdam

Research project location Afdeling Zorgsupport

Amsterdam UMC, location AMC Meibergdreef 15

1105AZ Amsterdam, The Netherlands

Time period

(3)

Abstract

Background and objectives

There is not much comprehensive knowledge available on how to analyse room usage on the hospital outpatient clinic, even though space is a vital part of the daily operations. The objectives of this study are to explore approaches that have been applied before to study this problem, to determine the factors that are taken into account, use this information to develop an analysis method for Amsterdam UMC, location AMC and to compare the factors included in this method to those found in literature.

Methods

A scoping review is performed to discover approaches and factors relevant to room usage analysis. After selection of a suitable approach, an analysis method is developed for the Amsterdam UMC, location AMC using an iterative approach and a pilot specialty. The first version of the developed model is then used to analyse other specialties to assess its applicability.

Results

The scoping review results in several approaches that are used to analyse room usage and a list of factors that were taken into account. A calculational approach was deemed most suitable for the goals of the analysis and a calculational model was developed to analyse room usage on a pilot specialty. With small alterations in input data and interpretation, the model could be applied to analyse other specialisms as well.

Discussion

We succeeded in developing a model based on appointment data that assists in reasoning about rooms on the outpatient clinic in Amsterdam UMC, location AMC. Not all factors found in literature are included. A suggestion for future research is to explore the relevance of factors in different approaches and contexts.

(4)

(5)

Samenvatting

Doel en achtergrond

Hoewel fysieke ruimte een belangrijk onderdeel is in de dagelijkse gang van zaken op de polikliniek, is er is weinig bekend over hoe kamergebruik op de polikliniek geanalyseerd kan worden. Het doel van deze studie is om te verken-nen welke aanpakken er beschreven worden in de literatuur en welke factoren er meegenomen worden in zo’n analyse, om deze informatie te gebruiken om een analysemethode te ontwikkelen voor Amsterdam UMC, locatie AMC en de relevante factoren te vergelijken met wat er in de literatuur gevonden is.

Methoden

Een scoping review is uitgevoerd om een overzicht te krijgen van de aanpak en factoren die gebruikt zijn om kamergebruik te analyseren. Op basis van deze uitkomsten is een rekenmodel ontwikkeld voor Amsterdam UMC, locatie AMC ontwikkeld met behulp van een iteratieve aanpak, gebruikmakend van een pilot specialisme. Het model is vervolgens gebruikt om het kamergebruik bij andere specialismes te analyseren en de toepasbaarheid te evalueren.

Resultaten

De resultaten van de scoping review hebben geleid tot verschillende approaches die gebruikt zijn om kamergebruik te analyseren en een lijst met factoren die daarin meegenomen zijn. Een rekenmodel werd als meest geschikt bevonden voor de analyse in Amsterdam UMC, locatie AMC en een rekenmodel werd on-twikkeld gebaseerd op afsprakendata om kamergebruik van het pilotspecialisme te analyseren. Met kleine aanpassingen kon het model ook toegepast worden op andere specialismes.

Discussie

Er is een succesvol rekenmodel ontwikkeld om kamergebruik te analyseren op de poliklinieken van Amsterdam UMC, locatie AMC. Niet alle factoren die gevonden zijn in de literatuur zijn geïncludeerd. Een suggestie voor vervolgonderzoek is om te verkennen welke factoren relevant zijn in welke aanpak en situaties.

(6)

(7)

Acknowledgement

Amsterdam, July 26, 2019 This thesis is the result of 8 months of hard work, during which I was supported by several people.

Edith, I am happy that we could continue our collaboration after my short internship in June 2017. Your enthusiasm for your job (and excel) is highly contagious. Thank you for your insights and support and for helping me grow professionally.

Marieke, thank you for giving the right hints to jog my thinking exactly when I needed it, for encouraging me and above all, thank you for the very useful feedback.

Finally, I would like to thank everyone who continued to believe in me when I did not do so myself, most importantly my boyfriend Matthijs and my friends Joanne, Sharon and Alma. Thank you for your mental support during the last eight months. Thank you to my parents who allowed me to move back in with them for eight months and provided me with a bike, clean laundry and dinner when I returned home late.

Finally, I would like to thank all people that made my time as a student worth-while.

Kind regards, Kyra de Lange

(8)

(9)

1

Introduction

1.1 Background

It has been a long time since healthcare was solely about caring for and curing people. Political, economical and social concerns have found their way into the field, each with a different view on the concept of ’optimal care’. What results is a complex system characterized by a multitude of different and often conflicting interests: medical specialists, patients, managers, governments, insurers and tax payers are just several of the stakeholders involved.

An example of a topic where these stakeholders have particularly conflicting views is the financial aspect of healthcare. Healthcare expenses continue to rise globally, due to both an increase in demand and an increase in costs. Multiple authors describe changes in demographics, such as the ageing population in many developed countries, as an important contributing factor to this trend[13, 7, 8, 33]. Secondly, due to advances in the medical field, diseases that used to be deadly are now chronic, with their patients needing care for a longer period of time. Lastly, those same medical advances have led to the development of costly new technologies and treatments, increasing costs as well as improving health [8].

There is a general agreement on what lies ahead. The developments described above will continue to drive up healthcare costs and demand. Additionally, healthcare positions itself more and more as a service industry. Patients are increasingly aware of their power as a customer and demand a high quality of service, such as short waiting lists to access care and low waiting times [33]. Governments and other policy makers see the need to control healthcare expenses and pressure the system to cut costs and improve quality. As a result, healthcare organizations such as hospitals feel the urge to organise themselves more efficiently in order to provide the best care with the limited resources available [14]. Naturally, these changes also effect the individual providers further down in the organization. Care needs to be faster, cheaper, better and easier accessible.

(12)

The need to organize care more efficiently is evident, whether it is to reduce the costs of the current level of production or to cope with the larger demand within the same budget. Due to the opportunities in this area and the vast size of the healthcare system in terms of money and manpower, operational research specifically in the domain of healthcare (health OR) and operations management applied to healthcare settings (health OM) have developed into fields of their own [7]. Examples of applications of operations research in healthcare are widely spread and the field is growing continually [7]. Topics often touched upon are for example forecasting of demand, managing and allocating resources, and managing and evaluation of performance. Yet, it is reported that many decisions in healthcare organizations are still made without the help of operational research [13]. From a scientific point of view, it is noted that the actual implementation in practice of health OM models developed in the scientific field is lacking [7]. It is vital that changes made to cut costs are well researched, to prevent implementing solutions that in the end do not reduce costs or increase efficiency in the first place. Another aspect, also cited by Brailsford and Vissers [7], is the balance between service for patients and efficiency for providers. How far can healthcare organizations go in improving efficiency when it inevitably affects quality of service for patients? It is impossible to answer this question without insight in the current operations and thorough research on the proposed changes and their effects. The fields of health operations management and research can provide guidance for policy makers, management and (groups of) individual providers to improve the healthcare system.

Hospital outpatient departments

This thesis focuses on operations management in a specific part of the healthcare system: the hospital outpatient department. Hospitals are traditionally divided into an inpatient and an outpatient department. The outpatient department is the place where patients can consults specialists, usually on an appointment basis, without being hospitalized [33]. Due to the wide range of specialties available at a hospital outpatient department, it is a highly dynamic environment. Patients with a wide range of medical conditions divided over a multitude of specialties walk in and out, needing different types of care. To deliver this care, specialized equipment might be needed. From a staff point of view, health care providers and supporting staff work together to provide care. Some patients need to see multiple care providers at the same time, or consecutively in a specific order. The demand for outpatient care is generally growing, due to both the

(13)

healthcare demand growth in general described earlier and a shift from inpatient treatment to ambulatory treatment [33]. Not surprisingly for such a complex environment, optimizing operations in the outpatient clinic has often been the subject of research, resulting in, for example, many simulation studies that cover different aspects of daily operations, for example doctor capacity [12] and no-show/cancellation behaviour [1]. In particular, improvements in appointment scheduling have been researched extensively [21, 9].

Space on the hospital outpatient departments

To deliver care, various resources are needed, such as staff and space. There are several types of space on the hospital outpatient clinic. Following a patient from the entrance, he typically enters a reception area, where registration takes place. He or she then enters the waiting area, either centralized or decentralized. Depending on the room policy, the patient is either escorted to a exam room by a support staff member, or he is called from the waiting room by the care provider to his/her room. Taking a physician’s point of view, he might not only be concerned with rooms where he or she has patient encounters, but also rooms where he can complete patient adminsitration (office space) or rooms for telephone consults. Additionally, conference rooms can be found on outpatient departments. This study focuses on the central aspect of this story: the room where contact with the care provider takes place and the patient receives ambulatory care. This room is denounced by many names. Some studies call this exam(ination) rooms, others say consulting or consultation rooms. Other studies name both, implying there is a difference between them. For the sake of this study, we define exam room as an enclosed room at the outpatient clinic where an encounter between a care provider and a patient takes place and care is provided. This includes rooms used by physicians, but also support staff such as nurses, if they have contact with the patient in said room. We do not mean rooms where patients make a new appointment with adminstrative staff, as no care is provided here. For the same reason, we also do not mean waiting areas. In the remainder of this thesis, the above definition is linked to ’room’. For now, no distinction is made between rooms with specific equipment, or rooms where patients receive treatment, and rooms where a patient receives care by means of a conversation with a care provider.

There is not much known in literature about capacity management of rooms in the outpatient clinic. There are multiple complicating factors imaginable when

(14)

reasoning about room capacity, such as no-shows or cancellation rates, room equipment or demand variability on different time scales. The global healthcare trends described earlier make reasoning about space on a strategic, long-term level even more challenging. Also applicable here is the earlier managed balance between increased efficiency and quality of service, not so much for the patient, but for the healthcare providers making use of the rooms. Their interests and those of management can be conflicting. Lastly, operations on the hospital outpa-tient clinic are closely linked to many other departments in the hospital.[33] If, and when, medical specialists will hold clinic hours is often dependent on sched-ules of other departments such as the operating theatres and wards. Efficiency on the hospital outpatient department has to compete with many other priorities.

A vital component of room capacity management on the outpatient department is room usage. [19] But how can hospitals reason about room usage? We can distinguish between the numer of rooms in use (capacity) and how well they are used (utilization). These two concepts are interrelated: apart from knowing the volume, a utilization percentage can show how much improvement potential there is available. It can be used to bridge the gap from knowing how many rooms are in use towards determining how many exam rooms you need. We deem room usage to consist of two components:

• How many rooms are currently used?

• How efficient are they used?

To our best of knowledge, there is not much comprehensive knowledge available on how to analyse room usage on the outpatient clinic in its entirety. Klarich et al. [19] describes considerations during room analysis, but does not apply it to a case study. Pauly en Jellinek describe considerations when planning a new ambulatory care facility, but they do not show how room usage is analysed. There are several examples available of studies evaluating rooms in for example a simulation setting [31, 23], but they limit themselves to a single clinic. However, there is no generalized knowledge available on the approaches that can be taken and what should be taken into account when doing such an analysis.

(15)

1.2 Research objectives and research

questions

The problem of room usage is currently urgent in the Amsterdam UMC, location AMC, a tertiary care hospital in the Netherlands, as they look to rebuild their outpatient department. This hospital has just merged with another tertiary care hospital in the same city. Insight in exam room capacity is not only needed for the construction of the new building, care is also bound to shift towards the other location, or from the other location towards the AMC location. In the case of the Amsterdam UMC, location AMC the two questions stated above need to be answered for all specialties on the outpatient clinic, within a time frame of 7 months due to the upcoming renovation.

However, to our best of knowledge, there is not much comprehensive knowledge known in literature about how to do this. The aim of this research is to learn more about the approaches that can be applied to this problem and to investigate what factors are relevant when reasoning about room usage, and to use this information to analyse room usage in Amsterdam UMC, location AMC.

To achieve this aim, the following objectives are set:

1. Explore the approaches that have been used by literature to analyse room usage

2. Determine what factors have been taken into account by those studies when analysing room usage

3. Use this to develop an analysis method that can be applied to the Amsterdam UMC, location AMC outpatient department

4. Compare the factors from literature with the factors included in the devel-oped analysis method

1.3 General approach and structure

To reach the objectives stated above, an analysis method is developed. Developing such a method is an example of design research as described by [35]. Design research is characterised by designing and studying artifacts (models, methods, algorithms) in their context (a real-world situation). The artifact is, in this case,

(16)

the method to be developed. The context is the Amsterdam UMC, location AMC outpatient department.

To structure the development of the model, the engineering cycle in Figure 1.1 is used. Design research focuses on the first three phases of the cycle, problem investigation, treatment design and treatment validation, being less concerned with actual transfer of the technology to the real-world situation. In other words, while validation of the model is part of the research project, the implementation in the actual problem context is not subject to research. The bullets under each phase give suggestions for research angles.

Fig. 1.1.: The engineering cycle [35]

The structure of the thesis follows the engineering cycle as follows: chapter 2 describes the problem investigation phase by conducting a scoping review. This review will result in the approaches applied and the factors taken into account, sketching an overview of the problem area. In chapter 3 the treatment design phase will be described, resulting in an analysis method based on one specialism on the department. In chapter 4, we test whether the model is also applicable outside the context of this pilot specialism. The results are then discussed in chapter 5, followed by a conclusion and suggestions for further research.

(17)

2

Analysis of approaches and

factors

2.1 Introduction

In the introduction, it was described that even though efficiency plays a large role in today’s hospitals, not much research has been done on the approaches applied to study room usage and the factors that are relevant when studying this problem. However, some studies have described their analysis of room usage in a specific context. This chapter aims to combine knowledge of these case studies on two levels: by reviewing their approaches and associated considerations, and by reviewing what factors they took into account. In addition, the definitions of room utilization are studied, as they appear to differ between studies.

The results of the review will subsequently be used to provide the foundation of room usage analysis in the Amsterdam UMC, location AMC. Based on the results of the literature review and the constraints of the specific context of this study, an appropriate approach will be determined to apply in the Amsterdam UMC, location AMC. The factors extracted from the studies and the analysis of the utilization calculation methods will be taken into account in the development phase described in chapter 3. In later chapters, the factors found in this literature review will be compared to the factors included in the analysis for the Amsterdam UMC, location AMC to reason about their applicabilities.

2.2 Methods

Munn et al. [22] suggest using a scoping review approach, as opposed to a systematic review, for studies that aim to identify, map, report or discuss certain characteristics or concepts in papers or studies, such as this one. Arksey and O’Malley [2] have described a framework to execute scoping reviews, which was further refined by Levac et al. [20]. This framework consists of the following five obligatory stages:

1. Identify the research question

(18)

2. Identifying relevant studies 3. Study selection

4. Charting the data

5. Collating, summarising and reporting the results The methods applied in each phase are detailed below.

Stage 1: identify the research question

Based on the goal of the review, the following research question is used to guide the scoping review process.

Review research question What approaches are used in literature to

study room usage in the outpatient department and what factors do these studies take into account when analyzing this?

Levac et al. [20] suggest also defining the purpose of the review in this stage in order to facilitate decision making in the selection and data extraction processes. The purpose of the review is twofold:

Review purpose

1. Make a justified decision on an approach to follow in the practice of the Amsterdam UMC, location AMC.

2. Learn more about the applicability of factors included in room usage analysis in the specific context of the Amsterdam UMC, location AMC.

Stage 2: Identifying relevant studies

An electronic search was performed using Ovid (using Pubmed and Embase databases) and Scopus. An exploratory search was done in Scopus to gather suit-able search terms and to manually compile a validation set of studies that should be found in the search. This validation set was used to test the completeness of the search query and to refine it where needed.

(19)

Determining the search strategy was done in an iterative manner, as described by Levac et al. [20]: searching the literature, refining the search strategy and review-ing articles for inclusion were done alternately. Alternative searches (narrow and broad) were compared, while assessing whether the broad search would lead to more relevant results by reviewing the studies that appeared in the broader search but not in the narrow search. Reviewing studies also led to the identification of additional search terms. Ultimately, a balance was found between the breadth of the search and the time and manpower available.

Stage 3: Study selection

The final search results were deduplicated using the built-in deduplicate function-ality of OVID and, after combination of the OVID and Scopus search results, the search for duplicates-functionality of Mendeley.

The screening process was done in Rayyan, an application to manage the literature review screening process [26]. Screening was done on title and abstract first, only removing studies that did clearly did not address the research question. Of the remaining studies, the full text was reviewed. The studies were first given a rating between 1 and 5 representing to what extent each study aligns with the review goal and research question. This rating was used to scope the amount of studies in a bottom-up approach by first including the highest rated studies and then extending the inclusion scope by reviewing the next highest rated studies. As the goal is to gather an overview of the available research on this topic, few selections were made. All types of studies were included (quantitative, qualitative and descriptive). No selection on publication year was made.

Stage 4: charting the data

Charting was done by one researcher. General characteristics of the studies that were charted were title, year of publication, study type and the object of study (setting). Additionally, the approach of each study was extracted, together with the methods for data collection. If named, the considerations from choosing that approach and/or data collection method were charted. If a study described how utilization was calculated or defined, this was recorded as well. Finally, the articles were read to extract the factors that were taken into account.

(20)

Stage 5: collating, summarizing and reporting the results

On suggestion of Levac et al. [20], the original fifth stage of Arksey and O’Malley [2] was divided into three steps: analysis, result reporting and considering the meaning of the results.

The analysis step starts with a descriptive numerical summary of the included studies, presented in a table with general study characteristics. This is followed by a thematic analysis focused on the review purpose. In line with the purpose of this review, the themes discussed here are 1. the study approaches, 2. the calculation of room utilization and 3. the factors included in the analysis.

The thematic approach will also be used in the reporting step. Results are pre-sented per theme; an overview of both the used approaches and the definition of utilization and a list of factors is reported in a table, with a textual explanation.

In the final part, the implications of the results are reviewed in light of the formulated purpose of the review and this study in its entirety. In this case, the overview of approaches will be used to identify the most suitable approach for developing a model in the specific context of the Amsterdam UMC, location AMC.

Suitable approach for Amsterdam UMC, location AMC

Due to the constraints of this research, the approaches identified in literature are evaluated based on the following aspects:

• Required time to apply the approach (should be possible within the time-frame of this research)

• Potential to meet the specific goals for the analysis method to be developed in this research.

(21)

2.3 Results

Stage 2: identification of relevant studies

The following query was used; the syntax used here is that of Scopus. The same query, adjusted to the syntax required by Ovid, was used in that database.

Query: ( ALL ( room* ) ) AND ( TITLE-ABS-KEY ( capacit* OR planning

OR utili* OR manag* OR allocat* OR number* ) ) AND ( TITLE-ABS-KEY ( ( outpatient PRE/2 department* ) OR ( outpatient PRE/2 clinic* ) OR ( ambulatory PRE/2 clinic* ) OR polyclinic* ) )

The search resulted in 3679 records. After deduplication, 2708 records re-mained.

Stage 3: study selection

The 2708 remaining records were screened for eligibility by reviewing the title and abstract. This screening excluded studies that did not correspond with the research question. When in doubt, the study was included for full-text review. Of the 131 records identified as relevant based on the screening, the full text was reviewed based on correspondence with the research question. 16 studies were included in the review.

Stage 4/5: charting, analysis and reporting

Given the overlap between the two stages, the results will be combined.

In table 2.1 an overview of the characteristics of the included studies is shown. Of the 16 included studies, 13 were case studies, studying room usage in a specific situation. This situation can range from a single clinic of one medical specialty to an entire outpatient department of a hospital. The most applied method in these case studies was discrete-event simulation, which was used by 9 studies. The other 4 case studies used other methods to analyze room usage This will be discussed more in detail in the theme ’study approaches’. Studies of room usage have been conducted over at least the past 50 years. Sumner and Hsieh [29] performed a simulation study of a orthopedic outpatient clinic in 1972, while

(22)

Tab. 2.1.: Overview of included studies in the review

Citation Title Year Type Object of study

Sumner and Hsieh [29] Long range prediction of examining

room requirements 1972

Simulation case-study

An orthopedic outpatient clinic Eady et al. [11] Characteristics of outpatient room

utilization 1983

Case study - other method

38 adult and pediatric clinics in the University of Michigan hospital Côté [10] Patient flow and resource utilization in

an outpatient clinic 1999

Simulation

case-study Local family practice

Okotie et al. [25]

The effect of patient arrival time on overall wait time and utilization of physician and examination room resources in the outpatient urology clinic.

2008 Case study - other method

Outpatient urology clinic, one single process (type of appointment) with one physician

White et al. [34] The effect of integrated scheduling and

capacity policies on clinical efficiency 2011

Simulation case-study

An outpatient orthopedic clinic Hulshof et al. [16] Analytical models to determine room

requirements in outpatient clinics 2012

Analytical methodology

Alternative

organizational policies Klarich et al. [19]

What ambulatory care managers need to know about examination room

utilization measurement and analysis

2016 General approach and methodology

Outpatient clinics in general

Ridley et al. [27] Development of an ambulatory care

space utilization reporting system 1982

Entire University of Wisconsin ambulatory facility (33 clinics) Isken et al. [17] Simulating outpatient obstetrical clinics 1999 Simulation - general Outpatient obstetrical

clinics in general Swisher and Jacobson

[30]

Evaluating the design of a family practice healthcare clinic using discrete-event simulation 2002 Simulation case-study Family practice healthcare clinic Santibáñez et al. [28]

Reducing patient wait times and improving resource utilization at British Columbia Cancer Agency’s ambulatory care unit through simulation

2009 Simulation case-study

Ambulatory care unit in large cancer center

Baril et al. [3]

Design and analysis of an outpatient orthopaedic clinic performance with discrete event simulation and design of experiments 2014 Simulation case-study An outpatient orthopedic clinic Norouzzadeh et al. [24]

Simulation modeling to optimize healthcare delivery in an outpatient clinic 2016 Simulation case-study An internal medicine outpatient clinic Vahdat et al. [32]

Decreasing patient length of stay via new flexible exam room allocation policies in ambulatory care clinics

2018 Simulation case-study

A cardiovascular outpatient clinic, but in general

Berg et al. [4] Improving Clinic Operational Efficiency

and Utilization with RTLS 2019

Single site clinic, three specialties, with 21 exam rooms Berg et al. [5]

Use of simulation to evaluate resource assignment policies in a

multidisciplinary outpatient clinic

2019 Simulation

case-study Multidisciplinary clinic

Berg et al. [4] uses novel technology such as real time locating systems to study room usage.

(23)

Tab. 2.2.: Utilization calculation

Citation Utilization definition/calculation

Sumner and Hsieh [29] Not defined

Eady et al. [11] Actual visits per day divided by potential visits per day Côté [10]

Weighted average of ratio length of time where zero, one, two or three exam rooms are occupied to the total length of time required to complete a shift

Okotie et al. [25] Ratio of total MD time in the exam room to the total patient time in the exam room

White et al. [34] Not defined Hulshof et al. [16] No use of utilization

Klarich et al. [19] Use/capacity

Ridley et al. [27]

(total new patients * average new patient exam time) + (total return patients * average return patient

examination time) / (total half-day sessions * number of assigned rooms * time per session) * 100%

Isken et al. [17] No use of utilization

Swisher and Jacobson [30] No use of utilization

Santibáñez et al. [28] Time patient is in the room divided by hours of operation of the room

Baril et al. [3] No use of utilization

Norouzzadeh et al. [24] Not defined Vahdat et al. [32] No use of utilization

Berg et al. [4] Portion of the room’s available time during which the room was occupied by a patient

Berg et al. [5] Not defined

As illustrated in the introduction, a key concept of room usage is’room utiliza-tion’, describing how efficient a room is used. Not all studies explicitly state what

they define as room utilization and how it is calculated. In table 2.2 is shown how the studies included in this review differ in defining and calculating room utilization. For example, Eady et al. [11] use the number of appointments as a measure of the realised production of the provider, calculating the utilization by dividing this by the potential production (in number of appointments) that could have taken place in the room. Berg et al. [4] defined room utilization as the portion of a room’s available time during which it was occupied by a patient. Santibáñez et al. [28] also consider room utilization as the percentage where either patient or patient plus physician are present in the room.

(24)

Klarich et al. [19] presents the following formula for utilization, which summa-rizes the above approaches:

% utilization = examination room use

available examination room capacity (2.1)

According to them, both numerator and denominator of this fraction can be described in multiple ways. Using appointment duration, the total appointment duration in a room (use) can be divided by the time a room is operational (capacity). Likewise, one could divide the number of realized appointments by the amount of appointments that can take place in a room. While numerous variations on this concepts exists, they share the essence of dividing room use by room capacity.

On the contrary, Okotie et al. [25] evaluate utilization on a different level, namely as the ratio of the time the care provider spends in the exam room (with the patient) to the total time the patient spends in the exam room. This measure describes a different aspect, namely how efficient the time of the patient is used. This utilization could be high, when the physician is present during most of the patient’s visit in the room, while the room can be empty most of the time. In table 2.3, an overview of the studies and theirapproaches is shown.

Discrete-event simulation is the most often used approach to study room usage. The considerations for a simulation approach were not always described, but multiple studies mentioned that discrete-event simulation allowed to model the complex situation at the outpatient clinic. Other studies described that this approach created a safe environment to evaluate multiple scenarios in a replication of the real system and could also facilitate discussion about operational changes and their impact with key stakeholders. Other methods applied are the presentation of key statistics in a report, such as Eady et al. [11], Okotie et al. [25], Ridley et al. [27], and Berg et al. [4]. Hulshof et al. [16] presents an analytical model to reason about room usage on a higher level, by determining the number of rooms required per physician in certain room allocation policies.

In table 2.4, thefactors studies took into account when modeling room utilization

or room usage are listed. Some factors were named by multiple studies. The most often included factor is the number of examining rooms (also called the examining room capacity). Most studies include this factor as a variable in their

(25)

Tab. 2.3.: Study approaches

Citation Approach Data collection Considerations

Sumner and Hsieh [29] Discrete-event simulation Literature reviews, interviews, forecasting

Often used methods (expert judgement and historical square foot per bed ratios) do not consider specific hospital characteristics and are based on existing hospitals Eady et al. [11] Statistical analysis Time measures by_{staff on paper forms} None

Côté [10] Discrete-event_simulation Tracking form

Simulation-based applications are well-suited to estimate and evaluate the potential effect of changes to a facility’s environment and they have a decided advantage in modeling flexibility since they do not require the restrictive assumptions or simplifying generalizations commonly found in analytic approaches.

Okotie et al. [25] Statistical analysis Observations None White et al. [34] Discrete-event_simulation Time measures on_{time stamp cards} None Hulshof et al. [16] Analytical modelling Not applicable None Klarich et al. [19] Descriptive Not applicable Not applicable

Ridley et al. [27] Statistical analysis Hand completion of_forms

Computations had been attempted on the basis of a gross outcome measure, the number of visits per examination room per day. This methodology was rejected because it was apparent that an ambulatory care visit is nog a standardized event; that is, visit length varies considerably by clinic and by type of patient. Goals: usable for each of 30 individual clinics, presentation per day of the week and monthly/year to date, understandable to providers and form a basis of discussion and negotation of room assignment, ability to change data file easily to reflect operational changes, limited need for ongoing clerical or data processing staff time, ability to perform additional studies, integration with other information systems of the hospital

Isken et al. [17] Discrete-event_simulation Not applicable

The complexity of outpatient clinics makes simulation an attractive component to analytical models. Features of these systems which can be difficult to model analytically include time of day and day of week dependencies in both demand and capacity, scheduled arrivals and complex patient routings and associated resource use. Simulation providers a modeling approach that can accommodate such complexities relatively easily. Swisher and Jacobson [30] Discrete-event simulation Literature, group of experts

Provide a tool for decision-making in the clinical environment, allows medical decision-makers a means of visualizing changes to the patient-physician encounter while allowing management scientists to study the effects of those changes on key decision variables and performance measures

Santibáñez et al. [28] Discrete-event_simulation Booking system, time_{study by surveyors} ’Considering the characteristics of the ACU process, we decided to use discrete event_simulation Baril et al. [3] Discrete-event_simulation

Time measures ’collected at the clinic’

Considering that our study includes many factors interacting between each other, we used discrete event simulation to test scheduling rules.

Norouzzadeh et al. [24]

Discrete-event simulation

Observations, manual form completion and extraction from EMR

Simulation modeling allowed the team to test and evaluate 39 alternative improvement scenarios and select the optimum setting in a safe virtual environment which exactly replicates the real system.

flexibility to model with high resemblance and reliability, capability to characterize complex situation, facilitate decision-making process, analyze all possible alternatives of all natures, safe environment

Vahdat et al. [32] Discrete-event

simulation Observation

With the focus on interdependencies among various types of resources and the dynamic reallocation of resources based on the definition of multiple interdependent system characteristics, a DES method is identified to be the preferable methodology for this study, rather than a queueing approach.

Berg et al. [4] Descriptive statistics, probability analysis

Real time locating system

RTLS provides an automated means of collecting operational data on clinic activity. Historically, such data has been challenging to collect due to inaccurate or biased estimates from stakeholders or time-consuming observational time studies which often result in small sample sizes.

Berg et al. [5] Discrete-event_simulation Historical data and_{expert opinion}

simulation model. The list of factors is grouped into categories in table 2.5, combining similar concepts used by studies.

Suitability of approaches for the Amsterdam UMC, location AMC outpatient department

As was shown in table 2.3, the most often used approach in room utilization studies is simulation. Another approach is used by for example Berg et al. [4], where existing data from their RTLS system is used to present data on room usage. Eady et al. [11] also describe an approach where collected data is used

(26)

to compose management reports. Both describe an approach where they model and visualise the room usage based on calculation methods (simplifications) and statistics/charts.

However, as described by Berg et al. [4], collecting data for simulation models is a time-consuming process. In many of the included simulation studies, building the model was preceded by an observation phase or time study to determine the range and distribution of model parameters. Sumner and Hsieh [29] have shown by a sensitivity analysis that it is important that these parameters are estimated accurately. Many simulation studies are executed on 1 specific clinic or specialism, see table 2.1.

The included simulation models in this review are used to evaluate multiple scenarios; some of them involve room capacity as a parameter, but all studies also include multiple other decision factors. Calculational and statistical approaches such as Berg et al. [4] and Eady et al. [11] make use of available or collected data and present this to inform multiple stakeholders in the organization and assist in decision-making.

(27)

Tab. 2.4.: Factors

Factor Cited by

Outpatient clinic demand (#patients,

#clinics) Sumner and Hsieh [29]

Patient arrival distribution (on

appointment time or not), arrival rate of patients

Sumner and Hsieh [29], Côté [10] Physician examining time distribution Sumner and Hsieh [29]

Number of physicians Sumner and Hsieh [29] , Norouzzadeh et al. [23] Presence/number of medical students or

residents Sumner and Hsieh [29], Santibáñez et al. [28] Physician arrival time (lateness) Sumner and Hsieh [29], Santibáñez et al. [28] Number of examining rooms, examining

room capacity

Sumner and Hsieh [29], Eady et al. [11], Côté [10], White et al. [34], Ridley et al. [27], Isken et al. [17], Swisher and Jacobson [30], Santibáñez et al. [28], Baril et al. [3], Berg et al. [5] No-show rate Sumner and Hsieh [29], Isken et al. [17]

Walk-in patients Sumner and Hsieh [29], Isken et al. [17] Operational hours Eady et al. [11], Ridley et al. [27] Average time patient is in the room Eady et al. [11]

Average number of visits per month Eady et al. [11] Working days per month Eady et al. [11] Appointment policy (scheduling),

scheduling template, appointment order, scheduling rules

White et al. [34], Isken et al. [17], Santibáñez et al. [28], Baril et al. [3]

Flow variability, patient flow patterns,

patient flow types White et al. [34], Isken et al. [17], Baril et al. [3] Preparation time Hulshof et al. [16]

Consultation time Hulshof et al. [16] Post-consultation time Hulshof et al. [16]

Travel time Hulshof et al. [16]

Number of new patients, number of

return patients Ridley et al. [27] Examination time new patients,

examination time return patients Ridley et al. [27] Time per session Ridley et al. [27] Number of sessions Ridley et al. [27]

Patient type Isken et al. [17]

Appointment requests Isken et al. [17] Number of scheduled patients Isken et al. [17] Support space (waiting areas) Isken et al. [17]

Number of providers Isken et al. [17], Berg et al. [5]

Amount of support staff Isken et al. [17], Swisher and Jacobson [30], Baril et al. [3], Berg et al. [5]

Exam component durations Isken et al. [17] Exam component resource requirements Isken et al. [17] Patient flow rules (walk-ins, late arrivals,

no-shows Isken et al. [17]

Appointment duration (adjustment) Santibáñez et al. [28], Berg et al. [5] Double bookings Santibáñez et al. [28]

Room allocation policy Santibáñez et al. [28], Norouzzadeh et al. [23], Berg et al. [5] Patient priority Norouzzadeh et al. [23]

Increase in patient volume Norouzzadeh et al. [23]

(28)

Tab. 2.5.: Grouped factors

Group Factors

Amount of staff

Number of physicians [29, 24] Number of providers [17, 5] Amount of support staff [17, 30, 3, 5]

Presence/number of medical students/residents [29, 24]

Amount of patients

Outpatient clinic demand (#patients, #clinics) [29] Patient volume [23]

Number of new patients [27]

Average number of visits per time period [11] Number of return patients [27]

Number of scheduled patients[17] Appointment requests[17]

Arrival of patients

Patient arrival distribution (deviation from appointment) [29, 10] Patient flow rules (walk-ins, late arrivals, no-shows) [17] Walk-in patients [29, 17]

No-show rate [29, 17]

Time spent in exam room

Physician consultation time [29] Patient preparation time [16] Post-consultation time [16] Exam time new patients [27] Exam time return patients [27] Exam component durations [17]

Appointment duration (adjustment/standardization) [28, 5] Average time patient room time [11]

Scheduling

Scheduling rules [3]

Appointment scheduling policy [34] Scheduling template [17]

Appointment order [28] Double bookings [28]

Patient types/patient flow

Patient flow patterns [17] Flow variability [34] Patient flow types [3] Patient type [17]

Opening hours of facility

Time per session [27] Operational hours [11, 27] Working days per month [11] Number of sessions [27]

Other

Number of examining rooms [29, 11, 10, 34, 27, 17, 30, 28, 3, 5]

Travel time [16]

Physician arrival time (lateness) [29, 28] Support space (waiting areas) [17]

Exam components resource requirements [17] Room allocation policy [28, 23, 5]

(29)

2.4 Discussion

Findings and interpretation

The results of this chapter show that several hospitals have attempted to study room usage and have applied multiple approaches to do this, such as simulation or calculational approaches. It can be expected that a variety of approaches have been tried to analyse room usage. The studies had different goals that ask for different approaches. In some cases, room utilization was an outcome measure and provided the basis on which several scenarios, for example different numbers of rooms, were evaluated. Discrete-event simulation is a suitable approach to evaluate multiple scenarios without having to implement them in practice, and has not only been applied in multiple aspects of the outpatient clinic [15], but also in multiple other areas of the healthcare system [18]. However, conducting them requires detailed data analysis, which is also mentioned by [18] as a reason they have not been conducted in more complex, interrelated settings.

These studies took a wide range of factors into account. Grouping similar factors together showed that there is a large overlap; however, many factors are only addressed by one or several studies, pointing at the complexity of the relation between room usage and other aspects of outpatient clinic operations. The many factors that we found are likely related to the wide range of possible aspects of the outpatient clinic operations that can be studied, which is also illustrated by Hong et al. [15]. Some studies focused on the influence of appointment scheduling on room utilization, while others focused on staffing levels or room allocation to providers. The right set of factors to include in a model will therefore depend on the key aspects to be studied.

There is not one generally accepted and applied definition of room utilization; the concept is not clear-cut. As a result, the different definitions found here all represent other aspects of the outpatient clinic, such as the efficiency of the room in terms of time, in terms of number of appointments or in terms of efficient use of patient time. This is important to keep in mind when defining this concept. The requirements of the analysis in the Amsterdam UMC, location AMC is that it assists in determining how many rooms are needed for each specialism on the outpatient clinic. A constraint is that the analysis applicable to all specialisms in the outpatient department of the hospital. In case of choosing discrete-event simulation as an approach, this would result in about 40-50 separate simulation

(30)

models to be developed, tested and finetuned, with a higher amount of parameters to be studied beforehand. Given the context requirements of the Amsterdam UMC, location AMC and the approaches found in literature, a calculational approach based on available data seems the most suitable method for applying in the Amsterdam UMC, location AMC. Studies that had a comparable goal were Eady et al. [11] and Klarich et al. [19]. To systematically apply such an approach to all specialisms on the outpatient clinic, a model is developed that encapsulates the calculational approach.

The importance of the findings is as follows. There are multiple succesful ap-proaches in literature to analyze and improve room usage. The found factors show the aspects that are potentially relevant when analyzing room usage. Other researchers can use this information to determine what factors they want to ad-dress when modeling a specific case. The lack of a universally accepted definition of room utilization implies that it is important for a researcher to specify and communicate what is meant with room utilization in his or her study.

Strengths, weaknesses and limitations

A strength of this review was the application of the scoping review approach. The broad approach of a scoping review was suitable to gather an overview of the available literature. The framework of Arksey and O’Malley [2] that was applied in this study is well-known and by taking the extensions suggested by Levac et al. [20] into account, the framework could be applied successfully. However, the scoping review methodology also has limitations. The framework of Arksey and O’Malley [2] contains an optional sixth stage, where other stakeholders are consulted for additional insights and perspectives This was not done in this review due to time constraints. Although deemed optional by Arksey and O’Malley [2], Levac et al. [20] consider this stage as compulsory. Although we do not think different approaches would have been found, it could have led to additional factors being extracted from the studies, as those factors were not always clearly marked as such by authors.

There were not many studies that reported room usage analysis on a large scale, which would align more with the specific context of the Amsterdam UMC, location AMC. Therefore, we also included analysis of single outpatient clinics and also family practices, as we suspect them to have similary operations to a single hospital outpatient clinic. Eventually, an approach was chosen based on the studies that did perform a larger-scale analysis.

(31)

Bias might be introduced because we used examples of room usage studies to determine a suitable approach, only considering approaches that had succes-fully been applied and published before. This way, unsuccesful approaches, or approaches not published, are not considered. By selecting an approach from already available literature we limit ourselves to approaches that have been suc-cesful in the past, but do not broaden the range of approaches that could be used to evaluate room usage.

Take-away message for next chapters

Within this study, the results of this chapter will be used as follows. The calcula-tional approach deemed most suitable will be applied to the Amsterdam UMC, location AMC outpatient clinic by developing a model based on that approach. The factors found here will be compared to the factors found during model de-velopment. This will be extended in chapter 4, where the model is tested on additional specialisms to investigate additional factors. The unclarity in room utilization definition means that extra effort will be put into defining this concept for this study.

(32)

(33)

3

Model development

3.1 Introduction

In the previous chapter we identified a calculational approach to be suitable. To systematically apply such an approach to all specialisms on the outpatient clinic, a model is developed that encapsulates the calculational approach. Multiple studies describe the advantages of a model to study room usage. For example, both Klarich et al. [19] and Eady et al. [11] mention that the analysis improves the insight and understanding of room utilization on not only the management level, but also by medical and nursing staff on the work floor. This chapter therefore describes the development of a model to analyse room usage in all specialties the Amsterdam UMC, location AMC via a calculational, statistical approach.

The previous chapter produced a list of factors that are relevant to studying room usage. The list of factors are potential elements to include when calculating room requirements and utilization. In this chapter, the mathematical foundation of the calculation model will be laid by selecting a calculation method for room usage. This method will then be tested in a first model version constructed using a pilot specialism. The purpose of this is to determine what factors from literature appear to have relevance in this context and what factors have not, and additionally to determine factors that were not identified in the literature review, but are shown to be relevant for the context of Amsterdam UMC, location AMC.

3.2 Approach

In figure 3.1 the basic structure of the model to be developed is shown.

We regard the development of the model to be similar to software development. We therefore used a software methodology to develop the model. The spiral model of software development of Boehm [6] was used to structure the development process of the calculation model. The model of Boehm consists of cycles (called spirals) consisting of roughly four phases: determine objectives, identify and resolve risks, development and test and plan the next iteration. A key aspect of this software process model is that it is risk-driven: an assessment of risks done

(34)

Fig. 3.1.: Model structure

in each cycle determines the steps to be taken in the current cycle and the ones that follow.

This approach was chosen because there was no current insight in the problem and literature could not grant comprehensive information on this; the approach allowed to wield out unsatisfactory designs early on and to minimize the chance of pursuing the wrong direction, while retaining the flexibility that was needed for a project with no prior known requirements. Choosing an iterative methodology allowed for actively involving key personnel on the work floor by producing multiple versions of the end product and regularly validating the results and assumptions in meetings with them. The spiral approach resulted in a gradual enlargement of the project size: while iterating, more experts were consulted and the model was made more complex. Using a risk-drive approach ensured that while developing on a single specialism, the risks that could be encountered while scaling up to the entire outpatient department were constantly kept in mind.

Below, the main activities within each cycle are described, including the methods that are applied in all cycles. Because the course of the development process is dependent on the risks and alternatives identified along the way, the methods that are specifically applied in each cycle are discussed together with the results in the next section.

(35)

Determine objectives

Each cycle started with determining the objectives of this cycle by the main researcher. The objectives were based on the outcome of previous cycles, with exception of the first cycle, which was started as a problem investigation.

Determine alternatives and identify and resolve risks

Based on the objectives, alternatives were identified to reach those objectives. Several methods were applied to come to those alternatives, which are described more in detail in the cycle descriptions in the next chapter. The risks in each cycle were to choose the wrong alternative. Therefore, the risks were in general resolved as follows: criteria were set by the main researcher and alternatives were evaluated with experts (either domain experts or data analysts) or by means of data analysis.

Development and test

Each cycle contained a development phase where the chosen alternative was implemented. Testing took place with domain experts to include operational experience.

Plan the next iterations

Based on the results of the development and test phase, a plan was made on what to do next.

3.3 Methods and results

In total, three cycles were executed. The first cycle concerned problem investiga-tion and choosing a data source In the second cycle, a calculainvestiga-tion method was determined. In the third cycle, the calculation method was evaluated based on the experience of domain experts.

The methods that were applied in each cycle differed and where partly dependent on the alternatives and risks identified in each cycle and the results of previous

(36)

cycles. In this section, the methods and the results applied in each cycle are discussed in combination, per cycle.

3.3.1 Cycle 1: problem investigation

Objectives, constraints and risks

The development process commenced with a problem investigation cycle where, as suggested by Klarich et al. [19], key objectives were to gain familiarity with the space and operational processes at the outpatient clinic. Additionally, suitable input data had to be selected to use in the model. Apart from the time frame (7 months) prescribed by the upcoming renovation, no explicit constraints were set at this point. The risk was to select a data source that would not allow us to model room usage accurately for all outpatient clinics within the given time frame.

Methods

To resolve this risk, data sources had to be identified and assessed. The model was developed based on a pilot specialty. This was a single specialty on the outpatient clinic, surgery, that was used as a first case to start development. To gain familiarity with the operations of this specialty, a team leader and a schedule administrator were interviewed together. A team leader is responsible for the daily operations and support staff on a group of specialisms at the outpatient clinic and therefore has a good overview of the many aspects of daily operation. The schedule administrator builds and manages the schedule blueprints of care providers and has knowledge of the administrative aspect of outpatient clinic operations.

Based on this interview and after consulting a data analyst experienced with hospital data, the data options were listed and assessed for their suitability to use in the model. Building forth on the constraints and risks, the data sources were evaluated on three terms: their scalability, how representative they were of reality, and their availability. Scalability means whether the data can be collected for all outpatient specialisms within the 7-month time frame by one researcher, without enormous effort of the domain experts. Representativity of reality deals with how well the data represents the actual event taking place. Availability concerns whether the data will be available for all specialisms. Evaluation took place based

(37)

on the gathered knowledge from the interview with the domain experts and experience of data analysts experienced with the available data in the hospital. We further evaluated the representativity of appointment data by screening it thorougly for unusual situations, for example patients having multiple appoint-ments at the same time (either with different or the same care provider) and care providers having multiple appointments at the same time. These various forms of ’duplicate rows’ were summarized into several examples that were presented to the domain experts to determine whether these were representative of reality or not. The ultimate goal was to have one record for each appointment that took place in reality.

Results

The interview with the domain experts resulted in the following information: • On the outpatient clinic, care providers handle appointments. A care

provider occupies a room and calls the patients from the waiting room to see in his/her room.

• There were no standard opening hours, except for the building opening at 7AM and closing at 6PM. However, clinics rarely started before 9AM and often later.

• Rooms were allocated to care providers per half-weekday in a weekly room schedule made in spreadsheet software managed by the team leader. Allocation could not be done on a smaller scale, as this would lead to an overly complex puzzle.

• Care providers were often overbooked, but by 4-4.30PM, the department was generally empty.

• Rooms were not dedicated to a certain specialism for the entire week. Although the specialisms within this group were generally allocated rooms close to each other, this was not a rule.

• While some specialisms had clinics all week, others only had clinic hours on several days of the week.

• There are some rooms that contained highly specialized equipment or were architecturally different, or both. An example is the outpatient surgery room.

(38)

Tab. 3.1.: Data sources and assessment

Alternative Scalability Representativity Availability

Appointment data + 0 +

Schedule blueprints - - +

Weekly room schedule 0 -

-• During the day, there was no registration of what rooms were actually in use. In the experience of the team leader, not all care providers on the weekly room schedule would actually be present. This was demonstrated by the fact that on paper, there weren’t enough rooms to fit all care providers, but in practice, there was always enough space.

• Registrations that were available were the database of all appointments, the weekly room schedule, the appointment schedule template and the physician schedule.

After consulting domain experts and a data analyst experienced with the available data within the hospital, there were several alternatives identified and consid-ered:

• Appointment data extracted from the hospital’s information system

• Schedule blueprints per care provider, made by the schedule administrator, which serves as a framework to planners to plan actual patient appoint-ments.

• Weekly room schedule made by the team lead or schedule administrator The meeting with domain experts revealed that an important issue for them was that although on the weekly room schedule a shortage of rooms occured on some days, there was practically never a problem in practice. They could not know whether the people on the schedule were actually occupying rooms they were assigned to, but from the above observation resulted the suspicion that they probably did not. Therefore, the schedule was not representative of reality. Given that the schedule was made by each team leader and this was not standardized, using this to analyze room usage would lead to hand-counting and interpretation of a variety of weekly room overviews. In this case, the room schedule was for a specific week in December 2018, and was not available for other periods. This led to a low availability.

(39)

The schedule blueprints formed the basis for appointment scheduling in the entire hospital could not be extracted from the hospital information system in an automated manner, which made this data source unscalable. In addition, the templates show the planned clinic hours, not the completed appointments, which was not representative of reality.

The appointment data could be extracted from the hospital information system in an automated fashion and included all specialties on the outpatient clinic, making extraction scalable and availability high. Investigating the representativity as described in the methods, we found that in some cases, patients or care providers were intentionally booked double; these appointments should both be counted, while in other cases a single appointment in reality yielded multiple records in the data. Consulting the domain experts led to enough insight to differentiate between those duplicates to be able to remove them appropriately. The data also included appointments that were no-show or cancelled. Another finding in this phase was that data on what room was used on what time on the appointment level was not registered automatically: appointments in the appointment data were not linked to rooms in any way. The weekly room schedule did include room numbers, but as described earlier no record was made of whether a care provider actually occupied the room he or she was assigned to.

Planning the next iteration

To finalize this cycle, it was decided to use the appointment data as input for the model. The weekly room schedule for each specialty was retrieved nonetheless for comparisons.

3.3.2 Cycle 2: calculation method

Objectives

The objective of cycle 2 was to determine the calculation method of the model. Both utilization and the number of rooms cannot be determined directly from the data source; therefore, a calculation effort needs to be made. The constraints here were to calculate the required number of rooms separately for each specialty but using the same method for all of them.

(40)

Methods

Based on the findings of the previous cycle, we constructed a baseline model in spreadsheet software. As the rooms were currently divided based on half-day, this baseline model presented ’room numbers’ per half day. The reason for this was that current operations had a clear morning and afternoon shift with a break in between; the number of active appointments clearly dipped here. In addition, while some appointments started before 9 or took place after 16.00, this was far from the peak point during the day; there would not be a shortage of rooms outside of these time periods. Another reason to produce a room number per shift, which was considered the lowest level of detail attainable, allowed for combining specialties with different variation patterns in the week to come to a lower total room number. In the baseline version, the number of rooms was equal to the number of providers with at least one appointment. This was deemed a very basic method of determining room numbers from appointment data and was used as a starting point for the investigation of more sophisticated calculation methods. As found in the previous cycle, room utilization could not be calculated per room due to the lack of data on specific rooms. Room utilization was therefore defined as follows: in a use divided by capacity formula as described by Klarich et al. [19], the use was defined as the total duration of appointments this specific care provider has in a specific half-day, and the capacity the total number of minutes that is available in one room in that day. Operational hours were set at 9AM to 12PM and 13PM to 16PM.

Based on this meeting and analysis of the data, a list of calculation methods was composed in a discussion session by the main researcher and fellow data analysts. The list was extended and refined after a meeting with fellow data analysts, also from the other location of the hospital.

To resolve the risk of choosing a calculation method that wasn’t representative of reality, the methods were worked out for an example data set and both validated by the domain experts and discussed in detail with fellow data experts. Using the base model, the calculation methods were applied to a test specialism from both locations of the hospital to assess their feasibility and to test whether the chosen method could be applied to the data of both locations. The results of this analysis were discussed with the group of data analysts to arrive at a calculation method to apply in the model. The main criterium used in this decision were the ability of the method to represent reality.

(41)

Tab. 3.2.: Calculation methods and assessment

Alternative Representativity

Weekly room schedule -Parallel appointments + Target utilization percentage -Counting the number of care

providers 0

Fixed and flexible care providers +

Results

The five methods established with data analysts were the following: 1. Weekly room schedule

2. Using the number of parallel appointments in 5-minute time periods 3. Dividing the total number of appointment minutes by the desired utilization

percentage

4. Counting the number of care providers

5. Fixed-flexible method: separate room calculation for high and low utiliza-tion care providers

Although the weekly room schedule was rejected as a data source in the previous cycle, it was included here for comparison reasons.

A case example can be found in figure 3.2, which will be used to illustrate the outcome of each method. This is a certain morning (9-12) where 5 care providers,

Fig. 3.2.: Example of the calculation methods

A to E, had appointments and therefore were present in the outpatient clinic that day. Using the five methods, the following is calculated:

(42)

Fig. 3.3.: Number of active appointments per 5 minutes, related to the example in figure 3.2.

Weekly room schedule The weekly room schedule provided illustrated the prob-lem with this registration: sometimes, care providers come on a walk-in basis to do one appointment and find a free room on the spot. For this example of five care providers, it could be possible that the weekly room schedule only has a room scheduled for care provider A and providers B to E came in on a walk-in basis. However, there are also care providers that have a room on the schedule, but do not use this room to its full extent. Care provider B could be someone who has a dedicated room in the room schedule, but only occupies it between 9.30AM and 10.00AM. Concluding, the number of rooms scheduled for this group of care providers could vary from 1 to 5. This illustrates how the weekly room schedule does not reflect reality in all cases.

Number of parallel appointments in 5-minute time periods The duration of ap-pointments was in 5-minute increments. Per 5-minute time point, the number of active appointments was counted. An appointment was considered active if it started before or on the time point, and ended after the time point. A 10-minute appointment starting at 10.00 would be counted at active at 10.00 and 10.05, but not at 10.10. Figure 3.3 shows the calculations per 5-minute time point for the case example in figure 3.2. To extract a required room number, a decision needs to be made on how to relate the parallel appointment data to a room number. An example of such a decision would be to set the room number at the maximum parallel appointments level within a shift; in the example, this would mean that 4 rooms were needed during this shift.

Dividing the total appointment duration by the desired utilization percentage The last option to be considered was to divide the total appointment duration by the

Analyzing room usage on the Amsterdam UMC, location AMC outpatient department using historical appointment data