Integration and its influence on patient flow performance in complex care: a single-embedded case study

1

Integration and its influence on patient flow performance in

complex care: a single-embedded case study

Master Thesis Supply Chain Management University of Groningen

Faculty of Economics and Business

By Sanne Velkers

S3770486

s.velkers@student.rug.nl

Supervisors University of Groningen: First supervisor: Prof. Dr. J. T. Van der Vaart

Second assessor: Dr. A. G. Regts

2

ACKNOWLEDGEMENT

3

ABSTRACT

This research aims to understand how complex care processes are integrated and whether integration improves patient flows, through analysing four cases with varying levels of detail and dynamic complexity. In a single-embedded case study of an oncology care process in the Netherlands, interview data is collected to uncover the integrative practices and influencing factors. Besides, archival data was used to evaluate the flow performance. This study showed that complex care processes are integrated through sharing planning/waitinglist information, a service level agreement, reserved capacity and combined appointments, although the level of detail and dynamic complexity influences whether all or some integrative practices are used and how it influences flow performance. Research showed that the case with the highest complexity has the highest level of integration. However, a lower degree of integration with either low detail or dynamic complexity does not necessarily imply low performance. Therefore, it appears that there are more factors of influence to this relationship, which should be included in further research. Through this study, hospital managers are provided with a clear understanding of integration in complex care and how it may help them improve patient flow performance.

4

TABLE OF CONTENTS

3.2.1 Description of oncology care process ... 14

3.2.2 Units of analysis ... 15

3.3 Data collection and measurement ... 17

3.3.1 Interviews ... 17

3.3.2 Archival data ... 19

3.4 Data analysis ... 19

3.4.1 Interviews ... 19

3.4.2 Archival data ... 20

3.4.3 Within and cross-case analysis ... 20

4. RESULTS ... 22

4.1 Within-case analysis ... 24

4.2 Cross-case analysis ... 29

4.2.1 Integrative practices and flow performance ... 29

4.2.2 Antecedents of integration ... 33

5. DISCUSSION... 38

5.1 Integration in complex care ... 38

5.2 Integration and flow performance improvement ... 39

5.3 Relationship between integration and flow performance ... 40

6. CONCLUSION ... 42

6.1 Theoretical implications... 42

6.2 Managerial implications ... 43

6.3 Limitations and further research ... 43

REFERENCES ... 45

APPENDICES ... 50

5

Appendix B: Data structure ... 60

Appendix C: Resource usage per case ... 62

Appendix D: Detailed overview of integrative practices ... 63

6

1. INTRODUCTION

Patient flows are an important determinant of the performance of healthcare delivery processes (Bhattacharjee & Ray, 2014). The speed at which patients move from one step in the process to the next one is referred to as patient flow performance (Drupsteen, Van Der Vaart, & Van Donk, 2013). In organising patient flows, different operational factors must be considered, such as increasing and variable demand (Gualandi, Masella, & Tartaglini, 2019), shared resources, the level of urgency (Villa, Barbieri, & Lega, 2009) and routing variety in care pathways (Vissers & Beech, 2005). The effect of these operational factors increases when patients become more complex (Leeftink, Bikker, Vliegen, & Boucherie, 2018). To achieve high flow performance for complex patient groups, two basic mechanisms are available to hospitals: prioritizing (Bhattacharjee & Ray, 2014) or capacity buffers (Roemeling, Land, & Ahaus, 2017). However, the amount of complex patients increases (Leeftink et al., 2018) and prioritizing more patients increases waiting times again. Besides, hospitals are forced to reduce costs (Giancotti, Guglielmo, & Mauro, 2017) and adding capacity leads to rising expenditures (Hans, Van Houdenhoven, & Hulshof, 2012). Therefore, other mechanisms must be considered to organise complex patient flows.

An increase in patient complexity requires a more integrated approach, where multiple disciplines together organise and optimise the patient flows (Leeftink et al., 2018). Previous research by Drupsteen et al. (2013) showed that integration in hospitals improves patient flow performance. Integration refers to the extent of how harmoniously different departments of an organization work together and the level of coordination in their activities (Barki & Pinsonneault, 2005). However, the study of Drupsteen et al. (2013) focussed on patient flows in orthopaedics, a relatively simple care process. The level of urgency, shared resources and routing variety are relatively low which makes it easier to integrate (Drupsteen et al., 2013). By building further on Drupsteen et al. (2013), this research aims to understand how complex care processes are integrated and whether this integration leads to better flow performance.

7 and Gimenez, van der Vaart, & van Donk (2012) conclude that more integration is not always needed or beneficial. Besides, they state that additional research is required concerning how specific types of integration cause improved performance (Gimenez et al., 2012). Furthermore, Drupsteen, van der Vaart, & Van Donk (2016) and concluded that more research is necessary concerning the factors of influence on integration in complex care processes.

Based on the above, it can be concluded that the literature is incomplete. Previous studies on integration in complex care have focussed on manufacturing/production environments or a simple care process, highlighting the gap for studies concerning integration in complex care. Second, studies have concluded that higher levels of complexity require greater levels of integration, that integration is more difficult or that this is not always needed. Therefore, it is unclear whether complex care involves high or low levels of integration and also how these levels were achieved. Finally, the literature provides evidence that integration in complex processes improves performance, although this relationship is not completely understood yet. Therefore, the main question of this research is: “How are complex care processes integrated

and what is the influence of integration on patient flow performance?”

To answer the main research question, this research must address the following. First, it explores how complex care processes are integrated by identifying the integrative practices and the corresponding influencing factors. Second, it tries to determine if and how integration influences patient flow performance by examining the flow performance of different process steps that are integrated. A single-embedded case study is performed in a hospital in the Netherlands and the necessary information is gathered through interviews and archival data. This research has two practical contributions. First, it addresses the gap concerning studies on integration in complex care and the unclarity about the degree of integration. Second, it addresses the question if and how integration leads to improved performance in complex care, which is unclear at this point. The practical contribution is that this research provides hospital managers with an understanding of the kind of integrative practices that can be used in complex care processes and how they are of use to improve performance.

8

2. THEORETICAL FRAMEWORK

This chapter contains the theoretical framework, where the relevant literature is discussed concerning complexity and integration and the conceptual framework is presented.

2.1 Complexity

This research focusses on an SC with a high level of complexity, where the SC concerns a flow of patients, rather than material or goods (De Vries & Huijsman, 2011). Complexity is an important theme in SCM literature since SCs have become increasingly complex over the last years (Bode & Wagner, 2015). In defining SC complexity, Bozarth, Warsing, Flynn, & Flynn (2009) incorporated the ideas of the systems theory. This theory defines the types and elements in the systems and their relationships (Chandra & Tumanyan, 2005) and states that systems with similar components are different due to the way they are arranged (Helou & Caddy, 2006). Accordingly, a complex system is defined as: “one made up of a large number of parts that interact in a non-simple way” (Simon, 1962, p. 468).

Bozarth et al. (2009) describe SC complexity as the level of detail and dynamic complexity exhibited by the products, processes and relationships in the SC. Detail complexity is defined as: “the distinct number of components or parts that make up a system” and the dynamic complexity as: “the unpredictability of a system’s response to a given set of inputs, driven in part by the interconnectedness of the many parts that make up the system” (Bozarth et al., 2009, p. 79). Detail complexity and dynamic complexity are highly interrelated. A larger number of varied elements implies greater possible interactions, a variety of behaviours and states of the system (Bode & Wagner, 2015). Serdarasan (2013) states that companies implement reduction strategies for detail complexity, for example reducing the number of products. Furthermore, they try to manage and adjust their operations in dealing with dynamic complexity, through SC integration, collaboration, visibility and information sharing (Serdarasan, 2013).

9 are different from each other. Finally, there is uncertainty, which concerns the ability to make a clear picture of the system (Isik, 2011).

In this research we regard the level of detail complexity to be influenced by the number of patients in the SC. In this respect, a higher level of patient volume increases the complexity of an SC (Isik, 2011). Also, the numerousness of the planning activities increases, as well as the required resources and costs (Bozarth et al., 2009). Complexity as a result of higher levels of volume can be addressed by implementing dedicated resources, which are for the use of one patient group only (Vissers & Beech, 2005). Having dedicated services reduces variability in treatment and length of stay, and allows for better prediction and coordination of all units in the hospital related to a particular class of patients (Green, 2012). Dedicated resources, in contrast to resources shared between different users (Van Donk & Van Der Vaart, 2005), have higher throughput and require less complex planning and control methods to use them efficiently (Gyulai, Kádár, & Monostori, 2014).

Given the complaints of the patient, which can be seen as the inputs to the system, the patient is routed through the system according in a random manner (Creemers & Lambrecht, 2011). In this research we consider the variety in patient routing to be a driver of dynamic complexity. The routing defines the different steps that are taken and the resources used. According to Vissers & Beech (2005), a high variety in the number of operations, duration of the operations and variety in routes results in less predictability. This, in turn, requires more flexibility in planning these operations for the patient. SCs with higher flexibility are more costly than those with lower levels of flexibility (Merschmann & Thonemann, 2011).

This research only focuses on two drivers of complexity, as the goal of this research is not to describe a complex system in detail, but to understand integration in a complex context. Therefore, including these two drivers seems appropriate for this research aim.

2.2 Integration

The main concept of this research is integration. Integration can be defined as: “the degree to which a manufacturer strategically collaborates with its SC partners and collaboratively manages intra- and inter-organization processes” (Flynn, Huo, & Zhao, 2010, p. 59). Integration can be defined according to its span, scope and intensity (Drupsteen et al., 2013).

2.2.1 Dimensions of integration

10 et al., 2013). This is defined by Stevens (1989), who states that no integration is characterized by activities with different responsibilities that are divided into separate, independent departments. Functional integration focuses on the inward flow of goods, but this flow is not well managed throughout the organization. In internal integration, the flow of goods into the organization until the end customer receives them is managed and is characterized by integrated planning and control. Finally, in external integration, the customers and suppliers are also included. This implies being customer-oriented rather than product-oriented, cooperating with the suppliers from the earliest stages in product development and involvement at all stages in the SC (Stevens, 1989). According to Drupsteen et al. (2013), functional integration concerns integration between two SC partners and internal integration between three partners. In this research, the focus is on integration within the organization.

Research showed the positive effect of high levels of integration on organizational performance. Both Flynn et al. (2010) and Leuschner, Rogers, & Charvet (2013) showed that internal integration directly improves performance. Pagell (2004) concluded that a lack of integration is a sign of working according to cross-purposes, leading to lower levels of organizational performance. Gimenez et al. (2012) analysed the relationship between integration, performance and complexity and found that when complexity is high, integration increases performance. However, when complexity is low, integration does not improve performance. Besides, more integration in higher levels of complexity is not always required and specific integration practices or patterns might be required (Gimenez et al., 2012). Shou et al. (2017) also showed that higher levels of product complexity and variety increased the need for internal integration, but their study did not include the effect of integration on performance.

The scope of integration refers to the aspects that are integrated (Drupsteen et al., 2013). These are physical flows, planning and control, organisation, the flow of information and product development (Van Donk & Van Der Vaart, 2004). The scope of this research is the planning and control function. Given the varying degrees of complexity and information requirements concerning planning and control decisions (Butler, Karwan, & Sweigart, 1992), planning is often hierarchically organized into strategic, tactical, offline and online operational planning (Hans et al., 2012).

11 share relevant information and there is a commitment between them, for example, an agreed quantity of products. In the third stage, the majority of the planning and control decisions are centralized (Van Der Vaart & Van Donk, 2004). Within the three stages, different integrative practices may be found. This is explained below.

2.2.2 Integrative practices

The study of Frohlich & Westbrook (2001) identified eight integrative practices for integration between manufacturers and customers or suppliers. These are access to planning systems, sharing of production plans, joint EDI access/networks, knowledge of inventory mix/levels, packaging customization, delivery frequencies, common logistical equipment/containers and common use of third-party logistics. The study of Drupsteen et al. (2013) found the following integrative practices: sharing planning and waitlist information (related to the transparency stage), cross-departmental planning (commitment stage) and combined appointments (integrative planning stage).

The study of Van Donk & Van Der Vaart, 2005) contains a framework of integrative practices relating to the level of uncertainty in volume and variety of products in the SC. When the levels of uncertainty are low, integrative practices are related to the physical flow of products and concern simple operating procedures and working together to optimize the control of inventories. When the uncertainty concerning volume is high, it is more difficult to allocate capacity and stocks are kept. When uncertainty in volume is low, buyers can predict capacity and suppliers can reserve capacity. Finally, for both high levels of uncertainty, capacity reservation and keeping stocks is not possible. Suppliers should be provided with adequate information by the buyers to organize its capacity as efficiently and effectively as possible (Van Donk & Van Der Vaart, 2005).

2.2.3 Antecedents of integration

Antecedents can be described as factors that enhance or impede the implementation of integration (Lee, Kim, Hong, & Lee, 2010) and help to understand the relationship between complexity and integration. Drupsteen et al. (2016) defined five antecedents that are initiating, facilitating or inhibiting integration.

12 of complexity, SCs are cut into pieces, leading to a fragmented network of semi-independent organizations and locally measured performance (Akkermans & Dellaert, 2005). Besides, it leads to misaligned incentives and conflicting goals, preventing the SC to operate as a system (Mathew, John, & Kumar, 2013). The second initiating antecedent is process visibility. According to Drupsteen et al. (2016), an understanding the total care process is regarded as a prerequisite for integration. Low levels of process visibility imply limited knowledge concerning the consequences of decisions for other departments and lack of understanding for which information must be shared or received. (Drupsteen et al., 2016). SCs must encourage function areas to share information and pursue objectives with a broader scope than single-function goals (Richey, Chen, Upreti, Fawcett, & Adams, 2009).

Two inhibiting antecedents are identified, which are shared resources and uncertainty/ variability. The ability of a supplying department (of shared resources) to organize its production capacity among different users determines the efficiency and effectiveness in the SC (Van Donk & Van Der Vaart, 2005). According to Drupsteen et al. (2016), there may be conflicting objectives between the department supplying the shared resource and the user of the shared resource. The supplying department is responsible for the utilisation level of the shared resource and the user for the flow performance of patients. This makes supplying departments more reluctant to allocate capacity (Drupsteen et al., 2016). When there is more uncertainty/ variability, is more difficult to predict and allocate capacity for the supplying department, because buyers cannot predict future demand for the shared resource given a certain volume of patients (Van Donk & Van Der Vaart, 2005). Another study investigated the degree of integration in relation to the type of business conditions (simple or complex) and types of resources (shared or dedicated). When business conditions are more complex and there are shared resources, more integration is needed but the presence of shared resources makes this more difficult (Van Donk & Van Der Vaart, 2004).

13 2.4 Conceptual model

The conceptual model is shown in figure 2.1. This model is very simple, although it captures the aim of this research: to understand integration in complex care processes and find out whether it improves flow performance. Based on the above-discussed literature, it is hypothesized that for very high levels of complexity (both detail and dynamic), integrative practices will mostly be seen concerning information sharing, whereas for a lower level of detail complexity will result in more reserved capacity. In care processes, a stock of patients cannot be kept, which leaves it unclear what kind of integration will be seen when there higher levels of dynamic complexity. It appears that higher levels of integration in more complex processes increases flow performance, however, this relationship is not entirely clear yet.

14

3. METHODOLOGY

In this chapter, the research design, data collection methods and measurements and data analysis are discussed.

3.1 Research design

This research followed an abductive approach, which leads to new insights about existing phenomena, by investigating them from another perspective (Kovács & Spens, 2005). This was appropriate for the aim of this research, as integration was studied in a complex care process by building further on earlier research of Drupsteen et al. (2013). Furthermore, the research method that is used is a case study. Not only is the case study an appropriate method for answering ‘how’ and ‘why’ questions (Yin, 2003), it also allows an in-depth inquiry into the problem (Yin, 2012), which is necessary given the current gaps and unclarity in the literature. The type of case study that is performed is a single-embedded case study (Yin, 2012). A better and in-depth understanding of the case can be gathered through analysing a single case, while by including embedded units the advantages of multiple-case study can be achieved. This is, for example, a higher level of transferability (Grünbaum, 2007).

3.2 Case description

The research was conducted in a hospital located in the Netherlands, which was selected through convenience sampling (Etikan, Musa, & Alkassim, 2016). This hospital was willing to participate in this research and is located in close geographic proximity. Furthermore, the care process that was analysed in this research is oncology. Different researchers stated that oncology is a complex care process (Drupsteen et al., 2016; Leeftink et al., 2018; Liang, Turkcan, Ceyhan, & Stuart, 2015). Not only from a medical perspective but also from an SCM perspective. In diagnostics, oncology patients require a variety of activities, which have precede relationships and involve resources from different specialisms (Leeftink et al., 2018). Besides, treatments which are often received from multiple departments for extended periods (Liang et al., 2015). This makes oncology an appropriate example of a complex care process.

3.2.1 Description of oncology care process

15 science (NSC) and radiological examinations (RAEX). NSC includes three activities, PET-, SPECT scans and sentinel node procedures. These are grouped as these individual procedures have relatively low occurrences The results of the examinations are discussed in the repeat outpatient consult (OPC2) and the treatment phase begins. This may include surgery (OR) and chemotherapy (CMT). After treatment, the patient must return to the hospital for a series of follow-up appointments. The patient can also receive more treatment or palliative care. Although there may be other diagnostic activities or treatment options than those included in this research, this research focussed on the most common activities and those performed in this hospital only. Furthermore, the referral process and follow-up phase were excluded from the research. The process is shown in figure 3.1.

3.2.2 Units of analysis

This research includes four embedded units, which are four different patient groups. This also means that there are four units of analysis. These cases were chosen based on their level of detail and dynamic complexity. The level of detail and dynamic complexity varies between the different groups. The groups were chosen in this way deliberately, to better understand integration in a more complex context, and if and how these complexity characteristics lead to different outcomes, as proposed by literature. This means that both theoretical and literal replication logic is used (Yin, 2003). The characteristics of the cases are shown in table 3.1. Figure 3.2 shows the volumes per patient group for the first visit, diagnostics and treatment. As it is uncertain whether patients have cancer up till diagnostics and not every patient receives treatment, these volumes are also included.

16 Table 3.1 Characteristics of cases

Case Type Detail complexity Dynamic complexity

1 Malignant neoplasm of breast High High

2 Malignant neoplasm of prostate High Low

3 Malignant neoplasm of ovary, corpus uteri and other and unspecified female genital organs

Low High

4 Malignant neoplasm of skin High Low

Case 1 involves patients with breast cancer. This patient group has high routing variety since patients can be referred to the hospital through three referral flows. These referral flows are arranged based on the level of urgency. For these patients, there are a variety of diagnostic and treatment options, which can also be combined, resulting in a large number of patient routings. Approximately 650 patients came to the hospital in the analysed period (January 6th 2019 until May 4th, 2020) and figure 3.2 shows that many patients were actually treated. This can be explained by the fact that there is a population screening for this patient group in. This implies that patients are already examined before the OPC1. According to the hospital, this group is a high volume group. In line with the hospital, this research uses the same classification.

The second case includes patients with prostate cancer. For this group, the routes of the patients are generally spoken the same. There are different diagnostic activities and treatment options, but not as much as in the first case, which implies a low level of routing variety. Again the same classification as the hospital is used in terms of volume, which is high.

Case 3 includes three smaller patient groups, which we refer to as gynaecologic cancer. These three groups were chosen as the individual groups would not yield appropriate results, as they are very small. For endometrial cancer, there is also a population screening. This case has the lowest volume of patients. Additionally, not all patients receive treatment in this hospital. For patients with ovarian cancer, the surgery happens elsewhere. The routing variability is high, as there are different diagnostic activities and treatment options.

17 3.3 Data collection and measurement

In this research, both primary and secondary data was collected. First, secondary data was used by reviewing the available literature on this topic, which increases the level of construct validity (Yin, 2003). Furthermore, primary data was collected through different data sources, which are interviews and archival data. The data from the interviews were compared with the archival data and vice versa. This means that the collected data is triangulated, resulting in higher levels of reliability, construct validity and more convincing and accurate conclusions of the case study (Yin, 2003).

3.3.1 Interviews

In this research, 14 interviews were conducted, shown in table 3.2. First, unstructured interviews were held with the project leader of VBHC and quality manager, to better understand the context of the research. These respondents provided information on the oncology care process, patient perspective and quality management and criteria.

Second, medical coordinators (MCs) were interviewed, who are physicians and heads of the department. They oversee different types of care paths within the department and are responsible for managing and organizing medical capacity. The unit heads (UHs), were also interviewed who responsible for a smaller part in the complete care process of the patient, for example, the outpatient clinic. The UH manages hospital capacity, for instance, equipment and

666 657 538 459 440 188 152 140 ₈₅ 2766 1801 1288 0 500 1000 1500 2000 2500 3000

OPC1 Diagnostics Treatment

Vo

lu

m

e

C1 C2 C3 C4

18 consultation rooms. Both the MCs and UHs provided data with regards to how patients move through the care path, the activities performed in different departments, integration of planning and obstacles they encounter regarding integration. From each patient group, both the MC and UH were included. Furthermore, UH of the MRI, inpatient services oncology (ISO), pharmacy and MC surgery were also interviewed.

Finally, the care process manager (CPM) was interviewed. This person is responsible for in making decisions at the strategic planning level about the type of care to be delivered, innovations and the position of the hospital in the market. Thereby, the CPM guards the quality of care and emphasizes high flow performance of all patient flows. This respondent provided information on the strategy and vision of the hospital, goals concerning oncology care, throughput criteria and the obstacles to integration and flow performance from their perspective.

Table 3.2 Interview respondents

Interview Function Date Duration (in min)

1 Project leader value-based healthcare 04-07-2020 20

2 Quality manager SONCOS norms 04-14-2020 20

3. UH breast centre 04-21-2020 60

4. MC mamma care 04-29-2020 15

5. UH outpatient clinic urology and dermatology 04-24-2020 40

6. MC urology 05-13-2020 30 7. MC dermatology 04-21-2020 30 8. UH gynaecology 04-23-2020 60 9. MC gynaecology 04-22-2020 30 10. MC surgery 04-22-2020 30 11. UH clinical pharmacy 04-22-2020 30

12. Senior MRI technician 05-07-2020 30

13. UH inpatient services oncology 04-23-2020 40

14. CPM 04-22-2020 40

19 (both Dutch and English). All interview questionnaires follow the same structure, where the questionnaires are divided into five sections: introductory questions, questions about the patient group and (part of the) care process, planning and control, integration of planning and control and concluding questions.

To determine the degree of integration, questions were asked about what kind of integrative practices there were with other specialisms and/or departments. If so, the respondents were asked about for example the kind of information that is shared, with whom and how often this is done or how much capacity is reserved per week. Questions were also asked concerning obstacles and challenges in integration, which were linked to the antecedents defined earlier.

3.3.2 Archival data

To calculate the patient flow performance of the different cases, patient transaction data was included. This data shows for a specific patient the diagnosis, the occurrence of and different steps that were taken and the dates of these steps. This data makes it possible to calculate the flow performance of the different cases and determine the waiting time between process steps. The SONCOS quality norms were also included in this research, to evaluate the actual flow performance with the norms specified. The SONCOS norms include among other things, the maximum waiting time for patients for different phases in the care process. To measure the effect of integration on flow performance, the flow performance of the steps with integration and those without integration were compared as well as flow performance between cases.

3.4 Data analysis

3.4.1 Interviews

With permission of the interviewees, the interviews were recorded. This made it possible to make transcripts of the interviews. In this research, all interviewees permitted recording. Unfortunately, one interview was not successfully recorded. As notes were made during the interview, data could still be collected and used.

The interview transcripts were coded in Atlas.ti. The literature provided guidance for the possible codes that could occur, which means that deductive coding is used. However, some codes could not be linked to the information provided in the theoretical framework, which means that inductive coding is also used (Fereday & Muir-Cochrane, 2006).

20 the respondents are assigned an objective code. In axial coding, these objective codes were linked together in new categories. Finally, selective coding was used, where the categories were linked to core concepts, either defined through literature (stages of integration, integrative practices and antecedents) or own interpretation (Gioia et al., 2013; Voss et al., 2002). According to Gioia et al. (2013), coding themes categories and concepts provides the basis for building the data structure, which gives a graphic representation of how the raw data is processed into concepts. The data structure is shown in Appendix B.

3.4.2 Archival data

The patient transaction data is analysed in Excel. First, the dataset was filtered for faulty data and non-relevant data was removed. In calculating how the flow performance, the INDEX and MATCH functions were combined. This formula returns for a specific patient, the date when a specific step occurred, for example, OPC1. By using this formula again to find the date for the MRI, the waiting time between the two steps can be determined. This was done for all steps in the process for all patients. By calculating the occurrence of waiting times, creating cumulative sums and the percentages accordingly, the flow performance is given. Furthermore, the usage of the different resources (shared and dedicated) is calculated, as well as the standard deviation (STDEV). This allows us to see how different the resource usage is per week, in other words, the volume variety for a given resource.

It should be noted that for one patient, a step can occur more often. However, we assumed that for each process step, an activity occurred only once for each patient. In calculating the flow performance for the diagnostic activities, the steps were excluded that did not have a first visit before the diagnostic activity or a repeat visit after the step, to make sure that the calculations referred to the diagnostic phase. For surgery and chemotherapy, only those were considered after the repeat visit, so this only refers to treatment.

3.4.3 Within and cross-case analysis

For the within-case analysis, a database is made for each case, whereby the data gathered from all data collection methods is collected and combined. The within-case analysis contributed to a better understanding of integration and its effect on flow performance for a given case with specific characteristics. This is necessary before generalized patterns can be found across cases (Eisenhardt, 2016).

22

4. RESULTS

In the within-case analysis, the cases are described and their flow performance and integrative practices are discussed. The cross-case analysis compares the cases and discusses the antecedents of integration that were identified during data analysis. To begin with, two figures concerning the flow performance and a table with integrative practices are shown. This is discussed in detail in the within-case analysis.

There are three throughput criteria for oncology patients, shown in table 4.1. The referral norm is not included, as the data did not include information regarding this.

Table 4.1 SONCOS norms (Stichting Oncologische Samenwerking, 2020)

Norm Complete within

Referral general practitioner and first consult < 1 week

First visit and treatment < 6 weeks

Diagnostic phase < 3 weeks

Figure 4.1 shows the flow performance concerning the time between OPC1 and treatment (OR or CMT). The shows the time in days (x-axis) and the percentage of patients who finished this part of their care path (the y-axis). At 42 days, the flow performance of case 2 is 36%.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 P er ce n tag e o f p atien ts f in is h ed Time in days C1 C2 C3 C4

23 Figure 4.2 shows the flow performance for the diagnostic phase, the time between all diagnostic activities and all treatment options.

The cases make use of different resources, shared or dedicated. The total and average resource use and STDEVs between weeks are calculated per case. This is shown in Appendix C. The integrative practices that were found are summarized in table 4.2. A detailed overview of this table is shown in Appendix D. For all cases, planning and waiting list information is shared between the OR and the groups and occasionally involving radiology during a tactical planning meeting (TPM). Furthermore, there is a service level agreement (SLA) regarding priority rules and throughput times. To meet the throughput criteria there are shortened access times for oncology patients at specific process steps, to make sure that they are seen as fast as possible. An example of where this could be seen is at pathology.

Table 4.2 Integrative practices

Integrative practice Type Stage Applies to cases

Specialisms have a tactical planning meeting with the OR planning and radiology for MRI (radiology only when necessary)

Sharing planning and waiting list information

Transparency 1, 2, 3, 4

Maximum waiting time OR is communicated

Sharing waiting

list information Transparency 4 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 450 P er ce n tag e o f p atien ts f in is h ed Time in days C1 C2 C3 C4

24 Scheduling details OR

communicated

Sharing planning

information Transparency 4

Formal agreements about priority and throughput times

Service level agreement

Commitment &

coordination 1, 2, 3, 4 Reserved capacity at OR Reserved capacity Commitment &

coordination 1, 3*

Reserved capacity at MRI Reserved capacity Commitment &

coordination 1, 2, 4 Reserved capacity at nuclear

medicine Reserved capacity

Commitment &

coordination 1

Mammography and echo appointments combined

Mammography and surgical consult combined

Combined appointments

Integrative

planning 1

First consult and echo are combined Combined appointments

Integrative

planning 2

Appointments for removing stitches and discussing diagnosis are combined

Combined appointments

Integrative

planning 4

* Exact amount of capacity could not be obtained during data collection

4.1 Within-case analysis Case 1: Breast cancer

This group is one of the spearheads of the hospital, which implies that this group may operate above the production budget. There is a specifically designed breast centre to accommodate all these patients. Comparing the flow performance of this case (by looking at figure 4.1) with the norms in table 4.1 shows that at 42 days, 74% of the patients started their treatment. Figure 4.2 shows that at 21 days, 16% of the patients completed diagnostics.

25 While there is not a throughput norm for OR and CMT, the UH said that it takes three weeks till a patient receives surgery after being diagnosed. At three weeks, the flow performance for OR (409 surgeries) is 13% and for CMT (194) 22%.

Integrative practices were found in all three stages of integration. First of all, most integration is seen between the breast centre and the OR, where information about planning and waiting times is shared. While volume uncertainty is low (STDEV 3.1 surgeries per week), two slots are reserved per day. 13% of the patients received surgery within three weeks. Secondly, capacity is reserved at the MRI and information is also shared with radiology when needed. Three slots are reserved per day, again with high volume uncertainty (2.2, and for 42% of the patients, the norm was met.

As a result of the SLA, the biopsy’s of these patients are examined before others. It was seen that for 92% of the patients this happened within three weeks. At nuclear science, there is also reserved capacity, 2 slots per day and again volume uncertainty is low (3.1). Only 16% of the patients had an examination within three weeks. Finally, mammographies and echo appointments, as well as mammographies and surgical consults, can be combined. The flow performance of the combined appointments is shown in figure 4.10, p. 32. This could not be calculated for mammographies and surgical consults. At three weeks, the flow performance is

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 P er ce n tag e o f p atien ts f in is h ed Time in days

MRI BIO ECHO CT NSC

26 48%. Combined appointments are also decoupled when the waiting time increases. For mammographies and echo’s alone flow performance is 76% and 91%.

Case 2: Prostate cancer

Like the previous group, this group is a spearhead of the hospital. Figure 4.1 and 4.2 showed that at six weeks, 37% of the started treatment. 9% of the patients complete their diagnostic phase within three weeks.

This group also uses all shared resources and has a dedicated resource which is echo equipment. The flow performance of the diagnostic activities is shown in figure 4.4. At three weeks, the flow performance is 52% for ECHO (318), 46% for BIO (267), 31% for CT (64), 29% for RAEX (42), 27% for MRI (150) and finally, 19% for NSC (161). With regards to treatment, the figure shows that at three weeks, the flow performance for OR (103) is approximately 7% and for CMT (81) 23%.

For this group, integrative practices were also found in all three stages. Information about planning and waiting times are shared between the urology department and the OR. 7% of the patients are operated within 3 weeks. Second, information is shared when necessary with radiology about the MRI and capacity is reserved. 3 MRI slots are reserved per day, while the volume uncertainty is low (1.7). It was seen that 27% of the patients completed this step within

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 P er ce n tag e o f p atien ts f in is h ed Time in days

MRI BIO ECHO CT NSC RAEX OR CMT

27 three weeks. Furthermore, the SLA also yields for this group. Accordingly, the flow performance for BIO is 46%.

Finally, there are combined appointments, for the first consult and echo. This occurred 259 times and the flow performance of combined appointments is 56% at three weeks. For echo’s alone, this is 52%.

Case 3: Gynaecology

According to figures 4.1 and 4.2, at six weeks, 80% of the patients started their treatment. The diagnostic phase is completed within three weeks for 38% of the patients.

This case uses all shared resources, although MRI and nuclear science only once. There are no dedicated resources for this case. Furthermore, not all patients receive surgery in this hospital. Figure 4.5 shows the flow performance for all process steps. At three weeks, the flow performance of BIO (100) and ECHO (76) is 64% and 34% for CT (59). The flow performance for RAEX (34) is 16%. MRI and nuclear science only occurred once and have flow performances of 0% and 100% respectively.

According to the MC, patients can usually receive surgery within two weeks after being diagnosed. At two weeks, the flow performance for OR (48) is 46%. At three weeks, this is 69%. For CMT (30), the flow performance at three weeks is approximately 69%.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 Perc en ta ge o f p at ie n ts fin is h ed Time in days

MRI BIO ECHO CT NSC RAEX OR CMT

28 Integrative practices for this group were found in the transparency and commitment & coordination stage. First of all, information is shared between gynaecology and the OR and there is reserved capacity and volume uncertainty is low. However, information about how much capacity this is could not be obtained. At three weeks, it shows that 69% of the patients received surgery. Furthermore, the flow performance for BIO, where priority is given, is 64% at three weeks. Finally, the UH of the gynaecology clinic also discussed that they are considering sharing capacity with their neighbouring policlinic, the children’s clinic. This is not included in this research as it is not implemented yet.

Case 4: Melanoma

Figure 4.1 shows that 68% of the patients waited six weeks between OPC1 and OR. Treatment for this group only includes OR, as melanoma patients do not receive chemotherapy. Figure 4.2 shows that 42% of the patients completed diagnostics within three weeks.

This case makes use of all shared resources, except chemotherapy and there are no dedicated resources. Furthermore, MRI and radiological examinations only occurred 3 and 4 times. According to figure 4.6, the highest flow performance is seen for radiological examinations (4), which is 50%. For CT (17), it is 25%, and for ECHO (79) it is 15%. BIO (1368) has 11%, NSC (72) has 5% and finally, MRI (3) has 0%.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 P er ce n tag e o f p atien ts f in is h ed Time in days

MRI BIO ECHO CT NSC RAEX OR

29 For surgery, the waiting time is usually two weeks according to the MC. The flow performance for OR (1188) at two weeks is 79%. At three weeks, the flow performance is 81%.

Integrative practices were found in all three stages for this case. In the transparency stage, information is shared between dermatology and the OR in the TPM. Additionally, the OR communicates the maximum waiting time and scheduling details with dermatology. Results showed that 79% of the patients received surgery in two weeks and that there is high volume uncertainty (9.9). Second, information is also shared with radiology when needed and capacity is reserved. Per week, 4 slots are reserved. However, these slots are not only for melanoma patients, they are reserved for ‘weak tissue’ patients. The MRI was used by this group only three times and none of the patients received their MRI within 3 weeks. Looking at biopsy’s, for which priority is given, for 11% of the patients biopsy was done within three weeks. Finally, there are combined process steps in the integrative planning stage. This implies that when the stitches are removed, the physician also discusses the diagnosis of the patient. This happens within the outpatient clinic dermatology. The flow performance for these steps cannot be calculated, as the data does not specify exactly when the stitches are removed in the process.

4.2 Cross-case analysis

This section further explores the relationship between integration and performance by looking at each integrative practice individually and the corresponding flow performances. Furthermore, this section contains the factors discussed by respondents that help or hinder integration, which are categorized in the antecedents.

4.2.1 Integrative practices and flow performance

First of all, for MRI information is shared with all cases when necessary and capacity is reserved for cases 1, 2 and 4. Figure 4.7 compares the flow performances and it shows that at three weeks, case 1 (3 slots reserved per day) has a flow performance of 42%, and case 2 (3 slots) 27%. For case 4 there are 4 slots per week, and at three weeks the flow performance is 0%. For case 3, only one MRI was taken within three weeks. According to respondents of cases 1 and 2, the MRI is the bottleneck:

“We want to grow and we have permission for that. But then I have more patients and they are in traffic because the MRI has an occupancy rate of 95% and cannot do my extra number

of mamma patients” (UH C1)

30 Integration was also found between the patient groups and the OR. Information is shared with regards to planning and waiting times for all cases. For case 1 and 3 capacity is reserved and additional information is shared with case 4. At three weeks, it shows in figure 4.8 that case 4 has a flow performance of 81%, case 3 69%, case 1 has 13% and case 2 only 7%.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 P er ce n tag e o f p aten ts f in is h ed Time in days C1 C2 C3 C4 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 P er ce n tag e o f p atien ts f in is h ed Time in days C1 C2 C3 C4

Figure 4.7 Flow performance MRI

31 For integration as a result of the SLA, we looked at BIO as shown in figure 4.9. All cases receive priority and at three weeks, case 1 has a flow performance of 92%, case 3 64%, case 2 46% and finally case 4 11%, with the highest amount of BIO.

The results also showed that appointments for mammography and echo’s are combined in case 1, 2 and 4 (for which flow performance cannot be calculated). Both cases have echo equipment as a dedicated resource and case 1 has their own mammograph, all of which are used in combined appointments. With combined appointments, the flow performance at three weeks is 48% for case 1 and 56% for case 2, shown in figure 4.10.

Finally, it was found that there is reserved capacity for case 1 at nuclear science, 2 slots per day. For case 1 the flow performance at three weeks is 16%, 19% for case 2, and 5% for case 4. Case 3 only had one examination at nuclear science. For case 4 there is no integration concerning this process step, although the MC discussed that for patients with a higher stadium melanoma, this step can be a bottleneck. Figure 4.11 shows the flow performance.

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 425 P er ce n tag e o f p atien ts f in is h ed Time in days C1 C2 C3 C4

32 For the process steps CT, RAEX and chemotherapy, no integrative practices were found. Their flow performances are shown in Appendix E. The flow performance at three weeks for

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 P er cen tag e o f p atien ts f in is h ed Time in days C1 C2

Figure 4.10 Flow performance combined appointments

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 25 50 75 100 125 150 175 200 225 250 275 300 325 350 375 400 P er ce n tag e o f p atien ts f in is h ed Time in days C1 C2 C3 C4

33 chemotherapy (case 1: 69%, case 3: 22%, case 2: 23%), CT (case 1: 36%, case 3: 34%, case 2: 25%, case 4: 19%) and RAEX (case 4: 50%, case 1: 29%, case 2: 16%) are low, but not necessarily lower than the steps with integration. However, these process steps also have lower occurrences.

4.2.2 Antecedents of integration

During data collection, respondents discussed different factors that were helping or hindering integration. These factors are grouped according to the antecedents.

Performance management

This research on focused performance measures concerning throughput times. The performance measures concern larger parts of the care path rather than individual process steps. This means that the departments are collectively responsible for performance. The SLA between the involved departments helps to meet these performance measures:

“All relevant parties try to meet the SONCOS norms. So at radiology, at diagnostics, they all make sure that the results are available as soon as possible (…) So if pathology receives a biopsy with the suspicion of cancer, then those always precede those that are not suspicious

of cancer. They can ensure short access or supply dates” (MC C3)

To meet the performance measures, information is shared between departments with regards to their planning and waiting times and if necessary, reserved capacity is adjusted:

“Every week in the TPM we [specialisms and OR] look at what do we see coming our way, how is the realization of access times per specialism and also per patient group. If I see we

have too little or too much [capacity] I make adjustments in the grids” (UH C2 & C4)

It seems that performance management positively influences information sharing and accurately reserving capacity.

Process visibility

The respondents also discussed their understanding and sight of the complete care process for the patient. The UH of C2 said that they are looking for a better connection between what is happening inside the outpatient clinic and on the outside. Another respondent commented:

“That coordination [concerning planning], I don’t have that with all the outpatient clinics (…) I am very curious how that might be disturbing in the chain or a bottleneck through

34 Low level of process visibility can be due to the lack of information sharing:

“If we get new registrations, that is always unexpected. We never know what we can expect. The oncologists, they already have sight on the fact that they will be seeing new patients. But we only see that at the time of registration. I think the process can be much faster so that we

could already have a place available for that patient” (UH inpatient services oncology)

Furthermore, the lack of process visibility makes it difficult to actively manage the capacity needed in the care path with all involved departments:

“I think the biggest part with regards to improvements is having more insight in numbers, with which you can have a more active planning with each other. Now it is more or less that we act on what we see coming our way, so this means that you are running behind the facts.

So what you want is to be more in front” (UH C2 & C4).

While cases 1, 2 and 3 speak of a low level of process visibility, for case 4 the level is higher:

“Dependent on the type of melanoma, the patient is seen a specific amount of times in the hospital for a certain amount of time. We plan and estimate because we already have the diagnose when the next control appointment is. So that is planned immediately” (MC C4)

For this case, the process seems much simpler and this case does not deal with other departments that much, which lowers the need for coordination between departments. Moreover, this makes combining appointments easier.

It appears that low process visibility negatively influences information sharing. Higher levels of process visibility would influence the commitment & coordination stage because departments could reserve capacity (more precisely). Finally, high process visibility positively influences combined appointments.

Information technology

The hospitals' information system is used to share information about waiting times, planning and can be used to schedule combined appointments. IT in case 4 enables the number of activities to be reduced:

“The patient does not see that the same day, but I can see when the patient has the appointment. Because if I call the patient the day after when giving the diagnose, I can

35 Respondents of cases 1, 2, 3 and OR indicated that there is a need for a dashboard:

“What I miss is a dashboard that states whether the planning is done correctly or not (…) I just want to hear well the planning is alright there are no waiting times. You want to have a

dashboard where everything is green and it is right” (MC surgery)

“So a kind of dashboard where every week or month you receive back what the waiting times

were, what the throughput times are, the average waiting time for MRI, so it is easier to turn the knobs” (UH C2 & C4)

The UH of inpatient services oncology said that the information system is not functioning properly. On the one hand, the system does provide oversight in which patients are allocated to a bed. Also, it occurs that appointments are delayed and the planners must change the complete sequence and related appointments at other departments manually:

“That is why they have a paper agenda in which they put stickers that say this patient is in bed 1 and this bed is occupied until 11 and after a new patient can be allocated to that bed (…) They [planners] won’t go over the whole treatment trajectory, because it is delayed with a week occasionally. The more they plan in advance, the more they have to restore if someone

is delayed by a week” (UH inpatient services oncology).

The above shows that IT facilitates information sharing between different departments and this could be increased when with a dashboard. Accordingly, capacity can be reserved more accurately. Bad functioning IT makes it difficult to plan sequences of appointments and leads to inefficient work procedures.

Shared resources

Different patient groups use multiple shared resources. The fact that these resources must be shared leads to integration, as agreements have to be made about the allocation thereof:

“There are also slots at the MRI and for fusion biopsy we have slots. Because we know what is coming in approximately every month, we have reserved places for it” (MC C2)

“For the MRI we have created specific grids so that our planners and secretaries know at

which times they can plan certain examinations” (Senior radiology technician)

36

“What we do at the breast centre is only mamma related echo’s. If we would do an echo of something non-mamma related, it will do something with our capacity. So we have agreed, only mamma-related echo’s because we do not have enough capacity for that” (UH C1)

In case 1, appointments with the mammograph and echo are combined and for case 2 combined appointments involve a first consult and an echo. Having dedicated resources makes integration easier, as the planners of the department do not have to coordinate with other departments for available slots.

Case 1 also has a mammograph and surgical consult combined. It happens that there is no waiting time for the mammography while there is for the surgical consult. When this happens, the appointments are decoupled, so that the mammography can take place as soon as possible. Shared resources positively influence reserved capacity, while they negatively influence combined appointments.

Uncertainty / variability

During the interviews, uncertainty in the care path of the patient and variability in demand was also discussed by the respondents. The MC surgery said that for the larger patient groups (cases 1, 2 and 4) there is less uncertainty in the demand for surgeries as the number of patients, in general, is larger:

“That is also the case for breast cancer. These numbers are so large, you can operate patients every week (…) it is mainly because the numbers are bigger, every week there are a

certain amount of patients” (MC surgery)

This makes it easier to reserve capacity for these groups. The UH of C1 said that the population screening causes uncertainty in the process as it is unclear how much capacity they need. According to the UH outpatient services, the fact that the population screening occurs and the possible increase in demand is not shared with them. However, the CPM concluded that the population screening, relating to cases 1 and 3, can be used as a predictor of demand rather than seeing it as a disturbance of the process:

“Our biggest predictor and that also yields for the mamma, are the population screenings. So at the moment, a patient arrives for a population screening, we already the percentage that

37 Finally, the respondents discussed the importance of looking at variability in demand not from one individual care path or department, but collectively to adequately manage capacity and meet the throughput criteria:

‘’You notice that for many unit heads it is difficult to understand (…) that there is still variability that they have to absorb. But everyone looks at that individually (…) I believe by looking more broadly at the group [oncology] you are less affected by variability (…) Also,

you are not looking ahead. I think we miss quite some information” (CPM).

“ If you would combine that [demand for MRI services of each group] with all four [breast- prostate-, lung- and colon cancer] big groups, you will get a situation like the larger the scale

the smaller variability. This can result in better alignment to reduce capacity loss at those follow-up diagnostic activities” (UH C2 & C4).

38

5. DISCUSSION

The results concerning integration in complex care, the effect of integration on flow performance and the relationship between integration and flow performance are discussed in comparison with the literature in this chapter.

5.1 Integration in complex care

The span of integration ranges from no integration to functional integration. Internal integration is only seen when the MRI joints the TPM. This means that the level of integration is low. Two integrative practices were seen in all cases; sharing planning and waiting list information and the SLA.

Case 1 (high volume and routing uncertainty) has six integrative practices. It has the most reserved capacity (MRI, OR and NSC) and the volume uncertainty is low compared to the amount of reserved capacity. This is in line with Van Donk & Van Der Vaart (2005), that capacity reservation is possible. Additionally, when the slots are not filled, they can be filled by acute patients. Even though there are combined appointments, they do not seem to occur that often for the type that was examined. This can be explained through the high level of uncertainty in patient routing or the differences in waiting times.

There are four integrative practices for case 2 (high volume, low routing uncertainty). For this patient group, capacity is only reserved at the MRI. However, both NSC and MRI have similar total and average occurrences and volume uncertainty, and it was said that the MRI is the bottleneck in the process. However, it has higher flow performance than nuclear science. It was also found that there are combined appointments for this group. For both cases 1 and 2 the combined appointments involve dedicated resources. Less complex planning and control methods are required for dedicated resources (Gyulai et al., 2014), which appears to make appointments easier to combine.

39 Finally, case 4 has six integrative practices in all three stages. The OR shares scheduling details and communicates the additional OR waiting time. There being high volume uncertainty, this situation corresponds with Van Donk & Van Der Vaart (2005). Furthermore, capacity is reserved at the MRI. However, the MRI was only used three times, which raises the question of whether capacity reservation is even necessary. It seems therefore logical that the reserved capacity is for a larger group, including these patients. Combined appointments were also found where only the physician is required, which makes integration easier.

This research did not find evidence of cross-departmental planning, in contrast to Drupsteen et al. (2013). In this case, the patient is planned by the secretary of the supplying department. Higher levels of complexity cause fragmentation (Akkermans & Dellaert, 2005) which can explain why the patients are planned this way. Finally, a new integrative practice was found, the SLA. The nationwide agreements are translated into prioritization rules and shorter access times. The research of Drupsteen et al. (2013) included a patient group with a lower degree of urgency, where such rules are not necessary.

5.2 Integration and flow performance improvement

In general, case 1 (6 integrative practices) with the highest level of complexity, has the highest flow performance, while case 4 (6 integrative practices) with lower complexity has a relatively low flow performance. Therefore, the conclusion of Gimenez et al. (2012), stating that integration in higher levels of complexity leads to higher levels of performance is supported. Furthermore, case 3 (high detail complexity), with the least amount of integration has higher flow performances than cases 2 and 4 (high dynamic complexity). This implies that the effect of integration on flow performance is not straightforward.

Concerning the OR, the flow performance for case 4 (highest level of information sharing) is the highest, implying that sharing more information leads to higher flow performance. It also seems that more integration is not necessary for this case. For cases 1 (high volume) and 3 (low volume) capacity is reserved and the flow performance is low for case 1 and high for case 3. It seems that reserving capacity works when detail complexity is high, but for smaller volumes only. Finally, case 2 has the lowest flow performance here and the least amount of integration here, which show that in general, integration improves flow performance.

40 For the other integrative practices, the performance effect is not quite clear. Concerning the MRI, capacity is reserved for cases 1, 2 and 4 and there is (almost) no integration with case 3. Cases 3 and 4 have limited occurrences, which make it impossible to compare cases. The results do show that case 1 (high use of MRI and high complexity) has a higher flow performance than case 2 (lower use and lower complexity), again in line with Gimenez et al. (2012). It contrasts with both Bozarth et al. (2009) and Gyulai et al. (2014), who state that more patients leads to more shared resource consumption and therefore lower throughput times. A similar result was found concerning biopsies. The flow performance of case 1 (highest complexity) is the highest, while cases 2 and 3 (lower occurrences) also have lower flow performance. Finally, the flow performance for case 4 is very low, while it has the most occurrences. Perhaps prioritizing that much patients limits the overall effect. It remains unclear why cases 2 and 3 have low flow performance here.

5.3 Relationship between integration and flow performance

The relationship between integration and flow performance is further examined by exploring the antecedents of integration defined by Drupsteen et al. (2016), which are also applicable and in line with the results of this research.

41 contrast to Drupsteen et al. (2016), it was found that shared resources initiate integration, as agreements are made concerning their usage.

42

6. CONCLUSION

This research aimed to answer the following research question: “How are complex care

processes integrated and what is the influence of integration on patient flow performance?”

This chapter answers this question, discusses the theoretical and managerial implications and the limitations and future research directions.

In general, more complex processes are integrated through four types of integrative practices: sharing planning and waitinglist information, an SLA, reserved capacity and combined appointments. The implemented practices differ concerning complexity levels. When the level of detail and dynamic complexity is high, all four integrative practices are used and higher levels of integration are seen. When only dynamic complexity is high, a lower level of integration is seen while combined appointments do not occur. A high level of detail complexity meant that all four above practices were used, although more information is shared when the total volume of patients is higher. Furthermore, when the volume uncertainty is higher, more integration is used or dedicated resources are implemented.

Overall, it can be concluded that integration improves performance in complex care processes. Higher levels of integration with higher levels of complexity (both detail and dynamic) lead to higher levels of flow performance. At the same time, a low level of detail complexity and a low level of integration can lead to higher flow performance. This means that only information sharing can improve performance, but in other instances, reserved capacity is also needed. Furthermore, more dynamic complexity requires more integration, although, with a lower degree of volume uncertainty, this is not necessarily true. The above suggests that the type and amount of complexity matters, as well as the integrative practice. While it is clear that the relationship between integration and flow performance is influenced by both detail and dynamic complexity, this relationship is not entirely straightforward.

Furthermore, this research showed that the current levels of integration are initiated by performance management and shared resources, they are inhibited by low levels of process visibility and uncertainty/variability in combination with shared resources. Finally, they are and can be facilitated more through IT.

6.1 Theoretical implications