Towards an efficient discharge process

(1)

Towards an efficient discharge process

Rolien Elsa Lucker

(2)

2

Towards an efficient discharge process

Student R.E. Lucker Studentnumber: 10565086 E-mail: rolienlucker@gmail.com Mentors N. Elsinghorst, MSc E-mail: NElsinghorst@chipsoft.nl ing.M. Schoutendorp E-mail: MSchoutendorp@chipsoft.nl ChipSoft B.V. Tutor dr. ir. T. Broens Faculty of Medicine

Department of Medical Informatics, Amsterdam UMC location AMC - UvA E-mail: t.h.broens@amsterdamumc.nl

Location of Scientific Research Project ChipSoft B.V.

Orlyplein 10

1043 DP Amsterdam The Netherlands

Practice teaching period November 2018 – July 2019

(3)

3

Preface

This thesis describes my scientific research project (SRP) called “Towards an efficient discharge process”. This SRP is the final product of the master Medical Informatics. This SRP was performed at the Kliniek department of ChipSoft B.V. in Amsterdam. The aim of this thesis is to understand the current discharge process and to determine whether the discharge process can be optimized using predictions based on patient characteristics.

I would like to take the opportunity to thank everyone who helped me create this thesis. First of all I would like to thank my tutor, dr. ir. Tom Broens, and my mentors drs. Nicole Elsinghorst and ing. Merel Schoutendorp. Thank you for your help when we decided to change the direction of the SRP and thank you for the meetings and feedback that made my work better. I appreciated your guidance and I enjoyed working with you. There is also a number of Medical Informatics teachers that I would like to thank: dr. Martijn Schut for the help with my statistics, prof. dr. Kitty de Jager for the feedback during the home coming days and drs. Marieke Sijm for the input at the start of my internship. Thank you to the aftercare employees and nurses that were available for an interview, you helped to make it possible to create this thesis.

I would also like to thank my friends, family and fellow students for their mental support and giving me feedback when I needed it. In particular I would like to thank Patrick for always standing behind me and being proud of me. I would like to thank Suzanne, Frank, Alma, Denise and Lotte for their help and feedback during this internship. There are also employees at ChipSoft who I would like to thank for their help: Hugo for showing me a number of functionalities and Laurens for helping me with SQL. Thanks to everyone who was my roommate and made sure I had a pleasant internship: Frederique, Jos, Mike, Nathalie, Percy, Tim and Wendy. Finally, thanks to all the other colleagues from ChipSoft that I have spoken to and from the department Kliniek in particular!

(4)

4

Summary

The demand for care in the Netherlands increases in the coming years. Amongst other reasons, this is due to the aging population. It is important to have an optimized discharge process, to prevent a big increase in the amount of bed-blockers. Bed-blockers are patients that are medically ready to be discharged, but are waiting for availability at an alternative location. The aim of this research is to understand the current discharge process and to determine whether this discharge process can be optimized using predictions based on patient characteristics. First, we want to get a better

understanding of the current discharge process. To do this, we perform semi-structured interviews with aftercare employees. We create a model of the current discharge process based on the information gathered here. Next we perform a literature review to identify patient characteristics that have the potential to predict the type of aftercare that is needed after admission. We also perform a data analysis using multinomial logistic regression using data from a hospital in the Netherlands. We conclude that there are four patient characteristics that influence the type of aftercare, namely whether it was an emergency procedure or not, the marital status, ASA class and age of the patient. We determine that the multinomial regression model that we developed can predict 65% of admissions correct. Our goal is to have at least 90% accuracy before usage in daily practice. Thus, we conclude that at this moment it is not possible to optimize the current discharge process based on predictions on these patient characteristics. Further research is needed to determine how we can accomplish more accurate predictions on the type of aftercare after

admission. When this is accomplished, it would be possible to use these predictions to optimize the discharge process.

Samenvatting

De zorgvraag neemt toe in Nederland. Eén van de oorzaken hiervan is de vergrijzende bevolking. Het is belangrijk om een geoptimaliseerd ontslagproces te hebben om een toename van het aantal bed-blockers te voorkomen. Bed-bed-blockers zijn patiënten die medisch klaar zijn voor ontslag, maar die wachten op beschikbaarheid van nazorg. Het doel van dit onderzoek is om het huidige ontslagproces beter te begrijpen en om te bepalen hoe het ontslagproces kan worden geoptimaliseerd door middel van voorspellingen op basis van patiëntkenmerken. Eerst willen we een beter inzicht krijgen in het huidige ontslagproces. Om dit te doen houden we semigestructureerde interviews met

nazorgmedewerkers. Met de informatie die we hier vergaren, maken we een model van het huidige ontslagproces. Vervolgens voeren we een literatuuronderzoek uit om de patiëntkenmerken te identificeren die mogelijk het type nazorg kunnen voorspellen dat nodig zal zijn na een opname. We voeren ook een data analyse uit met gegevens van een ziekenhuis in Nederland. We maken hier gebruik van multinomiale logistische regressie. We concluderen dat er vier patiëntkenmerken zijn die het type nazorg kunnen beïnvloeden, namelijk of het spoed was of niet, de burgerlijke staat, ASA-klasse en leeftijd van de patiënt. Het ontwikkelde multinomiale regressiemodel kan 65% van de opnames correct voorspellen. Ons doel voor gebruik in de dagelijkse praktijk is om een accuraatheid van ten minste 90% te hebben. We concluderen dus dat het op dit moment niet mogelijk is om het huidige ontslagproces te optimaliseren op basis van voorspellingen op deze patiëntkenmerken. Vervolgonderzoek is nodig om te bepalen hoe we nauwkeurigere voorspellingen kunnen maken voor het type nazorg na opname. Wanneer dit is bereikt, zou het mogelijk zijn om deze voorspellingen te gebruiken om het ontslagproces te optimaliseren.

Keywords

Aftercare; Discharge; Destination; Bed-blockers; Process modeling; Interviews; Predictions; Literature Review; Multinomial regression analysis

(6)

6

1. Introduction

The number of bed-blockers are on the rise in the Netherlands (1, 2). Bed-blockers are patients that occupy a bed in a hospital without a need for medical specialistic care (3, 4). For example, in 2017, the amount of bed-blockers increased from 152 to 248 at the Rijnstate hospital (1). Besides discomfort for the patient, this introduces unnecessary organizational and financial burden for the healthcare organizations. In 2018 it costs the Dutch government up to €269.82 per day that a patient stays in the hospital unnecessary (5). This emphasizes the need for an efficient discharge process. This thesis zooms into the problem of bed blockers and searches for ways to optimize the discharge process to reduce the amount of bed-blockers.

In this chapter we start by discussing the problem of bed-blockers. Second, we discuss the background, context and environment of this research. Then we provide the goal and research question of this research. Next, we present the approach and expected results. Finally, this chapter ends with an outline of the remainder of this thesis.

1.1 Problem of bed-blockers

Patients who have to stay in the hospital longer than medically necessary get a ‘verkeerd bed’ or ‘vergoeding vervallen ziekenhuisindicatie’ registration. These registrations were first introduced in 1987 (6). Patients get these types of registrations when their medical specialistic care indication has expired, and they have to stay in the hospital until they can be transferred. The difference between a ‘verkeerd bed’ and a ‘vergoeding vervallen ziekenhuisindicatie’ is visible in Figure 1. A ‘verkeerd bed’ registration is used for patients who have an indication for long-term care with stay. An example of a facility that provides long-term care with stay is a nursing home. A ‘vergoeding vervallen

ziekenhuisindicatie’ registration is used for patients who do not need long-term care with stay, but are waiting for another type of aftercare, such as homecare (3, 4). Health insurance companies have to pay for both these types of registrations. Both of these types of patients do not need medical treatment at the hospital anymore and are medically ready to be discharged, however these patients are unable to leave the hospital. In this way these patients block beds for patients that do need medical treatment. When a hospital is close to full capacity, this means that there are no beds available for new patients, increasing waiting times and accessibility to care (7). This is why we will refer to these types of patients as bed-blockers.

Figure 1: The difference between a ‘verkeerd bed’ and ‘vergoeding vervallen ziekenhuisindicatie’ registration

When a patient becomes a bed-blocker, this can have various negative effects. Firstly, for the patients, it can be disappointing that they have to stay in the hospital longer than necessary.

(7)

7

Secondly, they do not receive the type of care that fits their situation best, since the care given in an aftercare institution differs from the care given in the hospital. Thirdly, a delayed discharge increases the risk of developing medical complications (8). An example of such a complication is a nosocomial infection, also known as a hospital-acquired infection. Another negative effect is that this heightened risk of developing complications can then lead to an increased caregiver burden upon discharge. For other patients, who are waiting for medical treatment, bed-blockers also have a negative effect on the waiting times for admittance into the hospital. Additionally, bed-blockers increase the total health care costs, because a bed in a hospital is more expensive than a bed in a long-term care facility (9). A bed in a nursing home costs around €210 per day, compared to around €270 per day for a bed in a regular care unit of a hospital (5, 10). A bed at home costs even less. An efficient and patient-oriented discharge process is of great importance to shorten the length of stay in the hospital, reduce the number of bed-blockers and reduce the total healthcare costs.

A delay in the discharge process can have various reasons, such as delayed requests for aftercare or long waiting lists at the aftercare institutions. Examples of aftercare institutions are rehabilitation centres and nursing homes (11). One study found that of all the hospitalized patients waiting for a determined alternative level of care, 36 percent was waiting for palliative care, 33 percent was waiting for a place in a nursing home, 18 percent was waiting for rehabilitation and 12 percent was waiting for homecare (11). A recent article published by NRC showed that in the Netherlands, patients are waiting an average of five to eleven days for a place in a nursing home, five days for rehabilitation and two days for palliative care (12). Numbers about the waiting time for homecare were not available.

A hospital dealing with the problem of bed-blockers is the St. Antonius Ziekenhuis in Nieuwegein. To handle the amount of bed-blockers, they opened a special department for them (12). In this

department these patients can wait until they can be transferred to an aftercare institution. This example indicates the magnitude of the problem of blockers. To improve the problem of bed-blockers, a possibility is to optimize the discharge process. We believe that one possible way to optimize this process is to request the aftercare earlier. This could be realized if there is a way to predict the type of aftercare based on the characteristics of the patient. This will be the focus of this thesis.

1.2 Background, context and environment

Multiple examples in Dutch hospitals show that there has been an increase in the number of bed-blockers in recent years (1, 2). One of the most important reasons of this increase in the number of bed-blockers is that the Dutch population is aging. In 2016 the Netherlands was populated by 3.1 million people of age 65 and up, and the expectation is that this number will rise up to 4.7 million people in 2041 (13). This will lead to an increase in the amount of care that needs to be provided (14). On the other hand, there is a big shortage of nurses, with hundreds of unfulfilled job advertisements (15).

To cope with this problem, the Dutch government has made a lot of changes in the Dutch healthcare system in recent years to stimulate elderly people to live longer at home. The general law on

exceptional medical expenses (‘Algemene Wet Bijzondere Ziektekosten’) that was put in practice in 1967, expired on January first 2015 (16). This law covered extramural care expenses, such as long-term care with stay. What was first covered by this law, is now covered by the other laws. The government also decided to close residential care homes (‘verzorgingshuizen’). Of the 2000 existing residential care and nursing homes in 2014, 800 will be closed by 2020 (17). Additionally, to be approved for a place in a long-term care facility, elderly people need an increasing demand for care. In 2018, around 58,000 elderly people were living at home that would have been living in a long-term care facility ten years ago (18). In 2018 the Dutch government also developed the program ‘Langer

(8)

8

Thuis’ (19) to stimulate more elderly people to stay in their own living environment for as long as possible.

Since the changes in the care system for elderly people, there has been an increase of around 30 percent of elderly people that visit the emergency department (20). In Figure 2 a visual

representation of elderly people that visit the emergency department and become bed-blockers is visible. Around 800.000 elderly people visit the emergency department every year, of which 543.000 get admitted to the hospital (21). 59.3% of the elderly people that get admitted, do not have a need for medical specialistic care, but they get admitted because they are unable to return home. This means that 322.000 elderly people become bed-blockers every year.

Figure 2: a visual representation of elderly people that visit the emergency department and become bed-blockers (21)

The total amount of hospital admissions in the Netherlands has more than doubled from 1993 until 2012 (22). One area that showed an increased care demand was the orthopedics department. This is illustrated by the fact that there has been a 12% increase in the amount of hip fractures in the last five years (23). Besides, between 1995 and 2005 the amount of knee arthroplasties has increased with 196%, and the amount of hip arthroplasties has increased with 50% (24). Arthritis is the main reason for knee or hip arthroplasties, also called replacement surgeries. As people get older, the chance of developing arthritis increases. In Table 1 the prevalence of arthritis for elderly people can be found, grouped by age. Of people aged between 65 and 69 years old, around 21 percent has been diagnosed with arthritis and of people aged above 85 years old, this percentage is around 42 percent (25).

Table 1: prevalence of arthritis grouped by age(25)

Age 65-69 70-74 75-79 80-84 85+

Prevalence of arthritis (%) 21.2 25.8 32.0 37.0 41.5 The aging population in combination with other factors, such as the increase in the prevalence of obesity, causes the expectation that the amount of knee arthroplasties will increase with 297% and the amount of hip arthroplasties with 149% until 2030 (24). Because of this expected increase of patients in the orthopedics departments, it is very important to have an optimized discharge process

(9)

9

here, to prevent a big increase in the amount of bed-blockers. This is why for the scope of this research, we focus on the orthopedics department.

1.3 Environment of the research

This research is performed at the company Chipsoft. ChipSoft is a company located in the Netherlands and Belgium that develops a hospital information system/electronic patient record system called HiX (26). Healthcare professionals can be aided in the discharge process by an electronic patient record system. However, research has shown that this aid is not always provided and that there is still room for improvement (27). One of the goals of ChipSoft is to develop software that aids healthcare professionals as much as possible in providing care to the patient (28). This means that they want to provide software that supports the healthcare professionals in the total healthcare process from start to end, including the discharge process. It is important that the software aligns with the work of the healthcare professionals involved in the discharge process. When ChipSoft gets a better understanding of the current discharge process and the bottlenecks that healthcare professionals face, they can improve their software even further, so that it supports the involved healthcare professionals in the total care process.

1.4 Goal and research questions

The aim of this project is to determine how the discharge process can be optimized using predictions based on patient characteristics for patients in the orthopedics department of a hospital. To do this, we first need to get a better understanding of the current discharge process in the orthopedics department and the bottlenecks that are experienced here. Next, we need to determine whether it is possible to predict the type of aftercare that will be needed and which patient characteristics have the potential to predict this. This leads to the following research question and sub-questions: Research Question

What does the current discharge process in the orthopedics department of a hospital look like and is it possible to optimize this process using predictions based on patient characteristics?

Sub-questions:

1) What does the current discharge process for patients in the orthopedics department look like?

2) Which bottlenecks are experienced in the current discharge process? 3) Which patient characteristics influence the type of aftercare that is needed?

4) Is it possible to predict the type of aftercare that is needed based on patient characteristics?

1.5 Approach and expected results

In this section we describe the general approach and expected results of this thesis. In the

subsequent chapters of this thesis more detailed descriptions can be found on the used methods. To answer the first research sub-question, we conduct semi-structured interviews with aftercare employees working at different hospitals. An aftercare employee is a nurse working at the transfer office of the hospital. The aftercare employee mediates between the patient and the different aftercare institutions to arrange the type of aftercare that best fits the patient. With the information gathered during these interviews, we construct a model of the current discharge process. This model helps us understand the current discharge process. We also identify bottlenecks in this process to answer the second sub-question.

We then answer the third and fourth sub-questions by performing a literature review and a data analysis to see which patient characteristics have the potential to predict the type of aftercare that will be needed after admission to the hospital. We create an overview of the most mentioned patient

(10)

10

characteristics found in the literature that can influence the type of aftercare. Next, we perform a data analysis on patient data from a hospital in the Netherlands. We use data from patients admitted between 01-01-2012 and 01-01-2018. We include most of the patient characteristics that were mentioned in the literature to influence the type of aftercare and supplement this with some other patient characteristics. We determine which patient characteristics in our dataset have a significant relationship with the type of aftercare. We build a multinomial logistic regression model that includes these patient characteristics and we predict the type of aftercare for a set of admissions from a test dataset. We evaluate our model by comparing the predicted outcome variables to the real outcome variables and calculating the accuracy of the predictive model. A visible overview of the approach and expected results of this research can be seen in Figure 3; in the blue blocks the

methods planned on using are visible and in the white blocks the expected results.

Figure 3: overview of the approach and expected results of this research

1.6 Outline of this thesis

The remainder of this thesis is structured as follows:

 Chapter 2: in this chapter we discuss the methods that are used to get a better

understanding of the current discharge process and the bottlenecks that are experienced in this process. We present the results in the form of a model of the current discharge process and we present quotes about the experienced bottlenecks mentioned by the aftercare employees.

 Chapter 3: this chapter describes the literature review on patient characteristics that can predict the type of aftercare. We discuss the methods used and we present the results of this literature review.

 Chapter 4: this chapter describes the performed data analysis to determine which patient characteristics influence the type of aftercare and whether it is possible to predict the type of aftercare based on these patient characteristics. We discuss the methods that we used, the results and an evaluation of the data analysis.

 Chapter 5: this chapter presents the general discussion of this thesis. In this chapter we first discuss the main findings and answer the research question. We talk about the strengths and limitations of this research, we give some implications of our work and we end this chapter by giving some recommendations for future research.

Interviews •Model of the current discharge process Literature review •Overview of patient characteristics found in the literature that can predict the type of aftercare

Data analysis •Overview of

patient characteristics based on our data analysis that influence the type of aftercare •Evaluation of to what extend it is possible to predict the type of aftercare based on these patient characteristics

(11)

11

2. Current discharge process of orthopedic patients

2.1 Introduction

In this chapter we answer the first two sub-questions:

2) Which bottlenecks are experienced in the current discharge process?

We also gather information to determine whether aftercare employees think it is possible to predict the type of aftercare based on patient characteristics. We do this to get a first impression of the answer to the fourth research sub-question. In the following chapters this sub-question will be answered in more detail:

4) Is it possible to predict the type of aftercare that is needed based on patient characteristics? In this chapter we first describe the methods used to answer the mentioned research sub-questions. We describe how we performed semi-structured interviews with aftercare employees working at transfer offices of different hospitals located in the Netherlands. The first goal of the interviews was to create a model of the current discharge process from the perspective of the aftercare employee. The second goal of the interviews was to create an overview of the bottlenecks that were

experienced in the current discharge process. The third goal was to collect ideas about whether it would be possible to predict the type of aftercare based on patient characteristics. Next, we present the results. We end this chapter with a discussion of our findings.

2.2 Methods

Participants

Six aftercare employees working at different hospitals in the Netherlands were contacted in February 2019 by e-mail. We contacted aftercare employees working at hospitals that ChipSoft or the

University of Amsterdam had good relationships with. Procedure

We developed a list of 31 questions to ask during the interviews. The total set of questions that were asked can be found in appendix A. We divided the questions into four categories:

1. Introduction: we started with some introductory questions about the aftercare employee. We asked if they could introduce themselves and about their role in the hospital. We also asked introductory questions about the transfer office and the amount of patients that they process.

2. Current discharge process: next we asked questions about the current discharge process. The questions were based on templates that we found in the literature. The Business Process Analysis Questionnaire provides questions to obtain a comprehensive description of a business process(29). Additionally, CAP TODAY provides questions to ask during process mapping sessions (30). We used and adapted the questions that we found helpful from both templates to fit our specific situation. We also asked questions about the communication with the different aftercare institutions. We asked about the waiting times for the different types of aftercare institutions and at what moments they contact them to arrange aftercare for a patient.

3. Bottlenecks: when we had obtained a better understanding of the discharge process of the hospital that we were visiting, we asked questions about the bottlenecks that they

experienced in the current process and how these could possibly be solved. We asked which type of aftercare caused the most problems and where most improvements could be made.

(12)

12

4. Predictions: we asked whether they thought it would be possible to predict the type of aftercare. We also asked whether they believed this would only be possible for certain treatments and which ones these would be.

With aftercare employees who indicated that they wanted to participate, an interview was planned at a moment of their choice. The interviews were conducted at the working location of the aftercare employees. The participants received no reimbursement for participation. At the start of each interview, permission was requested from the aftercare employee to record the interview.

Subsequently, the questions from appendix A were asked in a semi-structured way. The interviews were conducted by the author of this thesis. The mentor of the author was also present during the first interview. The time it took per interview was approximately one hour. During the interviews notes were taken.

Analysis

The audio recordings of the interviews were transcribed manually in Microsoft Word 2016. The type of transcription that was used was the edited transcription (31). In this type of transcription some parts of the audio file are left out of the transcription, as long as the meaning of what was said did not change. For example, parts of sentences where the interviewee was searching for the right words before saying them could be left out. This made the transcript more readable, while the content stayed the same. A different transcription method that we considered is the verbatim transcription. In this type of transcription every spoken word, emotion and mumble is transcribed (31). This is a very time consuming method and we concluded that this would not add more value to the

transcripts in comparison to the edited transcription method. A less time-consuming method would be to only write a summary of what said during the interviews, but we concluded that this would not be sufficient for our research. This is why we did not choose these methods, but chose to use the edited transcription method.

We constructed a model of the current discharge process based the information that we gathered during the interviews. To develop the current process model, we used the Business Process Model and Notation (BPMN). BPMN is a flow chart method that can be used to map all steps in a business process from start to finish. BPMN provides a visual overview of the business process. Creating a visual overview of the process makes it easier to understand the process than a description in plain text (32). Some other methods that can also be used to make a visual overview of a process are the Value Stream Map (VSM), Program Evaluation and Review Technique (PERT), Unified Modeling Language (UML) and Event-driven Process Chain (EPC). We considered these different techniques and came to the following conclusions about them:

1. VSM: The goal of a VSM is to identify ‘waste’ in a process and then later remove this waste (33). Waste is identified as any activity that does not add value from the perspective of the customer, who would be the patient in this case. Another important aspect of the VSM is the lead time. This is the time it takes between initiation and completion of the process. This is calculated by adding up the time spent on each activity. We expected that it would not be possible to retrieve information about the time spent on each activity in the process. We also expected that the time spent per activity differed too much per hospital.

2. PERT: A PERT chart is a tool used for project management. Since we want to model a process and not a project, we did not find this technique suiting. Projects are things that are

performed only once, while processes are performed repeatedly (34). Another disadvantage of using a PERT chart is that it uses the time spent per task and as we mentioned before, we did not expect this to be possible.

3. UML: UML is another modeling language that can be used for modelling. A possibility is to use UML to construct an activity diagram for example. In an activity diagram, the flow of different activities and actions in a process is visualized. When we compare BPMN to UML,

(13)

13

we see that BPMN has a process-oriented approach, while UML has an object-oriented approach. This is why BPMN is more suitable for business process modeling , while UML is of better use for specifying, visualizing, constructing, and documenting software systems (35, 36).

4. EPC: The EPC notation can be used to model business processes and workflows. EPC is best used to model higher-level business processes, while BPMN is better for modeling lower-level business processes (37). Where EPC is better to obtain a general impression of a

particular process, BPMN is better on an execution level to obtain all the important details of this process. Since we feel like we want to model a low-level process, we determined BPMN to be a better fit than EPC.

BPMN provides an intuitive and easy way for non-experts to understand what is modelled. It has a has a simple, but powerful semantic (38). In table 2 an overview of some of the most important elements used in BPMN is provided.

Table 2: overview of elements in BPMN (39-42)

Element Meaning

Pool/swimming lane: a pool is the organization that assigns tasks to the lanes. The swimming lanes are the components responsible for executing the tasks. This can be departments or single persons.

Start event: the event that triggers the start of the process.

End event: the event that completes the process.

Task: a unit of work.

Sequence flow: connects the tasks in the right order.

Exclusive gateway: models a decision. The sequence flow whose condition is true is chosen and followed. Only one flow can be followed. Parallel gateway: two or more tasks are performed at the same time, in parallel. Intermediate conditional event: the process waits and continues when the condition is true.

We used an academic version of the web-based program Signavio (43). With this software it is possible to create a BPMN model. Because we expected that the current process differed per hospital, we first developed one model per hospital. We then combined the multiple models that we

(14)

14

developed into one business process model of the current discharge process that takes into account the differences between the different hospitals.

2.3 Results

Current discharge process

Five of the six contacted aftercare employees responded that they would be willing to take part in an interview. This means that the response rate was 80%. An overview of the characteristics of the hospitals where the aftercare employees were working is visible in Table 3. Two of the aftercare employees were working in an academic hospital and three were working in a general hospital. The included hospitals also differed in the number of patients that they treat on a yearly basis. The interviews were conducted from 6 February 2019 until 28 February 2019.

Table 3: characteristics of the hospitals where an interview took place (44) Hospital

number

Type of hospital Number of patients treated per year (N) 1 Academic 150.000-200.000 2 General 150.000-200.000 3 Academic 100.000-150.000 4 General 100.000-150.000 5 General 100.000-150.000

Based on the information gathered in the interviews, the current discharge process was modeled. The current discharge model is made visible in Figure 4, 5 and 6. For readability the model was split into three figures. We chose to only include stakeholders in the model that were involved at multiple moments throughout the process, such as the aftercare employee and the nurse. Some of the less involved stakeholders were left out of the model, such as the specialist of geriatric medicine and the Centrum Indicatiestelling Zorg (CIZ). This decision was made to improve the readability. We identified the following steps in the current discharge process:

1. Patient admitted for treatment: the process starts when the patient gets admitted to the hospital for treatment.

2. Perform anamnesis: the nurse performs an anamnesis where (s)he asks questions about the patient. These questions are about multiple subjects, like the health and the living situation of the patient.

3. Request aftercare: when the surgery is finished, the nurse creates an order for aftercare and sends this to the transfer office. This order should contain information about the expected discharge date and the expected aftercare that will be needed.

4. Receive aftercare request: an aftercare request is received by the transfer office. This request contains information about the patient and the type of aftercare that the nurse expects to be needed after admission. This request can be received through the electronic patient record system or through another system, like Point. Point is a digital collaboration platform that can be used for care transfer (45).

5. Check aftercare request and expected discharge date: the transfer office checks whether the expected discharge date is included and whether the request made by the nurse contains a clear description of the expected type of aftercare. This can be done by the aftercare employee or by a special dedicated front office that checks all incoming requests and then assigns them to certain aftercare employees.

(15)

15

6. Read information about patient: when the request is correct and complete, the aftercare employee reads all the available information about the patient to get a good understanding of the situation and the aftercare that is needed.

7. Re-read information about patient: the aftercare employee waits until the moment that the medical discharge date is known. When this moment is known, the aftercare employee again reads all the available information about the patient to see whether there have been

changes in the situation of the patient and the aftercare that is needed.

8. Contact nurse (and other specialists) for additional information: based on the available information of the patient, the aftercare employee decides whether it is needed to speak with the nurse (or other specialists) about the patient. Some hospitals always contact the nurse and others only when they feel that this is necessary.

9. Provide information: the nurse has a conversation with the aftercare employee and provides the additional information that is needed.

10. Contact patient and/or contact persons: the aftercare employee has a conversation with the patient and/or the contact persons of the patient when (s)he feels like this is necessary. It depends on the hospital how often and in which cases this is done. Some hospitals only visit certain patients, such as patients that need long-term care with stay.

11. Decide which type of aftercare is best: based on all the information that the aftercare employee gathered, (s)he decides which type of aftercare fits the situation best. For this research we defined three types of aftercare, namely: (geriatric) rehabilitation, long-term care with stay (in a nursing home) and homecare.

12. Contact specialist geriatric medicine: for geriatric rehabilitation a specialist of geriatric medicine gets involved to perform a geriatric assessment. It is defined by law that this is necessary (46).

13. Request indication at the CIZ: for long-term care with stay, an indication of the CIZ is needed. This indication can be requested through their website our through Portero. Portero is a system developed by the CIZ that can be used to submit indications for long-term care with stay (47). The CIZ has to approve all requests for long-term care with stay in the Netherlands (48). In some cases they can approve the request with the information provided and in other cases they need to visit the patient to obtain additional information. The agreement is that the CIZ should give a reaction within 48 hours whether the request can be approved or not (49).

14. Fill in needed forms: all the needed forms are filled in. It depends per hospital and per type of aftercare which forms these are.

15. Find fitting aftercare institution: a fitting aftercare institution is being sought. Some hospitals have an overview of aftercare institutions where there should be a place available. This can be in an application or something they created manually. Finding an aftercare institution for the patient can be done by contacting the aftercare institutions by phone or digitally. 16. Register patient to aftercare institution(s): send all the requested forms and register the

patient to the chosen aftercare institution. The forms that are needed differ per type of aftercare. For long-term care with stay, the indication of the CIZ needs to be provided to the aftercare institution for example.

17. Contact patient and/or contact persons: when the aftercare employee finds a fitting place for the patient, (s)he has a conversation with the patient and/or the contact persons of the patient to inform them about the type of aftercare that the patient will receive and which aftercare institution will provide this.

18. Contact nurse: the aftercare employee contacts the nurse of the orthopedics department by phone or digitally to inform them about the type of aftercare that is arranged and about the date that the patient is expected to leave the hospital.

(16)

16

19. Receive information: the nurse gets informed by the aftercare employee about the type of aftercare that is arranged and about the date that the patient is expected to leave the hospital.

20. Arrange medication and medical devices: if it is necessary, the aftercare employee arranges medication and other devices. A lot of times they are responsible for arranging intravenous medication for example.

21. Inform patient about discharge: the nurse informs the patient about the moment of discharge and about what the patient can expect after discharge from the hospital. 22. Patient discharged: the patient is discharged by the orthopedics department.

Every step of the process that the aftercare employee or nurse performs gets documented in the electronic patient record system. At the end of the process the aftercare employee also creates a report of everything that was done. In Appendix B the total process map can be found.

(17)

17

Figure 4: The current discharge process part 1

Figure 5: The current discharge process part 2

(18)

18 Bottlenecks

During the interviews we asked the aftercare employees about the bottlenecks that they

experienced in the current discharge process. We translated the quotes from Dutch to English. The original quotes in Dutch can be found in Appendix C.

The first important bottleneck that was mentioned, were the capacity problems at aftercare institutions that cause waiting lists. One aftercare employee mentioned that “people are waiting for months, because they are waiting for a permanent place in a nursing home”. Another aftercare employee mentioned that “the aftercare institutions do not have free space” and that “regularly there are patients for whom we have to call 14 aftercare institutions and still cannot find a place”. Another aftercare employee mentioned that “the core of the problem is the shortage in homecare capacity. The transfer from short-term care with stay and geriatric revalidation is now also clogged, because people cannot go home with homecare”.

The second bottleneck that we identified was “the communication” between the health care professionals involved in the process. In one hospital it was mentioned that the aftercare employees “are dependent on other people. Sometimes there can be a delay in receiving a letter from geriatrics for example” and that “it would be nice to get feedback quick. Sometimes it can take a long time”. A different aftercare employee also mentioned that they “have a lot of discussions with doctors because they made promises to the patient that we cannot make true” and that “there are general practitioners that refuse to let a patient return home” and that this causes them problems. Besides, there can be a problem with the family members of the patient. One aftercare employee mentioned that they “have a lot of problems with family members that do not agree with our decision. They say that the situation at home is no longer sustainable and that their family member needs a place in a long-term care facility, but there are waiting lists of months long and we cannot let a patient wait here that long”.

Another bottleneck that was mentioned was a problem with incorrect or incomplete aftercare requests. One aftercare employee mentioned that they “would like to see that the requests would be way more correct. You just want to have a first-time-right request.” She also mentioned that they “would like to see that the doctors better fill in the Provisional Dismissal Date, because that is what we work with”.

Something that they also mentioned what they would like very much is “preferred supply, so that you can ask aftercare institutions to reserve a number of beds, so that we can easily discharge our patients, regardless of their illness". Another aftercare employee confirmed this by saying that “if the aftercare institutions reserve beds for us that would of course also improve the process. The moment that we have that, that would be a luxury”.

Predictions

We asked whether the aftercare employees believed that it would be possible to predict the type of aftercare based on patient characteristics, and we received mixed reactions. We translated the quotes from Dutch to English again and the original quotes in Dutch can be found in Appendix C. Some aftercare employees reacted that they did not believe that it would be possible to predict the type of aftercare based on patient characteristics. One aftercare employee said that “it would be possible, but unfortunately not with orthopaedic patients”. A different aftercare employee said that they “do nothing until the care request is clear and the patient is ready for discharge. The majority of patients that we see are very vulnerable with multi-morbidity and we don't see the patients who just go home”. However, another aftercare employee thought that it would be possible. She said that “you know that someone who is over 80 years old, lives alone, has fallen and broken a hip, that he or

(19)

19

she must go to rehabilitation. Someone who is 50 years old who gets a planned hip surgery will not need any aftercare”. The factors that she mentioned to influence the type of aftercare are:

- Age

- Living situation

- Emergency/planned admission

Another aftercare employee reacted more nuanced and mentioned that “you can make a general estimation about the type of aftercare, but you do not know in what state the patient comes out of the surgery.” This was supported by another aftercare employee that said that “you can never predict the type of aftercare entirely, because it is always dependent on a number of factors ... a date of admission can change and a surgery date can also still change, so it depends on that.” However she added that “orthopaedics is the most structured and predictable department”.

2.4 Discussion

Main findings

In this chapter we answered the following sub-questions:

2) Which bottlenecks are experienced in the current discharge process?

In the results section we described what the current discharge process for patients in the orthopedics department looks like. We made one model which took in consideration the differences between the hospitals that were included. Some important differences that we found between the hospitals that were included are the following:

1. Some hospitals made a distinction between a front and a back office. The front office first screened all the aftercare requests on completeness and then assigned them to the different aftercare employees. In other hospitals they just had one transfer office where the aftercare requests were checked and handled by the same aftercare employees.

2. It differed per hospital which types of patients the aftercare employees visited. In some hospital they only visited patients that needed terminal care, while other hospital did not visited this type of patient and instead only visited patients that needed other types of aftercare. It also differed at what moment(s) the aftercare employee visited the patient. At some hospitals the aftercare employee visited the patient before arranging aftercare, so that questions could be asked to help in the decision of which type of aftercare fitted best. At other hospitals, the aftercare employee only visited the patient after the aftercare was arranged to inform the patient of what would happen next. It was also possible that the patient was visited at both of these moments.

3. A big difference was in the way that the aftercare employees looked for available places. In some hospitals, they called the aftercare institutions one by one until they found one that had a place available. Other hospitals worked with Point or Verwijshulp (45, 50). In Point, the aftercare employee can administer the patient to multiple homecare providers for example. When the first provider declines the patient, the request gets automatically sent to the next homecare provider. At Verwijshulp healthcare providers can get an overview of the available places in nursing homes and rehabilitation centres. In our opinion working with a system like Point or Verwijshulp saves a lot of time in searching for an available place. These systems are not available throughout the whole of the Netherlands however, so this is not always

(20)

20

In this chapter we also gathered information to determine whether aftercare employees think it is possible to predict the type of aftercare based on patient characteristics. We did this to get a first impression of the answer to the fourth research sub-question:

4) Is it possible to predict the type of aftercare that is needed based on patient characteristics? We described the bottlenecks that were mentioned by the aftercare employees. The biggest bottleneck that we identified, were the waiting lists at aftercare institutions. This is however not something that can be influenced by ChipSoft. Other bottlenecks that were mentioned were the communication with other people involved in the discharge process and the incorrectness or incompleteness of aftercare requests. We conclude that there is still room for improvement in the discharge process. Next, we mentioned the reactions of aftercare employees to whether they believe it would be possible to predict the type of aftercare based on patient characteristics. Some said that this would not be possible. However, other aftercare employees thought that this would be possible and mentioned some patient characteristics that could predict this. The patient characteristics that were mentioned were: age, the living situation of the patient and whether it was an emergency or elective procedure. The fact that this was mentioned strengthens our research goal to further examine whether it is possible to predict the type of aftercare based on patient characteristics. Relation to other work

In 2012 research was performed by Meijer about the discharge process at the VUmc in the departments of neurosurgery and orthopedics (51). A model of the discharge process of this department was developed. In this model the discharge process from the perspective of the nurses was presented. They also included a part of the discharge process from the perspective of the aftercare employee. We concluded that the general structure of the discharge process corresponds with the model that we developed in this thesis. In short, they described the discharge process as follows:

1. Request aftercare: when the surgery is finished, the nurse creates an order for aftercare and sends this to the transfer office.

2. Process request and plan a visit to the patient: the aftercare employee processes the request and schedules a consult with the patient.

3. Check necessary forms and visit department: the aftercare employee checks whether all the necessary forms are filled in completely and (s)he visits the department.

4. Consult with patient and decide which type of aftercare is best: when all the necessary forms are complete, the aftercare employee has a consult with the patient. It is decided which type of aftercare fits the situation best.

5. Register patient to aftercare institution(s): send all the requested forms and register the patient to the chosen aftercare institution.

6. Contact nurse: when a place is available at an aftercare institution, the aftercare employee contacts the nurse of the department to inform them about the type of aftercare that is arranged and about the date that the patient is expected to leave the hospital.

In 2011 research was also performed by van der Aa about the discharge process at the VUmc in the department of surgery (52). The discharge process of the surgery department and bottlenecks in this process were described. In Table 4 an overview is presented of the bottlenecks that were identified at the research of van der Aa and Meijer compared to the bottlenecks identified in this research. Van der Aa identified four bottlenecks in the discharge process from the perspective of the aftercare employee. Firstly, van der Aa mentioned that the nurse sometimes forgot to put in a request for aftercare and this caused a delay in the process. Secondly, the forms were not collected daily, which caused a delay in the discharge process. Thirdly, she mentioned that because there was not a

dedicated aftercare employee for every department, the nurses did not know who to contact when a problem occurred. Finally, she mentioned that a request for homecare should be put in before 11:30

(21)

21

and this was not always done on time. The bottlenecks that van der Aa identified were not found in this research.

Meijer identified two bottlenecks that caused a delay in the discharge process. The first bottleneck that she mentioned, were the capacity problems at aftercare institutions that cause waiting lists. This is a bottleneck that we also identified. The second bottleneck that Meijer mentioned is that the doctor or nurses did not complete the necessary forms in time. A bottleneck that we identified in this research was that the requests for aftercare received by the aftercare employee are incomplete or incorrect. We conclude that these bottlenecks have some overlapping.

Table 4: bottlenecks in the discharge process mentioned by different articles

van der Aa (52) Meijer (51) This research The nurse forgets to put in a request for aftercare x

The forms with requests for aftercare are not collected daily

x There is not a designated aftercare employee for every department

x Requests for homecare are not put out in time x

Incomplete or incorrect forms x x

Capacity problems at aftercare institutions x x

Communication problems x

Strengths and limitations

A strength of this research is that we validated the model of the current discharge process that we developed by other research. We concluded that the general structure of the discharge process in the VUmc as described by Meijer corresponds with the model that we developed (51). This indicates that the model that we developed of the current discharge process is generally the same for different hospitals in the Netherlands and strengthens our results. Some of the bottlenecks that we identified were also identified by Meijer, like the incomplete forms and the capacity problems at aftercare institutions (51). This also strengthens the credibility of our results.

The first limitation is that we only interviewed aftercare employees and no other healthcare professionals or patients involved in the discharge process. This means that for the model we only look from the perspective of the aftercare employee. To get a better understanding of the total current discharge process and all the different actors involved, it would be better to also interview other involved healthcare professionals and the patient itself. This would create a better overview of all the experienced bottlenecks from different points of view and it would give a broader

understanding of whether different healthcare professionals believe if it is possible to predict the type of aftercare based on patient characteristics.

A second limitation of this research is that we were not able to perform enough interviews to be able to draw objective conclusions about what the biggest bottlenecks were or whether it is possible to predict the type of aftercare or not. To be able to do this, we would need to perform more

interviews. However, a strength is that we the hospitals that we included are different types of hospitals. We included academic and general hospitals throughout different provinces in the Netherlands. They also differed in the amount of patients that they treat yearly and they used different electronic patient record systems. This increases the generalizability of the model that we created of the current discharge process. However, this is still not a complete representation of all the hospitals in the Netherlands.

(22)

22

Another limitation is that by performing interviews, there is the possibility of introducing interviewer bias (53). It is possible that the interviewee did not answer fully truthful, because (s)he might be ashamed of the actual answer and rather give an answer that sounds better. We tried our best to remain open and objective, but it is also possible that the interviewer steered the interviewee in a certain direction and influenced their answers in this way. This could have introduced bias in the results.

Recommendations for future research

We think that it would be interesting to combine the different views of all the healthcare

professionals involved in the discharge process (e.g., aftercare employees, nurses and doctors) into one model. We think it would also be very interesting to perform interviews with employees working at aftercare institutions, such as homecare providers, rehabilitation centres and nursing homes, to determine how their current admission process is. When the discharge process of the hospital fits well with the admission process of the aftercare institutions, we believe that this can improve the flow of patients from the hospital to the aftercare institutions and we believe that this will have a positive effect on the number of bed-blockers in the hospital.

We also recommend performing further research into the bottlenecks in the current discharge process. In related literature they sometimes identified other bottlenecks than we did. It is possible that we missed some of the experienced bottlenecks, due to the fact that we performed a limited amount of interviews. If it is possible to identify all the bottlenecks in this process, we believe that it is possible to eliminate most of them and minimize the length of stay in the hospital.

Besides, we recommend ChipSoft to explore the possibilities in developing a tool that provides insight in the available places at different aftercare institutions throughout the whole of the

Netherlands, and where the aftercare employees can immediately register a patient to an aftercare institution where a place is available. This tool could be similar to Point or Verwijshulp mentioned in this chapter before. Lastly, we think that it would be interesting to explore the possibility of reserving beds at aftercare institutions. If this is possible, it would save the aftercare employees a lot of time in searching for available places.

(23)

23

3. Literature review on patient characteristics that influence the

type of aftercare

3.1 Introduction

In this chapter we answer the following research sub-question:

3. Which patient characteristics influence the type of aftercare that is needed?

To answer this question we perform a literature review. The first part of this chapter describes the methods that were used to identify the patient characteristics that have a significant relationship with the type of aftercare. Next, we present the patient characteristics that were found in the literature in the results section. Finally, we end this chapter with a discussion of our findings.

3.2 Methods

The PubMed database was searched for articles describing the relationship between patient characteristics and the type of aftercare received. We chose to use the PubMed database, because PubMed is an archive containing biomedical and life sciences literature (54). A type of aftercare can be a place in a nursing home or rehabilitation centre, or it can be that the patient returned home with homecare for example. This is why we also refer to the type of aftercare as the destination of the admission, or the discharge destination. We searched the PubMed database for articles

containing terms like “discharge”, “destination”, “orthopedics” and “predict”. The search query that was developed and used is available in Appendix D.

Articles were assessed for inclusion in this literature review when there was a full text available and the language of the article was Dutch or English. In Figure 7 a flow-chart of the methods used to decide whether an article got included in the literature review is visible. First, the author of this thesis screened all the articles that were found using the search query described in Appendix D on the title and abstract. If the title and abstract suggested that the article described the relationship between patient characteristics and the type of aftercare, the full-text article was read. If this was not the case, the full-text article was not read. If the full-text article in fact described the relationship between patient characteristics and the type of aftercare, the article was included in the literature review. Otherwise, the article was excluded from the literature review.

Of all the articles that were finally identified for inclusion in this research the different study characteristics were determined. The study characteristics that were determined were: the year of publication, the country of publication, the study population, the sample size, the period of data collection, the study design, types of discharge locations that were predicted and the method used for data analysis. Then for all the included articles the patient characteristics that were mentioned to influence the type of aftercare were determined. Some of the patient characteristics were merged into groups. The patient characteristics that were merged together are described in the next section of this chapter. Next it was counted how many times each patient characteristic was mentioned. This is also described in the next section of this chapter.

(24)

24

Figure 7: flow-chart of the methods used for including articles in the literature review

3.3 Results

Number of articles found

For the relationship between different patient characteristics and the type of aftercare, 103 articles were found using the search query provided in Appendix D. In Figure 8 a flow-chart of the selection process of the articles for this literature review is visible. Based on the title and abstract 32 articles were selected for further review and the full-text articles were read. After reading the full-text articles, 16 articles were included in the literature review..

(25)

25

Figure 8: flowchart for the articles included in the literature review Study characteristics

The characteristics of all the studies included in this literature review can be found in Tables 5 and 6. All of the studies were published between 2004 and 2018. Most of them were performed in the United States of America, two in Japan and one in the United Kingdom. Almost all of the articles exclusively studied patients of the orthopaedics department, with only one study also including other departments. The study population of eight of the included articles were patients undergoing hip or knee surgery. The study population of seven of the included articles were patients exclusively

undergoing hip surgery. The sample size differed greatly between 54 and 107.300. The period of data collection differed between 1986 and 2016.

We classified all the studies that we found as cohort studies. In 15 of these studies, the data was collected retrospectively. This implicates that the level of evidence of these studies is at level 3. This is based on the levels of evidence as defined by Herrell (55). One study collected their data

prospectively. This implicates that the level of evidence of this paper is one lever higher at level 2. For the types of discharge locations that were predicted, 14 out of 16 studies compared two possible discharge locations to each other, meaning that they used a binomial outcome variable. This means that two possible outcomes were compared to each other. In these studies, home discharge was compared to discharge to an alternative location (56-69). This alternative location was formulated in different ways, such as “facility discharge”, “other discharge”, “elsewhere”, “non-home”,

“institution”, “extended care facility”, “post-acute care facility” and “alternative location”. Two articles predicted more than two types of discharge locations, meaning that they used an outcome variable that had more than two possible categories. Keswani et al. compared multiple outcome categories to each other (70). They compared: 1) home discharge to discharge to a skilled nursing facility, 2) home discharge to discharge to an inpatient rehabilitation centre, 3) discharge to a skilled nursing facility to discharge to an inpatient rehabilitation centre and 4) home discharge to

(26)

26

non-home discharge, where discharge to a skilled nursing facility and discharge to an inpatient rehabilitation centre are grouped together into non-home. Schwarzkopf et al. compared home discharge to discharge to a skilled nursing facility, and home discharge to home discharge with homecare (71).

When we look at the methods used in the articles found, we see that all of the articles use some form of logistic regression. In 11 articles they mention that they used multivariate logistic regression (56, 58, 60-63, 65-70). London et al. used Bayesian hierarchical logistic regression (59). Schwarzkopf et al. used multinomial logistic regression (71).

Table 5: study characteristics part 1 First author of

the paper Year Country Study population

Sample size Aharonoff (67) 2004 United States of

America

patients with hip fracture who were living at home before the injury

89.723 Barsoum(64) 2010 United States of

America

patients who underwent elective primary or revision total hip or knee arthroplasty

517 Bozic(68) 2006 United States of

America

patients who underwent primary or revision total hip or knee arthroplasty

7.818 de Pablo(58) 2004 United States of

America

patients who underwent elective primary or revision total hip arthroplasy

1.276 Gholson(56) 2016 United States of

America

patients who underwent primary total hip or total knee arthroplasty

107.300 Hagino(69) 2011 Japan patients with hip fracture who were

living at home before the injury

345 Halawi(63) 2015 United States of

America

patients who underwent primary total hip or total knee arthroplasty

372 Hayashi(62) 2016 Japan patients with hip fracture who were

living at home before the injury

54 Hyder(57) 2014 United States of

America

patients who underwent inpatient surgery

88.068 Inneh(60) 2016 United States of

America

patients who underwent primary or revision total hip or knee arthroplasty

7.924 Keswani(70) 2016 United States of

America

patients who underwent elective primary total hip or knee arthroplasty

106.360 London(59) 2016 United States of

America

patients who underwent total hip or total knee arthroplasty

14.315 Rudasill(61) 2018 United States of

America

patients who underwent elective primary or revision total hip or knee arthroplasty

392 Salar(66) 2017 United Kingdom patients with a fractured neck of femur

who were living at home before the injury

10.044

Schwarzkopf(71) 2015 United States of America

patients who underwent total hip arthroplasty

14.326 Sharma(65) 2018 United States of

America

patients who underwent total hip arthroplasty

(27)

27

Table 6: study characteristics part 2

First author

Period of data

collection Study design

Types of discharge locations that were predicted

Methods used for data analysis Aharonoff (67) 1986-1996 Retrospective

cohort

Home discharge vs other discharge destination

Multivariate logistic regression Barsoum (64) 2005-2007 Retrospective

cohort

Logistic regression Bozic (68) 2000-2002 Retrospective

cohort

Multivariate logistic regression de Pablo (58) 1995 Retrospective

cohort

Multivariate logistic regression Gholson (56) 2011-2013 Retrospective

cohort

Multivariate logistic regression Hagino (69) 1997-2008 Prospective

cohort

Multivariate logistic regression Halawi (63) 2012 Retrospective

cohort

Multivariate logistic regression Hayashi (62) 2009-2014 Retrospective

cohort

Multivariate logistic regression Hyder (57) 2011 Retrospective

cohort

Serial multiple logistic regression Inneh (60) 2011-2014 Retrospective

cohort

Home discharge vs other discharge destination Multivariate logistic regression Keswani (70) 2011-2013 Retrospective cohort Home discharge vs nursing home vs rehabilitation centre Multivariate logistic regression London (59) 2011-2013 Retrospective cohort

Home discharge vs other discharge destination Bayesian hierarchical logistic regression Rudasill (61) 2011-2012 Retrospective cohort

Multivariate logistic regression Salar (66) 2000-2012 Retrospective

cohort

Home discharge vs other discharge destination Multivariate logistic regression Schwarzkopf (71) 2010 Retrospective cohort Home discharge vs nursing home vs home with homecare

Multinomial logistic regression Sharma (65) 2014-2016 Retrospective

cohort

Multivariate logistic regression

Towards an efficient discharge process

Towards an efficient discharge process

Towards an efficient discharge process

Preface

Contents

Summary

Samenvatting

Keywords

1. Introduction

1.1 Problem of bed-blockers

1.2 Background, context and environment

1.3 Environment of the research

1.4 Goal and research questions

1.5 Approach and expected results

1.6 Outline of this thesis

2. Current discharge process of orthopedic patients

2.1 Introduction

2.2 Methods

2.3 Results

2.4 Discussion

3. Literature review on patient characteristics that influence the

type of aftercare

3.1 Introduction

3.2 Methods

3.3 Results