Factors that might influence non-adoption of a regional personal health environment: a qualitative study

(1)

Factors that

might

influence

non-adoption of

a regional

personal

health

environment

April 27th

2017

Stef Mooij – Scientific Research Project executed at Bernhoven Hospital

in the Netherlands to graduate from the Master Medical Informatics, at the University of Amsterdam.

A qualitative

study

(2)

(3)

Factors that might influence

non-adoption of a regional personal health

environment: a qualitative study

Student S. Mooij

Student number: 11162945 Email: s.mooij@amc.uva.nl

Mentors

J.E. Slot – Chief Information Officer Bernhoven hospital

M.G. de Waal – Team leader Innovation / Functional management Bernhoven hospital VE MICT Program ‘Droom’ Department of Innovation Tutor Drs. W.J.P.P. ter Burg Faculty of Medicine

Department of Medical Informatics, AMC-UvA Location of Scientific Research Project Bernhoven hospital

Department of Innovation Nistelrodeseweg 10 5406 PT Uden The Netherlands

Practice teaching period November 2016 – July 2017

(4)

(5)

Abstract

English

Background: Previous Personal Health Environments (PHEs) abroad have failed (Greenhalgh et al., 2010; Lau et al., 2015; Spil & Klein, 2014) because they were not being used (i.e. non-adoption). The PHE from Bernhoven Hospital and Synchroon foundation, collaboration with regional primary care physicians (PCPs), will be the first one of its kind in the Netherlands. The PHE will firstly be implemented for patients with chronic obstructive pulmonary disease (COPD) and rheumatic disease (RD). However, the factors that could influence non-adoption among these end-users are unknown. If the PHE of Bernhoven Hospital and Synchroon foundation is not adopted, it could lead to lead to losing self invested funds, governmental funds, funds from health insurance companies, reputational damage, negatively influence the intention to adopt future health information technologies among patients and healthcare staff, and patients and healthcare providers will not experience the benefits of a successful PHE.

Objective: The objective of this study is to examine the intended reasons for non-adoption of a regional PHE platform for hospital physicians, PCPs and for patients with COPD and rheumatic disease. Therefore the research question is as follows: What factors might influence non-adoption of a PHE platform among physicians, COPD and rheumatic disease patients?

Methods: An online desk research, followed by a Weakness-Based Language Processing (WBLP) method is used to find and organize factors for non-adoption among purposively sampled physicians, primary care physicians, COPD, and rheumatic disease patients. The WBLP method entails a focus group-like method with a main theme question similar to the research question. The WBLP meeting followed guidelines to organize, analyse, and prioritize the results from the Center for Quality Management (CQM, 1997).

Results: The WBLP meeting resulted in 23 themes that answered the main theme question. Priority points were given by the participants to 16 of 23 themes. A list comprising 16 potential factors that could influence non-adoption in the setting of Bernhoven Hospital and Synchroon foundation was generated with high and low prioritised potential factors for non-adoption.

Conclusion: Factors potentially affecting non-adoption were related to ‘not adding any or insufficient value for patients or physicians’, ‘not involving patients in the development process’, ‘physician concerns about more self management from patients’, ‘integration of the PHE’, ‘more work’, ‘financial barriers’, ‘contradictory or incomplete information in the PHE’, and ‘privacy and insight rights of the PHE’.

Discussion: The ability to generalise these results to other settings is low. However, if Bernhoven Hospital and Synchroon foundation implement mitigating strategies for non-adoption, they might have a better chance of preventing non-adoption.

Keywords:

Personal Health Record; Personal Health Environment; Factors for Non-Adoption; Health Information Technology

(8)

2

Nederlands

Achtergrond: Eerdere Persoonlijke Gezondheid Omgevingen (PGO’s) in het buitenland zijn al gefaald (Greenhalgh et al., 2010; Lau et al., 2015; Spil & Klein, 2014), omdat zij uiteindelijk niet of te weinig werden gebruikt (of non-adoptie). Het PGO van Bernhoven Ziekenhuis is een samenwerking tussen het ziekenhuis en de stichting Synchroon, een regionale samenwerking tussen huisartsen in de regio, wordt het eerste PGO van zijn soort in Nederland. De PGO zal als eerst worden geïmplementeerd voor een Chronic Obstructive Pulmonary Disease1 (COPD) en voor Reumatoïde Artritis (RA). Echter zijn de factoren die non-adoptie mogelijk kunnen beïnvloeden onbekend, ook voor deze gebruikersgroepen. Als de PGO van Bernhoven Ziekenhuis en stichting Synchroon niet wordt gebruikt, dan kan dit tot gevolg hebben dat zij zelf geïnvesteerde gelden, overheid gelden en geïnvesteerde gelden van de zorgverzekeraars verliezen of moeten terug betalen, reputatieschade kunnen oplopen, de toekomst van gezondheid informatie technologie negatief beïnvloeden en dat patiënten en zorgverleners nooit de voordelen van een PGO kunnen ervaren.

Doel: Het doel van deze studie is de voorgenomen redenen te onderzoeken voor non-adoptie van een regionaal PGO platform van specialisten, huisartsen en patiënten met COPD en RA. Daarom is de onderzoeksvraag als volgt: Welke factoren zouden non-adoptie van een regionaal PGO platform kunnen beïnvloeden onder specialisten, huisartsen en patiënten met COPD en RA?

Methode: Een online deskresearch, gevolgd door een veldonderzoek met de Weakness-Based Language Processing (WBLP) methode is gebruikt om factoren voor non-adoptie te vinden. De WBLP methode is een focus groepachtige methode waar de doelbewust gekozen deelnemers een hoofd themavraag beantwoorden die lijkt op de onderzoeksvraag. De WBLP bijeenkomst werd gedicteerd door richtlijnen van het Centrum voor Kwaliteit Management (CQM, 1997) om de resultaten te organiseren, analyseren en een prioriteit te geven.

Resultaten: De WBLP bijeenkomst resulteerde in 23 thema’s die de hoofd themavraag beantwoorden. Prioriteitspunten zijn gegeven door de deelnemers aan 16 van de 23 geïdentificeerde thema’s die mogelijk non-adoptie kunnen beïnvloeden in de setting van Bernhoven Ziekenhuis en stichting Synchroon. De lijst met mogelijke factoren voor non-adoptie is geordend van hoge naar lage prioriteit.

Conclusie: Factoren die mogelijk non-adoptie beïnvloeden zijn gerelateerd aan ‘het niet of amper toevoegen van waarde voor patiënt en professional, ‘niet betrekken van je patiënten bij de ontwikkeling van het PGO’, ‘zorgen van professionals over meer zelf regie van de patiënt’, ‘integratie van de PGO’, ‘meer werk’, ‘financiële barrières’, ‘tegenstrijdige of incomplete informatie in de PGO’ en ‘privacy en inzagerechten van de PGO’.

Discussie: De mogelijkheid om deze resultaten te generaliseren naar andere instellingen is beperkt. Echter, de PGO maakt een betere kans om non-adoptie te voorkomen als het Bernhoven Ziekenhuis en stichting Synchroon de aanbevelingen opvolgen om non-adoptie factoren te verlichten.

1

Er is geen Nederlandse vertaling voor deze chronische longziekte. In volksmond wordt de longziekte vaak genoemd naar zijn afkorting: COPD. COPD is een verzamelnaam voor chronische bronchitis en longemfyseem. Van: Longfonds.nl, geraadpleegd op 20-06-2017. URL: https://www.longfonds.nl/copd/alles-over-copd/wat-is-copd

(9)

3

Introduction

This chapter will give an introduction about the study at Bernhoven Hospital by describing the problem indication, scope, relevance, current state of the literature, objective, problem statement, research questions, approach and study outline.

Problem indication

In 2014 the Dutch minister of Public Health, Wellbeing and Sports wrote a letter to the Dutch Parliament, in which she envisions the future for Dutch healthcare. She wants to accelerate the motion of patients towards more self-reliance, self-management and self-care (Minister en Staatssecretaris van Volksgezondheid, 2014). She has formulated several objectives2 to be realised before 2019 to reduce the burden of chronic diseases. This has given the development of healthcare information technology in the Netherlands a boost, since smart Information Computer Technology (ICT) solutions can support patients to take control over their own health, in their own environment (Krijgsman et al., 2016a). A program called the ‘VIPP-program’ (Appendix A) has been brought to life as a consequence of the letter from the Dutch minister. The VIPP-program can be translated as ‘accelerating information exchange between patients and professionals’ (In Dutch: ‘Versnelling Informatie-uitwisseling Patiënt en Professional’). The program aims to help care institutions through information provision and funding to have a secure patient portal and/or a link to a personal health environment (PHE) where the care institution can exchange health data (NVZ, 2017).

Scope

Bernhoven Hospital

One of the hospitals that make use of VIPP-program funding is Bernhoven Hospital in the province of North-Brabant in the Netherlands. Bernhoven Hospital is a mid-size, regional hospital (Bennemeer & Peeters, 2015) with roughly 400.000 patients in their catchment area. The hospital focuses on general hospital care in the region, close to the patient. The hospital aims to be the most people oriented hospital in the Netherlands, and has been awarded the hospitality award in 2014 and 2016. The region - in which Bernhoven Hospital operates - is a rural area with a diverse population, and a relatively large elderly community of twenty-six percent being older than 60 years of age (CBS, 2016).

2

The first objective that is relevant for this study states: Before 2019, 80% of the chronically ill in the

Netherlands should have direct access to certain medical information, vital functions, test results, and can use these accordingly with mobile applications or internet applications.

The second relevant objective states 75% of the chronically ill and vulnerable elderly can (if he or she wishes) perform independent measurements.

(10)

4

In most Dutch hospitals, physicians work in a partnership with the hospital and are not on the payroll. The partnering physicians get a ‘fee for service’, meaning the physicians get paid when they do surgery for example. However, Bernhoven Hospital is unique in several aspects. For instance, most of the physicians active chose to be on the payroll and two of the members of the board of directors are physicians themselves. Bernhoven Hospital holds the dogma ‘Doctor in the lead’ (Witman, Smid, Meurs, & Willems, 2010) in high regard. Additionally, Bernhoven Hospital collaborates with the situated primary care physicians who are working together under the care foundation ‘Synchroon’ (meaning synchronous in Dutch).

This rather unique organisational structure in the Netherlands makes Bernhoven Hospital the perfect candidate for an ambitious program called ‘Droom’ (Dutch for ‘dream’). For this program the hospital has formed a partnership with the two biggest health Insurance companies in its region. Bernhoven Hospital strives for value-based healthcare. The goal of value-based healthcare is to transform the financial model of healthcare from a production based system to a value based one, where quality of healthcare is the drive for financing, not the quantity of it (Porter & Teisberg, 2006). To successfully establish a value-based healthcare system, both the organisation and regional approach need undergo profound changes. Key element in value-based healthcare systems is an integral approach in the entire region (Porter & Lee, 2013; Suter, Oelke, Adair, & Armitage, 2009). This will result in an alignment of the information and healthcare needs of the patient, with the combined efforts of regional primary care physicians and the Bernhoven Hospital (Porter & Teisberg, 2006).

The PHE platform is a project that will support alignment of information and healthcare needs of the patient. This platform will be implemented in multiple phases divided in time boxes. The first two time boxes will deploy the PHE focussing on rheumatic disease and chronic obstructive pulmonary disease (COPD). Eventually, the end result should be a lifelong, integral health platform for all inhabitants of the catchment area of Bernhoven Hospital. The goal of the PHE is to enable patients towards more self-reliance, self-management and self-care to take control over their own health, in their own environment. This way the PHE should support and enhance self-management of patients, and redirect care from hospital care to primary care, to decrease unnecessary care. The PHE of Bernhoven Hospital will be an online platform for patients, where care providers can communicate with the patient and other providers (after consent), and where patients can manage, share or add health information. Therefore, functionalities will be added to give online access to test results, appointments, lifestyle programs, but can also be used for self management by filling in questionnaires or other health monitoring tools integrated in the platform. This method offers a more holistic view of the patients’ health data (HIMMS, 2007; Kim et al., 2009). The PHE will primarily contain information from the Health Information System of Bernhoven Hospital and primary care physicians (PCP) affiliated with Synchroon foundation. However, the integration of other care providers in the PHE will be realised in a later phase of the implementation. Ultimately, the PHE can be joined by other existing information systems (e.g. from the physiotherapist) to form an integrated approach with one user interface for patients.

(11)

5 The end-users of the PHE will be all inhabitants of Bernhoven Hospitals’ catchment area and their care providers. However, the end-users for the first two time boxes will be patients with COPD and rheumatic disease, Bernhoven Hospital physicians, and primary care physicians affiliated with Synchroon foundation.

The expected benefits of a PHE for the above mentioned end-users are numerous. However, previous PHEs abroad have failed (Greenhalgh et al., 2010; Lau et al., 2015; Spil & Klein, 2014) in the past because they were not being used (i.e. non-adoption). Furthermore, Andrews et al. (2014) argues that “understanding perceptions of non-users regarding having a personally controlled electronic health record, particularly in the early stages of its implementation, can provide important insights for those involved with the continued roll-out of this initiative” (p.898).

Therefore the scope of this study was non-adoption among rheumatic disease and COPD patients, their primary care physician and hospital physician for a regional personal health environment platform.

Relevance

The PHE from Bernhoven Hospital will be the first one of its kind in the Netherlands. Therefore the risk factors for non-adoption of a regional project Bernhoven Hospital is working on are unknown. Investigating, mapping and controlling the risk factors are important when implementing a large scale PHE, since non-adoption could have large consequences. Therefore, this study is of practical relevance for Bernhoven Hospital. As well as for other healthcare organisations that wish to implement a PHE platform in the future, especially if that healthcare organisation wishes to collaborate with the PCPs in their region. With this study future implementers have a preliminary overview of reasons for adoption among end-users and recommendations how to mitigate non-adoption.

Current state of the literature

There is a scarcity of literature on factors affecting non-adoption of a PHE. One reason for scarcity is the fact that PHE as a health information technology is a new phenomenon. Studies that have been conducted are either about adoption a different health information technologies like portals (Gagnon et al., 2014; Kruse et al., 2015; Tulu et al., 2016; Van der Vaart et al., 2014), or are of technical nature, neglecting the socio-technical aspects of health information technologies (Urbauer et al., 2015; Wiesner et al., 2014). The second reason is the pro-innovation bias in the literature. Greenhalgh and colleagues (2010) mention the importance to study non-adoption to overcome the pro-innovation bias in the current literature. In a later study, Andrews et al. (2014) also underlines the importance of determining factors that inhibiting non-adopters from using a technology such as the PHE. “There is definitely a gap in our understanding as to why people are choosing not to use PHEs” (Nguyen et al., 2016, p.467). To my knowledge, only three studies have been conducted to investigate non-adoption among PHEs (Greenhalgh et al., 2010; Lau et al., 2015; Spil & Klein, 2014). However, all studies have been done after the PHE failed in stead of investigating non-adoption before implementation as a strategy to decrease the risk for non-adoption. Other studies that investigated negative factors influencing intended PHE use only sought negative correlations for one or two factors among a wide range of factors used for adoption (Andrews et al., 2013; Cocosila et al., 2014; Gagnon et al., 2016).

(12)

6

Problem statement

The funding for the PHE comes from several directions. First, Bernhoven Hospital and Synchroon foundation are investing in the PHE. Secondly, funds from the multiannual agreement with the two largest healthcare insurance companies (CZ & VGZ) are being used for the PHE project. Additionally Bernhoven Hospital makes use of the government VIPP-program for funding of the project. The stakes are high, because non-adoption of the PHE can have multiple serious consequences. If the PHE is not used by its end-users, funding would be lost from the multiannual agreement and the Bernhoven Hospital and Synchroon foundation. Additionally, one of the objectives from the VIPP-program that need to be reached is at least 10% adoption rate among patients. If the PHE fails to reach that due to non-adoption, Bernhoven Hospital will have to repay the government for the used funds to finance a part of the PHE project. Moreover, if the PHE fails to be adopted by its end-users, the Bernhoven Hospital could suffer from reputational damage among patients. Furthermore, failure of the PHE project could negatively influence the intention of patients and staff members to adopt future health information technologies, because of the negative experience with the PHE. Finally, the patients and their healthcare providers will not experience the benefits of a successful PHE if it is non-adoption occurs. Therefore it is essential for Bernhoven Hospital to investigate what factors can cause non-adoption among end-users and how to mitigate or avoid those risks.

Objective

The objective of this study is to examine the intended reasons for non-adoption of a regional personal health environment platform for hospital physicians, primary care physicians and for patients with chronic obstructive pulmonary disease and rheumatic disease.

Research questions

Main research question

What factors might influence non-adoption of a personal health environment platform among physicians, chronic obstructive pulmonary disease and rheumatic disease patients?

Sub questions for research

I. What is the definition of a personal health environment used by Bernhoven Hospital and how does it differ from definitions used in the literature?

II. What is known in the literature about adoption and non-adoption of a personal health environment among end-users (i.e. physicians, primary care physicians, patients)?

III. Which method is suited to assess the factors for non-adoption from multiple stakeholders’ viewpoint within the timeframe of this study?

IV. What factors might influence non-adoption among patients (COPD and rheumatic disease specific, if possible)?

V. What factors might influence non-adoption among hospital physicians and primary care physicians?

VI. What factors might influence non-adoption according to the development team?

(13)

7

Approach

An online desk search, followed by a Weakness-Based Language Processing (WBLP) method is used to find and organize factors for non-adoption among physicians, primary care physicians and two types of chronically ill patients. This is done in one meeting with all the above mentioned stakeholders. The results will be analyzed afterwards and practical implications or recommendations will be made following the results.

Study outline

In the next chapter section (methods) I describe how I conducted my preliminary research and my field research by clarifying the instrument, procedure and participants. Following the methods is the theoretic framework section where I explain core concepts like PHEs, non-adoption and the Weakness-Based Language Processing. In this section I answered my descriptive sub questions for research. The results section describes the characteristics of the participants and factors that could influence non-adoption found by Weakness-Based Language Processing. In the discussion section I summarize the principal findings, reflect on strengths and limitations of this study, discuss my results in relation to other studies, and end with recommendations to mitigate non-adoption factors for Bernhoven Hospital.

(14)

(15)

9

Methods

In this chapter I discuss my methodology for this study divided in two main components: a preliminary search part and a field research part. In the preliminary search part I discuss an online desk search and in the field research I discuss the instrument, participants and procedure.

Preliminary search

This study was written between November 2016 and June 2017 at the Information and Innovation department from Bernhoven Hospital. I gathered preliminary data through an online desk search from internal organisational databases as well as external online databases. The preliminary data was used to answer the first three descriptive sub questions for research.

Online desk search

The goal of the online desk research was to get answers to sub questions one, two and three for research. I sought help at the medical librarian in the Academic Medical Center library in Amsterdam (Joost Daams) for conducting an online desk search. The medical librarian is experienced with helping students with their online desk search for their master theses. The results from the online desk search are discussed in the theoretical framework section. Additionally, results from sub question two will be compared with the results from the field research in the discussion section. This step was done to find support for my findings of my field research.

Identification of relevant studies

I started the online desk search with a scoping search via Google Scholar search engine3 to get a first understanding of the topic. After the scoping search I made a list with key terms used in PHE medical informatics literature. The list of key terms was checked and augmented by the medical librarian. Databases PubMed and Scopus (1999-2017) were searched with the key terms “personal health record” and “adoption” or “non-adoption” with their synonyms in an OR-construction. All used synonyms and abbreviations can be found in Appendix B. The online desk search was deliberately kept broad, since PHEs as a topic for scientific literature is relatively new and underreported. Scientific articles, reports, and whitepapers were included in the searches.

(16)

10

Selection of studies

The identified articles were screened on title and abstract. Studies were included for reading full text when they met the following criteria: (1) describing a existing standalone or integrated PHE or concept for a standalone or integrated PHE; (2) gathered data from either people, patients, healthcare providers of experts; (3) gathered data from actual PHR users or non-users; (4) describing factors that influence adoption or non-adoption, either by an existing theoretical model or emerging themes. Articles reporting tethered PHRs (Patient Portals) and EMRs were excluded despite being a well researched information technology in medical informatics. Tethered PHRs were excluded since the technology differs from the PHE such as Bernhoven Hospital is implementing and therefore results are not generalizable to PHEs. Tethered PHEs offer a ‘view’ for patients of their medical data that is kept at that organisation or care institution, while a PHE aims to be a data management and self management solution for people (not only patients).

Standalone PHEs are included, since they can be connected to an EMR (Gaskin et al., 2011). Studies that were reported in Dutch or English were included. Additionally, I searched the reference lists of included articles and used the “related articles” function in PubMed to find additional relevant publications that were missed in the initial search. The quality of the study was not an inclusion criterion, since there were relatively little hits within the online desk search.

Extracting information from the included studies

I extracted information from included full text articles and created a tabular representation in a Microsoft Excel spreadsheet. General citation information (Author, year published, country); sample (size and population); theoretical model used (if any); methods; outcome; discussion and overall findings by the authors that could be used for answering my research questions were summarized in the Excel spreadsheet to create an overview which author investigated what, by whom and when.

(17)

11

Field research

I chose a qualitative research design to answer the sub questions four, five, six and seven. I used a Weakness-Based Language Processing method to gather, analyse, and interpret data from my field research. I answered my main research question combining the data from the online desk research and the field research to build on or add knowledge to the existing knowledge base.

Instrument

I sought an instrument that could uncover factors relating to complex behaviour or motivation, like adoption or non-adoption of a PHE. Because non-adoption is a negative approach of exploring factors related to PHE behaviour, I used the Weakness-Based Language Processing method from the Center for Quality Management (CQM) (CQM, 1997) to gather data in my field research. WBLP is a focus group-like method that can be used when the issue at hand (i.e. non-adoption of a regional PHE) appears large and complex. This is the case given the different end-users, and the fact that several PHEs have already failed to be adopted. WBLP is a method to answer a single theme question from multiple points of view in one group meeting, working together to find consensus on the most important answers to the theme question. I also chose the WBLP method because it is a time efficient method, being only one session. An additional benefit compared to separate interviews or focus groups, is that the WBLP method also reaches consensus among the participants about the risks for non-adoption and their priority. Moreover, it has been shown in the past that non-adopters often do not want to be interviewed (Greenhalgh et al., 2010). Therefore a different approach like the WBLP method might be needed to explore non-adoption. Compared to focus groups the WBLP method creates more uniform quality language due to its semantic guidelines. One drawback of the method is that the CQM advices to use an experienced facilitator for the language processing method. Fortunately an experienced facilitator with over ten years of experience could attend the group meeting (Jan Eric Slot). The facilitator is not part of the group of participants, but is only there to facilitate the process for the participants. Conclusively, the WBLP method is unique in combining three fundamental elements of language analysis into a sequence of steps. These three elements are essential for groups working with qualitative data to solve problems (i.e. non-adoption of the PHE). I. Preparing the teamwork environment; the WBLP method utilizes an efficient physical set-up for

teamwork and prepares the group to work together. Their recommendations for the entire set-up are tested and improved with every new draft and can be found in the theoretical framework. The predetermined seat arrangement, table position, and writing supplies, help create an organized environment for enhanced concentration. This can be especially helpful to increase the efficiency for beginning groups.

II. Making uniform quality language data; qualitative data can be ambiguous, since the meaning of language can be perceived differently by people. Semantic rules are used to make qualitative language data uniform and suitable for analysis.

III. Structuring the data; the group follows specific steps to build a common understanding, develop new insights, and reach consensus on the conclusions. Additionally, participants rate the degree of importance to constructs of the data4.

(18)

12

Participants

Sampling procedure

Participants for the WBLP meeting were selected using a non-probability purposive sampling method. “Purposive sampling is often used when a diverse sample is necessary or the opinion of experts in a particular field is the topic of interest” (Martínez-Mesa et al., 2016, p. 328). Purposive sampling was used because not many people have expertise about PHEs, since it is a relatively new information technology in the Netherlands. The selection criteria for participation comprised: (1) having self-reported knowledge about the PHE type Bernhoven Hospital wishes to implement. The WBLP method recommends that participants need to have knowledge and understanding about the subject covered (i.e. PHEs); (2) Having support from the group of people they represent; (3) being able to sympathize with the group of people they represent; (4) self-reported discussion with colleagues or fellow patients about PHEs and the problems facing around adoption by colleagues or fellow patients. Exclusion criteria included limitations that would prevent the participant from actively participating in the WBLP meeting, for example cognitive, hearing or visual limitations. Sample size

The sample size for a WBLP meeting is usually restricted to eight participants for beginning groups (CQM, 1997). Through the purposive sampling methods, the amount of participants per domain was determined to be roughly equal. Subdivided in to their knowledge domain, they could formulate the opinions of the patients (28%; n=3); physicians (36%; n=4), or development team (36%; n=4).

Participation

Eleven people were selected to join the WBLP meeting and to provide input. Oversampling was done because I expected three to four invites to be turned down. One invitation was turned down (9%). Ten out of eleven invited people attended the WBLP meeting (91%). One person from the patient domain could not attend the meeting. However she did provide her input via e-mail. This data was not lost and was put together with the other input from the WBLP meeting participants.

Participants

Participants for the patient domain were recruited via umbrella organisations. The organisations were the Dutch Patient Federation (‘Patiëntenfederatie Nederland’), Rheumatic Foundation (‘Reumafonds’), and the Pulmonary Foundation (‘Longfonds’). These umbrella organisations provide ‘expert patients’ that are familiar with speaking for their patient group and discuss with healthcare organisations about the problems they encounter in terms of for instance illness management. Participants for the physician domain were recruited via Bernhoven Hospital and the Synchroon group for primary care physicians. The physicians from the Bernhoven Hospital were working for Bernhoven Hospital at the time of the meeting and were in different ways involved in the PHE project. Participants for the development team were recruited via Bernhoven Hospital. They were recruited because they are responsible for the PHE project. They should align the software developer capabilities with the end-users needs and are responsible for program management. They frequently discuss the PHE project with different involved parties and can provide a different perspective on reasons for non-adoption. The diversity of the participants supports the examination of different perspectives and alternative answers (Murphy et al. 1998).

(19)

13

Procedure

Preparatory to the WBLP meeting

The participants received an invitation via e-mail (Appendix C). The invitation was made as an infographic, using the online program Pictograph. After participants agreed to join the WBLP meeting, they received a reminder to send their four answers to the theme question at least two days before the meeting. The amount of answers to the theme question (Fig. 1) was limited to four answers, so participants would consider what they found important answers to the theme question. Additionally, four answers gave the participant insight in expected amount of answers and keep the information workable during the WBLP meeting. By reviewing given answers before the meeting, I could decrease time during the meeting to make uniform quality answers. This measure was necessary even though participants received semantic guidelines with examples specific for this meeting (Appendix D).

“What can hold back the patient or healthcare provider from using the Personal Health Environment for COPD or rheumatic disease?”

Figure 1. Theme question to which each participant of the Weakness-Based Language Processing meeting provides four answers.

The WBLP meeting

The WBLP meeting was held in a nearby event center called ‘de vrije teugel’, which is often used by the Bernhoven Hospital to arrange meetings (appendix E). The meeting was held from 17.30 to 19.00 PM. Later on the day was the best possibility to get as many accepted invites as possible, hence almost everyone works from 09.00 to 17.00. The actual room set-up for WBLP meeting matched perfectly with the room setup guidelines (CQM, 1997). Furthermore, an experienced facilitator was present. The facilitator had no input in the answers given by the participants. His purpose was to ensure a good atmosphere and group dynamics in a way that every participant provided input during the process.

During the meeting, completing all stages from the WBLP guideline would be too time-consuming. The facilitator and I decided in advance that it would be feasible to at least reach stage two, and begin with stage three. Normally a whole WBLP meeting lasts 4-6 hours (CQM, 1997). Stage three up to five would be conducted afterwards. The reward for participation in the meeting was (I) a great experience (the WBLP method); (II) a refund for travelling expenses; (III) and including dinner because of the timing of the meeting.

(20)

14

After the WBLP meeting

After the WBLP meeting all first level aggregations were complete (stage two). This meant that all initial answers were aggregated to transcending themes called ‘first level themes’. The first level themes were numbered 1-23 so that they could easily be scored in a later stage. The participants were informed that the facilitator and me would aggregate the first levels again to new transcending themes called ‘second level themes’ (stage three). The second level themes were recited with the letters A-I so they can be easily kept apart (Fig. 2). A diagram of the initial answers, the first- and second level themes was made in Microsoft Visio (stage four). The diagram was send via e-mail to all participants for verification of the results.

Reaching higher abstraction levels Aggregate first level themes to second level themes Aggregate answers to first level themes

Figure 2. Visual representation on how the WBLP method allows to reach higher levels of abstraction through aggregation of answers from lower level of abstraction.

Additionally I asked the participants to give a priority score to rate the most important first level themes. Each participant could assign a total of 10 points. The points were distributed in 4, 3, 2, and 1 point, with 4 points being the highest priority. The CQM normally gives 6 points (3, 2, and 1) to each participant. However, since the sample size (11) was above the usual number of eight participants, the amount of answers to the theme question would automatically be higher as well. It seemed reasonable to also increase the amount of points that could be given by the participants, so that the points can be evenly distributed as in a situation with less answers and less points. A record was kept of all priority scores in a Microsoft Excel spreadsheet to analyze which participants prioritize what risk factors as most important for non-adoption of the PHE. The priority scores will indicate the most important themes related to the theme question.

In the next chapter I elaborate upon definitions, characteristics, benefits and adoption rates of PHEs, and the room set-up, semantic rules and five stages of Weakness-Based Language Processing in more detail.

(21)

15

Theoretic framework

In this chapter I describe the three main subjects around this study: personal health environments, adoption of personal health environments and Weakness-Based Language Processing.

Personal Health Record and Personal Health Environment

Definitions and synonyms

In recent years, there has been an increasing interest in personal health records (PHRs). The increase of interest is not just from the Dutch government, but is seen worldwide, with publications mostly coming from Australia, Europe, North America and the United Kingdom. In the Netherlands, the term ‘(personal) health record’ (PHR) has a bad reputation due to problems with the adoption of a national electronic health record in 2009 (Van Der Maat, Reisma-Van Rooijen & De Jong, 2010). Therefore, the term personal health environment (PHE) from the Dutch Patient Federation was adopted instead of PHR by Bernhoven Hospital. The term PHE was chosen to prevent association with the bad reputation of the national electronic health record. In this study, the definition of the Dutch Patient Federation (DPF) will be used: “An universally accessible, understandable for layman, user friendly and lifelong tool to collect, manage and share relevant health information and to support self management by controlling health and healthcare through standardized data collection for health data and integrated digital care services; The PHE is managed by the patient or legal representative; Is secured in a way that privacy is guaranteed and is in no way a legal medical file, unless stated otherwise” (Bierma, Heldoen, Pluut, & ter Brake, 2013, p. 18) (Appendix F). Numerous other terms are used in the literature in addition to the term personal health environment. For example personally controlled electronic health record (PCEHR), interactive preventative health record (iPHR), personal health record-based self management system (PHRSMS), and personally controlled personal health record (PCPHR) to name a few. The most commonly used term in the literature is personal health record (PHR). To specify that a PHR is in digital format, not paper format, the term electronic personal health record (ePHR) is often used.

(22)

16

Characteristics

Several health information technologies in the literature fall under the heading of ePHRs. Gagnon and colleagues (2016) recently distinguished four subcategories of ePHR types (Table 1). Subcategory IV resembles most with the definition for a PHE used at the Bernhoven Hospital. Therefore I compared subcategory I-III with subcategory IV in this paragraph. At the end of this paragraph I present an overview of the characteristics for all subcategories (Table 2).

A standalone ePHR (I) differs from an integrated ePHR (IV) in many ways. Often a standalone ePHR is from a commercial vendor (e.g. Google, Microsoft) and is not tethered to medical software like an integrated ePHR. The similarity between a standalone ePHR and an integrated ePHR is that both solutions can be used by both patients and healthy individuals. A tethered ePHR (II) is often called a patient portal and is always bound to one organisation or care institution (Heldoen, Herk, & Veereschild, 2011). The tethered ePHR only shows data from one organisation, while the integrated ePHR can show information from multiple care providers (e.g. hospital and primary care). Therefore, a patient portal can be a subject of the integrated ePHR. For example, the patient portal of the primary care physician can be a subject of an integrated ePHR. Secondly, tethered ePHRs only show information coming from EMRs, not from healthy individuals which are the case for the integrated ePHRs. Lastly, tethered ePHRs often lack self management tools and only show medical information and sometimes have appointment making or medication refill functionalities. The differences between an un-tethered ePHR (III) and an integrated ePHR are almost the same differences as a standalone ePHR. The un-tethered ePHR has the only benefit that it can sometimes be shared with healthcare providers. Un-tethered ePHRs have much less value for patients, since it can never be a complete health record because professional data is missing.

Table 1. Description four subcategories of ePHRs according to Gagnon et al. (2016, p.426) ePHR types

I

“A standalone ePHR that allows the collection of health information on a portable media device or a website that generally enables viewing and managing the data by the patient only.

II

A tethered ePHR that is managed by a facility or institution where patients can access and update their health information from the facility or institution’s EMR with various degrees of control.

III An un-tethered ePHR that does not communicate with an EMR, but allows patients and sometimes the healthcare providers to have access to the record.

IV The integrated ePHR that makes it possible to gather and view data from multiple sources, such as an EMR, patients, or healthcare providers”

(23)

17 Table 2. Summary of characteristics ePHRs

Standalone ePHR (I) Tethered ePHR (II) Un-tethered ePHR (III) Integrated ePHR (IV)

User(s) Patient and

healthy individual Patient

Patient and healthy individual (sometimes HCP) Patient, healthy individual, and HCP Provider Commercial vendor Facility or care institution Commercial vendor Facility or care institution Data source(s) Patient or healthy individual (sometimes EMR)

1 EMR Patient or healthy

individual Patient, healthy individual and >1 EMR Additional functionalities (e.g. appointment making, drug reorder) No Yes No Yes Self management

tools Often Sometimes Sometimes Yes

Abbreviations: ePHR, electronic personal health record; PHE, personal health environment; HCP, healthcare provider; EMR, electronic medical record

Benefits of PHEs

The number of publications with PHRs as subject has increased rapidly. Many studies have been done about the perceived or expected benefits from PHEs by either patients or healthcare professionals. The expected benefits in the literature are numerous: better emergency care when all medical data is bundled (Tang, Ash, Bates, Overhage, & Sands, 2006); higher patient satisfaction (Pagliari, Shand, & Fisher, 2012); increasing trust in their medical doctor (Fisher, Bhavnani, & Winfield, 2009); higher patient satisfaction (Pagliari, Shand, & Fisher, 2012); better communication between the patient and the physician, by studying their health data before an appointment and to come up with relevant questions (Fisher, Bhavnani, & Winfield, 2009; Kahn et al., 2010) or to fill in questionnaires before an appointment Kaelber & Pan, 2008; Wald et al., 2009); lower healthcare costs through an electronic consultation hour as an accessible way for patients to communicate with their physician (Wakefield et al., 2010); improve disease- or medication management by sending reminders or alerts (Wright et al., 2012); PHE data for research purposes, for example effectiveness of certain treatment or public health (Weitzman, Kelemen, Kaci, & Mandl, 2012), and finally PHEs could reduce the burden of information management in chronically ill patients (Ancker et al., 2015).

When patients collect the data of different caregivers in their personal health record, a more comprehensive health record will be the result (Berry et al., 2011; Tang et al., 2006; Turvey et al., 2012).

(24)

18

However, it is not proven that this results in better quality of care since systematic reviews conclude that very few empirical studies have been done to investigate clinical- or patient outcomes. Those few studies conducted have trouble showing statistical significant impact (Kalra & Fernando, 2013). Additionally, no large, long-term studies have been done to determine the effects of PHEs. The speed at which PHE functionalities change and new functionalities are being added makes it hard to determine the impact of a PHE in healthcare (Price et al., 2015).

PHE adoption by end-users

Adoption factors

Adoption of PHEs (Kool, Verhoef, & Kremer, 2014) and e-health in general (Krijgsman et al., 2016b) is low in the Netherlands, partly because awareness is low in general. Awareness of PHEs is often found to be low pre-implementation and sometimes even after go-live (Armani et al., 2016; Greenhalgh et al., 2010; Lehnbom et al., 2014; Spil & Klein, 2014). Among all potential PHE users, chronically ill patients are probably the most important end-users for PHEs (Price et al., 2015). Chronically ill have the highest probability of being ‘eager adopters’ (Determann et al., 2016). Other potential eager adopters are elderly, young mothers and patients with reduced mobility (Gagnon et al., 2016). Elderly tent to value the ability to try a technology and to observe someone else using it. Being able to observe and try the PHE has the potential to develop computer self-efficacy in older adults and therefore increase the perceived ease of use (Emani et al., 2012; Logue & Effken., 2012). Contrarily, younger adults value the ability to customize a PHE and to access it in various ways (e.g. tablet, smart phone) more than other users (Nguyen et al., 2016; Rief et al., 2017). Other important end-users of PHE technology are healthcare providers. Healthcare providers play an important role in uploading medical data in the PHE as well as stimulating the use among patient users (Lehnbom et al., 2014; Logue & Effken, 2012; Cabitza et al., 2015).

Perceived value is often mentioned as a factor for adoption of PHEs (Agarwal et al., 2013; Andrews et al., 2014; Emani et al., 2012; Gagnon et al., 2016; Tang et al., 2006; Yau et al., 2011). For example Andrews et al. (2014) found perceived value to be the most influential factor on intended adoption among sample of middle-aged general population in Australia. If the patients are not aware of the relative advantage of PHE use, the PHE quickly loses effectiveness or value (Sartain et al., 2014). Furthermore, the PHE should create value for future investors, in order for it to be sustainable (NPCF, 2015c). With fewer users, the PHE becomes less interesting for future investors. Therefore non-adoption should be avoided if a sustainable business model for the PHE is the goal. In order for the PHE to be successful in adoption rates, it needs to be used by both the patients and healthcare providers. Another often mentioned adoption factor is perceived usefulness or perceived job enhancement (Agarwal et al., 2013; Andrews et al., 2014; Chung et al., 2016; Cocosila et al., 2014; Gartrell et al., 2015; Hsieh et al., 2016; Liu et al., 2013). Perceived increase in workload might have a negative influence on perceived usefulness (Yau et al., 2011). Moreover, interviews with the most successful PHE providers in the USA showed that perceived increase in workload for patients and healthcare professionals was found to be the strongest barrier to adoption (Wells et al., 2015).

(25)

19 Several studies investigated perceived ease of use as a facilitator for intention to use a PHE (Andrews et al., 2014; Chung et al., 2016; Gartrell et al., 2015; Hsieh et al., 2016; Liu et al., 2013). Gagnon et al. (2016) argue that a PHE should be easy to understand and navigate for all user groups. They found perceived ease of use as most frequent mentioned facilitator for PHE adoption among a group of experts.

Adoption rates

The importance of this study lies in the low adoption rates of existing PHEs (California HealthCare Foundation, 2010; NPCF, 2015c), failed PHEs (Greenhalgh et al., 2010; Lau et al., 2015; Spil & Klein 2014) and low reported intention to use in the Netherlands (Determann et al., 2016). Although several PHEs are available in both the Netherlands and worldwide, adoption rates are disappointing. A large national survey in the Netherlands (n=11.241) among a patient panel reported only 7% of survey participants uses or has used a PHE (Van Der Steen & Van Haastert, 2015). Seven percent is still lower that the target adoption rate of 10% from the VIPP-program. Additionally, the 7% reported usage is low, because among the panel members 79% has at least one chronic condition. It is assumed that people with chronic conditions could benefit the most of PHEs and therefore most likely contains the largest group of PHE adopters (Cocosila & Archer 2014; Gee et al., 2015; Harrison et al., 2015; Kerai, Wood & Martin 2014; Logue & Effken, 2012; Price et al., 2015). One could speculate that for the general Dutch population, PHE usage is lower since the rate of chronically ill is lower than in the panel sample.

The development and usage of PHEs is generally lower in the Netherlands than abroad. For instance the United States has a more mature PHE market, with more PHE platforms. Despite the more developed market however, the adoption rates are low (Bates & Wells 2012; California HealthCare Foundation, 2010). A national survey (n=1849) in California, United States reported only 7% PHR adoption rates among their sample (California HealthCare Foundation, 2010). Note that survey was done prior to the survey in the Netherlands. It can be assumed that the adoption rate in the United States has risen since 2010 (Bates & Wells, 2012; Ford, Hesse & Huerta, 2016).

Australia has launched a national PHE in 2012, but stayed well under their adoption target for the first year (Kerai, Woods, & Martin, 2014). The PHE was named Personally Controlled Electronic Health Record (PCEHR). Adoption remained low, especially under the healthcare providers with only 140 views per month nationwide (Lehnbom, Douglas, & Makeham, 2015). The Australian government decided to change the image and to make some changes. The government went from an opt-in model to an opt-out model and changed the name to lose the negative image and boost adoption rates (Australian Digital Health Agency, 2016; Jolly, 2015). Furthermore, Canada has also shown slow progress in adopting PHEs as well (Gagnon et al., 2016).

(26)

20

Previously failed PHEs

Despite the slowly growing adoption rates abroad, several PHE projects unfortunately did not survive the subsidy phase. Meaning they did not create enough value for either patients, professionals, investors or all of the before mentioned (NPCF, 2015c).

The first example is personal electronic health record ‘HealthSpace’ from the UK. HealthSpace UK had low adoption rates and eventually failed. In three years time (2007-2010), only 0.13% of invited participants opened an account in HealthSpace, while 5-10% was anticipated in the original business case (Greenhalgh et al., 2010). Awareness was low in the UK, therefore Greenhalgh and colleagues interviewed formal users of HealthSpace to find more answers. The authors report that patients who did use HealthSpace, ran into a cumbersome and bureaucratic registration process, found usability poor (no user-centred design), functionalities did not align with their workflow and their healthcare data was not automatically uploaded to the PHE. The policy makers hoped by implementing HealthSpace, the perceived benefits of a PHE and the adoption rates would automatically follow. Unfortunately for all involved parties, this was far from reality.

Secondly, Google Health was introduced in 2008 by technology giant Google, but was prematurely retired in 2012. Google Health suffered from low adoption rates and limited use among users (Brown & Weihl, 2011). Google Health was also adopted by the Lucile Packard Children’s Hospital at Stanford University to create an interoperable PHE for their patients (Gaskin et al., 2011). However, with the fall of Google Health, the Lucile Packard Children’s Hospital also had to find another solution and could only provide lessons learned, just like Greenhalgh and colleagues (2010). Spil and Klein (2014) speculate that Microsoft’s HealthVault will also eventually fail for the same reasons Google Health failed: low trust, high risk in commercial PHE vendors and ignoring perceived usefulness and relevance for patients when they designed their PHE. However, currently Microsoft HealthVault is still active as a standalone personal health record.

A third example of a failed PHE is the web-based Personally Controlled Health Management System that was deployed in Australia for adults with Asthma. Even though this was a well-designed PHE and tested in a randomised controlled trial, 81% did not access the PHE or accessed it only once (Lau et al., 2015). Asthma patients from the study reported that their reasons for non-adoption were discouragement of healthcare professionals, time consumption of managing their PHE, wrong information at the wrong time, and no customization of information to their healthcare situation (Lau et al., 2015).

Finally, the Dutch Patient Federation reports that the PHE called ‘Medlook’ failed after fifteen years on the market (NPCF, 2015c).

Conclusively, PHE adoption is currently low and slowly rising, while some PHEs have failed. Additionally, a recent survey in the Netherlands report that a large part of the population is not planning on using a PHE for the next two years (Determann et al., 2016). It is not self-evident that implementing a PHE will automatically mean adoption by its end-users, even when there are numerous expected benefits. Non-adoption of PHEs is a potential risk for failure and the reasons in the literature for non-adoption remain underreported (Lau et al., 2015). Furthermore, there is no established methodology to investigate the potential risks for non-adoption (Greenhalgh et al., 2010).

(27)

21

Weakness-Based Language Processing

Weakness-Based Language Processing (WBPL) is a method from the Center for Quality of Management (CQM) that is headquartered in Cambridge, Massachusetts in the USA, with satellite locations all over the USA and the rest of the world (CQM, 1997). This method was first developed by professor Shoji Shiba of Tsukuba University in Japan and has evolved and improved six times from the starting point in 1987 up to 1996 by members of the Center, internationally known behaviourists, anthropologists, linguists and quality management thought leaders. WBLP method is a tool for organizing qualitative data and creating insight with a group of people. The method is used to find weaknesses within a (future) process or product and to prioritize the weaknesses with ratings of importance. The goal of this method is to analyze a theme question with a group, to determine the weaknesses in a clear and non-ambiguous way and to emphasize fact-based language in one group session.

The upcoming paragraphs are written in chronological order. The CQM describes a physical set-up for teamwork that should be arranged before the meeting. Secondly, the participants should be familiar with the semantic rules that help create uniform, quality answers to the theme question. If these two preceding elements have been met, the meeting can start. In total five stages for execution are described with an example of a WBLP diagram.

Physical set-up for teamwork

The WBLP method perceives a proper physical set-up as one of the three fundamental elements5 for problem solving. The CQM provides several recommendations regarding the room setup (Fig. 3) and the materials needed (Table 3). The CQM describes the physical setup and materials in great detail, making it easier to adhere to their well experienced guidelines and leaving less room for mistakes regarding the setup.

First, a large enough room should be reserved for the meeting to fit in all the materials. It should contain a table and chairs, with a wall chart. The table should be just large enough, so that everyone can face the wall chart and sit shoulder to shoulder. The chairs ideally do not have armrests. CQM advises to have roughly eight participants in the meeting (CQM, 1997).

(28)

22

Figure 3. Schematic representation of a room setup according to the Center for Quality Management (1997, pp. 25-26)

Table 3. Materials needed to facilitate the WBLP method (CQM, 1997, p. 25) Meeting materials

- One large box that can hold all the materials (size of a shoebox)

- 1 roll of ¾ inch masking tape - 1 roll ¾ inch invisible tape - 2 packages of 3 x 5 inch Post-its - 12 black medium-point pens - 12 red medium-point pens - 12 blue medium-point pens - 4 black flip chart markers - 4 red flip chart markers - 4 blue flip chart markers - 4 green flip chart markers

- 20 each of ¾ inch red, blue and green dots (sticky dots)

- 2 glue sticks - 2 pencils - 1 large eraser

- 1 roll 1 inch white correction tape - 1 roll 1/3 inch white correction tape - 1 pair of scissors

- 1 sheet of paper about 35 x 45 inch or 2 large 27 x 35 inch flip chart sheets, taped together so the total is about 3 x 4 feet

(29)

23

Semantic rules of the WBLP method

The Weakness-Based Language Processing method uses semantic guidelines to convert the language of affection to the language of report. These semantic guidelines have the goal to create uniform understanding about the meaning of an idea, without losing the essence of the original thought. The following paragraphs will give short explanation about the semantic guidelines with examples. Language of report and language of affection

All languages have two types of languages, the language of report and the language of affection. In normal circumstances (e.g. a conversation), these two types of languages are used at the same time. The language of affection contains a sentiment, for instance “I approve,” or “I disapprove”. But the language of report is based on facts. When a problem needs to be solved, the language of report needs to be used according to the WBLP method.

Judgement and inference

The basic elements of language of affection are judgements and inferences. These are statements that cannot be verified. An inference is a statement about the unknown made on the basis of the known. A judgement is an expression of the writer’s approval or disapproval.

Inference example

The COPD patient is angry that he can not get digital access to his data

Language of report (without inference)

The patient banged his fist on the table of the doctor and threw the paper chart in the air

The second sentence can be verified, in contrast to angry. The first sentence is therefore an inference, while the second sentence is the language of report

Judgement example

The COPD patient has a wonderful coping style

The language of report (without judgement)

The COPD patient has quit smoking and start exercising

Not everyone would say the patients’ coping style is wonderful, but the second sentence can be verified.

Sometimes a judgement and inference cannot be separated. Therefore they can be seen together as an opinion. The more important part is to clarify the facts that are behind this opinion. This can be done using the interrogatives what, when, where, who and how.

(30)

24

The ladder of abstraction

Hayakawa (1990) explains a ladder of abstraction in his book. This ladder of abstraction (Fig. 4) is used in the WBLP method to steer answers from participants to low as possible abstraction level, and to find transcending themes. For example, if you start with a Rheumatologist, he or she has a name, like John. This is one step down on the ladder of abstraction. If you go one step up the ladder, the Rheumatologist is a Specialist, which is always a Physician and a healthcare provider and so on. The difficulty lies in going one step down or up, without skipping abstraction levels. For example go from the level of Rheumatologist to healthcare provider.

Figure 4. Ladder of abstraction example, inspired by the example from the Center for Quality Management (1997, p.16)

Two-valued orientation and multi-valued orientation

The easiest two-valued orientation is saying the something is “good” or “bad”. However, reality is often somewhere in between good and bad. A multi-valued orientation is a way to factually state the area around good and bad.

Two-valued orientation example

The adoption rates of the PHE are good

Multi-valued orientation example

The adoption rates of the PHE are 80% among COPD-patients

(31)

25

Five stages of execution

A WBLP meeting is guided by five stages and a set of semantic rules to attain clear answers to the thematic question at hand. During the first stage, the meeting starts with a warm-up phase. There is a ten minute open discussion where every expert in turn gives his or her thoughts on the theme question without judgement or critiquing. Next, every group member receives about four Post-its cards and is invited to write one concrete answer to the theme question, in a full sentence per card. This is done individually, without talking to each other. Cards are posted immediately after writing, so that every other expert can read it, to avoid writing a similar card. The goal of stage one is to attain clear, non-ambiguous answers to the thematic question.

During the second stage, all cards are scrubbed one at a time by the whole group. Meaning: the author of the card explains his or her card. If the card is not clear to all, the card will be edited in a way that it will be clear to all group experts. This process is repeated for every card written. Note: this is not about if the card is right or wrong, but whether the answer to the theme question is clear and is at a low level of abstraction. Additionally, the scrubbed cards are grouped together where possible so that overarching themes can emerge in the next stage. At the end of stage two, after initial grouping, the team checks for omissions. If the team unanimously agrees on a thought, the thought is added. The goal of stage two is create full understanding about the cards and group them at the lowest level of abstraction.

Stage three is all about extracting common thoughts in the formed groups and reaching higher abstraction levels. The team writes a title for each group, to reach a higher level of abstracting one step at a time. The group titles get scrubbed and new groups are formed, in the same way as in stage two. These steps will be repeated until there are no more than five elements (grouped and ungrouped cards) left. Next, a layout is formed and a few causal arrows are drawn where possible, and groups are positioned so that relationships are visible. The goal of stage three is to reach the highest abstraction level, step by step and to order and create an overview in the card layout.

In stage four, a diagram is made with the group. Going from a low abstraction level to a high abstraction level, all cards are put on the board and marked by different colours to create an overview (Fig. 5). The goal of stage four is to create a visual understanding about the abstraction levels, so that all the answers and their groupings can be evaluated and rated in the next stage. Stage five is the evaluation stage and the final round. Each participant gets three coloured voting stickers in red, blue and green. They respectively represent three points (most important), two points and one point (least important). The points are counted and on the board a visual presentation is made (by e.g. cross checking) to highlight the most important answers to the theme question. Relationships of the most important answers are drawn to other answers if necessary. The goal of stage five is to reach consensus with the group to which answers on the theme question are deemed most important.

After a WBLP meeting, there are multiple answers to one thematic question with a clear overview what is meant and causal- or interrelationships between topics. All answers are rated of importance, reaching consensus among a group of experts to which items should be dealt with first.

(32)

26

Figure 5. Example of a WBLP diagram (CQM, 1997, p.52).

Text upper left corner: main theme question of a WBLP meeting; Grey rectangle: answer to the theme question given by a participant; white rectangle with ‘1st-level title’: first aggregation of answers to a transcending theme (i.e.: 1st-level title/theme); grey line: all answers that form a new transcending theme (i.e.: 1st-level title/theme); blue line: all answers and first level themes that form a new transcending theme (i.e.: 2nd-level title/theme); red line: All answers, 1st- level themes and 2nd-level themes that form a new transcending theme (i.e.: 3rd-level title/theme); black arrows: causal relationship between themes; black arrow with two perpendicular lines: contradiction between themes