Data warehousing in healthcare: assessing information needs concerning patient portals

(1)

Data warehousing in healthcare: assessing

information needs concerning patient portals

R.D.Duijm

1

_{Master of Medical Informatics, University of Amsterdam}

(2)

Project information

Name student: R.D.Duijm BSc Student number: 10175369

Email address: r.d.duijm@amc.uva.nl rduijm@chipsoft.nl Department of Data Warehousing, ChipSoft

Department of Medical Informatics, Academic Medical Centre

Mentor: L.E.C. Leloup MSc Email address: LLeloup@chipsoft.nl Department of Data Warehousing, ChipSoft

Tutor: Dr. ir. R. Cornet PhD Email address: r.cornet@amc.uva.nl Department of Medical Informatics, Academic Medical Centre

Period

Eight months starting from the 1st of November 2017. Finalized on the 4th of June 2018

(5)

Summary

The implementation of patient portals is an important step toward self-management, shared decision making, patient empowerment, and patient-centered medical care. Patient portals provide opportunities to gather and analyze data. The main goal of this thesis is to assess the feasibility of answering information needs using data warehousing. A scoping review, interviews, and regression analysis were performed. Additionally, reports on patient portal use were created.

A scoping review identified stakeholders, information needs and functionalities about patient portals. As a result, 192 information needs and 222 functionalities were clustered using concept mapping. The three information need cluster names are clinical, management, and use information. The six functionality clusters are: access to medical record, administration, communication, patient entered information, reminders, and self-management. These clusters were used to map customers’ information needs and functionalities.

Interviews were conducted to identify the available functionalities and their information needs with their importance and urgency. Ten interviews were conducted with hospital employees, a patient, and a patient federation. The most important information needs for the hospital employees were focused on complying with an acceleration program (VIPP), proof of effectiveness, and patient characteristics. The patient was focused on personalized information and the patient federation was concerned about privacy and sensitive informa-tion. However, the realization of a personalized patient portal was expressed as a positive development. Answering important information needs was expressed to be necessary and of high urgency.

One important information need was about the effect of patient portal use on no-shows in the hospital. The regression analysis showed that patient portal use seems to have a significant inverse association on the odds of a no-show. Additionally, age, gender, the number of appointments, the number of days between planning and appointments, and appointments on Wednesdays and Thursdays (compared to Mondays) were found to have a significant influence on the odds of having no-shows.

A report on the use of a Dutch patient portal was created to answer fifteen information needs. Various difficulties and suggestions were expressed. The results of this thesis can be used as a guideline to answer information needs about patient portals using data warehousing. Analyzing data from patient portals using data warehousing can create unique and valuable answers and insights to information needs.

(6)

Samenvatting

De implementatie van patiëntenportalen is een belangrijke stap in het realiseren van zelfmanagement, gedeelde besluitvorming, versterking van de positie van de patiënt en patiëntgerichte zorg. Ook bieden patiëntenportalen een kans om data te verzamelen en te analyseren. Het doel van deze thesis is om de haalbaarheid van het beantwoorden van informatiewensen door middel van datawarehousing te beoordelen. In deze studie zijn een scoping review, interviews en regressieanalyse uitgevoerd. Ook is een rapport over het gebruik van een patiëntenportaal gemaakt.

Een scoping review was uitgevoerd om stakeholders, informatiewensen en functionaliteiten van patiëntenportalen te identificeren. 192 informatiewensen en 222 functionaliteiten zijn geclusterd middels Concept Mapping. De drie clusters voor de informatiewensen zijn: klinische-, management- en gebruiksin-formatiewensen. De zes clusters voor de functionaliteiten zijn: toegang tot medisch dossier, administratie, communicatie, patient-ingevoerde informatie, herinneringen, en zelfmanagement. Deze clusters worden gebruikt om de informatiewensen en functionaliteitwensen van klanten te groeperen

Er zijn interviews gehouden om functionaliteiten en informatiewensen met hun urgentie en belang te identificeren. In totaal zijn 10 interviews gehouden met werknemers van ziekenhuizen, een patiënt en een patiëntfederatie. De belangrijkste informatiewensen voor de werknemers waren gericht op het voldoen aan de VIPP eisen, bewijs van de werking van patiëntenportalen, effectiviteit en patiëntkarakteristieken van gebruikers. De patiënt richtte zich vooral op gepersonaliseerde informatie en de patiëntfederatie uitte bezorgdheid over het gebruik van gevoelige informatie, maar zag het nut wel in een gepersonaliseerd patiëntenportalen. De interviews lieten zien dat het belang voor het beantwoorden van belangrijke informatiewensen hoog is.

Eén belangrijke informatiewens is om de invloed van patiëntenportalen op de kans op no-shows te bepalen. De logistische regressie liet zien dat het gebruik van patiëntenportalen een significante inverse associatie heeft met betrekking tot de kans op no-shows. Ook liet de regressie zien dat leeftijd, geslacht, het aantal afspraken, het aantal dagen tussen de planning en de afspraak, en afspraken op woensdagen en donderdagen (in vergelijking tot maandag) een significante invloed hadden op de kans op no-shows.

Als laatste is een rapport gemaakt over het gebruik van een patiëntenportaal. Vijftien informatiewensen zijn in dit rapport beantwoord. De resultaten van de thesis kunnen gebruikt worden als leidraad om informatiewensen over patiën-tenportalen te beantwoorden door middel van data warehousing. Het analy-seren van patiëntenportaaldata door middel van datawarehousing kan unieke en waardevolle inzichten creeëren om informatiewensen te beantwoorden.

(7)

Chapter 1

Introduction

Enabling patients to access their personal medical data is an important step toward self-management, shared decision making, patient empowerment, and patient-centered medical care [1]. Uptake of these concepts contributes to enabling self-directed behaviour changes for patients [2]. Examples of these be-haviour changes are: stop smoking, eating healthier, and increasing medication adherence. Additionally, the decision-making power is redistributed between the patients and physicians. Researchers studied the opinions of various patient groups on their needs and wishes for being enabled in their healthcare. According to Bastiaens et al., the elderly patient wishes to be informed and to have a caring relationship with healthcare providers [3]. One of the main reasons for shared decision making and patient empowerment is to increase the effectiveness of interventions and behaviour change by motivating the patient internally instead of externally; Also known as intrinsic and extrinsic motivation respectively. Intrinsic motivation is motivation from within the person. An example of intrinsic motivation is: "I want to quit smoking in order to be healthier". Extrinsic motivation is motivation through external reinforcement. An example of extrinsic motivation is: "My partner wants me to quit smoking". The combination of high intrinsic and medium extrinsic motivation is proven to be most effective for behavioral change [4]. Therefore, enabling patients to access their medical data can increase the healthcare quality, motivation, and feeling of control for patients.

Engaging patients in their medical care is not a new concept, but it is a fairly recently-studied concept. Moreover, enabling patients in their medical care is considered to be covered by the four biomedical ethics. The four widely accepted principles of biomedical ethics are autonomy (the patient has the right to choose their treatment), beneficence (a practitioner should act in the best interest of the patient), non-maleficence (do not cause harm) and justice (fair and equal distribution of scarce health resources) [5]. Enabling patient empowerment complies with three of the biomedical ethics principles: autonomy, beneficence and non-maleficence: patients are more prominent actors in their own healthcare, practitioners are able to adapt their treatment plan to the patients’ wishes, and patients have information on their own health and can therefore influence the treatment plan to avoid errors. Stiggelbout et al. even argue that justice might be enhanced when patients choose to avoid procedures

(8)

after addressing the potential benefits and risks [6]. Additionally, the care can be personalized when the patient has a say in the treatment of his or her disease. Information technology (IT) can support patients in their healthcare process. Moreover, patients actively search for information on their disease online [7]. Nictiz (The Dutch knowledge centre for ICT in healthcare) published a yearly eHealth monitor stating that over 50% of the patient population uses the internet to look up medical information [8]. The healthcare professionals are cautious about the information that the patient acquires and uses during the consultations. The result is that the patient has more knowledge on his or her disease, but the quality of the knowledge can be low [9]. This can influence the patients’ ideas and knowledge on their disease or medical complication. As a result, healthcare providers are concerned with the quality and accuracy of the online information that is being used by the patients [10]. Therefore, the healthcare providers have a new role as educator. A solu-tion could be a platform with verified and research-based knowledge for patients. One technology that could help with the above arguments is the electronic patient portal. Many functions that are possible within a patient portal can enhance the communication between the patient and the healthcare provider. Only 1% of the Dutch patient population used a personal online health record in 2016 [8]. However, the number of hospitals that provides patients with patient portals tripled between 2015 and 2016 to 23% in the Netherlands. All of these patient portals have the functionality of showing the electronic medical record (EMR) to the patients. However, not many controlled studies are conducted to prove the effectiveness of patient portals [11, 12]. Still, more and more hospitals make patient portals available for their patients. This is because the Dutch association of hospitals initiated a program that motivates hospitals and other health institutes to create forms of data exchange between patient and professional (VIPP) (Dutch: [13]). More information on VIPP is provided in Chapter 2: Background.

Patient portals show opportunities to gather data. Nictiz stresses the impor-tance of gathering and analyzing the data from the patient portals in order to increase the quality of the healthcare that is provided in the Netherlands. In fact, many boards of directors of hospitals want to know more about the user statistics of their patient portals (Dutch: [14]). However, not only boards of directors have a need for information. Other stakeholders have a need for information as well. Abernethy et al. published an exploratory study on using patient-reported data [15]. They concluded that electronic patient-reported outcomes can be used for real-time research data. Data warehousing can be used for data analysis on the gathered data from patient portals. Often, the data warehouse enables the creation of reports and cubes which can visualize and summarize the answers to the stakeholders’ questions of the patient portals. However, little research is published on the use of data that is gathered through patient portals.

In this study, stakeholders will be sought and their wish for information will be addressed. Additionally, the aim is to address and fill the knowledge gap on the use of data gathered through patient portals. Lastly, various information needs

(9)

will be answered using data warehouse principles and architectures.

1.1 Research objectives

The main research objective is to identify wishes and needs for information con-cerning patient portals. A scoping review will be conducted to map the knowl-edge and needs concerning patient portals. Next, interviews will be conducted to identify the current needs in the Netherlands concerning patient portals. Data that is acquired from patient portals will be used to answer to the information needs and wishes. More concrete:

• What is already known in literature about:

– Possible stakeholders concerning patient portals – The available functionalities in various patient portals – The information needs concerning patient portals

• How does the literature compare to the current setting in the Netherlands? – Comparing the results from literature to insights resulting from

in-terviews

– Identifying the current state of affairs within hospitals concerning information needs in patient portals

– Which information needs are most important

• Which information needs can be satisfied in the current system?

– Assessment of information needs through data warehouse reporting and statistical analyses

– Reflecting on the strengths and weaknesses of the used data ware-house principles and architecture for answering various types of ques-tions

• Which recommendations can be made to answer to the stakeholders’ in-formation need?

1.2 Chapter organization

The second chapter will provide background knowledge on the aspects used in this research. The third chapter will discuss the scoping review that has been conducted in order to identify use, functionalities, and the need for information concerning patient portals.

In the fourth chapter, the conducted interviews will be discussed. Additionally, the results from chapter three and four will be compared and an overview of the wishes or needs for information from various stakeholders will be noted. Chapter five focuses on answering one information need that was requested often. Chapter six reflects on the use of data warehousing for answering certain information needs. This chapter will be based on the methods, results,

(10)

and conclusions of the previously written chapters. The goal is to determine which questions can be answered by using data warehousing within the current system of "Zorgportaal". Zorgportaal is the integrated patient portal created by ChipSoft. It is integrated with the hospital information system called HiX. In chapter seven, a set of information needs that were identified through inter-views will be addressed. The result will be a report that answers information needs on the use of the patient portal. The experienced difficulties will be dis-cussed. Chapter eight gives an overall conclusion of the study and chapter nine shows the appendices.

References

1. Blumenthal, D. & Tavenner, M. The "meaningful use" regulation for elec-tronic health records. N Engl J Med 2010, 501–504 (2010).

2. Anderson, R. M. & Funnell, M. M. Patient empowerment: myths and mis-conceptions. Patient education and counseling 79, 277–282 (2010). 3. Bastiaens, H., Van Royen, P., Pavlic, D. R., Raposo, V. & Baker, R. Older

people’s preferences for involvement in their own care: a qualitative study in primary health care in 11 European countries. Patient education and counseling 68, 33–42 (2007).

4. Curry, S., Wagner, E. H. & Grothaus, L. C. Intrinsic and extrinsic moti-vation for smoking cessation. Journal of consulting and clinical psychology 58, 310 (1990).

5. Beauchamp, T. L. & Childress, J. F. Principles of biomedical ethics (Ox-ford University Press, USA, 2001).

6. Stiggelbout, A. M. et al. Shared decision making: really putting patients at the centre of healthcare. Bmj 344, e256 (2012).

7. McMullan, M. Patients using the Internet to obtain health information: how this affects the patient–health professional relationship. Patient edu-cation and counseling 63, 24–28 (2006).

8. Krijgsman, J. et al. More than Technique: ehealth Monitor 2016. The Hague: Nictiz and NIVEL (2016).

9. Eysenbach, G., Powell, J., Kuss, O. & Sa, E.-R. Empirical studies assessing the quality of health information for consumers on the world wide web: a systematic review. Jama 287, 2691–2700 (2002).

10. Fox, S., Rainie, L. & Horrigan, J. The Online Health Care Revolution: How the Web Helps Americans Take Better Care of Themselves. Wash-ington, DC: Pew Internet & American Life Project; November 2000. PIP_Health_Report. pdf (2006).

11. Ammenwerth, E., Schnell-Inderst, P. & Hoerbst, A. The impact of elec-tronic patient portals on patient care: a systematic review of controlled trials. Journal of medical Internet research 14 (2012).

12. Goldzweig, C. L. et al. Electronic patient portals: evidence on health out-comes, satisfaction, efficiency, and attitudesa systematic review. Annals of internal medicine 159, 677–687 (2013).

(11)

13. The Dutch association of hospitals. Versnellingsprogramma Informatie-uitwisseling Patient en Professional: de patient meer inzicht in zijn eigen zorg. https : / / www . vipp - programma . nl/. [Online; accessed 12-01-2018]. 2017.

14. Pluut, B., Peters, E., Sinnige, J. & Schreuder, C. Het gebruik van patientportalen: een verkennende studie https : / / www . nictiz . nl / SiteCollectionDocuments/Whitepapers/Whitepaper_het_gebruik_van_ patientportalen.pdf. [Online; accessed 06-11-2017]. 2017.

15. Abernethy, A. P. et al. Electronic patient-reported data capture as a foun-dation of rapid learning cancer care. Medical care 48, S32–S38 (2010).

(12)

Chapter 2

Background

2.1 Patient participation

Patient participation is involving patients in their own care. This broad concept can be split into three main categories: patient self-management, shared decision making, and patient empowerment. These categories overlap at various aspects and have slight nuances that distinguish them from each other. Additionally, patient-centered healthcare will be discussed. This concept is not part of the broad concept patient participation, because it does not necessarily let the patient take an active role in their healthcare. However, this concept will be explained, because it is often mentioned in literature and it is an often discussed concept in modern healthcare. In addition, patient participation can support patient-centered healthcare. Numerous articles use these concepts with various definitions and meanings. Therefore, the knowledge provided here is an interpretation and deduction of the latest ideas of these concepts. The four concepts will be discussed below.

Patient participation is a fairly new, but often studied concept. Originally, patient participation is advocated as a way to improve patient safety [1]. It is the result of the evolution of patients being passive recipients of medical care to being active co-producers of health. Patient participation has been successful in various medical fields [2]. Patients believed they had more choice within their treatment. Additionally, higher levels of quality of life were found with patients who participated actively in their care process. Furthermore, patients with certain characteristics can be more motivated to participate in their care process. Examples of these patient characteristics are comorbidity, lack of confidence, lack of medical knowledge, or various sociodemographic parameters [1].

Patient self-management

Patient self-management is a concept that covers the day-to-day decisions that are made by the patient. These decisions influence their health. The patient can be educated in self-managing his disease. This is often done for patients with chronic diseases. The goal is to give the patient the knowledge and decision capabilities to let the patient live with the best possible quality of life [3]. A feeling of responsibility should be created for both the patient and the health

(13)

professionals in order to enable the patient’s self-management skills. Various responsibilities have been presented for self-managing patients with chronic dis-eases: using medication properly, changing behavior, adjusting for social and economic consequences, coping with emotional consequences, and interpreting and reporting symptoms accurately [4]. These tasks require a fair amount of knowledge and responsibility. Additionally, the disease cannot be managed by the patient on his own. The medical expertise of medical professionals is needed. Therefore, another responsibility is needed: the patients’ responsibility of com-munication. All in all, patient self-management requires dedication, knowledge, responsibility, time, and professional guidance.

Patient empowerment

Patient empowerment starts when the healthcare provider acknowledges the fact that the patients are in control of their care [5]. It is designed to enable self-directed behavior change. The most important misunderstanding about patient empowerment is the idea that the patients need to be motivated, con-vinced and/or persuaded to be empowered. Anderson et al. state that patient empowerment does not include doing something to the patient. It does, how-ever, include possibilities and tools for the patient in order to help the patient to reflect on their knowledge and experience of the disease. Four fundamental components to the process of patient empowerment are mentioned in literature: Understanding the patients’ role, the patients should acquire sufficient knowl-edge to communicate with the healthcare provider, patient skills, and enabling a facilitating environment [6]. The World Health Organization (WHO) defines patient empowerment as: "A process in which patients understand their role, are given the knowledge and skills by their health-care provider to perform a task in an environment that recognizes community and cultural differences and encourages patient participation." [7]

Shared decision making

Shared decision making is an approach where clinicians and patients share the best available evidence when faced with the task of making decisions, and where patients are supported to consider options, to achieve informed preferences [8]. Various benefits for shared decision making were found: Patients 1) learn about their health, 2) recognize that decisions need to be made, 3) understand the pros and cons, 4) have the information and tools, 5) are better prepared to talk, 6) collaborate with their healthcare team, 7) and are more likely to follow through on their decision. Patient autonomy should be realized in order to en-able shared decision making. Patient autonomy is the ability for the patient to make his or her own decisions. The goal is to incorporate patient preferences in the patients’ treatment. Shared decision making can be measured through qualitative research methods. For example, patients can be asked if they feel informed and empowered [9]. Various decision aid tools are available that can help by the shared decision making process. Examples of decision aid varia-tions are web-based tools, videos, and pamphlets [10]. Shared decision making is a key component of patient-centered healthcare according to the National Learning Consortium of the United States of America [11]. The concept of patient-centered healthcare will be discussed below.

(14)

Patient-centered healthcare

Patient-centered healthcare is slightly different from the concepts discussed above. Patient-centered healthcare is care that is tailored to an individual pa-tient. The Institute of Medicine defines patient-centered healthcare as: "Provid-ing care that is respectful of, and responsive to, individual patient preferences, needs and values, and ensuring that patient values guide all clinical decisions [12]." Patient-centered healthcare is not just about tailoring the medical di-agnoses to the patients’ wishes. Additionally, communication and relationships between the patients and the caregiver are important. Emotional, mental, social and financial perspectives should be taken in to account. The Picker institute conducted research on the meaning of patient-centered healthcare. Eight di-mensions of patient-centered healthcare were reported as a result from focus groups that were part of the Picker institute’s research [13]. The eight dimen-sions are: 1) respect for patients’ values, needs and preferences, 2) information and education, 3) access to care, 4) emotional support, 5) family and friends involvement, 6) a continuing and secure transition between healthcare settings, 7) physical comfort, and 8) coordination of care. Rathert et al. conducted a sys-tematic review on patient-centered healthcare and outcomes. They concluded that there was stronger evidence for a positive influence on self-management and satisfaction for patients who were treated according to a patient-centered treatment [14]. Davis et al. predict modern medicine to be patient-centered [15].

2.2 The Dutch healthcare system

The Dutch healthcare system is based on the Bismarck model [16]. This means that the Dutch healthcare system is based on insurance and competition. The Dutch healthcare can be roughly divided into three groups: Primary, secondary and tertiary care. The primary care is accessible for every citizen. It consists of the general practitioner (GP), the dentist, physiotherapist, pharmacist, home nursing, and social services. The GP has a central role in the primary care. The GP acts as a gatekeeper: The GP can refer a patient to secondary or tertiary care, with the goal to reduce unnecessary costs. Secondary care is specialized care. The specialized care can be provided in hospitals and clinics. Tertiary care is highly specialized care that can be provided when the primary and secondary care are not sufficient. Academical hospitals and highly specialized clinics provide tertiary care.

The Dutch healthcare system is financed by Dutch citizens through premiums, and the Dutch government [16]. All Dutch citizens are obliged to have a healthcare insurance. The insurances are provided by insurance companies who have the task to keep the healthcare quality high. Additionally, competition between insurance companies should reduce the healthcare cost. However, healthcare costs are increasing drastically throughout the years [17]. In 1998, 10,4% of the gross domestic product was spent on healthcare. This increased to 13,8 in 2016.

(15)

However, according to the Euro Health Consumer Index (EHCI), the Dutch healthcare system has proven to be effective compared to other western coun-tries [18]. The EHCI ranks health systems in various councoun-tries based on 48 indicators. The indicators are divided into six subdisciplines: patient rights and information, accessibility, outcomes, range and reach of services provided, prevention, and pharmaceuticals. The Dutch healthcare system has been in the top three of the EHCI since 2005 [18].

2.3 Patient portals

Patient portals are applications that allow patients to interact and communicate with their healthcare institution and provider. Patient portals are mostly applications that enable the interaction and communication via the Internet. The common functionality for most patient portals is to provide health infor-mation to the patient. The most common functionalities in existing patient portals allow patients to renew prescriptions, make appointments, view generic and/or personal health information, and send messages in a secure environment [19, 20]. However, not only patients have benefit from using patient portals. Various other stakeholders can have benefit from using the patient portal. The Dutch association for hospitals (NVZ) published a vision document concerning Dutch healthcare in 2020 (Dutch: [21]). Two chapters in this vision document discuss making patients partners in their healthcare, and using ICT systems more intensively in healthcare. An acceleration program (VIPP) was set up as a result. This acceleration program has two main subjects: patient and information, and patient and medication. The first goal for patient and information topic is to make it possible for every patient to download his or her medical data before the first of July 2018. Secondly, before 31 December 2019, every patient must be able to access a patient portal in which the patient can upload medical data. For patient and medication, every healthcare provider should be able to access a digital and up-to-date medication overview of their patients before the first of July 2018. Secondly, medication prescriptions should be digitalized before 31 December 2019. Lastly, the healthcare provider must be able to provide a digital medication overview to the patient when dismissing the patient. This means that a huge (financial) boost is provided in order to create and implement patient portals in the Netherlands.

More specifically, the topic "patient and information" is split up into three modules: A1, A2, and A3. A1 states that the patient has to be able to download personal medical information. A2 states that a secured patient portal has to be made available wherein patients 1) can view his personal medical data, 2) can download the data in a structured way, and 3) can share his personal medical data with other health institutions. A3 states that three out of five goals should be reached. The goals are: 1) A functionality to add output from an eHealth applications with a use of five percent, 2) a patient portal use of 25 percent of the patients, 3) implementation to a data exchange standard (MedMij (Dutch: [22]), 4) provide an overview of care providers who accessed the patient’s medical record from the last 180 days, and 5) provide an overview of the patient’s medication known to the hospital in which the patient can request medication and request a change to the medication. Next, the patient

(16)

and medication topic consists of two modules: B1 and B2. B1 states that the hospital must be able to show an up-to-date overview of the medication as a part of the medication process in clinical and out-patient setting. B2 states that 30 percent of the prescribed medication recipes must be online sent. Lastly, at patient dismissal, a medication guideline should be provided according to the current standards (Dutch: [23]). Table 2.1 shows the specific norms of the VIPP.

(17)

Table 2.1: The VIPP requirements Module Norm

A1

• A patient must be able to download: – Elements of the basic care data set – Letters from specialists

– Discharge letter

– Laboratory and radiology results

– Used implants of a pacemaker, hip or knee prosthesis, breast implant or pelvic implants • The digital available information should be readable on a computer

• A hospital-wide procedure for the processing of digital requests should be made available to the patient • The download should be made available within three workdays after the download request

• The download should be made available for every patient who requests the download A2

• The basic care data set structurized according to a Nictiz standard • Dismissal letter

• Letters from specialists • Progress letters

• Used implants

• Medical data is provided through a secured patient portal • Medical data is downloadable in the patient portal • Medical data is available after seven days

• At least ten percent of the patients who had (diagnosis and treatment relevant) contact with a hospital should have logged in in the last 30 days

A3

• The patient portal has the functionality to add output from an eHealth application

• The number of unique logins is at least 5 percent of the patient population who use eHealth interventions in the last 30 days

• 25 percent of the patient population have to be logged in in the last 30 days • The MedMij standard is implemented and in use

• The patient portal shows an overview of care providers who accessed the patient’s medical record from the last 180 days

• An overview of the patient’s medication should be available. Additionally, sending change requests should be possible in the patient portal

B1

• The requested information on medication is at least 70 percent of all clinical patients in the last 30 days • The requested information on medication is at least 25 percent of all outpatient in the last 30 days B2

• The standard for electronic presciption is implemented in the EHR

• The number of electronically sent prescriptions is 30 percent of all prescriptions at minimum • Up-to-date medication overviews are standardized according to medication guidelines • The standard for medication appointments is implemented in the EHR

• The total number of medication appointments that is registered in the last 30 days is at least 80 percent of the number of medication appointments scheduled at dismissal

(18)

2.4 Data warehousing and business intelligence

The principle of data warehousing is bringing all sorts of data together in one system. The reason for this is to analyze this data. The main difference between data warehouses and databases is the use of the data. For databases, it is important to provide a basis to make systems that make many transactions run efficiently. This database type is a so-called online transaction processing (OLTP) database. Data warehouses are built for analytics. This form is called an online analytical processing (OLAP) database. It is optimized to support analytical queries. Data from a source system is transferred to a data warehouse. This is done in order to store history and in order to reduce the necessary processing power on the source systems to do the analyses. Setting up data warehouses is costly, it is therefore of utmost importance that the data warehouse gets used to its full potential in order to create profit. This can be done through business intelligence (BI). BI is the act of transforming data to in-formation. Originally, BI is used for analyzing data on customers, competition, economic statuses, and trends. The goal is to generate information from this data to support the decision making process within companies. Company man-agement used to ask questions such as "What has happened?". These questions can often be answered when the data warehouse has been implemented cor-rectly. This, however, is not the full potential of the data warehouse. Questions such as: "Why has this happened?" or "What will happen?" can be attempted to be answered using data warehousing. This requires a significant amount of knowledge and skill of the BI employees. Customer behavior, preferences, and value can be understood by analyzing data. According to Watson et al., acquiring the data is the most challenging aspect of BI [24]. This is because of political issues, poor data quality, and legacy technology. Additionally, a clear understanding of the assumptions that are made during the extract, transform and load (ETL) phases is of utmost importance. Metadata can help to keep these assumptions clear. Metadata is data that describes characteristics of data. One common structure that is frequently used for data warehousing is Data Vault. This structure is created by Dan Linstedt and it is based on the Capability Maturity Model Integration(SEI/CMMI) level 5 best practices which is the highest level of appraisal for organization maturity [25]. According to Dan Linstedt, Data Vault is a detail-oriented, history-tracing and uniquely linked collection of normalized tables. These tables support multiple functional business domains. It is a hybrid approach that combines the third normal form (3NF) and dimensional modelling. 3NF is a way of normalizing a database to reduce duplication of data and to ensure data integrity. Normalization is organizing database columns and rows in order to reduce data redundancy and in order to improve data integrity. Dimensional modelling is a design technique for databases that surround a fact database table by dimension tables in a star scheme based on the Kimball approach [26]. However, the principle of dimensional modelling has not been discovered by a single person. General Mills and the Dartmouth College used the terms "dimensions" and "facts" for the first time in the 1960’s. Next, AC Nielsen invented the concept of conformed dimensions. The goal of dimensional modelling is to enable the possibility to answer questions for end users. These dimensional models are called data

(19)

marts. The Data Vault model consists of three defined components: Hubs, Links and Satellites. Hubs are database tables that represent a business entity. Links are database tables that model relations between hubs. Lastly, satellites are database tables that contain relevant data from the business entities. The strong points of Data Vault are: 1) History can be easily stored and used for analyses; 2) Adding components to the model can be easily realized; 3) Data Vault enables real-time data loading. The most prominent weak point is the possibility of the rapid proliferation of database tables. Therefore, performance tuning of queries is absolutely necessary.

It is difficult to find structural advantages and disadvantages of data warehouse environments. This is, because the most advantages and disadvantages lie in the way a data warehouse is realized within an environment. However, for every data warehouse implementation, various things have to be kept in mind. Firstly, the data quality. The data warehouse will be ineffective when the data that is being stored is of low quality, even when an appropriate way of implementing a data warehouse is chosen in a certain environment. Another determinant is the data itself that is being stored. The data warehouse will be ineffective when the wrong data is being stored. No useful business intelligence practices can be done with the wrong data. Another important determinant is data complete-ness. Of course, analyses can be done on incomplete data. However, this will sketch a skewed and wrong image of the actual situation. Still, various trade-offs have to be made even if the data is correct, complete, and of high quality [27]. Firstly, the data to be stored should be decided upon. The decision has effect on the time it might take to clean up data sets or the expenses that have to be done to keep the quality of the data high. Secondly, timeliness is a topic that should be decided upon. Higher costs will be expected when real-time analyses are to be realized. Additionally, choices on which parts of the data should be available in real-time should be made. Data warehousing can be used for decision making, reporting, and operations. The data warehouse structure will be more complicated when more than one of the uses of data warehousing is chosen [27]. Next, semantics can introduce interpretation errors. An exam-ple when semantics can be a problem is when data about an identical topic is measured and stored in various ways. Various data warehouse strategies can be chosen when the topics above are taken into account. One example is the choice between a star or snowflake schema. When a star schema is chosen, the dimension tables are not normalized. Disadvantages of non-normalized tables are: Higher chance of data integrity problems, and more space will be used when storing these non-normalized tables. This is, because it is more likely to have more redundant data in non-normalized tables. However, the query per-formance is higher, because fewer joins have to be made than when a snowflake schema is chosen. Additionally, the queries that are written will be easier to interpret. Next, the time to load data into the schema is lower in a star schema. Lastly, more efficient navigation is realized when using a star schema. This is, because every dimension is joined to one fact table. This makes the joins more representable for what the goals of the join are: A join with one fact and one dimension table can already create useful information. As mentioned earlier, the advantage of using a snowflake schema is the fact that the tables can be normalized. This will reduce redundancy and lower the chance of having data integrity problems. Additionally, the snowflake schema’s structure has a higher

(20)

flexibility: Dimension tables can be adapted more easily without changing the whole structure. Those changes will have a lower impact than changes that are done in a star schema.

2.5 Scoping review versus systematic review

A scoping review will be conducted in this thesis. A scoping review differs in various concepts from systematic reviews: Systematic reviews are based on a systematic and transparent approach in order to define a research question. Additionally, a search for studies, a quality assessment, and an attempt on summarizing findings quantitatively or qualitatively is done when conducting a systematic review. A clear definition of the scope of the research has to be set. Therefore, in order to conduct a high-quality systematic review, a clear view of the uncertainties, and (gaps in) knowledge on the research topic is obligatory in order to conduct a systematic review [28]. A useful way to collect, assess, and organize important background data is by using the scoping review tech-nique if no clear view of the research topic is present. The goal is to acquire knowledge on the research topic that is being studied. Scoping reviews can be published on their own or they can act as a foundation of knowledge for future research that will be done in that specific research area. In this thesis, a scoping review framework created by Arksey and O’Malley will be used [29]. Addition-ally, recommendations on this framework will be used where possible [30]. This framework is originally created in order to identify gaps in existing literature. However, this framework can also be used to summarize and disseminate re-search findings. The framework consists of six steps. These steps are shown in Table 2.2. Note that the study selection step comes after identifying relevant studies. This enables the researcher to get an idea of the available evidence first. Afterwards, the researcher can narrow the scope. Step six is an optional step. This step is added, because Arksey and O’Malley gathered valuable information while discussing their findings with practitioners and consumers.

Table 2.2: Arksey and O’Malley’s scoping review framework # Step name Step description

1 Identifying research question Maintain a wide approach to generate breadth of coverage 2 Identifying relevant studies Decide on the coverage and databases

3 Study selection Set (post-hoc) inclusion and exclusion criteria

4 Charting the data Chart general information and specific information of the studies

5 Collating, summarizing and reporting results

Report data related to research question

6 Optional consultation This optional step consists of including practitioners and consumers in order to discuss the findings

(21)

2.6 Study setting

The information needs will be answered using the data that has been provided by a Dutch hospital. The data is stored in a database that stores both data from the (patient) portal platform (Zorgportaal) and from the hospital information system (HiX). Normally, this data is then prepared by making various pre-calculations. Thereafter, the data is being extracted, transformed and loaded in a raw data vault and a business vault. Note that the raw data vault includes data with minimal transformations. The only transformations that have been done are the transformations within HiX. Next, the data is moved into data marts. These data marts are a collection of fact tables and dimension tables that are created for a specific information need or wish. An example is: "how many interventions where done in the last year at my hospital?". The data marts are star schemas. The dimension tables are set to be conformed dimension tables. This means that type two dimensional techniques apply. Type two dimensional techniques are focused on managing history by not overwriting changed values in the database, but creating so-called time slices of data. Information that is needed to answer some of the end-users’ questions is stored in the data marts. Then, these data marts are used to generate visualizations of the information through cubes and or reports. Cubes are used when the user should be able to adapt the input forms of the data. Reports are mostly standardized visualiza-tions of predetermined inputs. No data marts are available for analyzing data from Zorgportaal. Therefore, most queries will firstly be run on the raw data vault. Thereafter, possible data marts will be created in order to be able to analyze data from Zorgportaal.

(22)

References

1. Longtin, Y. et al. Patient participation: current knowledge and applicability to patient safety in Mayo Clinic Proceedings 85 (2010), 53–62.

2. Street Jr, R. L. & Voigt, B. Patient participation in deciding breast cancer treatment and subsequent quality of life. Medical Decision Making 17, 298–306 (1997).

3. Bodenheimer, T., Lorig, K., Holman, H. & Grumbach, K. Patient self-management of chronic disease in primary care. Jama 288, 2469–2475 (2002).

4. Holman, H. & Lorig, K. Patient self-management: a key to effectiveness and efficiency in care of chronic disease. Public health reports 119, 239– 243 (2004).

5. Anderson, R. M. & Funnell, M. M. Patient empowerment: myths and mis-conceptions. Patient education and counseling 79, 277–282 (2010). 6. Angelmar, R. & Berman, P. C. Patient empowerment and efficient health

outcomes. Financing sustainable healthcare in Europe: new approaches for new outcomes 1, 3 (2007).

7. World Health Organization. WHO Guidelines on Hand Hygiene in Health Care: First Global Patient Safety Challenge Clean Care Is Safer Care; 2, Patient empowerment and health care. https://www.ncbi.nlm.nih.gov/ books/NBK144022/. [Online; accessed 13-11-2017]. 2009.

8. Elwyn, G. et al. Implementing shared decision making in the NHS (2010). 9. Ting, H. H., Brito, J. P. & Montori, V. M. Shared decision making.

Cir-culation: Cardiovascular Quality and Outcomes 7, 323–327 (2014). 10. Spatz, E. S. & Spertus, J. A. Shared Decision Making. Circulation:

Car-diovascular Quality and Outcomes 5, e75–e77 (2012).

11. National Learning Consortium of the United States of America. Shared decision making https : / / www . healthit . gov / sites / default / files / nlc _ shared_decision_making_fact_sheet.pdf. [Online; accessed 10-11-2017]. 2013.

12. Baker, A. Crossing the quality chasm: a new health system for the 21st century. BMJ: British Medical Journal 323, 1192 (2001).

13. Gerteis, M., Edgman-Levitan, S., Daley, J. & Delbanco, T. Through The Patient Eyes 1993.

14. Rathert, C., Wyrwich, M. D. & Boren, S. A. Patient-centered care and outcomes: a systematic review of the literature. Medical Care Research and Review 70, 351–379 (2013).

15. Davis, K., Schoenbaum, S. C. & Audet, A.-M. A 2020 vision of patient-centered primary care. Journal of general internal medicine 20, 953–957 (2005).

16. Daley, C., Gubb, J., Clarke, E. & Bidgood, E. Healthcare Systems: The Netherlands. London: Civitas Health Unit (2013).

(23)

17. CBS. Zorguitgaven - kerncijfers https : / / opendata . cbs . nl / statline / # / CBS/nl/dataset/83037NED/table?ts=1516095700784. [Online; accessed 16-01-2018]. 2018.

18. Bjornberg, A. Health Consumer Powerhouse - Euro Health Consumer In-dex 2016 https://healthpowerhouse.com/media/EHCI-2016/EHCI-2016-report.pdf. [Online; accessed 16-01-2018]. 2016.

19. Turvey, C. L. et al. Transfer of Information from Personal Health Records: A Survey of Veterans Using My HealtheVet. Telemedicine and e-Health 18, 109–114 (2012).

20. Ross, S. E. et al. Adoption and use of an online patient portal for diabetes (Diabetes-STAR) in AMIA annual symposium proceedings 2006 (2006), 1080.

21. NVZ. Zorg voor 2020 http : / / www . zorgvoor2020 . nl/. [Online; accessed 06-11-2017]. 2016.

22. MedMij. Welkom bij MedMij https://www.medmij.nl/. [Online; accessed 06-02-2018]. 2018.

23. Ministry of Health, Welfare and Sport. Handboek VIPP Eindtoets https: //www.rijksoverheid.nl/binaries/rijksoverheid/documenten/rapporten/ 2016/12/27/handboek-vipp-eindtoets/Handboek+VIPP+eindtoets+ter+ publicatie2.pdf. [Online; accessed 06-02-2018]. 2016.

24. Watson, H. J. & Wixom, B. H. The current state of business intelligence. Computer 40 (2007).

25. Inmon, W. H. & Linstedt, D. Data architecture: a primer for the data scientist: big data, data warehouse and data vault (Morgan Kaufmann, 2014).

26. Kimball, R. & Ross, M. The data warehouse toolkit: the complete guide to dimensional modeling (John Wiley & Sons, 2011).

27. Ballou, D. P. & Tayi, G. K. Enhancing data quality in data warehouse environments. Communications of the ACM 42, 73–78 (1999).

28. Armstrong, R., Hall, B. J., Doyle, J. & Waters, E. ’Scoping the scope’ of a cochrane review. Journal of Public Health 33, 147–150 (2011).

29. Arksey, H. & O’Malley, L. Scoping studies: towards a methodological framework. International journal of social research methodology 8, 19–32 (2005).

30. Levac, D., Colquhoun, H. & O’Brien, K. K. Scoping studies: advancing the methodology. Implementation Science 5, 69 (2010).

(24)

Chapter 3

Scoping review

3.1 Introduction

The availability of patient portals is steadily increasing throughout the years [1]. More data is stored about the use of patient portals. Therefore, more data can be analyzed and used in order to gain valuable insights into the use, effectiveness, efficacy, and statistics. Additionally, business intelligence questions could be answered with the data from patient portals, management questions can be discussed, and outcomes can be measured. This enables a new era of the use of data from patient portals in healthcare. It is still to be studied if addressing these topics is valuable.

Stakeholder analyses are processes that assess influential (groups of) people and organizations involved in any process. According to Johnson and Scholes, the definition is: "Those individuals or groups who depend on the organization to fulfill their own goals and on whom, in turn, the organization depends" [2]. It is important to know who your stakeholders are as a company in order to identify the level of power and interest those stakeholders have over your product or solution [3]. Nutt found that half of the 400 strategic decisions that were made failed because of poor or lack of stakeholder analyses. By analyzing your stakeholders closely, one can understand how decisions are taken by those stakeholders in a specific context. Stakeholder analysis aims to understand stakeholders from the perspective of an organization. Next, values will be determined for the stakeholders that have been identified. The main question will be: What is the value of this particular stakeholder? In other words: Why would the stakeholder use or be involved with a patient portal?

Information wishes and needs, but mostly functionalities can motivate and engage the stakeholders in using the patient portal. Note that not only the patient should be satisfied with the patient portal. Additionally, the other users such as caregivers, family members, and hospital management should feel that their wishes and needs are covered by the patient portal. Functionalities can enable the patient in accessing, viewing, downloading, entering or sending data or information within the patient portal. Additionally, data can be gathered on the use of these functionalities. This data can be used to answer certain

(25)

(statistical or prognostic) questions.

In 2017, a total of 28 patient portals are implemented in the Dutch hospitals [4]. The most common functionalities of patient portals in the Netherlands are shown in Appendix 1 [4]. The most common functionality of the Dutch patient portals is the functionality to access personal data. All Dutch patient portals (N = 28) provide this functionality. The functionality that provides the patient access to the admission and discharge letters is the second most common functionality. Additionally, access to medical data, laboratory results, appointment overviews, and forms and/or questionnaires are other common functionalities. The least common functionality is the input for self-monitoring and the ability to adapt and/or add data in the medication overview. In this chapter, a scoping review will be conducted in order to identify and in-terrelate 1) the current possible stakeholders for patient portals, 2) the available functionalities of patient portals, 3) the needs and wishes for information from those stakeholders.

3.2 Method

The scoping review framework created by Arksey and O’Malley will be used [5]. Additionally, recommendations made by Levac, Colquhoun, and O’Brien will be used where possible [6].

The MedLine database was sought. Additionally, Google Scholar was used to orient and find keywords for the search in the MedLine database. The scope of the scoping review is patient portals, stakeholders, and information needs. This also includes data usage and data warehousing. Search terms relating to ePHR, phr, patient portal, personal medical record, information need, data warehousing, functionalities, usage, level of use, and data analysis were used. The search queries are shown in Table 3.1. An article was included when it is written in English, and when it addresses information assessment of patient portals, stakeholders, or patient portal functionalities. Qualitative, descriptive, and quantitative studies of any type were included. After conducting the literature search, the duplicate studies were removed. Secondly, the articles were screened on title and abstract. Thirdly, the included articles based on title and abstract were fully screened. This process is shown in Figure 3.1.

(26)

Table 3.1: Search queries Query number Query

Query 1 (("Patient Portals"[Mesh]) AND "Data Collection"[Mesh]) NOT "Ethics"[Mesh]

Query 2 ("health records, personal"[MeSH] AND "Health information management"[MeSH] AND "Databases as Topic"[MeSH]) NOT "Ethics"[MeSH] NOT "Law Enforcement"[MeSH] NOT "Pri-vacy"[MeSH]

Query 3 ("patient portal"[TIAB] OR "patient portals"[TIAB] OR "Patient Portals"[Mesh] OR "PHR"[TIAB] OR "ePHR"[TIAB] OR "personal health record"[TIAB] OR "Patient Access to Records"[MeSH]) AND ("information needs"[TIAB] OR "health information technology"[TIAB] "Strategic information management"[TIAB] OR "Management information"[TIAB] OR "Business intelligence"[TIAB] OR "KPI"[TIAB] OR "Key performance indicator"[TIAB] OR "performance indicator"[TIAB] OR "data warehouse"[TIAB])

Query 4 ("Health Records, Personal"[MeSH] OR "PHR"[TIAB] OR "ePHR"[TIAB] OR "patient por-tal"[TIAB] OR "patient portals"[TIAB] OR "Patient Portals"[Mesh]) AND ("Functionali-ties"[TIAB] OR "Functionality"[TIAB] OR "Usage"[TIAB] OR "Use"[TIAB]) AND ("Expecta-tion"[TIAB] OR "Expectations"[TIAB] OR "Attitude"[TIAB] OR "Attitudes"[TIAB] OR "Per-ception"[TIAB] OR "Perceived"[TIAB] OR "Perceptions"[TIAB] OR "Determinants"[TIAB] OR "Determinant"[TIAB] OR "Satisfaction"[TIAB]) NOT ("environmentally"[TIAB] OR "Require-ment"[TIAB] OR "Requirements"[TIAB] OR "Communication"[TIAB] OR "Design"[TIAB] OR "Usability"[TIAB] OR "Literacy"[TIAB] OR "Behavior")

(27)

Study characteristics were gathered. The study characteristics consist of: the year the article was published, the country in which the research was done, the study design, the main objective, the measurements chosen, the outcomes, and lastly the type of effect. The type of effect can be negative, neutral or positive. Additionally, the information needs, patient portal functionalities, and possible stakeholders were noted while the articles were fully read.

The found information needs and patient portal functionalities will be clustered through concept mapping [7]. Microsoft Excel 2016 and R version 3.4.3 were used for the concept mapping process [8, 9]. This process consists of six steps: 1) Preparation, 2) Generation of statements, 3) Structuring of statements, 4) Representation of statements, 5) Interpretation of Maps, and 6) Utilization of Maps [7]. The preparation consists of selecting participants, and developing a focus. The focus was already set and decided on the literature research. The generation of statements has been done by the studies that have been included. This is because an implication or guess of the meaning of the concept introduces bias. Next, these statements were structured by the participants in groups that they seemed fit. Two lists of concepts were mapped: One for the functionalities, and one for the information needs. The structuring of the statements was done in Excel 2016 [8]. Next, analyses have been done in the representation of statements phase. First, matrices were created to see if two concepts were mapped in the same group by one participant. This has been done for all participants. Second, a summation of the matrices was established. Third, in R, a two-dimensional non-metric scaling technique was used in order to determine the distances between the elements in the matrix [10]. Fourth, the distances that were found through the scaling technique was used for (bottom-up or agglomerative) hierarchical clustering [11]. The elbow method, the silhouette method, and the gap statistic method were used to guide two researchers to determine the right number of clusters [12–14]. In short, the elbow method calculates a percentage of variance of the clusters. The elbow method determines whether the increase of information (decrease of variance) increases with a lower amount: The model does not get much better per extra cluster. The silhouette method looks at how closely the data can be matched to its determined cluster. Secondly, the silhouette method looks at how loosely the data is matched to its neighboring clusters. The gap statistic method simulates (by using bootstrap) various numbers of clusters on the dataset. Next, it calculates the variance within the clusters. Then, it compares this variance with a distribution with no logical or obvious clustering. The gap statistic is considered to be a more sophisticated method, because it provides a statistical procedure to formalize the elbow and the silhouette heuristic. A right number of clusters was determined by visually inspecting the result and by using the results from the three methods described above in order to get an idea on the number of clusters that are necessary. Therefore, these results were used as input for discussion on the right number of clusters. The clusters were named based on inspection of the clustered concepts. The interpretation and inspection of the clusters that were created by concept mapping was done by two researchers. This was done by schematically analyzing the concepts that change clusters during the analyses. The cluster analyses were done from clustering the concepts in one group up to 20 groups for both the information needs and functionalities.

(28)

The structured information need concepts and patient portal functionalities were used to create insights on the scopes of the published (and included) articles. In addition, SNOMED CT was used to create a hierarchy of the stakeholders that were identified in the review [15].

3.3 Results

The literature search identified 160 articles. A total of 128 articles were ex-cluded. The exclusion reasons are shown in Table 3.2. The most common reason for exclusion was the fact that the studies did not have patient portals in their scopes (N = 52). Furthermore, studies solely on prototyping and/or acceptance, studies that the author could not get access to, and studies without a clear research question, scope, or goal were excluded (N = 17, N = 12, and N = 11 respectively). Ten studies focused solely on communication between patient and caregiver. These articles did not specify that their study was on communication through patient portals. Therefore, these articles were excluded.

Table 3.2: Exclusion criteria

Exclusion criterium N

Studies not about patient portals 52 Studies solely on prototyping and/or acceptance 16

Unavailable studies 12

No proper research 11

Studies solely on communication 10 Studies solely on IT architecture 6 Studies solely on implementations 4 Studies solely on feasibility 4 Studies solely on usability 3

Non-English studies 3

Studies solely on portal usage 2

Studies solely on ethics 1

Studies solely on workload 1

Duplicates 1

Cost effectiveness/efficiency 1 Studies solely on a single outcome measure 1

Study characteristics

The study characteristics that were gathered are shown in Appendix 2 [16–47]. Seven studies were published in 2017. Additionally, 16 studies were published in 2016. Three studies were published in 2015. Next, six studies were published in 2014 or before. The earliest published study is from 2010. Eight studies used quantitative methods, whereas 18 studies used qualitative methods. Further-more, three articles used mixed methods. Lastly, one study described a protocol, one conducted a literature search, and one study described an implementation process of a new patient portal in their current software and hardware archi-tecture. 18 studies showed positive results, 13 studies reported neutral results,

(29)

and one study reported negative results towards patient portals. 62,50% (20 out of 32) of the studies were conducted in the United States of America (USA). Additionally, five studies were conducted in Europe. The other studies were conducted in Canada, Argentina, Australia, Iran, Japan, and Russia.

Stakeholders

A total of 35 stakeholder types were identified. The stakeholder types have been mapped using SNOMED CT concepts. The mapping is shown in Appendix 3. The arrows represent so-called "IS A" relationships. White concepts are concepts that have been identified using literature. Additionally, these stakeholder types were concepts in SNOMED CT. Blue concepts are identified stakeholders, but these stakeholder types were no SNOMED CT concepts. The blue concepts have been mapped where appropriate. Grey concepts are concepts that were not found in literature. However, these concepts are necessary in order to map other stakeholder types properly. A total of ten stakeholder types were identified under the SNOMED CT core concept "Person (person)" (Figure 6.2). Six out of ten stakeholder types were no patients, but family members, volunteers, government managers, elderly, and children. The four patient types that were identified were the older patient, the overall patient, the low health literate patient, and the maternity patient.

Figure 6.3 shows the stakeholder hierarchy for the SNOMED CT core concept "Social Context (social concept)". A total of 23 stakeholder types were iden-tified that are mapped in under this core concept. Three patient groups were mapped in this hierarchy (Hispanic patients, patients with a black ethnicity, and patient organizations and support groups). Most stakeholder types that have been identified in the literature are healthcare professionals (N = 12). Four stakeholder types are professional scientists/engineers/technologists, namely: laboratory technicians, vendors, BI consultants, and researchers. Two stakeholder types could not be mapped within the two proposed hierarchies: BI departments and health insurers. This is because BI departments are no social concepts or persons. Additionally, ’health insurers’ is a vague concept. Health insurers could be companies, governments, or actual individuals. As shown in Appendix 3, the most prominent stakeholders that were found in literature were patients and caregivers. Various stakeholders have not been studied extensively. Moreover, various stakeholders have not been studied at all. The managerial, administrative, and clerical stakeholders have been mentioned once.

Functionalities

Appendix 4 shows a list of all unique patient portal functionalities that have been identified in the literature. A total of 222 functionalities were identified. 124 of these 222 (55,86 %) functionalities were unique. The functionalities pro-vide the ability for the users to view, access, manage, edit, download, copy, add or adapt data within the patient portals. Various studies mentioned identical functionalities of patient portals. The number of times a functionality is men-tioned by different studies is shown in Table 3.3. The most often menmen-tioned

(30)

functionality is "Access to health information". This functionality has been mentioned by 12 out of 32 studies (37.5%). Next, appointment booking (N = 9), medication list viewing (N = 9), Communication with care provider (N = 7), and viewing laboratory results (N = 7) are mentioned most often. On average, an included article mentioned 6,9 functionalities (SD = 4,70). The fewest number of functionalities mentioned is 0 and the maximum number of functionalities mentioned is 16. 86 out of 124 (69,4%) functionalities have been mentioned once.

Table 3.3: Duplicate functionalities

Functionality Mentioned

by number of studies

Access to health information 12

View laboratory results 12

Appointment booking 9

View medication lists 9

Communicate with care provider 7

Requesting prescription refills 6

Secure messaging 6

Provide decision support 5

View allergies 4

View immunizations 4

Access to findings and medical images 3

Access to medical record 3

Accessing radiology reports 3

Educational materials 3

E-mail functionality 3

Portal based reminders 3

Register / write progress notes 3

Viewing radiology images 3

Access control 2

Appointment details 2

Communication with glucometers 2

Exchange medication information 2

Following referrals 2

Functions related to preventive care 2

Information delivery 2

Interoperability 2

Messaging 2

Notifications for physicians 2

Nutrition support 2

Overview of healthcare workers and their roles for the patient 2

Provide administrative information 2

Reminder E-mails 2

View Clinical notes 2

View demographic information 2

View information on patient’s condition 2

View patient history 2

(31)

Most functionalities that are available in the patient portals are specifically cre-ated for the patient. However, some functionalities have benefits for caregivers and other stakeholders as well. An example is the diary or self-monitoring functionality. These functionalities enable the patient to enter (medical) information into the patient portal. Then, this can be analyzed and used by the caregivers in order to increase the quality of the treatment.

Various differences were found when the Dutch patient portal functionalities were compared with the found functionalities through the literature search. E-consults, video E-consults, and teleconsultation is mentioned twice in the studies found during the literature search [41, 47]. This is different from the Dutch patient portal functionalities. 17 out of 28 Dutch patient portals provide func-tionalities concerning E-consultation. Additionally, the questionnaire or form functionality has been mentioned once in literature. 23 Dutch patient portals provide this functionality. Lastly, the functionality that enables the patient to enter (medical) data is available in only 4 Dutch patient portals. This function-ality has been mentioned 8 times in the literature.

Concept Mapping

Seven participants were sought to systematize the functionalities that were identified in the included literature. The goodness of fit is considered to be fair for the functionality clustering (Stress = 0.1030) [10]. Figure 3.2 shows that the elbow method determined that four clusters were optimal whereas the silhouette model determined it was two clusters, and the gap statistic method determined it was seven clusters. The total number of clusters that is needed is 6. That means that the gap statistic method was closest to the researchers’ interpretation of the clusters. The final functionality clustering is shown in Appendix 6. A total of 13 clusters were necessary in order to cluster all functionality concepts appropriately to their contexts. The strongest cluster was determined at the cluster analysis that created two groups. All participants clustered these concepts in the same group. This cluster is called: Access to medical record. The patient-entered information cluster was determined after the six-cluster analysis. Next, the reminders cluster was de-termined after ten-cluster analysis. The administration cluster was dede-termined at the cluster analysis that created 11 groups. Lastly, the communication and self management clusters were determined at the 13 cluster analysis. The biggest cluster consists of 41 concepts whereas the smallest cluster con-sists of seven concepts. The average cluster contains 20.5 concepts (SD = 10.78). Figure 3.3 shows the clustering determined by machine learning (left) and the proposed clustering by the researchers (right). Note that the position of the concepts on the plane did not (and should not) change. Near or full agree-ment was found for the clusters: Access to medical record and patient-entered information. The researchers changed the other determined clusters based on context.

(32)

Figure 3.2: Cluster analytics for the functionalities

Figure 3.3: Functionality: Concept clustering

Information needs

Appendix 5 shows a list of all unique information needs that have been found in the literature. In total, 192 information needs were found. 129 of these 192 (67,2%) are unique information needs. The majority of information needs that were found came from caregivers. Table 3.4 shows how often an informa-tion need is meninforma-tioned. Informainforma-tion needs on clinical outcomes are meninforma-tioned most often (N = 9). These clinical outcomes are requested by both patients and healthcare professional. Next, patient demographics are mentioned 6 times. The need for information about patient demographics was expressed by healthcare professionals. These patient demographics can be used for (scientific) research. Additionally, it provides information about what type of patient uses the pa-tient portals. Next, information on administrative efficiencies are requested 6

(33)

times. These administrative efficiencies are based on concerns from healthcare professionals. Various groups of healthcare professionals have concerns on the potential increase of their (administrative) workload. The information need about cost reduction has been found 5 times. 95 out of 129 (73,6%) informa-tion needs that have been found have not been meninforma-tioned multiple times. On average, an included article mentioned 6 information needs (SD = 4,30). The range of the number of the mentioned information needs is between 0 and 18.

Table 3.4: Duplicate Information Needs

Information need Mentioned

by number of studies Clinical outcomes 9 Patient demographics 6 Administrative efficiencies 6 Cost reduction 5 Number of logins 4 Quality of care 4

Change in workflow for physicians 4

Proportion of patients using the prescription reauthorization 3

Patient-physician interaction 3

Provider workload 3

Adoption rate of patient portal 2

Effect of physiological measures 2

Effect of patient portal on treatment plan 2

Frequency of use 2

Improvement in quality of care 2

Medical error reduction 2

Medical research 2

Medication adherence 2

Number of days used (last date - first date) 2

Patient characteristics 2

Patient education level 2

Patient portal satisfaction 2

Association between treatment and time to readmission 2

Patient safety 2

Portal usage 2

Proportion of patients using the messaging functionality 2 Extracting unstructured clinical note information 2

Prognosis 2

Quality of life 2

Security of patient portal 2

Disease progression 2

Treatment plan information 2

Optimizing performance 2

Data warehousing in healthcare: assessing information needs concerning patient portals