An analytical capability roadmap leading to value-based healthcare

An analytical capability roadmap leading to

value-based healthcare

Master Thesis - MSc Business Studies

Aug 30, 2015 Arne Waalkens Student nr: 5699630

Supervisor: E. Peelen

1. Contents

2. Introduction... 6

3. Problem statement ... 9

4. Theoretical background ... 10

4.1. Why providing value in healthcare ... 11

4.2. Definition of customer value ... 12

4.3. Valuebased healthcare ... 13

4.4. The origin of data analytics ... 16

4.4.1. Business Analytics and Intelligence ... 17

4.4.2. Healthcare value through data analytics ... 20

4.4.3. Defining analytical maturity ... 22

4.5. Maturity models ... 23

4.6. Shortcomings of BI maturity models ... 23

4.7. Delta model ... 24

4.8. Evaluation of analytic models ... 27

4.9. Change management ... 30

4.10. Conclusion ... 32

5. Development of a maturity model, results and research design ... 33

5.1. Maturity and dimensions ... 33

5.2. New model specifics ... 34

5.2.1. Involving physicians ... 35 6. Research method ... 37 6.1. Research problem... 37 6.2. Research approach ... 37 6.3. Research data ... 38 6.4. Methods ... 39 6.5. Results ... 39 6.6. Empirical findings ... 39 6.6.1. Hospital description ... 40 Hospital 1 ... 41 Hospital 2 ... 46 Hospital 3: ... 50 Hospital 4: ... 53

7. Final maturity model ... 56

7.1. Perspective through D.E.L.T.A stages ... 58

7.2.1. Summarizing analytic maturity plot... 64

8. Discussion ... 65

8.1. Relevance ... 67

8.2. Limitations ... 68

Appendix ... 69

Meetbaar beter initiative ... 70

TDWI’s Business Intelligence Maturity Model... 71

Gartner’s Maturity Model for Business Intelligence and Performance Management ... 73

Preface

With this Master Thesis I conclude the research I have conducted being a Master Business Study student at the University of Amsterdam and an employee at Accenture. There have been many hours, days, weeks and months of hard work poured into the finishing these theses, and with pride I am writing the final sentences.

There have been many people that have contributed and helped me in various ways during this process and without whom I would not have completed my thesis. A view I would like to mention specially.

First and foremost my thank and appreciation for the hospital employees that made them self-available for discussion and interviews allowing me to gather the crucial material to test my model and gain deeper insight into the status of data analytics within hospitals.

Secondly I want to thank my mentor at the UvA for the commitment and valuable guidance during the writing process and ensuring the academic quality of this paper.

There are many more that have contributed with time and energy and I want to thank them all for their involvement and introducing me into a new and developing domain of healthcare. I hope you enjoy reading my thesis,

Thesis outline

This thesis aims to provide insight into how hospitals develop analytic maturity in order provide value-based healthcare. In this thesis a maturity model will be designed to map the current analytical maturity of hospitals and provide insight in the current analytic maturity status, and create an overview of the next steps in the transition process of the hospital organization towards the envisioned maturity end state.

Firstly an introduction to the increasing popularity and usage of data analytics is given, followed by examples of application in the healthcare market and the relation to specific challenges that the healthcare market is facing. This leads to the main problem in chapter 3 and research questions that are followed by an extensive literature review in chapter 4 to provide the necessary context around the main topic and provide the theoretical foundation of the model that is proposed. This model is subsequently applied within four case studies on each element of the model (Data, Enterprise, Leadership, Target, Analysts and Physicians) in chapter 5. In chapter 6 the research method is discussed and the results of the case study and proposed maturity model are described in chapter 7. Lastly the key recommendations are given in chapter 8 and limitations and future research suggestions identified.

2. Introduction

It’s a digital world according to Gartner (2010) and with the increasing technological advances that are the prime enablers behind new digital business opportunities the world has seen, there is an explosive growth in the availability of accessible data and data technology. Their disruptive power stems from merging virtual and physical worlds, the growth of intelligence everywhere and the emerging new realities of IT (Gartner IT technology trends 2015). The new opportunities generated by the Data Revolution help us gain new insights in, and understanding of processes and organizations, new understanding of hidden values. It also brings a new perspective to excising and new challenges (Chen, 2014). This development has gained considerable attention from industries, governments, pubic media, and scientific journals (Davenport, 2010; McAfee, 2012; Lohr, 2012; Chen 2012).

The importance and relevance of big data is beyond all doubt where (Gartner, 2010) stated, “Information will be the 21th Century oil” and big data is seen as means for strategic competitive advantage (Fisher; 2009). Although big data has been around for several of years with the build up of data from sensors, phones, televisions, public infrastructure like roads, buildings, utilities produces like water and gas meters have been available and flowing for many years. The change however is not the quantity of data but that we only recently are able to do something with it, that’s the big data revolution. (Mayer-Schönberger, 2013).

Business leaders and governments are looking for ways to identify new opportunities for growth of business or creation of jobs. With the emerging of data analytics a new key source of opportunities has gained attention (GITR, 2014). Companies like Google, Netflix and Amazon make use of big data provides to gain insights into their customers, the market and internal operation by developing algorithms that can predict the type movies or books a customer would be interested in. They do so by combining navigation data and purchase history (Mayer-Schönberger, 2013; Netflix, 2015).

Traditionally the private sector has been a front-runner in the field of analytics use for functional or strategic purposes (Chen, 2014; Murdoch, 2013). A contributing factor is the availability of data and experience of the market in using data in functional and strategic application. Analytics has gained a lot of traction within companies that are known for high availability of data such as telecommunication, financial and insurance services.

This is the opposite of the healthcare sector where despite the availability of (medical) data, the investments for analytics have been marginal (Raghupathi & Raghupathi, 2014; Murdoch, 2013). According to Groves (2013) hospitals have underinvested in the infrastructure that supports analytics and big data despite the abundance of medical data, which is generated from medical procedures and are stored in EMR’s.

However, changes are being noticed and the overall healthcare market trends are driving the push for insight driven health. There are several contributing factors to this change:

First the increase of data availability caused by the digitalization of medical records (Kiers, 2010). The adoption of Electronic Medical Records (EMR) created an abundance of digital medical data and provides an interesting data opportunity for hospitals to gain deeper insights into the healthcare business processes, treatment cost, quality and effectiveness.

Second, advanced technologies such as easy access to high CPU power, machine learning; data integration and cloud storage enable vast amounts of (real-time) data to be processed, shared and analysed (Gartner 2014).

Third, the rules and regulations of the Dutch government are stimulating healthcare reform through policies, create a drive for data standards, sharing of performance and data force cost mandates. Hospitals in the Netherlands are obligated by law to share performance data on healthcare treatments with the Dutch Quality Institute for the purpose of providing transparency and insight for healthcare consumers on quality of care. This forces hospitals to create insights into their healthcare processes and use the available data to comply with the quality measurement standards. The outcomes are not only used by consumers, but also by hospitals as competitive market information between hospitals.

Lastly the changes in demand from the customers and stakeholders force hospitals to adapt and change their strategy and operational processes. On the one hand patients are demanding higher quality of care, insight into treatment value and outcomes related to quality of life of the patient and more accountability from physicians (Cortada, 2013). At the same time healthcare insurance companies demand transparency on costs and treatment outcome. The shifting demographic combined with the medical advances resulting in an expansion of treatment options and an increase of lifespan.

With the increase of life expectancy the cost of healthcare having skyrocketed and further growing each year, the Dutch the health expenditure in 2014 was 14,5% of the GDP and rose 1,8% to 95 billion euros (CBS, 2015). The expectation is that with the increasing demand for healthcare caused by the aging population, the budget required to provide the expected level of healthcare will be continuously expanding (Berwick, 2008)

The way hospital organizations operate has changed significantly within the increasingly more complex market. The traditional view on healthcare delivery is no longer valid as it is more focused on volume based and non-collaborative interaction between insurance payers and care providers instead of focusing on treatment effectiveness and value (Groves, 2013). This shift towards a value based healthcare system requires hospitals to have in-depth knowledge of their current performance and operation.

Several scholars have underlined the need to shift from volume to a value-based healthcare system (Porter, 2010; Groves, 2013; Brown 2003; Cortada, 2013). The need for insights on performance, treatment effectiveness, quality and patient satisfaction support the insight driven health trend and enable hospitals to transform and innovate their operation towards a health and sustainable future (Raghupathi &Raghupathi, 2014).

Healthcare data contain facts on economic performance as well as all treatment activities and when analysed and processed could provide valuable information for operational, tactical and strategic purposes. According to the Gartner Hype cycle (Gartner, 2013), Analytics is only recently expected to reach the “plateau of productivity”, meaning that analytics has yet to reach full maturity but the potential has been recognized.

Scholars (Groves, 2013; Cortada, 2013; Raghupathi &Raghupathi, 2014), leaders in healthcare and consulting firms (IBM, 2014; SAS, 2015, have recognized the potential contribution of data analytics towards an insight driven healthcare system. This thesis aims to provide further insight into the benefits and potential application of data analytics within hospitals.

3. Problem statement

Although the importance of data has been recognized in the literature, health analytics is still in an emerging state (Raghupathi &Raghupathi, 2014) and many healthcare organizations have not yet implemented BI systems (Hanson, 2011).

In the literature there is a considerable gap in the application and implementation of data analytics in healthcare. In this thesis an extension to the literature will be made based on scientific research on analytic maturity and development phases of data analytics.

The gradual process of increasing the maturity of analytics requires more than the availability of data; it requires integration in the decision-making process, analytical capabilities, a well-defined outcome measurement framework and complete organization support (Davenport, 2010). The use of maturity models in the Business Intelligence field is not uncommon. Although there has been limited research on the factors that contribute to successful

implementation of BI in the healthcare domain (Foshay, 2013), some authors have recognized that the excising models lack industry depth (Chuah & Wong, 2011; Rajteric, 2010;

Lahrmann, 2010).

The applications of data analytics in healthcare are numerous; from quality and process optimization to cost control and fraud detection (Groves, 2013; Raghupathi & Raghupathi, 2014). As previously identified, prior research (Hanson, 2011; Chuah & Wong, 2011; Rajteric, 2010; Foshay, 2013; Lahrmann, 2010) suggests that current excising maturity models lack industry depth and are not aligned with the specifics of an healthcare organization which leaves potential unused.

In order to successfully bring value to the patient’s through insights gained from data analytics, the analytical maturity of the organization influences the effectiveness of the healthcare organization in creating the insights (Davenport, 2010). The main purpose of data-analytics is the creation of value for the organization. Therefor the scope of this thesis is focused to providing value to the patients in a clinical setting.

The goal of this thesis is to provide the industry depth that is currently lacking for analytic models, and therefor the scope of the research on analytics maturity models is drawn around hospitals. To facilitate better alignment with the current developments in the Dutch healthcare

(Kwaliteitsinstituut Nederland, 2015), the focus of this thesis is aimed at defining how data analytics can be used in enabling value-based healthcare, as an answer for the increasing healthcare costs, and tension field between the call for transparency on quality, insight on value and cost effectiveness. The following research questions will function as the foundation of this thesis:

 How does a healthcare industry specific maturity look like?  How mature is the current use of analytics in hospitals?

 How can hospitals (gradually) develop analytic maturity for the sake of the volume to value transition?

In the Netherlands there are two types of hospitals, the academic hospitals and common (regional) hospitals. The difference between the two is that academic hospitals that provide next to the same care as a regional hospital are also university medical centre for scientific research and teaching institute for the development of new medical technologies or treatments (Kwaliteitsinstituut Nederland, 2015). The regional hospitals also have treatment centres for specific illnesses but are not generally a teaching hospital.

This theses focus on regional hospitals for two reasons. The first is that regional hospitals are currently participating in value-based initiatives (Meetbaar Beter, 2015) and are collaborating with other regional hospitals. The second reason is that regional hospitals do not have a history of using data for analysis as academic hospitals have with PhD studies and their role in providing and developing new treatment procedures. This combination of factors creates an interesting challenge for the regional hospital in which a maturity model could provide

guidance in the maturity growth process.

The definition interpretation of value is described in par 4.2. In this thesis a theoretical maturity model is developed for the purpose of plotting the analytical maturity status of hospitals and providing guidance and focus in the process of organizations that are aspiring analytical growth.

4. Theoretical background

4.1. Why providing value in healthcare

This thesis focuses on how analytics could add value to the transformation from a volume-based healthcare system towards a value-volume-based healthcare system. The term value is used in many different forms and shapes in the literature. Defining value for healthcare requires looking at the patient value and health industry value. The patients are becoming more organized and demanding higher level of quality and more personalized care. The healthcare industry on the other hand is demanding a better control of costs, which are steadily growing each year. This thesis focuses on how analytics could add value to the transformation from a volume-based healthcare system towards a value-based healthcare system. The value definition requires more elaboration since it is being applied in financial, human resource and many other domains (Wikstrom & Normann, 1994). How value needs to be interpreted and how it can be described within the outline of this thesis.

The governmental organization Zorginstituut Nederland falls under the responsibility of the Ministry of health, wellbeing and sport (Ministerie van VWS), and has the responsibility of improving the quality of care and to enable access to information on the quality of delivered care for the public. The Zorginstituut Nederland is responsible for facilitating discussions between care organizations on quality of care, how to define, standardize and measure quality definitions as well as sharing best practices.

The level of quality of care is determined by instruments, which consist of quality indicators and validated questionnaires. Measuring instruments bring systematic and reproducible results for a transparent image of the quality. Care providers are obligated to supply data based on the measuring instruments and this data is made public by the Zorginstituut Nederland, not only for research purposes but also for patients via a simplified version of the quality standards overview.

The data on quality standards are used by patients in the decision-making process when comparing the quality of care between different care providers. Health insurance organizations use it in their process of purchasing of care and contract negotiations with care institutions. By creating transparency on the quality of care not only for care institutions or healthcare insurance companies but also for patients, is a strong example of a patient-centric adaption or value-based view on healthcare.

Another task of the Zorginstituut Nederland is to advice the Ministry of VWS on the basic health insurance package selection. Determining the type of care that will be covered in the basic health insurance package. This is done on basis of evaluation of the factors of necessity, quality of life and (cost) effectiveness. Looking at the perceived effectiveness of treatment and possible cost while taking into account the quality of life aspect relates value to what (Zeithaml, 1988) describes “perceptions of what is received and what is given”. The evidence-based approach, which is rooted in the selection process, causes for regular public debate when specific medication is not covered due to lack of health contributing evidence. It has been a long standing discussion which arises from time to time in the media when specific medication is taken out of the standard insured care package and has to be paid by the patients themselves. Usually the medication is used by a small number of patients for treatment of rare diseases, which makes the medication expensive.

The value in this discussion touches on a fundamental right of a patient to receive the right care and the impact of the value-based care discussion, which creates a tension field because of the impact on patient lives. It is Porter who states, “Cost reduction, without regard to the outcomes achieved, is dangerous and self-defeating, leading to false “savings” and potentially limiting effective care. A focus on value, not just costs, avoids the fallacy of limiting treatments that are discretionary or expensive but truly effective” (Porter, 2010, p.2).

Focusing only on cost management and not quality has led to a narrow-minded efficiency drive in the healthcare domain. Cutting cost without driving for quality resulted in decisions made which were unhealthy for patients and medical staff (Elza, 2013). Porters argument is that the focus on value of care and not only on cost will provide a better solution to the value-based challenge by focusing effort on treatments that have proven to be effective and create patient value and cutting treatments which haven’t proven to be effective.

4.2. Definition of customer value

The concept of customer value has been described in the literature from many different angles but with major commonalities that stand out. For example, customer value is something that is perceived rather than determined by the seller and it focuses around the trade-off that the buyer receives and what he or she gives to acquire the product or service (Woodruff, 1997).

This trade-off plays a pivotal role in the research on perceived value (Zeithaml, 1988). Dodds (1991) describes the perceived value as a trade-off between perceived quality and perceived psychological and monetary sacrifice. From a patient perspective Porter defines value as the outcomes achieved per dollars spent (Porter, 2011).

Measuring the actual patient satisfaction on received healthcare quality is difficult because the patient is usually unable to grade or identify the related activities within the whole care process. The quality can therefor be deducted from indirect observable or measurable items by the patient. The tree main factors which are related to hospital care are described as the three A’s of healthcare by Chow (2009).

 Affability, or interpersonal manner, referring to smooth interaction and communication between the medical staff and patients,

 Accessibility and availability, referring to physical access, waiting lists, home visits, waiting time, etc.

 Ability, referring to the perceived technical quality of care and outcome.

For this thesis a combination of value definitions described in prior literature is used to define the customer value.

The customer value definition used in this thesis is the trade-off between the level of affability, accessibility and ability as benefits of service, and the needed financial or human resource investment to enable the delivery of the service.

4.3. Value based healthcare

It were Porter & Teisberg (2006) who described the ‘value-based healthcare system’ in their book “Redefining Health Care: Creating Value-based Competition on Results”. In their book the authors outline that value is not purely aimed at cost reduction, but is intended to achieve the best possible outcomes in an efficiently was a possible (Lee, 2010). Porter & Teisenberg under scribe that physicians are the main driver behind changes in healthcare reform and that the goal of the healthcare system is not to minimize costs, but to deliver value to patients.

The solution, according to the Porter & Teisenberg, is to introduce a competition framework in healthcare “based on patient value over the full cycle of care —from prevention and diagnosis through recovery or long-term disease management” (p3.,Porter, 2011). With the result that, “value should be the preeminent goal in the healthcare system, because it is what

ultimately matters for customers (patients) and unites the interests of all system actors” (p3. , Porter, 2011).

The core focus of maximizing value for patients is about achieving the best outcomes at the lowest cost (value-based healthcare). This is a completely different approach as to the supply-driven health care system which is a focuses on providing care to patients-needs regardless of the out-comes (patient-cantered system).

Improving the value for patients is the goal (or must be the goal) of healthcare providers. The patient value is defined as the trade-off between the level of affability, accessibility and ability as benefits of service, and the needed financial or human resource investment to enable the delivery of the service. Which is a different approach in the patient- cantered system where the treatment of the patient was the main focus, regardless of the outcomes or impact on the patients’ quality of life.

Value based healthcare stimulates a higher efficiency per euro spent on healthcare and focus the reimbursement system on the right incentives that are aimed at patient value (Porter, 2011). The value based healthcare models try to align the payment and objective measures of clinical quality.

The volume-based model that has dominated the healthcare provider landscape is Fee for Service (FFS). This model generally describes the arrangement how healthcare provider performs treatments or tests on a patient in return for payment from the health insurance or the patient self. It has been acknowledged that the FFS model stimulates the volume and intensity of healthcare service with the result of an increase of patient admissions, testing and treatments. This movement has been responsible for the increasing cost of healthcare the recent years. (Fischer, 2009; Berwick, 2008).

By transforming the reimbursement from FFS towards to a bundled Fee For the Care (FFC) of medical condition including treatment, testing and admission costs (Porter, 2011) and aligning provider payment with the value in a value based performance model, instead of providers being paid by the number of visits and tests they order (FFS), their payments are now based on the value of care they deliver (value-based care).

The operationalization of the new healthcare model, which relates to Porter’s value principal is the Triple Aim framework (Berwick, 2008). The triple aim framework has been adopted in policies by different governments and healthcare authorities like the United States of America and the Dutch National healthcare institute, as a means to get the rising healthcare cost under control. The triple aim framework focuses on simultaneous realization of improving the customer experience, improving healthcare of a defined population and lowering of the cost per capita. This is different from the current coefficient (epidemiological) paradigm in health care, where these three objectives are often pursued separately and monitored. In the current paradigm cost is being pursued, while the Triple Aim thinking is that by improving the perceived quality of care the other objectives will be achieved.

There are tree preconditions defined by Berwick (2008) for reaching Triple Aim status:

 Identification of a population. Described by Hardin, 1968 a “tragedy of commons” where the individual needs deplete the benefits of the population. One of the biggest failures of the current health care system that the triple aim theory tries to resolve according to Berwick, is that the current healthcare system focuses predominantly on the needs of individuals instead of the needs of a population that is defined as a group of individuals characterized by their specific needs that must be addressed in order to provide the best quality of care.

 Recognize limitations of governmental and financial abilities where, according to Berwick, budget constraints and policy levers must be identified that insist upon principles of health equity.

 Identification of a single integrator that takes responsibility for three targets in conjunction with the population. An integrator that is able “to focus and coordinate services to help the population on all three dimensions at once.” Berwick (2008). The integrator is well positioned in order to make connections in the healthcare continuum, between communities and their community resources, between patients their healthcare providers, and health insurance organizations.

To achieve the Triple Aim goals requires healthcare organizations need to have insights in their past and current performance and the ability to implement and execute interventions to improve. The ability to combine administrative, financial, medical and patient satisfaction data requires healthcare organizations to have a strong analytics capability, data foundation and the right tools and processes to be able to continuously measure performance.

As Berwick (2008) describes the biggest obstacles blocking a value based healthcare system are technologies that have limited impact on outcome, facilitating system integration or creating competition. Measuring health quality and costs: A value-driven system, as described by Porter (2011) and Berwick (2008), is driven by transparency. Stakeholders should therefore cooperate, and demand insights in the delivered value and costs of healthcare providers.

Even though the work of Berwick and Porter refers to a broader value-enhancement strategy, they have two important effects on the relevance of Big Data. First, they drive investments in IT-infrastructure that are also needed to make effective use of Big Data, since the higher levels of system integration and transparency require changes in IT-infrastructure. This trend also manifests itself in the Dutch healthcare system, where hospitals are currently investing in infrastructure to provide the insights in quality and costs as required by the Dutch Quality institute (Kwaliteitsinstituut Nederland).

Secondly, the Triple Aim strategy clearly drives the demand for deeper insights in delivered value (medical effectiveness, outcomes, customer satisfaction, etc.) and costs, empowering hospitals to optimize their value delivery in order to compete. As noted earlier, several authors believe Big Data analysis could play an important role in delivering those insights.

4.4. The origin of data analytics

The term Big Data is not new, it made its first appearance in a Silicon Graphics (SGI) by John Mashey with the title “Big Data and the Next Wave of InfraStress” (Diebold, 2012). The fast increase of information is received a lot of attention in the past and is not a phenomenon of the digital times: in 1944, the book “The Scholar and the Future of the Research Library” mentioned that the size of American University libraries was doubling in size every sixteen years (Rider, 1944).

Since 2010 Big Data has received increasingly more attention. Even resulting in what could be called ”Big Data Hype”; after surveying more than 4.000 IT professionals, research concluded that Business Analytics, driven by the growing trend of Big Data, is seen as one of the four major technology trends by Gartner (2013). The expectations of global big data technology are to grow with 18% from $14.26B in 2014 to $23.76B in 2016 (Forbes 2014).

4.4.1. Business Analytics and Intelligence

In his article Chen (2012) distinguish three generations of Business Analytics and Intelligence (BI&A). First, BI&A 1.0, which refers to the application of database management, that heavily depended on “various data collection, extraction and analysis technologies” (Chaudhuri 2011; Turban 2008). The BI&A 1.0 systems “are mostly structured, collected by companies through various legacy systems, and often stored in commercial relational database management systems (..) the analytical techniques commonly used in these systems (..) are grounded mainly in statistical methods developed in the 1970s and data mining techniques developed in the 1980” (Chen, 2012,p. 1166,).

Secondly, BI&A 2.0 is aimed at Web intelligence, web analytics, and user-generated content collected through Web 2.0-based and crowd sourcing systems. BI&A 2.0 has been of importance in academic terms, by generating in research regarding website design, product placement optimization, customer transaction analysis, market structure analysis, and product recommendations (Chen, 2012).

The BI&A 3.0 generation is emerging with the steep rise of mobile device technology. The ability of mobile devices, and – in the light of the ‘internet of things’, an increasing number of other objects – to support “highly mobile, location-aware, person-cantered, and context-relevant operations and transactions will continue to offer unique research challenges and opportunities throughout the 2010s” (Chen, 2012, p. 1166,). The difference between the second and third generation of BI&A is that the latest and upcoming generation of BI&A does not only add volume but also great variety in the types of data where the second generation was mainly while the second primarily grew of the new and sudden rise of Big Data availability by large volumes of data.

Even though Chen, recognizes that BI&A 2.0 has gained significant attention from both industry and academia, the emerging research on BI&A 3.0 is still at a very early stage and predominantly exploratory and sometimes conceptual in nature.

The use of data for analytics and use of complex algorithms on data sets is not new. Also the term “Big Data” has received a lot of criticism because the quantification of “big” is a relative term (Jacobs, 2009). The general consensus is that the term refers to a series of significant

development in the data analytics field, which impact the decision-process and steering toward a value-based type of operating.

Significant in this trend is the rapid growth of data that are becoming more and more accessible trough the emerging of new technologies and services. Not only computers are connected to the Internet; but with the explosive growth of mobile devices and house hold appliances that are being connected to the Internet, setting off a new area of connectivity called ‘Internet of Things’. This new trend of connected devices refers to an increasing variety of devices that produce data and are able to send this data through via the Internet (Chui, 2010). These devices communicate by retrieving, storing, sharing and sending data, therefore causing a vast increase in data volumes; in the last two years, more data was produced than in the 2.000 years before (Hilbert & Lopez, 2011).

Due to the reality that wearable’s are getting smaller and can even be integrated in clothing initiated another trend that is following and is leaning heavily on big data is the Quantified Self (QS) (Appelboom, 2014). Trends were individuals are using self-tracking devices to gather of biological, physical, behavioural or environmental information within a group or as individual (Swan, 2013). The opportunities for big data application are numerous from models to support data collection, integration with EMR or healthcare platforms to combine different data sets and provide extended analytic opportunities to create a variety of views and different insights to the data. QS applications provide meaningful insights in behaviour baseline, variability or pattern recognition for the individual or a group.

The growing number of devices that collect, share and store data has led to a growth in the variety in type of data that is available. As argued in the literature, the sheer volume is not of particular importance, but it’s the variety in types of data and relation to other data. The growing ability and efforts to consolidate, mine and aggregate data, “Big Data is fundamentally networked”. Allowing industry and researchers to discover patterns that were invisible in the past (Boyd & Crawford, p1., 2011). The term ‘Big Data’ is therefore referred to the increasing ability of expanding variety, since it allows the ability of combining data sources and searching for patterns, that are not visible without a wide variety of available data.

Secondly the term Big Data is connected to the trend of increasing amount of data creation within the different industry domains. In specific industries like, oil and gas, transportation, construction, automobile industry, machines and sensors are continuously measuring data points and generating large quantities of data streams containing values over a short period of time (Travis 2013; McAfee & Brynjolfsson, 2012). The amount of data that was created in 2 days in 2013 is the same as the total amount that was created by humans up till 2003, which was estimated at 5 Exabyte’s (1018 bytes). According to (Sagiroglu, 2013) the digital data lake in 2012 was expanded to 2.72 zettabytes (1021 bytes) and the prediction is that it is currently in 2015 reaching about 8 zettabytes of data.

Parallel to this trend new technological improvements have been developed to focus on the processing and analysis of data (Mayer-Schönberger and Cukier, 2013). These emerging new technologies make it possible to process large amounts of data collect, combine and discover data patterns and transform the data into information and new insights into data. This is a new development, which is a giant step forward in technology and capability when comparing it with the old ways of data mining. It creates new value for value-based decision-making processes. Having access to information based on not only a single data source but also a combined data source creates new value and reliability for the analytic outcomes. Using the newly harvested information in decision-making processes can provide companies with new insights enabling process optimization and more effective levels of operational excellence and higher numbers of customer satisfaction.

Although the term “Big Data” does not include the critical aspect of the trend that organizations also developed the capability to make use of the newly gained insights and have aligned organizational and human resource capability. As Spil (2002,. p.1) argues, “Although the new technology has a high potential, a great deal of the solution will be of an organizational nature”. The availability of the technology is only part of the Big Data trend according to McAfee & Brynjolfsson (2012). Organizations have updated and rejuvenated their culture and operational and strategic leadership and governance structure to adopt and absorb the information created by Big Data and use the new insights in their daily and strategic operation. The biggest challenge of big data is not technical or data related but the “lack of understanding of how to use analytics to improve the business” (LaValle, 2013)

Without preparing the organizational structures for the application of Big Data in their decision-making process to an evidence-based model, it is not possible to reach its full potential. Previously organizations tended to fall back on leadership instincts when information was scarce or unclear or they would focus on what could be derived from available data instead of focusing on the quality of the data source. A change in organizational and leadership thinking on the capability of data and how to use this in decision-making process is needed (McAfee & Brynjolfsson, 2012).

4.4.2. Healthcare value through data analytics

The scientific studies that are published on healthcare analytics are from recent publication dates, which is not unusual in a relatively young research field like analytics.

Consultancy firms who are operating at a more practical domains use invest highly in analytic capability. There is however very little scientific literature on analytics, most articles on analytics and their value contribution have been written by consultancy firms and are largely based on use-cases and practical viewpoint and less scientific (Groves, 2013). Although there is a lack of scientific research in the analytics field, the existing literature emphasizes its potential and value for the different domains. Based on the investment growth from consultancy and corporate businesses (9.83 bil. dollars (Forrester, 2014)) and the recognition from scientific studies, it can be said that the potential use of analytics is deemed to be significant. More studies on the application of analytics within healthcare domain from scholars as well as consultancy firms are needed to gradually extend and deepening the knowledge base from a theoretical as well as a practical standpoint.

The Gartner hype cycle has positioned big data analytics still in the “through of disillusionment” phase which is before the plateau of productivity, meaning that there is still a lot of research to be done before the true value of analytics can be determined for general use (Gartner, 2014).

The application of data analytics by healthcare organizations has unlocked and created new insights from information. The newly gained insights have two spinoff effects according for Cortega (2012) they drive clinical and operational improvements. Based on the concept of value based healthcare, Groves (2013) describes five pathways how big data could add value; right value, right care, right provider, right living and right innovation.

The first pathway “Right value” focuses on continue enhancements while “reserving or

improving its quality“. By implementing measures for cost effectiveness of healthcare processes or prevention of fraud, limiting bureaucracy and waste in the process of reimbursements

The “right care” pathway focuses on getting the right treatment in time. To optimize the treatment requires coordination between all caregivers involved in the treatment process of the patient. Standardization of processes optimize efficiency and allow a focused effort towards the same goal based on shared information within the care continuum of the patient starting from prevention, diagnosis, treatment and after-care. To prevent inefficiencies like miss-treatment or over-treatment, standardization of processes (based on patient characteristics) is a key aspect. At the same time big data enables identification of patient groups with the same characteristics, which enable better customization of processes around the specific patient group, and thereby matching treatment to patient needs. Several countries have started with country or region based EMR implementation, but this has not been successful till this date. Collaboration among healthcare organizations around a single patient care process on smaller scale has seen small successes within pilot regions.

The “right provider” pathways, focuses on consolidating and centralizing specific and complex care in order to gain process efficiencies, quality increase and cost reduction benefit. The Dutch government is enforcing changes in the distribution of care facilities and reorganizes the healthcare landscape. By relocating complex care to a limited number of hospitals (centralization), which are identified via quality measurements and experience in the specific complex care. By increasing quality trough repetition the cost will decrease by specialization, which in turn result in optimization and cost benefit. The same goes for the decentralization of less complex care that is being transferred from hospitals to general practitioners, which operate on a lower cost level. By using big data analysis on quality, performance and production indicators, it can be defined which hospital has the right infrastructure and resources for complex or general care.

The last pathway defined by the authors is “Right living” which refers to the value of analytics in the prevention process where life-style and behaviour are related to patient medical data. Illness prevention has gained significant attention from governments, as it is a

very efficient way to cut healthcare cost. Information on how to interpret the right living pathway assists and encourages the patient making life style changes and live healthy as possible.

The “Right innovation” pathway discusses the exploratory innovation of “Big Data” where the discovery of unexpected correlation creates opportunities for new innovations and healthcare optimization.

From a healthcare continuum that stretches from prevention, diagnoses, treatment and aftercare, the scope of this theses focuses around value creation in the diagnoses and treatment section. There is an overlap with the definition indicated by Groves (2013) with the pathways of “right value”, “right provider” and ”right care”. These pathways are in line with the maturity model that is developed within these theses, and focused on the value-creation within hospitals primary process (diagnose, treatment). The other two pathways “right innovation” (that focuses on the research perspective) and “right living” (which focuses on the home situation of the patient) are more focused on external value creation and go outside the scope of this thesis.

4.4.3. Defining analytical maturity

Maturity models have been used for many years to visualize roadmaps or growth models (Lahrman, 2010). The first maturity model for Business intelligence (BI) have been created by commercial organizations like SAS (2009) and Hewlett Packard (2009) in which they described the vision and specific characteristics how and organization could come to BI maturity. A maturity model describes the phases of maturity an organization has to go through in their effort of reaching an end goal or ideal situation at which the use of data has been integrated at a technical (with the right technical infrastructure and analytical tools) and organizational level (by integration into the decision making processes). The maturity model assists the organization in this process of maturation (Lahrmann, 2010).

The need for maturity models in healthcare is that healthcare organizations are increasingly asked by internal and external stakeholders to do more with less. Having access to key information is critical for evidence based decision-making processes (Foshay, 2013). The use of BI systems has been proven to be effective in providing value to organizations. The implementation of BI using maturity model could provide hospital organizations with the

right framework to guide the BI implementation process.

4.5. Maturity models

Maturity models give organizations insight into the roadmap of improvement and can be used as a basis for comparison. Guided by the maturity model an organization can evaluate itself against the roadmap and determine the necessary actions to create forward progression reaching the envisioned end state (Wendler, 2012). The maturity model assists the organization in creating a common understanding of the roadmap and terminologies involved during the implementation process to various stakeholders (Klimko, 2001;Weerdmeester, 2003). The two main characteristics of maturity models can be identified. The first is that the maturity models are based on a forward transition process, focused on an end-state of maturity. For each step towards a higher analytic maturity the specific characteristics of the maturity stage are described.

Secondly a maturity model can be identified by the clear and measurable stages that have phase dependent characteristics described which are stretched over a time scale (Wendler, 2012; Klimko, 2001). During the process of development an organization can measure and determine it’s position on the maturity roadmap and identify which actions are required in order to make the transition to the next level or the maturity mode.

4.6. Shortcomings of BI maturity models

To manage the required changes at technical and organizational level in order to effectively apply “Big Data” in the decision making process, a variety of maturity models have been developed to guide the process of capability development (Foshay, 2013; Rajteric, 2010; McAfee & Brynjolfsson, 2012). The current maturity models do not provide the framework for healthcare organizations to guide implementation process. Researchers like Lahrmann (2010), Chuah en Wong (2012b), Rajteric (2010) en Aho (2009) have compared several different maturity models with the aim of determining a standard maturity model, the conclusion however was that due to the limitations discovered in the excising maturity models it was not possible to construct a standard maturity model.

According to Lahrmann (2010) the majority of maturity models tend to address their focus on the technology capability side like software applications or platforms, hardware infrastructure

and data usage, and place less emphasize the organizational side like human resources, organizational structure and culture. In reference Lahrmann (2010) identified thirteen dimensions that could be characterized by maturity models: Applications, Architecture, Behaviour, Change, Data, Efficiency, Impact, Infrastructure, Organizational structure, Processes, Staff, Strategy and Users.

For this thesis the lack of industry specific alignment and focus on the organizational side of the maturity model is of specific interest. Research has identified the relation between acceptance of change (Kotter 1995; 2002) by users and the level of readiness of the organization play a critical role in the success of the transition towards data usage in decision-making processes.

Research shows that hospital organizations differ from other types of organizations in this respect for various reasons. Spill, 2002 concludes that BI within hospital environment is complex because of the different interests that have a major impact for example research, education and interest of the patient but also the complexity of the IT application landscape. Also the hospital organizations are characterized by their unique culture, (Scott & Mannion, 2003, Berwick 2008) consisting of various subcultures of specialism. Another reason for the misfit of maturity models is that existing BI maturity models are vague and not concrete on maturity level of a capability (Foshay, 2013; Lahrmann,2010).

4.7. Delta model

Davenport and Harris (2007) have developed a D.E.L.T.A maturity model focused on the use of analytics within an organization. This model distinguishes five stages: ‘Analytically Impaired’, ‘Localized Analytics’, ‘Analytical Aspirations’, ‘Analytical Companies’, ‘Analytical Competitors’ and links criteria to each stage, based on five dimensions: Data, Enterprise, Leadership, Targets and Analysts (Harris, 2010).

The model does incorporates the organizational side in the dimensions with it’s emphasize on the importance of an enterprise perspective on data analytics (Davenport; 2010). Also Davenport recognizes that the ability for an organization to grow in data maturity is highly depending on the leadership that can enable the change environment by activating the right people; influence the culture and redirect monetary resources. But the model doesn’t take into account the industry specific contextual elements that influence the growth process of analytic

projects. The model is independent of the BI&A generation and has been used by analytic project as a guiding framework identifying the stages and objectives for guiding analytic projects (Raber, 2013;Harris, 2011). An overview of the framework can be found in table 2. In the following section, each paragraph further elaborates on five components.

Data

Analytics is about fact-based decision-making and data is there for a prerequisite for analytics. There are different data structures identified by Davenport (2010), structure, uniqueness, integration, quality, privacy, governance and accessibility of data.

Enterprise

Davenport (2010) recognizes that using analytics value is optimal when applied on an enterprise level. Enterprise as a factor is about the scope of the application of Analytics. Integration of data analytics with the enterprise business strategy and operational goals is vital for it’s success since the authors state that analytics strategy addresses the key issues that impact the core of the organization.

Leadership

The involvement of leadership is the most important factor determined for its success according to Davenport. As stated by Davenport (2010, p.57,), “leaders have a strong influence on culture and can mobilize people, money, and time to help push for more analytical decision-making”. They recognize that for Analytics to be fully embedded in the organization support by leadership is a key factor.

Target

To effectively use analytics it should be linked to the strategic organizational goals, which can be strategic in nature, but also tactical or operational focused.

Analysts

The definition of analysts “workers who use statistics, rigorous quantitative or qualitative analysis, and information modelling techniques to shape and make business decisions”, according to Davenport (2010, p.91). For its success it is vial that the organization has access to analysts who can work with the data and analytical models. A schematic overview of the D.E.L.T.A model is given below in table 2.

Table 1: D.E.L.T.A - framework (Davenport, 2010) Success Factor Stage 1 Analytically Impaired Stage 2 Localized Analytics Stage 3 Analytical Aspirations Stage 4 Analytical Companies Stage 5 Analytical Competitors Data Inconsistent, poor quality

and organization; difficult to do substantial analysis; no groups with strong data orientation.

Much data useable, but in functional or process silos; senior executives don’t discuss data management.

Identifying key data domains and creating central data repositories.

Integrated, accurate, common data in central warehouse; data still mainly an IT matter; little unique data.

Relentless search for new data and metrics;

organization separate from IT oversees information; data viewed as strategic asset. Enterprise No enterprise perspective

on data or analytics. Poorly integrated systems.

Islands of data,

technology, and expertise deliver local value.

Process or business unit focus for analytics. Infrastructure for analytics beginning to coalesce.

Key data, technology and analysts are managed from an enterprise perspective.

Key analytical resources focused on enterprise priorities and differentiation.

Leadership Little awareness of or interest in analytics.

Local leaders emerge, but have little connection.

Senior leaders recognizing importance of analytics and developing analytical capabilities. Senior leaders developing analytical plans and building analytical capabilities.

Strong leaders behaving analytically and showing passion for analytical competition.

Targets No targeting of opportunities.

Multiple disconnected targets, typically not of strategic importance.

Analytical efforts coalescing behind a small set of important targets.

Analytics cantered on a few key business domains with explicit and ambitious outcomes.

Analytics integral to the company’s distinctive capability and strategy.

Analysts Few skills, and those attached to specific functions.

Unconnected pockets of analysts; unmanaged mix of skills.

Analysts recognized as key talent and focused on important business areas.

Highly capable analysts explicitly recruited, developed, deployed, and engaged.

World-class professional analysts; cultivation of analytical amateurs across the enterprise.

4.8. Evaluation of analytic models

The maturity models are used as a tool in describing, explaining and evaluating the maturity transition process. All maturity models are based on the idea that changes that occur over time can be regulated and predicted (Rajteric, 2010).

The challenge for organizations in using business analytics maturity models is that the benefits are difficult to proof. Organizations struggle to make the return on investments explicit during the transition phases in the maturity model (Rajteric, 2010).

The majority of development models for BI, known as BIMM models (“Business Intelligence Maturity Models”), are focused on commercial private organization. There are several BIMM models described by Eckerson (2009) en Lahrmann (2011) but not all the BIMM models are usable in a hospital environment because of its complexity and culture (Spill, 2002). The complexity is the result of the different interests that play a dominant role within the organization; research, education, patient interest, but also the complexity of the hospital system infrastructure EMR, HMR and ERP software systems collaborating with department specific software (Berwick 2008; Groves, 2013).

Differences between commercial and hospital organizations that impact the maturity model application are according to Mettler en Vimarlund (2009):

o There is not a single point of responsibility within hospitals management layer. The responsibility is spread between the clinical and administrative processes that have a combined end responsibility;

o Hospitals operate in an environment with many actors (Spill, 2002) each having their own needs, like health insurance companies, government agencies, patients and general practitioners;

o Value indicators are difficult to define because of the complexity of the care giving profession. The treatment is not mechanical of purely systematic, a large portion of the actions are based on human interaction with a large variety.

gathering towards information creation and starting follow-up processes based on the information (Foshay, 2013; Spil, 2002).

The complexity within hospitals and the differences between private organizations block the process of getting from information to knowledge and redesign of processes. Therefor a maturity model could support hospitals in taking predefined steps and control the process of maturity towards the envisioned goal.

Reaching analytic maturity in combination of having health data recorded in EMR’s could pose a lot of potential for hospitals (Rajteric, 2010). For a complex hospital organization reaching analytic maturity would bring internal and external insights in guidelines, rules and regulation policies, medical protocols and quality levels for patient treatments and patient satisfaction scores. Also financial insights in treatment costs, effectively of treatments, budgets and return rates on investments.

With the new insights management and physicians have better control over their processes. It provides the opportunity of creating transparency and enable evidence-based medicine (Foshay, 2013; Sackett, 1996).

Many analytic maturity models have difficulty aligning the roadmap of maturity growth with the dimensions described by Lahrmann (2010) like data, users, strategy and organizational structure and strategy. At the same time the biggest challenge for healthcare organizations developing value-based healthcare programs has been the distinctive organizational culture (Berwick, 2008; Groves, 2013).

Although the D.E.L.T.A model provides the most coverage of the dimensions of the organizational analytic maturity, it is not a complete fit. For example the leadership dimension does not recognize the impact on change and cultural process when there is misalignment with staff not directly involved with the analytic process.

Other models like the Business Intelligence and Performance Management Maturity Model from Gartner describe the levels of maturity but not any specific dimensions. The model (developed by Rayner en Schlegel (2008) and later refined by Hostmann en Hagerty (2010)) is based on the view that business aspects are driving the business intelligence growth in the organization rather then technology elements. Although the organizational awareness levels are described in five stages of maturity from unaware,

tactical, focused, strategic and pervasive, the model lacks the specific dimensions that Lahrmann (2010) describes, detailing the technical components supporting the roadmap towards the envisioned state. The five stages are described in the appendix 1 The TDWI’s Business Intelligence Maturity Model developed by Wayne Eckerson in 2004 focuses on the technical aspects and organization needs during the BI maturity process. The models consist of eight key areas: Scope, Sponsorship, Funding, Value, Architecture, Data, Development and Delivery. Using a scale of five grades (Infant, Child, Teenager, Adult, and Sage), the eight aspects are valued and graded (Eckerson 2007b). The five levels of maturity provide a solid basis for assessment of maturity from the technical viewpoint. However the model does not touches the dimensions from Lehman, 2010 on the cultural and organizational view. The five levels are described in appendix 3.

Up till now the literature review the groundwork for a new model that aligns with the specific characteristics and challenges of hospitals that are in the process of

implementing or increasing their analytic capability.

The foundational model used in this thesis is the D.E.L.T.A model from Davenport although it has two shortcomings. The model is lacking a change management dimension to guide the business analytics implementation of a business analytics in a complex environment such as a healthcare organization. Also the model does is not aligned with the industry specific needs and lacks there for practical applicability.

Berg (2001) describes the importance of change management in hospital environment with the implementation of patient care information systems (PCIS). A clear aspect Berg identifies which aids to failure of implementation of PCIS, is a tendency in healthcare organization to classify PCIS implementation as a purely technical

operation and thereby not taking into account the impact on people, organization and culture (Berg, 2001). The success of the implementation of an EMR requires strong collaboration with the clinical staff and depends on several factors such as training in operational use of the system, attention to process changes and providing information about the project and possible impact at forehand of implementation (Poissant, 2005;

Aarts, 2004). Thereby recognizing the importance of involving the physicians during the implementation and change process.

4.9. Change management

The interest of organizations to use and apply “Big Data” in their decision making process require more than facilitating the technical capabilities. Scholars like (Kiron and Shockley (2011) have focused on the cultural change that is required to evolve to “analytics maturity”. In order to have analytic maturity to be come part of the

organizational culture, it is recognized that the analytical skills and required technology infrastructure can only lead to value if the organization as a whole is orientated towards evidence- based decision making.

The traditional top-down management paradigm is challenged when implementing data analytics for evidence-based decision making processes. Although healthcare and hospital organizations have a high degree of protocols and standardization,

traditionally it are the expert-physicians or expert managers who make largely intuition- or experience-based judgments, which are not wrong in itself but are susceptible for bias.

When successfully adopting evidence based decision-making processes and data analytics-based culture will require an organizational management paradigm shift. There are several power dynamic aspects that will be impacted when an organization changes from the traditional power hierarchies of the organization decision-making process. The access and rights to information, authorities on decision rights and assessment systems will change. Adopting data analytics results not only in technical changes but also at the level of cultural, organizational architectural and management control.

For organizational and cultural changes it is recognized that the change management mechanism is well suited to guide an organization trough the transition process.

In the process of reaching analytic maturity an organization aims at a future state where decisions will primarily be based on evidence, fact based analytical insight. Indirect multiple sub-organizational processes are involved in the evidence based

decision process, such as data collection, information sharing, knowledge sharing, collaboration, communication and decision making. The challenge for any

organization that is reorganizing and aligning sub-organization processes is to prepare for and overcome organizational resistance as the changes impact excising processes and stakeholders.

Kotter’s (1995) change management model gives clear guidance on how to implement organizational change in a controlled manner. Organizational change can touch upon the identity of the organization. Redefine “who we are” as an organization is about transforming its identity, which affect both who the organization as a collective is perceived and the individual members sense of their own identities in relation to the organization. Collective and individual identities are highly resistant to change, influence each other and their interactions are critical (Fiol, 2002).

Kotter (1995) describes eight steps though which the change is guided. The change model is useful in the communication to stakeholders and provides a clear activity plan. The downside of the eight steps are that they are nonlinear and when the change has set off, it’s difficult to steer the process or loop back to earlier steps in the model.

Kotter (2002) reorganized the eight-step model into three phases. In the first phase the focus is on creating a sense of urgency and general sense of awareness of the need for change. This step is about creating a vision and communicating this vision to the stakeholders, which should result in generating buy in and increase the level of acceptance. Kotter (2002) describes that the first phase is about determining the impact of the change and getting a read on the expected resistance and origin of the resistance. The second phase is focused on determining the level of resistance and validate if the need for change is urgent enough to go trough with the change and face the resistance.

The creation of a team and empower the team to act on the vision getting rid of obstacles, chancing systems or structures and encouraging risk taking and non-traditional ideas. The team should have credibility with peers and consist of change agents that know the organization and understand organization specific tension points that may arise during the change.

The final stage is focusing on the team to create a change plan and come up with short-term wins, which can be consolidated quickly before setting them as the standard.

4.10.

Conclusion

From the literature review it can be concluded that data analytics provides a major opportunity for hospitals (Cortada, 2012; Raghupathi & Raghupathi, 2014; Groves, 2013). Also based on the literature of (Lahrmann, 2010; Rajteric, 2010) it can be recognized that the current BI maturity models are inadequate when applied to the healthcare domain and lacking healthcare specific depth and a clear roadmap is non existing. According to Rajteric (2010) the maturity models do not extend all factors of the maturity process but focus on specific parts. The maturity models supports the organization in evaluating its business intelligence maturity and identify known and less known areas that need focus of attention in order to improve maturity as a whole. For healthcare organization the change management aspect is important in order to gain momentum and create enough buy-in to gain acceptance from the participants. The impact of change management coordinating the involvement level of medical staff is vital to the success of data analytics within the hospital organization. This emphasized the importance of adding change management elements to the analytic maturity model.

In line with the identified gap in analytic maturity model for hospitals the opportunity arises to extent current theoretic knowledge on this front. The important factor that change management elements have in the maturity model is recognized in the literature by its usefulness of managing organizational resistance and have a broader focus then development of technical capability Rajteric (2010) conclusion that asks for maturity models that have business specific factors included in order to have a better alignment with the true users of business intelligence systems.

Hence, in this thesis, building upon the literature that was reviewed in this chapter, the existing D.E.L.T.A-model is extended with a change management dimension, and provided with more concrete targets, based on change management models.