Big Data in Cardiology. Predicting, Preventing and Managing Diseases

Table of Contents

Abstract

Keywords

1. Introduction
1.1 Purpose Statement and Research Questions
1.2 Topic Justification
1.3 Scope and Limitations
1.4 Definition of Big Data
1.4.1 Healthcare Big Data Sources
1.4.2 Techniques and tools to analyze Healthcare Big Data
1.4.3 Application of Big Data in Healthcare

2. Methodology
2.1 Methodological Tradition
2.2 Methodological Approach
2.3 Data Collection
2.4 Methodology Applied for literature review.
2.4.1 Inclusion and Exclusion Criteria
2.4.2 Search Procedures
2.4.3 Methods for Data Analysis
2.4.4. Validity, Reliability and Generalizability
2.4.5 Ethical Considerations

3. Results
3.1 Cardiology
3.2 Results from Comprehensive Literature Review.
3.2.1 Prevention
3.2.2 Prediction
3.2.3 Management of Disease
3.2.4 Future Trends and Directions
3.2.5 Challenges
3.3 Results from Qualitative Interviews
3.3.1 Theme 1
3.3.2 Theme 2
3.3.3 Theme 3
3.3.4 Theme 4
3.3.5 Theme 5
3.3.6 Theme 6

4. Discussion
4.1 Part 1
4.2 Part 2
4.3 Part 3
4.4 Part 4

5. Conclusion

6. References

Acknowledgement

I would like to express my sincere gratitude to Linnaeus University, Sweden for providing me the opportunity to take part in this Master’s Program which has widened my horizon and helped me to understand the world of Information Systems and Informatics better. I would like to thank to all teachers who helped me in this journey. I would also like to thank my supervisor Dr. Krenare Nuci for providing valuable support and guidance. Finally, I would like to thank my family who is a constant source of motivation.

Abstract

The statement ‘Data is the new oil’ has been broadly acknowledged due to its wide-ranging importance. Utilizing big data offers a variety of benefits. Although the health sector was late in terms of exploiting the benefits of big data, currently, the adoption is accelerating. Healthcare is increasingly becoming an information science and the implementation of electronic medical records (EMR) and other information systems is growing rapidly. The patient data originating from smart devices and other sources like genomic databases are supporting the healthcare sector offering better healthcare delivery and increasing efficiency, hence saving costs. This study was conducted to analyze this process closer focusing on a case of Cardiology. Conducting a comprehensive literature review and qualitative expert interviews, the impact of big data in the field of Cardiology was explored. The result of the study shows that big data can play a positive role in three aspects: prediction of disease, prevention of disease and management of disease. Big data enables us to build models that can be used to predict the occurrence of disease. Based on this information, actions can be taken to prevent the disease. Data also helps to manage the disease by offering helpful insights. Medical personnel can retrieve the patient data, with the help of AI, they can make faster decisions allowing them to spend more quality time with the patients and reduce cognitive errors. Through the interviews, it was understood that even though the positive role of big data has been acknowledged, the implementation is still a challenge due to various limitations. The challenges lied mainly on technical know-how and domain knowledge. Further challenges were data security and privacy issues that need to be addressed to mitigate the risks that can be caused by them. The examples of big data implementation in various cases like in heart failure prediction or prevention shows a positive picture. The overwhelming majority of case studies analyzed in this regard show an optimistic picture. Due to growing importance and use of smart devices, IoT, genomics and the recent developments in the field of ICTs, it is expected that big data will not only leave a positive influence on the field of Cardiology, it will also change the way medicine is practiced and healthcare is offered.

Keywords: Big Data, Healthcare, Health, Cardiology, Cardiovascular Diseases, Medicine, Information Systems, Information and Communication Technology

List of Tables

Table 1: Healthcare Data Sources

Table 2: Predictive Tools for Healthcare

Table 3: Research Concepts

Table 4: Types of Research

Table 5: Characteristics of Qualitative and Quantitative Approach

Table 6: Types of Case Studies

Table 7: Information about the interview participants

Table 8: Symptoms and Diagnostics in Cardiology

Table 9: Subject Classification of Publications

Table 10: Further Categorization of Publications

Table 11: Stakeholders Role in Healthcare Big Data

Table 12: Identified Themes

List of Figures

Figure 1: Literature Review Process

Figure 2: Visualization of search procedure

Figure 3: Model for Data Analysis

Figure 4: List of body sensors

Figure 5: The sensor workflow model

Figure 6: Links between issues and themes

1. Introduction

The field of Cardiology is one of the largest in Medicine and Cardiovascular Diseases (CVDs) along with Cancer are the leading cause of pre-mature mortality worldwide. Hence, the use of recent technology like big data is prominent to reduce the disease burden and relax healthcare systems. However, assessing the role of big data in detail in the first hand is important before moving forward, which was the driver of motivation to write a thesis which aims to study the impact big data makes in the field of Cardiology and health sector in general. The Covid-19 pandemic is the depiction of the health sector worldwide facing underlying challenges. Even the well-established healthcare systems like those in Germany and Scandinavia are struggling to survive the stress test protecting human lives from the deadly disease. The pandemic shows the short falls of our society where the health sector is at the epicenter as it absorbs the outcomes generated by socio-economic interplay. For example, job loss due to pandemic causes mental health issues in individuals who take refuge in healthcare systems, businesses attempt to stockpile equipment like masks, sanitizers etc. leaving the health system a shortage of such materials that are needed to run daily activities and to save lives. In such times, due to the growing prevalence of disease, additional pressure is exercised on the health sector. In addition to it, ageing society, rising cost of healthcare delivery, medical staff shortages, new but expensive treatments, growing prevalence of non-communicable diseases like cancer and cardio-vascular diseases, existence of communicable diseases in developing countries in a greater extent, climate change, sedentary lifestyle etc. are putting pressure in health sector. Hence, healthcare budgets are increasing in the OECD countries with the United States spending 17% of total budget for healthcare and most western and northern European countries including Sweden spending more than 10% (Mikulic, 2018) to fulfil the public mandate of providing quality care. Costs however are increasing at an unsustainable level. The recent experience is, injecting financial means into the system alone to improve care and topping up manpower did not generate desired outcome. Hence, to mitigate challenges faced by the health sector, the role of innovation, technology and holistic thinking remains prominent.

Recent developments in healthcare technologies or technologies that can be implemented in the health sector have opened potentials for improvement. They include robotic surgery, biomedical sensors, 3D printing, artificial intelligence, internet of things (IoT), smart wearables, telehealth, precision medicine, cardiac pacemaker, artificial inhalers, artificial organs as well as process innovation and information and communication technologies. Most importantly, such technologies are able to generate vast amounts of data that can be used to optimize the processes and support decision making which will impact the quality of care delivery. However, the health sector has established its image as a laggard in terms of utilizing the full potential of technologies especially the use of data (Wang et al, 2018) in comparison to other fields like finance and commerce. The process of digitization in healthcare is also slow and has other constraints (Zetterberg & Ingmanson, 2016). Several factors might be influencing this, for instance the characteristics of data in terms of variety, velocity and veracity, human capital needed to analyze and make use of such data and general lack of workforce skilled in both aspects. In addition to this, legal and technical barriers for adoption also remain a challenge. Implementing the information systems (IS) is difficult, often requires a huge change within the industry and might end up in a failure (Beynon-Davies, 2013). Nevertheless, the scope of technology especially big data has been widely acknowledged.

The role of data in the healthcare domain can be in early detection of diseases, tracking biomedical information, reduction of medical errors, operational efficiency, improvement in decision making etc. (Raghupati & Raghupati, 2014) which can reduce cost and improve the quality of care generating a win-win situation for both individuals and healthcare providers. The development of data mining tools and applications offer an optimistic picture. Also, the web has emerged as a platform that houses data that can be analyzed to generate valuable information. Due to volume and variety of data, not only the public sector but also the private sector is increasingly establishing their position in this field. For example, health related data from social media are increasingly used for various purposes. Flu trends can be depicted from search keywords which led Google to launch Google Flu Trend. This motivated other firms to enter the race. Apple launched the ‘Researchkit’ as an open source app development tool which can be used in hiring participants for randomized controlled trials in the process of drug development or in other types of medical research. Also, the electrocardiogram (ECG) sensor was integrated into the Apple Watch which can measure artillery fibrillation, for which there is now no necessity to visit the doctor. Similarly, other firms are heavily investing in such technology because healthcare is one of the largest sectors and the demand for healthcare goods are comparatively inelastic. Furthermore, healthcare products not only generate commercial value but also have societal importance, as when one person uses ECG from a smart wearable device instead of visiting a doctor, the waiting time for another patient goes down. Google and Apple are collaborating to develop a Bluetooth based technology to trace contacts during Covid-19 pandemic is an example where a health-related technology and data is both of commercial and social value (Apple Press Report, 2020) At least in some places, the public sector has acknowledged the value of data in healthcare and the technological know-how of private sector in generating information from it. They have partnered with firms in the form of public-private partnerships to support the greater public good. The Human Genome Project launched in the US with the help of public funding aiming to sequence 3 billion genes was completed in 2003 opening a vast reservoir of data that can be used for medical research. Especially in Cancer research, the knowledge about gene mutation can help in understanding the disease better in a particular person. This project is the biggest source of genetic big data that has revolutionized the field of precision medicine. Apart from this, other health related big data can be used in a similar manner. The data are for instance electronic health records, clinical data extracted from randomized controlled trials, prescription information, web searches, sensor data from smart sensors, genotyping, radiology images, public health data, population information etc. (Raghupati & Raghupati, 2014).

This thesis seeks to answer the questions about the role of big data in healthcare focusing in the field of Cardiology where the impact in predicting, preventing and managing diseases will be studied and discussed. Also, the challenges will be mentioned. The next section will elaborate more on the purpose and the methodologies adopted.

1.1 Purpose Statement and Research Questions

The research focuses on the role of big data in the healthcare domain. The aim is to explore and acquire knowledge and impact of big data in the field of Cardiology and see how data can support in improving the health and contribute to lower the disease burden. This will be done from a health provider’s perspective. Hence, it will cover the whole process from prevention to discharge involving care delivery, management processes and involves issues surrounding the topic for other stakeholders like patients, government and private sector apart from the health sector. However, the perception from the side of other stakeholders is beyond the scope of this study.

A mixed-method qualitative approach will be used performing comprehensive literature review and semi-structured interviews with primary focus on Cardiology. The definition, characteristics and issues about big data will be dealt before approaching their role in Cardiology.

Hence, the research question is: What is the role of big data and how is it impacting the field of Cardiology in terms of predicting, preventing and managing the diseases?

1.2 Topic Justification

In Europe and worldwide, the cost of healthcare is increasing over the last decade. As a consequence, the health sector is under scrutiny and faces pressure to focus on cost-efficiency. Gaining efficiency however is easier said than done. Efficiency can be achieved from the trajectory of multiple factors. IS has an important role to play in this process. The evidence shows that implementation of IS has improved efficiency and the quality of care (Harrison & Palacio, 2006). The authors highlight that switching the procedures to digital recordings like electronic patient records, patient history, diagnostic information record etc. enable a faster retrieval of information saving time for medical personnel and improves the quality of care. Furthermore, to achieve this, it remains vital to understand the health sector. The scope of data for this goal is immense to untap the perspectives that are not seen otherwise. Data that is collected from individual units within healthcare, can be analyzed to study and evaluate the situation to improve the procedures and care delivery. Big data has potential to support in improving patient care, healthcare decision making, self-care, awareness building and to contribute in data pooling for further research (Feldmann et al., 2012). On the other hand, efficiency should not just be the goal of a healthcare system, inclusiveness and equality is important as well. Here too, big data offers support in terms of data mining, evaluation and support in decision making. It has been widely acknowledged that data is not only supporting in above mentioned points, but also to gain insights, detect inefficiencies, ineffectiveness, help in prevention and more (Raghupati & Raghupati, 2014).

The field of Cardiology, combined with the above listed diseases in Table 8, causes the highest mortality rate. Hence, it is of utmost priority to use all tools and technologies available to mitigate this prevalence. Big Data technology offers such potential. Collecting data from the disease lifecycle, from detecting symptoms through diagnostics and recovery can be used to understand the disease better and the environment of it where intervention can be planned afterwards to prevent it. As heart health is strongly correlated with physical activity, the role of smart wearable technology remains vital. Furthermore, environmental factors like infrastructure that allows physical activities are equally important as other public health components like hygiene, air pollution control etc. It can be concluded that Cardiology hence is related to all aspects of health: environment, technology and medicine. In such a condition, information system development to acquire big data can be fruitful for all aspects. Hence, if efficiency and care quality is to gain, it is important that we understand the interconnections between all the aspects and data can be the best tool to achieve this.

1.3 Scope and Limitations

For this research, the medical personnel are pre-selected based on the fact that the author knew them personally. Medical personnel who work as a Cardiologist, a Nurse in Cardiology department of a hospital, Physical activity officer working for a healthcare institute, public health expert and a digital policy making officer were chosen to include the whole picture that is either directly related to Cardiology or a field that has a direct impact were interviewed. Due to time and resource constraints, the number of interviewees were limited. The interview took place in German and Nepalese language which the author and interviewees spoke. However, the notes were taken in the English language as well as the master thesis. Hence, it is possible that much information might be lost in translation. It is equally important that the author takes a great deal of carefulness while interpreting the information given by the interviewees so that information is not mis-interpreted or wrongly translated.

For the comprehensive literature review section, only the publications published after 2008 were chosen. Given the rapid change in the field of big data and information systems, the tools and technologies used before that period has become obsolete. Hence, for the relevance, only newer papers were chosen. It might have missed some papers that might be relevant even though they were published before 2008. For information verification purposes for the grader, the author also only reviewed English language journal articles published in high ranking journals in developed countries. Publications from developing countries journals were not taken. All the interviewees except the urban planner were based in Europe. Hence, their views are limited to healthcare systems in Europe. A bigger sample with interviewees practicing in developing countries would add to information diversity. Furthermore, only qualitative studies were chosen due to methodology used for this research. Expanding it to quantitative studies might have added additional perspectives. This thesis is structured as follows:

Abbildung in dieser Leseprobe nicht enthalten

After this section of introduction, where the chosen topic was justified, and the scope and limitation of the thesis was mentioned, the next part deals about the methodology adopted for this research. The findings of comprehensive literature research are presented after that. Following the literature review, I will move to the empirical results collected from qualitative interviews and discuss the results and finally conclude the thesis.

1.4 Definition of Big Data

We live in an information age. The information is growing at an unprecedented rate, supported by massively growing data behind that. The data collected from large set of sources contributes to the already existing pool and becomes big data. Not only the data, but also the sources are growing. Due to the growth of sensors, IoT, smart devices and other technologies that are capable to capture data, the volume and variety of data is also growing. Although there is no universally accepted definition of big data, which is due to its changing nature, several entities define it in various ways. For example, Gartner Inc. defines Big Data as:

“a high volume, high velocity and high variety information assets that demand cost-effective, innovative forms of information processing for enhanced insight and decision making”.

The European Commission defines it as:

“large amounts of different types of data produced from various types of sources, such as people, machines or sensors. This data includes climate information, satellite imagery, digital pictures and videos, transition records or GPS signals. Big Data may involve personal data: that is, any information relating to an individual, and can be anything from a name, a photo, an email address, bank details, social media posts, medical information, or a computer IP address “

Volume, velocity, variety, veracity and value are five characteristics of big data (Raghupati & Raghupati, 2014). Feldmann et al. (2012) define the 5Vs as:

Volume: Quantity or scale of big data. From terabytes (1000 GB) to zettabytes (1 billion TB)

Velocity: Speed and analysis of real time processing and streaming.

Variety: Different forms of data (noisy/unstructured data, structured/cleaned data)

Veracity: quality, relevance and significance of data

Value: the value that can be added by big data

In the health sector, Raghupati & Raghupati (2014) define big data as a set of large and complex data that originate from electronic medical records (EMR), genomics, clinical research, medical imaging, social media etc. The changing nature of information and communication technologies (ICTs) might add other aspects of big data sources in healthcare (Mehta & Pandit, 2018). One such recent development is the internet of things (IoT) that gained prominence in the recent years. IOTs are capable of recording data of any nature, including health that can be used for various reasons. IOTs related to healthcare are for instance, smart wearables that can record physical activity data, health-related sensors, air quality sensors, sleep monitor, vital signs capturers and other medical devices that are capable of capturing biomedical information of a patient. It is estimated that more than 2 billion smartphones will come to the market until 2020 (a prediction made in 2012) that will contribute hugely to collect health related data (Feldmann et al., 2012). When these data are fed into the system, insights can be gained that was not yet possible. From prevention to treatment optimization and precision medicine, data plays an important role from where information and knowledge can be retrieved. At this point, it remains vital to define what big data in healthcare is, what are the sources of big data in healthcare, techniques to analyze them, application of them in different healthcare domains and the challenges present.

1.4.1 Healthcare Big Data Sources

The source of healthcare data can be divided into two big categories, namely payer-provider related big data and genomics-driven big data (Miller, 2012) where the source of the former is mainly electronic health records (EHR) whereas the latter comes from genetic databases such as the human genome project. The EHR data can be retrieved from the records of hospitals, clinics, health insurance companies, pharmacies, medical device companies, tech firms, research institutes etc. whereas genomic data (and other ‘omics’ data such as microbiomics, proteomics, metabolomics etc.) can be stored in special care centers, research institutes and even the same institutes as EHR (Raghupati & Raghupati, 2013) (Mehta & Pandit, 2018). In the few recent years, smart technologies like wearables have gained clear advantages in terms of their potential of data collection and management. Fitbit, Apple Watch, Android Wear and smartphones themselves are able to collect massive amount of health data from the users and store it into their servers or on the devices (Feldmann et al., 2012). Even though, many of smart wearable devices were initially thought for life-style gadgets, they emerged as a major health watcher and supported the health sector to collect important health data that can be used for meaningful purposes. A further optimization was achieved since smart wearables are capable of performing tasks that needed medical personnel’s intervention in the past. The electrocardiogram (ECG) sensor integrated in the Apple Watch can for example test for atrial fibrillation from anywhere minimizing the requirement to visit healthcare center (Bumgarner et al., 2018). The abnormalities can be detected beforehand and the series of events that can lead eventually to the disease are can monitored in advance. Such data is fed into electronic health records if the user allows it through configuring right settings. The number of such devices has grown so rapidly that they are now one of the largest sources of big data in healthcare. Another recent development is the social and professional media that emerged from the late 2000s which possess a reservoir of healthcare big data related to behavior and lifestyle of individuals (Mehta & Pandit, 2018). Facebook, twitter, linked-in are some of the popular sites that contain most data. The device manufacturers that support social media are emerging as stakeholders that also contain big data (Bumgarner et al., 2018) The Table-1 summarizes the different types of healthcare data sources:

Abbildung in dieser Leseprobe nicht enthalten

Table 1: Healthcare Data Sources (Source: Mehta & Pandit, 2018; Rumsfeld, Joynt & Maddox, 2016)

1.4.2 Techniques and tools to analyze Healthcare Big Data

Health data are heterogeneous and multi-dimensional. The table above shows that the data come in the form of prescriptions (which are often handwritten), they can be genomic data and some data can also be in the form of images, like X-Rays and MRIs. Hence, tools and techniques applied to analyze them also differ. Furthermore, due to the multi-dimensional nature, data integration also remains a challenge. Also, software used to analyze them is scarce (Taglang & Jackson, 2016).

Due to the evidence that applying big data analytics in health sector can improve the quality of care, optimize the processes and save costs (Raghupati & Raghupati, 2014) incorporating the techniques to do so in the decision making and new developments of disease specific techniques as well as existing techniques are being developed, improved or updated. According to Dash et al., (2019) most of the big data implementation in healthcare has been in predictive analysis. Mathematical models and algorithms that are built on historical data are able to predict the future events or medical incidences. The cost-saving factor is hence led by prevention of disease in the first place. Furthermore, it can be in using the algorithms to predict readmissions, adverse events or recurring incidences (Wang et al, 2015). There are several predictive analytic techniques that can be used in healthcare, which are for instance:

Abbildung in dieser Leseprobe nicht enthalten

Table 2: Predictive Tools for Healthcare (Source: Wang et al.,2015; Raghupati & Raghupati, 2014)

These techniques have been applied in many contexts within healthcare and beyond. Substantial number of other researches have been carried out to examine their feasibility. Going into detail about them would be beyond the scope of this research. However, it is worth acknowledging that the use of predictive analytics has helped to gain insights that would have otherwise been impossible. This field has emerged in the trajectory of Statistics, Information Systems, Mathematics and Computer Science. Innovative tools that emerged from their combination have enabled them to create value in the health sector and beyond. Raghupati & Raghupati (2014) have mentioned the most important tools that are used in health data analytics, namely: Hadoop, MapReduce, Pig, Hive, Jaql, Zookeeper, HBase, Cassandra, Oozie, Avro, Mahout, MongoDB etc. In addition to this, programming languages like Python, R, Java etc. are also used in big data analytics (Wang et al., 2015) With the development of new data sources and data, other techniques and tools might emerge in the future. Although there is no paucity of tools and techniques, challenges remain, for example to clean the data, to compute missing values, to reduce noises etc. (Dash et al., 2019). Due to frequent updates and features enhancements though, the tools are getting better. Hadoop for instance has the capability to apply distributed processing, storage and computation which enables the analysis of large datasets. MapReduce, which lies within Hadoop infrastructure, helps to simplify the data structures by mapping and reducing them and distributes the work among nodes and clusters. Both works together enabling increased signal processing and tolerating higher velocity of data (Mehta & Pandit, 2018). The goal was set to reduce the volume of data down to a manageable level and to distribute the load across many computers to carry out parallel processing. Map reduce creates the data map conjoining the key-value pairs and in second step, reduces the data by combining them key-values into smaller set of tuples (IBM, 2020)

1.4.3 Application of Big Data in Healthcare

Big Data offers enormous potential in healthcare by supporting decision making, improving care delivery, helping in prevention and optimizing the services (Feldmann et al., 2012). As mentioned in the introduction section, healthcare costs are likely to rise due to sedentary lifestyle, rising life expectancy and expensive treatment methods. Data is a boon for reducing cost in these circumstances. Healthcare expenditures are unsustainably high in many of the OECD countries. An extra dollar used for healthcare is the money lost for education, economy or other sectors upon which the healthcare system lies as a good healthcare system can only be financed by a good economic system and by well-educated and skilled workforce. When this balance is disturbed when expenditure in one sector is unsustainable, it will start to shake the fundamentals of the society creating room for turbulences. Hence, cost-efficiency analysis (CEA) is now the fixed part of reimbursement process of new treatment methods in many healthcare systems around the world, like in the UK, Netherlands or Denmark (Quentin et al., 2018). This argument plays as a strong point for using big data related technologies to reduce the healthcare cost. Although cost has been the central focus of big data implementation, it is not the only one. Other goals are to improve health, prevent the disease from happening and citizen empowerment by giving the power to the citizens to manage their own health through smart devices (Quentin et al., 2018). In a broader spectrum, big data has been applied in public health, environmental health, genomics, elderly care, prevention medicine as well as in individual medical domains like cancer, cardiovascular diseases, mental health etc. Application of big data with data mining algorithms can help in pattern recognition and facilitate machine learning and artificial intelligence (Mehta & Pandit, 2018). Combining big data, data analytics and machine learning contributes to early detection of disease, health state of individual or groups and facilitates personalized predictions (Dash et al., 2019). By this, healthcare institutions are supported by reduced waste, improved care and costs saved. Private firms using healthcare big data are offering smart devices that can manage individual health, provide recommendation and encouragements to undertake healthy behavior and alert in times of adverse events (Feldmann et al., 2012). In addition to that, data integration between the stakeholders offer better healthcare management, improve process efficiency, improve disease surveillance, enhance quality and reduce cost (Quentin et al., 2018). Due to such applications in all levels, it has been possible to manage personal health better, to identify adverse outcomes, identify inefficient treatment methods and manage population health efficiently (Raghupati & Raghupati, 2014)

According to Sukumar et al. (2015) big data integration into healthcare offers answers to eight important questions – 1. How will healthcare costs rise in the future? 2. How does policy change in healthcare impact behaviors and costs? 3. How do costs vary according to geography? 4. Can fraud be detected with the help of data? 5. Are there early signs of any epidemic/pandemic? 6. How are the choices of healthcare providers by the patients determined? 7. Why do health outcomes differ according to providers? 8. Which treatment is most cost-effective? The answers to these questions lead back to the goal of applying big data, namely saving cost, improving care quality and efficient decision making.

2. Methodology

This chapter deals with the research methodologies adopted to answer the research questions. The methodological tradition will be elaborated followed by the methodological approach and the philosophical underpinnings. It also includes the methods for data collection and the choices used for inclusion and exclusion criteria. The policies taken for the search procedure is described and visualized. Finally, the methods for data analysis is described which also includes the ethical considerations applied in the thesis.

2.1 Methodological Tradition

The aim of this section is to introduce research concepts. Research concepts are formed based on respective worldviews that can be termed as ontology and epistemology. They are supplemented by the respective research paradigm, where a paradigm is defined as an intellectual framework embodying a tradition of scientific research and theories (Creswell & Creswell, 2014). The research design is based on the paradigm.

In the field of scientific research, ‘Ontology’ refers to concepts that aim to answer questions referring to ‘what exists in reality?’. It is based on three underlying assumptions, namely i) the belief that there is only one single reality, ii) the belief that there are multiple realities and iii) the belief that the reality is constantly changing (Creswell & Creswell, 2014). The way how the question is posed is the main determining factor of choosing appropriate research paradigm according to the author which also have to be linked with epistemology.

Whereas ontology refers to ‘what exists in reality’, ‘Epistemology’ refers to ‘what can we know?’ and ‘how can we acquire knowledge’. Hence, these two terms are strongly linked to each other. Epistemology is also based on three assumptions, namely i) knowledge is measurable, ii) reality needs to be interpreted and iii) reality should be examined in a best way possible employing right tools (Creswell & Creswell, 2014).

Abbildung in dieser Leseprobe nicht enthalten

Table 3: Research Concepts (Source: Fletcher, 2017)

From both of the definitions above, it can be summarized that ontology can be described as the reality perceived by the researcher whereas epistemology seeks to find and claim the knowledge from the reality. Based on these theories, three paradigms can be adopted, which are positivism, constructivism and pragmatism. Positivism embodies objectivist approaches where research that are more quantitative in nature base upon, whereas constructivism embodies subjectivist approaches that are used in qualitative research types and pragmatist approach embodies both objectivist and subjectivist approaches and is mostly used in mixed-method research types (Boehme et al., 2012). Information Systems utilizes the paradigms based on these theories to seek knowledge. For the purpose of this research, the constructivist approach is most suitable as the research is based on literature review which analyses the findings of other researchers and qualitative interviews which seeks to understand the topic from the viewpoint of experts. However, many of the literature chosen for analysis include a large degree of quantitative data such as the health-related information. Hence, it will not be wrong to tell that this research falls somewhere between a constructivist and pragmatist approach.

2.2 Methodological Approach

This chapter will mention different methodological approaches that were considered to answer the research questions as well as the reasons behind the choice. The strengths and characteristics of these approaches will also be elaborated which underpins the decision of choice.

The primary goal of a research is to ‘increase the stock of knowledge’ and to improvise new applications from it. Hence, research should be novel, creative, systematic, transferable and reproducible (Creswell & Creswell, 2014). Various processes and steps are used to collect, analyze and understand the topic where, in first step, a research question is posed to clarify what the researcher is seeking to understand, in a second step, the relevant data to understand the topic is collected employing the appropriate method and finally the data is analyzed to extract the information which evolves as knowledge (Boehme et al., 2012) As listed in Table-4, different types of research can be used to advance knowledge, such as:

Abbildung in dieser Leseprobe nicht enthalten

Table 4: Types of Research (Source: Creswell & Creswell, 2014)

There are three approaches to research: qualitative research, quantitative research and mixed methods. Qualitative methods encompass the use of qualitative data to understand and explain social phenomena and includes the methods of data collection like participant observation, interviews (structured, semi-structured or unstructured) document analysis etc. and are mainly designed to help the researchers understand people in their natural contexts (Creswell, 2017) Qualitative methods are subjective in nature, which means that the viewpoint of the interviewee plays a central role. As a result, in health sector, it is often seen with skepticism (Malterud, 2001). However, Kaplan & Maxwell (2005) argue that social and institutional contexts are largely lost when the textual data are quantified, hence the strength of qualitative method lies therein that it allows the researcher to go deeper into the thematic along with other data collected during the process. Furthermore, the advantage of qualitative methods is also that it is useful in contexts where there is general lack of understanding of the phenomenon and exploring it would add value to this understanding and enhance the system itself (Story et al., 2017). In the field of Information Systems, it remains vital to understand how the user perceive the system. Based on the information acquired from this, product and service enhancement could be realized best. Hence, according to Kaplan & Maxwell (2005) qualitative research is especially beneficial in the development of explanations of the actual events and processes that led to explicit results which on one hand can help in identifying potential problems in the initial phases as they are forming, thereby providing feedbacks to improve the system whereas on the other hand, increasing the credibility and usefulness by taking individual’s perspectives into consideration. The various types of qualitative research are for example i) case study research, ii) action research, iii) design science research, iv) ethnography and v) grounded theory (Creswell, 2017). In addition to it, there is also a multimethod approach which adopts more than one method. This research falls under case study research as it focuses on the case of Cardiology within the healthcare domain. This is elaborated further later.

The second type of research approach is quantitative methods where the researcher undertakes empirical measurements of variables in order to study how they affect each other. How a dependent variable (Y) is affected by an independent variable (X) is the primary goal of a quantitative research (Creswell, 2017). Based on the data, a model is built which includes variables and intercept/slope and this model can be used to predict the change in dependent variable if there is a certain percentage of change in independent variable (Livari & Huisman, 2007). Quantitative methods are objective in nature and are used for theory generation in most cases or for hypothesis testing where a hypothesis is either accepted or rejected (Castellan, 2010). A mixed method approach includes both qualitative and quantitative methods and uses them to answer the research question. The Table-5 below summarizes the characteristics of both qualitative and quantitative approach.

Abbildung in dieser Leseprobe nicht enthalten

Table 5: Characteristics of Qualitative and Quantitative Approach (Source: Creswell, 2017)

As this research will be of qualitative nature, it is also important to talk about the different types of it. According to Livari & Huisman (2007) the various types of qualitative studies are for instance: Case study, action research, design science, ethnography and grounded theory. In some cases, more than one particular type is used. Within the qualitative methods, this research which deals with the role of big data in the healthcare domain with particular focus on cardiology would fit into the Case Study best. The tools used are semi-structured interviews and document analysis in the form of literature review. According to Yin (2003) a case study is appropriate when i) the aim of the research is to answer ‘why’ and ‘how’ questions, ii) the nature of the case cannot be manipulated, iii) the contextual conditions are relevant to study the phenomenon, iv) the boundaries between the phenomenon and context are not clear. The study of the role of big data in Cardiology sought to investigate how the data can be implemented in this domain which is strongly determined by the established procedures, nature and type of disease and the tools used in this process. The case could not be considered without the context as the context differs significantly even without a closely related medical domain. It is within the setting that the data can be utilized for optimization. The research would be biased if the context is not considered. Due to the unique nature of healthcare, careful consideration is imminent before applying any economic and business theories in it. Whereas the rule of demand and supply is a powerful and mostly used mantra in classical economics, it has less sense in healthcare. Patients do not demand more healthcare if the price is cheaper and the demand does not get less is the price is higher as health-related demands is based on needs and not on wishes (Arrow, 1967). While talking about data, one type of data plays a vital role in one particular domain within healthcare whereas other types are practical elsewhere. Genomics is mostly used in Cancer research and precision medicine whereas biomedical information collected through Pedometer and Accelerometer would be most appropriate in Cardiovascular research. Hence, a case by case focus is important in health-related study giving qualitative research in the form of Case Study various advantages. Yin (2003) has summarized various types of Case Studies, as mentioned in Table-6 below:

Abbildung in dieser Leseprobe nicht enthalten

Table 6: Types of Case Studies (Source: Yin, 2003)

This research would fall under the category exploratory case study as it is aimed at exploring the intervention of big data in the field of Cardiology and the outcome is not clear-cut because there are many contextual issues that need careful consideration and multiple perceptions of parties involved. Furthermore, evaluation of literature and qualitative interviews are subjective in nature where the author puts main focus on the information provided by the authors and interviewees. The explanatory and descriptive case studies do not fit best in a research of this type as the intention of the research is not to understand the causal link, as explanatory case study describes. However, descriptive case study could have been used but as the research will talk about the future potentials of big data too, one cannot say that it has been analyzed in a real-life context.

Even though case study is considered as a part of qualitative approach, it can also be used in quantitative methods especially where objective examination is essential. Furthermore, in contrast to other methods like grounded theory, there are no strict methodological requirements in a case study giving the researcher own initiative and creativity to design and conduct the research (Meyer, 2001). However, Meyer (2001) also emphasizes, as Crowe et al. (2011) that this can also be a disadvantage and subject to criticism if performed with a sub-optimal quality. Hence, special clarification should be provided in terms of the choice of design, data collection, analysis and validity and reliability. Starman (2013) has emphasized that other advantages of a case study is its conceptual validity, that there are topics where researchers are interested to understand but are hard to measure, possibilities to derive new hypotheses that can be added to existing knowledge, they can explore causal mechanisms and model complex causal relations. At the same time, the author also mentions some drawbacks that come with this. They are for example the constraints in generalization of findings, existence of confirmation bias i.e. there is a likelihood that the study is done to justify researcher’s preconceived notions infuriating to develop general propositions or theories (Starman, 2013). The limitations and biases of case studies can largely be reduced by respondent validation ( i.e. cross checking the findings and researcher’s interpretation) and by offering transparency during the whole process, like in the choice of respondents, methods used in data collection and analysis and interpretation of results (Mays & Pope, 2000). This contributes to the trustworthiness of results. Health sector enjoys the benefit of having similar characteristics internationally in terms of the nature of disease groups and treatment methods. Hence, the impact of a certain factor, like big data can be generalized to a high extent if the fundamentals are in place. Although there might be differences in terms of healthcare systems, other socio-economic factors and the infra-structure, the potential of big data can be largely generalized if the prerequisites are fulfilled. As a result, the drawbacks of a case study methodology might have little impact in this research.

2.3 Data Collection

This section will highlight the methods that were used in order to collect the data for this research. The method for interviews and literature review will be presented separately for the sake of clarity.

In qualitative research, data can be collected in the form of interviews, observations, ethnography, document analysis etc. (Creswell, 2017). This research is based on two forms, interviews and document analysis which includes comprehensive literature review. There are three types of interviews in research methodology: structured interviews, semi-structured interviews and unstructured interviews. A structured interview consists of close-ended questions with less room for flexibility allowing only pre-coded answers, whereas unstructured interviews consists of open-end questions allowing the interviewee to deviate to some degree within the subject area. The semi-structured interviews compose ingredients of both structured and unstructured formats (Jamshed, 2014). According to Mason (1994) no interview lacks any structure, hence, the expression ‘unstructured’ is vague. So, the types could be better expressed as ‘lightly structured’ according to the author. In this research, a lightly structured method was embraced to give more flexibility to the interviewees as the topic of big data within the healthcare domain is relatively new and to allow the respondents to express in their own way without holding on their responses.

Table-7: Information about the Participants

Abbildung in dieser Leseprobe nicht enthalten

2.4 Methodology Applied for literature review

It is standard to follow a procedure in research. The processes carried out based on an established procedure will ensure quality and validity of the research. For the literature review, the author follows the following review process suggested by Creswell (2017) which contains four different steps, portrayed in the Figure-1 below:

Abbildung in dieser Leseprobe nicht enthalten

Figure 1: Literature Review Process

2.4.1 Inclusion and Exclusion Criteria

It is standard to determine the criteria to include or exclude publications based on their distinct features which should fulfill the aim of the research question in order to answer it at the same time (Patino & Ferreira, 2018). Inclusion criteria can be defined as the major characteristics or criteria that a particular publication fulfills which the researcher aim to study and can be used for this purpose for further analysis whereas exclusion criteria is the set of standards determined beforehand by the researcher, who, purposefully justifies why publications that contains a predefined characteristics should not be considered for further analysis because it would either cause bias in the research, or is beyond the scope of the study or simply cause conflict of interest (Salkind, 2010). The aim of both inclusion and exclusion criteria is to fulfill the scientific requirement of the research, to contribute towards internal and external validity and remain in-line with the pre-defined objectives and research questions (Garg, 2016).

As the focus of this research is on the role of big data in the field of Cardiology, the inclusion criteria were the selection of papers that dealt about the role of data in Cardiology concerning the use, trends, potentials, issues, challenges, topics and future directions. The field of Cardiology includes a wide range of diseases as mentioned above which has the heart in the epicenter. Hence, papers focusing on any of these diseases were taken into consideration. Furthermore, also papers that dealt with Cardiology in combination with other diseases were also considered, for example Cardiovascular diseases. Apart from the diseases, other aspects in Cardiology includes disease management, the role of smart wearable technologies and public health. Papers addressing them were included in the inclusion criteria. For the initial understanding of big data, papers dealing with general information and features of big data were also included. Due to the relevance and the changing nature of big data, only publications after the year 2005 were included as the technology that existed before that period has become obsolete and has been replaced by other advanced methods. The majority of ICTs that are used in data analysis today were developed after this period and the use of data in Cardiology remained minimal. Hence, including the resources and technologies that existed from the post 2005 period would contribute to internal validity of the research.

Exclusion criteria for this research was any publication that dealt with big data in health but was not focusing on Cardiology or related areas. Also, papers that dealt with only another sub-field of health that was not related to Cardiology was excluded. Papers published before 2005 were excluded due to less relevance. In addition to it, publications in other languages than English were excluded. It cannot be denied that non-English language papers have high quality, but for the sake of comparability, verifiability and the limitation of author’s understandability, they were not considered for the review. Masters and PhD dissertations were excluded. Finally, news and magazine articles, publications that were not moderated by an expert panel were also excluded because they do not represent the scientific basis that is the requirement for a consideration in any academic paper.

2.4.2 Search Procedures

Abundant research has been carried out about big data and analytics. A big number of them have chosen the healthcare domain and many have focused on Cardiology and fields related to Cardiology. Big data in the field of Cardiology means not only the medical aspect of it rather also the care process, technology that are related to heart health, prevention aspects, public health, policies and more. The PRISMA model has been used for the selection process. In order to ensure the relevance, the author adopted following steps:

Step 1: The key words, “big data” / “big scale”, “health” / “healthcare” and “Cardiology” were used in the respective search engines. Results comprising all three words in each set were then used as the main keyword to search for research articles published after 2005.

Step 2: Publications that had the title or summary which contained the combination was selected as a potential set of papers to be reviewed. They were verified again by going through the abstract or summary and the keywords were checked. This strategy is particularly useful in order to omit the papers that included the keywords but did not mean them. After this process only the paper qualified to be a candidate for the research.

PhD and master’s Dissertations: Prior to the start of this dissertation, the author searched for masters and PhD dissertations in the area of big data in healthcare to gain knowledge about methodologies used and the outcome of the research. Here, the focus was not necessarily in cardiology rather all fields of medicine. Google search engine was used to look for publicly available dissertations. However, the primary goal of this task was only to gain understanding of the thematic and not to use for analysis for this research. Two dissertations were thoroughly analyzed but not used for literature review.

Scholarly Databases: There are a large number of scholarly databases. Some are subject specific, for example ASCE Library which offers publication in the field of civil engineering or AULIMP which offers resources related to military science. While searching for databases offering journal articles, the author found at least 47 different databases that were related to health sciences, computer sciences or multi-disciplinary. Among them, five most popular were chosen to use for this research and they are:

- Scopus
- Science Direct
- Web of Science
- Ebsco
- IEEExplore

The following Figure-2 visualizes the search procedure. PRISMA model is used for this purpose as it is the most used flow diagram to visualize the search process. The strength of this model is that it starts from the beginning when all available publications are considered and then narrows down to the ones adopted for the review. It is intriguingly simple to use, and the template is readily available from the official website.

[...]