See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/334029064

The emerging role of cognitive computing in healthcare: A systematic literature

review

Article  in  International Journal of Medical Informatics · June 2019

CITATIONS READS

24 3,373

3 authors:

Rajat Kumar Behera PRADIP KUMAR BALA

KIIT, Deemed to be University Indian Institute of Management Ranchi
20 PUBLICATIONS   98 CITATIONS    63 PUBLICATIONS   563 CITATIONS   

SEE PROFILE SEE PROFILE

Amandeep Dhir
Universitetet i Agder
156 PUBLICATIONS   3,985 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Food waste View project

Software Engineering View project

All content following this page was uploaded by Rajat Kumar Behera on 12 July 2019.

The user has requested enhancement of the downloaded file.

 1

The emerging role of cognitive computing in healthcare: A systematic

literature review

This is uncorrected proof of work that has been published in the following source:

Citation: Behera, R. K., Bala. P. K., & Dhir, A. (2019). The emerging role of cognitive
computing in healthcare: A systematic literature review, International Journal of Medical
Informatics, 129, 154-166, https://doi.org/10.1016/j.ijmedinf.2019.04.024

Notice: Since this is uncorrected proof of the work, copy editing or changes made during the
final publication are not reflected here. The final version of this document is available at:

https://www.sciencedirect.com/science/article/pii/S1386505619301911

© 2019 Elsevier B.V. All rights reserved.

 2

The emerging role of cognitive computing in healthcare: A systematic literature review

Rajat Kumar Beheraa, Pradip Kumar Balab, Amandeep Dhirc,d

a,b
Indian Institute of Management Ranchi, Ranchi, Jharkhand, India
c
Aalto University, Espoo, Finland
d
Optentia Research Focus Area, North-West University, South Africa
a
rajat_behera@yahoo.com,bpkbala@iimranchi.ac.in,c,damandeep.dhir@aalto.fi

Abstract

To assist medical professional in better treatment of diseases, and improve patient outcomes, healthcare has

brought about a cognitive computing revolution. The cognitive computing system processes enormous

amounts of data instantly to answer specific queries and makes customized intelligent recommendations.

Cognitive computing in healthcare links the functioning of human and machines where computers and the

human brain truly overlap to improve human decision-making. In regard to this convergence, this

systematic literature review (SLR) provides comprehensive information of the prior research related to

cognitive computing in healthcare. The SLR focused on methods, algorithms, applications, results,

strengths, and limitation using different research articles collected from leading international databases

using linear and citation chaining search. The main outcomes of the SLR include proposal on future

research direction, challenges faced by researchers, capabilities and the impact of cognitive computing on

healthcare outcome and a conceptual model, showcasing the better utilization of cognitive computing in

healthcare domain. This study concludes with managerial implications, limitations and scope for future

work.

Keywords: Cognitive Computing; Cognitive Computing in Healthcare; Systematic Literature Review

1. Introduction

Digital healthcare has changed rapidly with an increase in the use of electronic health data

produced by medical devices during clinical meetings or events [12]. But, this massive electronic health

data remain largely underused and there is an urgent need to convert this raw data into meaningful,

expressible and time-limited information [6-8]. However, there is lack of supply in data analysts and
 3

scientists, due to which, it cannot meet the demand of ever growing volume of this Big Data [45] [87].

The possible solution is to train the computer systems to perform human work and to facilitate the

management of large volume data, and cognitive computing is a possible alternative.

Cognitive computing is derived from cognitive science and Artificial Intelligence (AI) [2] and is

the development of computer systems modeled on the human brain [91]. Cognitive computing embodies

major brain behaviors of natural intelligence, including perception, attention, thought, etc., and has the

characteristic of integrating past experiences into itself as an emerging paradigm of intelligent computing

methodologies [92]. Cognitive learning is the function used to simulate cognitive processes such as

thinking and remembering operations and can be regarded as a mathematical tool for cognitive computing

[94]. By analyzing the cognitive mechanism, building cognitive computing system, and performing

cognitive processes, cognitive operators allow human thought processes to be simulated (e.g. perception,

attention, and remembering something) [93]. Cognitive computing systems in healthcare collect

individual, clinical and social data from different healthcare sources to improve patient engagement [3]

and the multidisciplinary combination of technologies such as Machine Learning (ML), Big Data

Analytics (BDA), AI, Natural Language Processing (NLP), Data Visualization (DV) and Deep Learning

(DL) allows such systems to figure the type and symptoms of a disease from data [1]. In such systems, [4-

5] the cooperation between machine and human beings is intrinsic and ensures that healthcare receives

more data which can be used to solve complex issues. Cognitive computing enables healthcare

professionals to acquire the best judgments from worldwide renowned medical practitioners and reach to

the remote locations and save clinical studies to match more patients to life and its promising value in

healthcare has generated an increasing interest of researchers from academia and industry [63].

Cognitive Informatics (CI) is a trans-disciplinary investigation of computer science, information

science, cognitive science, and intelligence science that investigates the brain and natural intelligence's

internal information processing mechanisms and processes, as well as their cognitive computing

engineering applications [89]. Studying human intelligence and its problem solving applications is a topic

covered in many disciplines, including philosophy, math and logic, neuroscience, psychology, cognitive
 4

science, computer science, etc., and CI and granular computing (GC) are two studies with different

emphases on human intelligence and human-inspired problem solving, wherein CI is the study of natural

intelligence and its mechanisms for processing information and GC explores various levels of granularity

in human-centered perception, problem solving, information processing, as well as their implications and

applications in knowledge-intensive intelligent systems design and implementation [88]. In CI,

objects/attributes are regarded as synapse-connected neurons, and the relationship represents the synapse

in CI, and the brain generates new synapse or relationship between the existing neurons to represent new

information [90].

Application of cognitive computing in healthcare is still in its infancy stage as there have been only

a few literature reviews, but none of them have provided any crucial insights on this emerging domain. As

a consequence, it becomes difficult and confusing to understand and apply the potential value of cognitive

computing on improving the quality of patient care. In addition, researchers may find it challenging to

track and use, such as its capabilities or impact on healthcare. An SLR is, therefore, conducted to capture

relevant literature from diverse sources with the aim: i) to present the published academic research work

on cognitive computing in terms of method, algorithms, applications and results used in the healthcare

industry; ii) to explore the emerging areas of cognitive computing in healthcare; iii) to present the future

direction of cognitive computing research in healthcare; and iv) to propose a conceptual model to

understand the impact of cognitive computing on healthcare organizations‟ performance.

The review paper is structured as follows: Section 2 discusses the background and motivation for

this study. Section 3 discusses the methodology used to carry out this review of literature. Section 4

focuses on analysis and discussion. Section 5 concludes the study with the managerial implication and

limitations. Table A1 captures the selected papers considered for the review. Table B1 in Appendix B

refers to the categorization of research articles and Table B2 refers to the current research on

methodology, algorithms, applications, and results.

 5

2. Background and motivation

There have been three areas of computing, namely: i) tabulating era, ii) programmable era and iii)

cognitive computing era [11] [19]. The tabulating era was the first era of computing, wherein data were

fed into mechanical systems and the computing was primarily performed by calculators, tabulating

machines, and vacuum systems [20]. This era started in 1900 and ended in 1950. The programmable era

was the second era of computing and was completely controlled by the programming, as inflected on the

system [20]. It was a paradigm shift from mechanical systems to the electronic systems wherein enormous

improvement happened in storage and performance benchmarks. Such computing is primary performed

by mainframe, personal computer, and smart computer machines. This era started in 1950 and is still

existence to date. The cognitive computing era evolved in 2011, wherein systems were developed to

understand the way humans operate, through senses, learning and experiences. The main driver of this era

was the sudden exponential increase in the amount of unstructured data, and then understanding, learning

and communicating with people in natural language rather than software code, and being able to extract

meaning and learn from large amounts of unstructured visual, verbal and numerical information and help

people make complex decisions based on them [20]. This era led to the establishment of automated IT

systems that can solve problems without the need for human assistance.

A cognitive computing environment requires sufficient data to detect patterns or anomalies and to

ensure that the analytical results are reliable and consistent [18]. The discussions about cognitive

computing always refer to Big Data (BD) and predictive analytics. BD facilitate the storage of large

amounts of data, and predictive analysis gives the ability to predict what will happen, whereas, in

comparison, cognitive computing gives the ability to learn from further interactions and suggest best

actions. In short, cognitive computing is a technology that carries three core traits, i) NLP, ii) assertion

and recommendations, and iii) continues to learn.NLP focuses on enabling computers to understand and

process human languages, bringing computers closer to understanding language at a human level.

Organizations can deploy cognitive solutions for item or product recommendation with continuous

learning and continuous improvements.

 6

Despite the development of digital technology, the healthcare industry is experiencing major

difficulties as presented in Table 1.

Table 1: Difficulties experienced in healthcare

Sr # Difficulty Description
1 Incomplete [22] Despite the use of optical character reorganization, the digital system
digital platform still has no access to all relevant data because not all of it is in a digital and
readable format.
2 Lack of [22] Vendors such as Amazon, IBM, Microsoft, and Google are known to
cloud-adoption provide the best cloud-based service. However, there are still some
healthcare organizations that are reluctant to transfer their data to the cloud
and, instead, resort to on-site solutions which may have limited abilities and
are potentially more complex.
3 Ever changing Healthcare is highly regulated industry where regulatory and reporting
and evolving requirements continue to increase and evolve with government policy. Such
regulatory requirements need quality reports around measurement like readmission,
requirements safety, and patient experience, and heavily influence pricing and financial
information to the public [23]. Such government-imposed regulations only
add burden.
4 Inconsistent [24] Many clinicians from different groups may have inconsistent views of
variable treatment for the same condition. Such treatment may not always lead to
definition personalized treatment and personalized care plans.
5 Privacy and Medical records are normally targeted by cyber thieves and the stealing of
security such health or identity data costs companies or individuals [25] and causes
potential damage.
6 Limited use [22] The most popular application in healthcare tends to be advanced image
processing and predictive analysis. However, much more can be offered by
the digital system. Interactive bots, NLP, ML and DL are just a few
examples in which only a limited number of hospitals participate.

In recent years, cognitive computing has been one of the most popular trends in healthcare

technology [13] and plays a significant role in improving communications between people, and machines

and has prompted the development of new models for human-machine interactions; the objective is to

transform data into time-bound actionable insights for improved healthcare outcome throughout the
 7

patient's endways journey [21]. Another prospective benefit is the magnifying of trust in people in that

computing devices can provide truthful responses within an acceptable trust range [14]. Rapid

development of BD, ML, DL, and NLP techniques and the ability to handle large amounts of structured,

unstructured, semi-structured data from heterogeneous sources makes cognitive computing a low complex

task [15], for instance healthcare [16]. As a result, in future engineering systems, cognitive computing and

relevant technology will play an important role [17]. Cooperation between humans and the machine is

innate in a cognitive system, which allows healthcare to gain greater value for solving compound

problems from data [37-39].

3. Research Methodology

The methodology for performing SLR is presented in Figure 1. It is broadly classified into i) research

context, and ii) research area. The research context covers i) current research by providing information in

terms of methods, algorithms, application, results, strength, and limitations using different research

articles. Three types of articles were examined, namely, literature review, empirical studies and

mathematical modeling in the digital databases using linear and citation chaining search, ii) future

research direction, iii) cognitive computing capabilities, iv) cognitive computing impact in healthcare, v)

challenges currently faced by researchers, and vi) conceptual model. Future research directions are based

on the following perspectives: research optimization, personalized knowledge-based medicine, better

disease management, prevention of disease outbreak, fraud detection and prevention in real time,

cognitive Internet of Things (CIoT), optimal treatment trial and visual analytics. The research area is on

healthcare.
 8

Fig. 1. SLR Map. (Source: self-compilation by authors)

An SLR collects empirical data using a formal protocol [46-48] and is typically the collection of

research studies in a variety of fields [49-51] to provide the reader with a broad spectrum of knowledge

on underlying study of current research. The current study conducts SLR on the basis of the guidelines

outlined by [9-10]. There are three different stages, expressly: i) planning the review, ii) conducting the

review, and iii) reporting the review. Each stage is divided into several steps and the process is presented

in Figure 2 from which the tasks from each stage can be easily comprehended.

Fig. 2. SLR organization. (Source: self-compilation by authors)

 9

3.1. Planning the Review

3.1.1. Research Question: Gather current research in terms of method, algorithms, and applications in

cognitive computing with application in healthcare and then present the results.

3.1.2. Inclusion Criteria: The inclusion criteria (IC) used to select the literatures to be included in the

review is presented in Table 2.

Table 2: Inclusion criteria for selection of literatures

IC# Description
IC 1 The used keywords are: “Cognitive Computing”, “Cognitive System”, “Cognitive
Informatics” and Healthcare, “Health care”. The operators of search syntax are OR,
AND.AND operator signifies that both keywords must be present in the search queries
and OR means that at least one keyword must be present in the queries searched.
IC 2 Studies published before December 2018.
IC 3 Studies published in English.
IC 4 Studies limited to document type of journal articles.
IC 5 Include abstract-based studies.
IC 6 Include full-text-based studies.

3.1.3. Exclusion Criteria: The exclusion criteria (EC) used to determine literatures for exclusion in the

review is presented in Table 3.

Table 3: Exclusion criteria to omit literatures

IC# Description
EC 1 Eliminate duplicate studies with matching title and/or Digital Object
Identifier (doi)
EC2 Eliminate studies based on quality evaluation questions and which is
discussed in the Review Protocol Evaluation stage.

3.1.4. Digital Database: The digital databases used to collect the data for the review of papers are i)

Scopus, ii) DBLP, iii) PubMed, iv) ScienceDirect, v) Springer, vi) Sage vii) Taylor & Francis, and

viii) Emerald.

3.1.5. Review Protocol Development: Scopus was first considered to extract data from these digital

databases as: i) compared to other digital databases, the extensive number of studies in connection
 10

with this study are indexed, ii) it is the leading digital literature database reviewed by peers, and

iii) it has extensive scientific and interdisciplinary information. DBLP, PubMed, ScienceDirect,

Springer, Sage, Taylor & Francis, and Emerald have also been browsed for the papers and

included those that are not referred to in Scopus. Furthermore, additional relevant papers matching

to the context of this study were included on the basis of full text with citation chaining search.

Citation chaining included relevant papers using both backward and forward approach in the

literature studies from the above digital databases.

3.1.6. Review Protocol Evaluation: With a view to support the criteria of inclusion, and exclusion, and

the selection of research data, it is vital to examine and evaluate the quality of studies. Indeed, the

objective of quality assessment is to make sure that the results of the study are suitable and

impartial [52]. Thus, a number of quality evaluation (QE) questions were identified in order to

improve this study. Previous SLRs [52-54] inspired the design of this review. An exclusion

criterion, i.e. EC 2 is the composition of QE questions and is presented in Table 4.

Table 4: Composition of QE questions

QE# Description
QE 1 The study contains evidence which is quantitatively or qualitatively analyzed. The probable answers
are: “quantitative research (+2)”, “qualitative research (+1.5)” and “no evidence (+0)”.
QE 2 The study unequivocally examines the benefits and limitations. The probable answers are: “yes
(+2)”, “no (0)” and “partially (+1)”. The score is partial if only one of the study's advantages or
challenges is reported.
QE 3 The output of the study is justifiable. The probable answers are: “yes (+2)” and “no (0)” and
“partial (+1)”. The score is partial if only very limited techniques are explained or one of the
techniques used is not detailed.
QE 4 The study was published in a reliable and recognized source. The probable answers are as follows:
(+2) if the summation of citations number and H Index is exceeding 100
(+1.5) if the summation of citations number and H Index is exceeding lying between 50 and 99
(+1.0) if the summation of citations number and H Index is exceeding lying between 1 and 49
(+0) if the summation of citations number and H Index is 0
QE 5 The study compares the proposed method with other methods and the probable answers are: "yes: +
1," "no: 0".
 11

To ensure the efficiency of this study's results, the quality score is considered a norm for exclusion

and, consequently, only the appropriate studies have been selected with a quality score of meeting or

exceeding fifty percent of the ideal score of 9. Table A6 of Appendix A provides a complete list of

the selected studies with quality evaluation score.

In Figure 3, various stages for data selection are presented, where Si, 1≤ i ≤ 6 represents sequential

stages and the description of each stage can be easily comprehended from it. Oval callouts linked to

each stage are the applicability of respective inclusion criteria(s) and exclusion criteria.

Fig. 3. Research Data Selection Process (Source: self-compilation by authors)

3.2. Conducting the Review

3.2.1. Search Syntax

Table 5 shows the search syntax that was utilized in the SLR for selecting the research papers.

Table 5: Search Syntax of the selected research papers

Data Source Search Syntax

Scopus (TITLE-ABS-KEY ("cognitive computing*") OR TITLE-ABS-KEY ("cognitive
system*") OR TITLE-ABS-KEY ("cognitive informatics*") AND (TITLE-
ABS-KEY ("health care *") OR TITLE-ABS-KEY ("healthcare *") AND
(LIMIT-TO (LANGUAGE ,"English") AND (LIMIT-TO
(EXACTKEYWORD,"Cognitive System*") OR LIMIT-TO
(EXACTKEYWORD,"Cognitive Computing") OR LIMIT-TO
 12

(EXACTKEYWORD,"Health Care") OR LIMIT-TO (EXACTKEYWORD

,"Healthcare*") AND(LIMIT-TO (DOCTYPE,"ar")ORLIMIT-TO
(DOCTYPE,"re")ORLIMIT-TO (DOCTYPE,"ip")
DBLP cognitive system* and healthcare type:Journal_Articles:
cognitive computing* and healthcare type:Journal_Articles:
cognitive Informatics and healthcare type:Journal_Articles:
PubMed (Cognitive Computing[Title/Abstract] OR Cognitive System[Title/Abstract] OR
Cognitive Informatics[Title/Abstract]) AND (healthcare[Title/Abstract] OR
health care[Title/Abstract])
ScienceDirect ("cognitive computing" OR "cognitive system" OR "cognitive informatics")
AND (healthcare OR "health care")
Springer (Cognitive Computing* OR Cognitive System* OR Cognitive Informatics*)
AND (healthcare OR health care)
Sage "cognitive computing" OR "cognitive system" OR "cognitive informatics" AND
health*
Taylor & Francis ("Cognitive Computing" OR "Cognitive System" OR "Cognitive informatics")
AND Health*
Emerald ("cognitive computing" OR "cognitive system" OR "cognitive informatics")
AND health*

The various attributes of the search syntax are: Scopus:(i) TITLE-ABS-KEY: the keywords

selected are searched for in the title, abstract and keywords of the paper, (ii) AND: operator that means

that both keywords in the searched item should be present, (iii) OR: operator that means that one of the

keywords in the searched item should be present, (iv) LANGUAGE: papers written in English are

addressed in this study, (v) LIMIT-TO(DOCTYPE): for this study, only journals were considered. Here,

'ar' stands for article in the journal, 're' stands for article in the review,' ip' stands for article in the press

(vi) LIMIT-TO (EXACTKEYWORD): include the papers matching to the exact keyword, and (vii) * is

the wildcard character that represents one or more character. The search syntax for others can be easily

seen from Table 1.

3.2.2. Data Synthesis

A qualitative meta-synthesis technique is used so as to gain in-depth understanding of method,

algorithms, applications, results, strengths, and limitations of current research. Figure 4 presents the
 13

output of the meta-synthesis technique. The figure depicts i) search strategy, including literature linear

search and citation chaining search, ii) number of studies at each stage of the process, and iii) the

summary of studies selected for synthesis.

Fig. 4. Output of qualitative meta-synthesis technique (Source: self-compilation by authors)

In stage 1, a total of 16,631 papers have resulted, out of which 224 were from Scopus, 5 from DBLP,

13 from PubMed, 840 from ScienceDirect, 11,197 from Springer, 2,111 from Sage, 2,112 from Taylor &

Francis, and 129 from Emerald with the selected keywords, English language and restricted till December

2018. The other numbers pertaining to each stage can be easily understood from Figure 4. Thirty-two of

the papers were found to be relevant to this study. The list of selected papers considered for this study is

shown in Table A1 of Appendix A; the details of the selected papers are given in Table A2 of Appendix

A; the breakdown of selected papers considered for review by year and digital database is given in Table

A3 of Appendix A; the breakdown of the selected papers‟ source H-Index is shown in Table A4 of

Appendix A; and the breakdown of the selected papers‟ source % of international collaboration is

presented in Table A5of Appendix A. It can be observed from: 1) Table A1 that the study by [80] has the
 14

maximum number of citations followed by the study by [85]; 2) Table A2 shows that 2018 contributed

the maximum number of papers followed by 2017; 3)Table A3 shows that Scopus leads the indexing

database followed by ScienceDirect; 4) Table A4shows that the study by [80] leads H-Index followed by

the study by [57]; and 5) Table A5 shows that average international collaboration leads by ACM

Transactions on Internet Technology and is followed by OMICS: A Journal of Integrative Biology.

3.3. Literature review reporting

This section presents the results of this study.

3.3.1. Data Formatting/Organization to demonstrate finding

A search comprising of literature linear and citation chaining was performed in digital databases, as

shown in Figure 3. These papers were thoroughly analyzed to select only the most relevant articles and,

finally, thirty-two articles were included in the study. QEs were applied to the studies to evaluate the

quality of the selected papers on the basis of full texts. Data of such articles were classified, organized and

formatted to demonstrate the finding.

3.3.2. Communication of finding

Figure 5(a) shows the word cloud on selected studies keywords and Figure5 (b) shows the word

cloud on selected studies article title. The keywords word cloud gives an overview of the keywords of the

selected articles and the article title word cloud gives an overview of the titles of the selected articles.

Figure 5(a) and (b) depicts that “cognitive computing” had been closely associated with health and

indicates that this study is unbiased, and trustworthy.

Fig. 5.(a)Word cloud on selected studies keywords; (b) Word cloud on selected studies article title
 15

The paired word analysis had been applied over the keywords to better understand which

keywords are being used together mostly. The analysis depicts the words that have been used often are: i)

cognitive computing with thirteen repetitions, ii) health care or healthcare with six repetitions, iii)

artificial intelligence with five repetitions, iv) Big Data with four repetitions, and v) cognitive systems

engineering, machine learning, personalized medicine with three repetitions each. Thus, this indicates that

cognitive computing and healthcare are strongly connected and this study is unbiased, and trustworthy.

Figure 6 depicts the distribution of articles by 1st author‟s country and it indicates that the United

States has contributed the highest number of articles by 1st author among other countries.

Fig.6: Distribution of paper authors by country

Figure 7 showcases the distribution of papers by year of publication and digital databases, and it

indicates 2018 has recorded the highest number of articles, and Scopus has the highest number of articles

for this study.

MDPI

2 Taylor &
1 2 Francis
1 1 Springer
1
2 3
1 ScienceDirect
1 8
1 4
1 3 PubMed
1

2015 2016 2017 2018

Fig. 7: Distribution of papers by year of publication and digital database

 16

Figure 8 showcases the distribution of articles by subject area and it indicates the majority are

from medicine, computer science, and engineering.

Fig. 8: Distribution of papers by subject area

Figure 9 showcases the citation count of the selected articles over the publication year and it

indicates that 2017 is leading in terms of the quality of citations. However, the quality of citations of 2018

and the other years may lead to growth in the days to come.

Fig. 9: Quality of citations of the selected articles for this study

Figure 10 showcases the distribution of papers by study type. Four study types, namely

experimental, conceptual, review and theoretical, are considered to categorize the articles.
 17

Fig. 10: Distribution of articles by study type

4. Analysis and Discussion

Cognitive computing can investigate a variety of different data types and their interpretation to

generate rich insight [26]. Cognitive computing includes a variety of tools and techniques, including BD,

Predictive Analysis, IoT, ML, NLP, Probabilistic Reasoning and DV [27]. Some of the cognitive system's

key features are: learning skills, knowledge improvement without reprogramming, development and

hypotheses analysis. These processes can be categorized as: i) observation, ii) interpretation, iii)

evaluation, and iv) decision [28].

The healthcare industry involves many different players who support patient wellbeing and

treatment. The data managed and used by various healthcare players are presented in Table 6.

Table 6: Different players in healthcare and data management role

Sr # Player Description
1 Patient Produces health data such as personally identifiable information, test results,
and previous medical history in non-digital format.
2 Healthcare Person or companies provides healthcare services and produces a) patient
providers medical records, b) data from medical devices and sensors, c) records of
hospital admissions, d) books of medical text, e) medical journals articles, f)
clinical research studies, g) regulations reports, h) billing, and i) cost data.
3 Pharmaceutical Data to support pharmaceutical research, clinical trials, drug efficacy,
firms healthcare provider‟s prescriptions.
 18

4 Healthcare payers Institutions that pay healthcare providers, including insurance companies,
private employers, the government and individuals, and generate data on
billing and usage review.
5 Government Produce regulatory data.
regulatory
services
6 Healthcare data Produce data on the use and effectiveness of prescription drugs, health
service providers terminology taxonomies and software solutions for the analysis of health
data.
7 Healthcare Individuals or companies providing health-related advice or reports for the
information betterment of society
service
8 Healthcare A group of researchers which works or leads independent research and
research center development of healthcare projects in a variety of topics and produces
research and reports data
9 Medical device Produces reports and research data. A patient‟s personally identifiable
manufacturer information can be characterized by the profiles a) core profile, b) health
profile, c) lifestyle profile, and d) social profile having full access to
records. This limits the information the system can see and causes the
medical record to be incomplete and is presented in Figure 11.

Fig. 11. Patient personal identifiable information profile

Core profile comprises of i) demographics like name, address, date of birth, contact number, and

marital status, etc.; ii) life changes like employment, divorce, and marriage, etc.; iii) family relations like

spouse, daughter/son, and grandchildren. Health profile comprises of medical report and biometric data.

Lifecycle profile comprises of health habits, hobbies, obesity, etc. Social profile comprises of wellness

activities, social interactions, chaplain visit, social services, etc.

 19

It can be easily made out and understood from Figure 12 that all the players have access to

different data sources and the government primarily controls and manages the regulatory requirements.

Fig. 12. Access to data by different players of healthcare industry

Figure 13 represents a conceptual model showcasing the better utilization of cognitive computing. In

the conceptual model, healthcare data generated by different players are in structured, unstructured and

semi-structured form and hosted in cloud with data-as-a-service offering.

Fig. 13. Conceptual model for better utilization of cognitive computing

Different components of the conceptual model are shown in Table 7.

Table 7: Conceptual model components

Component Description
Analytics services Services such as a) descriptive analytics, b) diagnostics analytics, c)
predictive analytics, d) prescriptive analytics, and e) cognitive analytics
describing what the best action is.
Continuous learning The distinct tasks are a) model specification, b) hypothesis generation, c)
 20

processing process hypothesis scoring, and d) model verification. The model has to learn
with distinct task continuously.
Presentation and The advantage of cognitive computing is that the healthcare professional
visualization of gets cognitive insights from all these types of data more easily and acts
information confidently and optimizes their decision-making to generate new data,
leading to future innovation.

Feasibility analysis: A huge volume of data, such as digital images from CT scans and MRIs,

medical device reports, patient medical records, clinical trial results, and billing records, are created and

managed by the healthcare ecosystem. Such data exist in a variety of formats ranging from manual paper

records and spreadsheets to unstructured, semi-structured, structured, and streaming data format. Some of

them are well-integrated, but most of them are not. As a result, significant challenges are posed due to the

vast amount of generated data and their analysis. Therefore, cloud computing and a distributed

architecture is the basic model needed to make cognitive computing operational on a large scale in a cost-

effective manner. The need to find patterns and outliers in structured, semi-structured and unstructured

data can help to improve patient care, which is one of the persistent challenges. Additionally, the

healthcare professionals get the required insights from all types of data to act confidently and optimize

their decision-making using different analytics, such as descriptive, diagnostics, prescriptive, etc., with a

continuous self-learning process by combining different technologies such as BD, ML, AI, DL, NLP, etc.,

thereby enabling significant improvements in outcomes.

4.1. Current research areas

Research articles from digital databases are categorized and presented in Table B1 of Appendix B.

Dimensions used for categorization of the journal articles are presented in Table 8.

Table 8: Journal articles categorization

Category Category Details

Settings Clinical–studies were conducted in real-world clinical setting.

Simulated - studies were conducted with simulated settings
Data collection methods One or more of the methods, such as surveys, interview or observation, etc.
 21

Participants Doctors, nurses, patients or other staff

Output What has been proposed by the authors, e.g. system or framework or functional
prototype or strategic solution, etc.

Current research in terms of method, algorithms, applications, results, strength and limitations are

presented in Table B2of Appendix B at an abstraction level. From the current research, the conclusion can

be drawn that cognitive computing in healthcare is an encouraging topic and is used to mend doctor-

patient gaps.

The authors‟ opinions of cognitive computing system in healthcare are that: i) it is a data driven

approach and can offer a better user Quality of Experience (QoE) in an emergency case; ii)it holds the

possibility of accurate, problem-list-centric patient record summarization, conceivably leading to greater

efficiency, better support for clinical decisions and improved patient care quality; iii) historical data

analytics can be used by clinics and hospitals to optimize the allocation of resources and workflows; iv) it

improves usefulness and enhanced usability of orders for electronic consultation; v) it affords intelligent

decision-making and decision support; vi) it relies on tried-and-tested ML and DL algorithms; vii) it

adopts to a cognitive system engineering process approach; viii) it has cost-effectiveness, i.e. lower costs

at the same quality of care or increased costs and improved quality of care at an acceptable incremental

cost per incremental unit of quality of care; ix) it affords a cultural change in the practice of medicine, i.e.

physicians and intelligent recommendation to facilitate on how patients are to be medicated; x) it assists

in the creation of individual treatment plans and, thus, improves the experience of patients and doctors;

xi) it helps tired radiologists to find anomalies of interest in images quickly.

4.2. Future direction of research

Cognitive computing has huge potential to transform the entire value chain of healthcare, from

discovery of drug to personalization of patient, improved clinical outcomes, more efficient management

of public health, and more wide-ranging recompense of care. It is exciting to perceive how cognitive

computing can further improve and expedite the union between clinicians, policy makers, players and
 22

researchers for cost optimization, reduction of risks and improvement of personalized care. The future

directions are presented in Table 9.

Table 9: Future directions

Sr # Research direction Description
1 Research Researchers from both industry and academia can a) predict disease trends in modeling future
Optimization demand and costs; b) improve the design and analysis of clinical trials, accelerate research times
and the accuracy of results; c) enable drug developers to reduce the risk of new drug products
and to translate research results into industry and regulatory practices methods and technologies.
2 Personalized Researchers can explore a) medical treatment tailored to the individual medical attributes of each
Knowledge-Based data-rich patient record and then examine the links between genetic variation, predisposition to
Medicine disease and specific drug responses, which will allow early detection and diagnosis before the
symptoms and reduce complication; b) adjust the use of genetic variation therapies and adjust
medicine doses to reduce side effects.
3 Better disease Researchers can explore on a) integrating personalized knowledge-based medicine and
management evidence-based care to better manage the disease; b) remote monitoring of patient in real-time to
monitor adherence to prescription and improve future treatment options; c) use advanced
analytics in patient profiles to detect anomalies and identify high risk patients with a specific
disease.
4 Prevent disease Researchers can explore in data patterns to determine possible outbreaks of infectious diseases
outbreak and the efficacy of vaccination programs, trends, such as alcohol-related emergency room visits
and accident injuries at home, that are typically more difficult to analyze.
5 Fraud detection and Researchers can investigate in real-time fraud, anomalies in the system of refunds and regulatory
prevention in real breaches.
time
6 Cognitive Internet IoT-based systems that implement real-world applications are constantly evolving and
of Things (CIoT) generating startling requirements, [40] requiring coordination of management processes and
interaction with people to learn from their intelligence and present more precise analytics. As a
result, a new IoT era named "Cognitive IoT" [41] was announced. Because mobile health
applications and wearable devices are increasingly used in daily lives [42-43], Cognitive IoT is
becoming one of the most popular trends. Researchers can explore on common Cognitive IoT
sensor of human body for real-time sensing of body signals and how it can helpful for the
doctors for real-time diagnosis and treatment service. [2] common Cognitive IoT sensors are i)
brainwave sensor- such sensor is placed on the scalp to detect brainwaves; ii) blood pressure
 23

sensor–a non-invasive sensor designed to measure systolic, diastolic and mean arterial blood
pressure utilizing the oscillometric technique; iii) electromyography sensor- measures the
electrical activity of muscles; iv) blood glucose sensor - measures the level of glucose in the
interstitial fluid and changes it into an electrical signal. The signal represents the amount of
sugar in the blood; v) electrocardiogram sensor- detects the electrical and muscular functions of
the heart; vi) skin sensor- measures a pertinent parameter of the skin (e.g. moisture, sebum); vii)
blood saturation sensor–monitors the oxygen saturation of patient‟s blood; and viii) motion
sensor- detects movement alerting medical staff to the patient's movement in the hope that the
patient will not fall. Figure 14 represents the layout of common human body sensors, which can
be especially applicable to the elderly and patients with chronic diseases. [44] Consumers can
even use IoT fitness trackers to monitor patients with health problems, such as cardiological and
oncological patients.
7 Optimal treatment Researchers can explore different ways to increase the transparency of different services and
trial drugs due to the wide existence of different healthcare providers‟ practices, results and costs.
This can be done by analyzing the results of a wide range of patients and then comparing the
effectiveness of various interventions and, thus, reducing the effects of over-treatment and
under-treatment. This will also enable patients to interact and make the best decisions with
clinicians.
8 Visual Analytics Researchers can explore a) modernistic techniques for visualizing data and its application to
healthcare data; b) areas and healthcare players that require more attention to patients, clinicians
and researchers; c) how storytelling can be applied by making a two-part effort to obtain
valuable insights. The data that feed the chart, graph or interactive dashboard must first be
timely, detailed and completely reliable, and, secondly, the visualization must present the
information in a clear, attractive and intuitive manner while adhering to the best practices of
cognitive computing, dashboard and scoring methodology.

Fig. 14. Common human body sensors layout

 24

4.3. Challenges faced by researchers

The challenges faced by the researchers are presented in Table 10.

Table 10: Challenges faced by researchers

Sr # Description
1 [18] Cognitive computing system builders need to collect sufficient relevant knowledge to
be useful and to represent it in a sense that adds to the knowledge of the system. To capture
and represent the knowledge of healthcare, experts who know vocabularies and rules of
healthcare are required; furthermore, they need to explain the codification of machine
learning and deep learning. In order to build the cognitive computing system, taxonomies
and ontologies with focus on a specific area of knowledge should be defined. However, they
argue that it is not possible to acquire enough knowledge to design a cognitive computing
system that replicates the healthcare industry, not even with the assistance of healthcare
industry experts.
2 [31] To succeed with deploying cognitive computing applications, it is essential to have
clarity regarding responsibilities between human users and the cognitive computing
application. Although a cognitive computing system assumes some prior responsibilities, it
also creates new tasks for humans, such as training and sustaining the cognitive computing
application. Without supervision, a cognitive computing system will lose relevance over
time and be able to handle fewer of its assigned tasks successfully since healthcare products
and customers change with time and new policies and rules for business might be required.
Thus, healthcare encounters the challenge of ongoing supervision of the cognitive
computing system, to monitor the performance and regularly recalibrating it to result in
correct outputs. Supervision is necessary to maintain and manage the quality, as well as to
ensure that the cognitive computing system retains its accuracy and relevance.
3 [32] Explains that cognitive computing systems need to be provided with related
dictionaries and thesauruses to enable language understanding. Interpretation implies the
ability to understand data, to derive the significance of sentences and paragraphs in a
language that goes beyond the definitions of terms. Cognitive computing systems differ
from keyword search and text analysis by being able to understand verbs, adjectives and
prepositions, allowing them to understand what language actually means rather than just
what it says
4 The language makes possible [33] for cognitive computing systems to help to understand
the world better, as well as to engage with us. But language in this case is not as simple as,
 25

for example, English or Chinese. The special language in healthcare can consist of chemical
symbols, medical images, or be embedded in legal terms. The challenge is to teach systems
structure, vocabulary of spoken and written languages, as well as business-specific words
and concepts.
5 A cognitive computing system can show [34] its full capability only through sufficient
training. To do this, question-answer pairs needs to be produced in natural languages.
Producing a sufficient number of question-answer pairs requires a lot of work. Training
cognitive computing system can be tedious and not effective if certain guidelines are not
observed.

4.4. Cognitive computing capabilities

Cognitive computing systems can help us to overcome our limitations in many situations [33]. Even if

we are in the early stages of cognitive computing, the next decade will stretch the limits of what is

possible with new software and innovations [18]. The possibilities of cognitive computing systems are

almost endless, its ability to handle complex tasks such as NLP, data mining, classification, and

knowledge management enables the cognitive computing system to make very sophisticated tasks and

answer complex questions [35].Through implementing cognitive computing, healthcare personnel can

focus on more important business-oriented or strategic initiatives and save hours of staff time and also

reduce the actual detection and resolution time [36].The different capabilities that [29]differentiate

cognitive computing systems from traditional programmed computing systems are presented in Table 11.

Table 11: Cognitive Computing Capabilities

Sr # Capability Description
1 Understand Similar to humans, cognitive computing systems understand medical
images, natural languages, and other unstructured data, in addition to
structured data like hospital EMR and claims/reimbursement.
2 Reason Cognitive computing systems can reason, grasp underlying concepts, form
hypothesis, and infer and extract ideas.
3 Learn With each data point, interaction and outcome, a cognitive computing
system develops and sharpens expertise, so it never stops learning.
4 Interact With the abilities to see, talk and hear, cognitive computing systems
naturally interact with humans.
 26

4.5. Cognitive computing systems’ impact on healthcare

Cognitive computing systems process massive amounts of data to understand, reason and then learn

from it and interact with the healthcare providers to develop enhanced treatment or care plan for the

patients. Figure 15 represents the impact of cognitive computing on healthcare. Such system takes various

forms of data input, such as claim, patient-generated, clinical, medical literature, patient profile, medical

imaging and, after processing, produces cognitive insights and probabilistic recommendation.

Fig. 15. Impact of cognitive computing systems on healthcare

The impacts in terms of cognitive insights and probabilistic are presented in Table 12 and are

based on the study [30].

Table 12: Cognitive Computing Impact

Sr # Impact
1 Provide timely insights for individuals.
2 Proactively identifies at-risk patients.
3 Predicts patient health needs and cost care.
4 Supports personalized medicine and clinical decision-making.
5 Recommends intervention based on probability of success.
6 Improves patient engagement and communication across care settings through the
use of technologies like mobile health app and wearable sensors.
7 Through the use of NLP technology, correlates relevant clinical studies for
knowledge-driven and data-driven support.
8 Helps healthcare players to better understand the population.
 27

9 Help researchers to analyze genomic data.

10 Help researchers to tie makers of certain diseases to many environmental and
personal factors that affect an individual‟s health.
11 Guides physicians to provide precision medicine that varies with each patient.
12 Merges advanced image and textual processing with visual reasoning abilities that
can identify the relevant information in the image.
13 Leads to innovate ways of using intensive care unit (ICU) data and to integrate
mobile monitoring data with EHRs while giving feedback to students.
14 Improves predictive modeling used in health risk stratification and health financial
risk management.
15 Helps patients to optimize their help through personalization information and
social support.
16 Converts unstructured documents into structured data, which helps to improve
communication in care settings.

The impact success can be measured by plotting an iron triangle wherein one side of the triangle

represents efficiency in time of treatment or diagnosis, another side represents cost saving and the final

side represents the personalized care. The three sides of the iron triangle should focus on managing and

exceeding patient expectation. The iron triangle is presented in Figure 16.

Fig. 16. Iron triangle representing the impact of cognitive computing in healthcare

5. Conclusion

This paper analyzes the existing academic literature in the field of cognitive computing and

healthcare. We followed the systematic literature review approach in this study and used digital databases

such as DBLP, PubMed, ScienceDirect, Springer, Sage, Taylor & Francis and Emerald to extract the

information. The SLR discussed methods, algorithms, applications, results, strengths, and limitations of

thirty-two articles using linear and citation chaining search.

 28

The review of prior literature shows that cognitive computing is the buzzword in healthcare and can

be seen alongside of AI, ML, DL, BD, and BDA. As a result of an explosion in data creation, it is

virtually impossible for a human to keep a tab on all the latest developments for decision-making

processes. Cognitive computing deals with complex situations characterized by uncertainty and

ambiguity, i.e. deals with problems of a human nature. Cognitive computing systems often weigh

conflicting evidence and propose a response that can be considered best instead of right. From this

summary, it can be argued that cognitive computing in healthcare is promising. Cognitive computing

systems are capable of capturing the process of human thinking and then learning from the errors when

they are committed by the system.

The study presented the future research directions, challenges faced by the researchers, capabilities

of cognitive computing and its impact on healthcare. The domain of cognitive computing in healthcare

will be incomplete without harnessing the benefits of cloud adoption. A conceptual model has been

proposed in this study that needs to be tested and verified to validate this model.

The key findings for researchers from this study are: 1) Academic literature has been present in this

combined field of cognitive computing and healthcare since 2013, but the emphasis has been on the year

2015. 2) While the future for cognitive computing may be very promising, some significant hurdles still

need to be overcome. 3) To date, international collaboration has hardly existed and is needed to highlight

the contexts and trade-offs in such research explorations between localization versus globalization. 4)

Most of the experiences reported come from the United States, where the healthcare system is organized

in a peculiar way, which is quite different from most other countries.

5.1 Managerial Implication

For the past few years, healthcare industry leaders have understood that, if unique data are captured

before the competitors find it, they can have competitive advantages. Slowly, the industry has started

integrating data across silos such as claims, patient-generated health assessments, clinical, and the most

important medical literature. The leaders of the industry understand that, if a meaningful relationship or

patterns are extracted from such data, information can be turned into knowledge to anticipate the change
 29

and to shape the future. In Figure17, such progress is presented with a technology driver. Progression

from data to information occurs through analytics and the progression from information to knowledge

occurs through cognitive computing. In reference to healthcare, the anticipated changes might be the

implementation of emergency department (ED) improvement strategies anticipating challenges and, then,

taking steps to prevent them. In relevance to ED, the future can be shaped better with better patient care

by not operating on a „first come, first served‟ basis, rather with patients being categorized and attention

given to the patients who need urgent help first.

Fig. 17. Progression of data to knowledge through analytics and cognitive computing

By means of advanced analytical algorithms and by combining diversified healthcare data,

cognitive computing systems uncover insights that were earlier beyond computational capabilities.

Without cognitive computing, people required to discover patterns and insights manually. Even with

plenty of time, researchers may miss the patterns and insights in health records. By contrast, if such large

data are processed using a cognitive computing system, the knowledge that an army of resources would

have required can be gained.

Cognitive computing systems facilitate sharing of knowledge through open question answering

(Open QA) system. It solves the situation in which ambiguity and uncertainty exist and attempts to

imitate the human brain‟s mechanism.

There are many limits to traditional analytics - problems need to be predefined, there‟s no

capacity for handling ambiguity, semantics for structured and unstructured data must be known, and

interaction with the end user is through formal digital means. Cognitive computing, however, opens up

possibilities where machines can learn new problem domains, reason through hypotheses, resolve

ambiguity, evolve towards more accuracy, and interact in natural means. This creates vast opportunities

for complex problem solving across all players and reduces medical treatment cost. It allows doctors to
 30

better understand what tests are to be performed to better understand the patient's health problem,

diagnose further problems and diseases, find appropriate solutions and provide the best possible care.

Hospitals can determine which patients are more likely to develop a certain disease or disease. Post-

discharge results may be controlled and the number of readmissions significantly reduced.

5.2 Limitation

The different limitations of cognitive computing in healthcare are: First, limited risk analysis:

whereas unstructured data are criticized for their lack of organization and difficulty in translating into

electronic health records, cognitive computing systems fail to analyze the risk that such data lacks. This

includes socioeconomic, cultural, political, and human factors, e.g. children in lower incomes households

are more likely to develop chronic problems, and taking all children with chronic problems, poor children

are more likely to have adverse health outcomes. In such cases, for complete risk analysis and final

decision-making, human intervention is necessary. Second, meticulous training process, i.e. initially,

cognitive computer systems need training data to fully understand and improve the process. The

painstaking process of training cognitive computing systems is most likely the reason for its slow

adoption. In addition, it is made even worse by the complex and costly process of using cognitive

computer systems. Third, lack of automated critical decision: cognitive computing systems complement

individual intelligence and analysis but depend on humans in taking vital decisions, and are lacking an

automated critical decision-maker.

 31

Appendix A

Table A1: Papers considered for this study

Study # Reference 1st Author No. of Citations
Country (As of Dec last 18)
S1 Hossain & Muhammad, (2018) [55] Saudi Arabia 10
S2 Mezghani et al., 2018 [56] France DNA
S3 Uddin & Hassan, 2018 [57] Norway DNA
S4 Chen et al., 2018 [58] China 20
S5 Bini, 2018 [59] United States 2
S6 Jayanthi et al., 2018 [60] India DNA
S7 Carbonaro et al., 2018 [61] Italy DNA
S8 Coccoli & Maresca, 2018 [62] Italy DNA
S9 Ahmed et al., 2017 [63] United States 10
S10 Leyens et al., 2014 [64] Netherlands 21
S11 Nemeth et al., 2016 [65] United States 2
S12 Amuthadevi & Monicka, 2016 [66] India DNA
S13 Kolker et al., 2016 [67] Turkey 7
S14 Robson & Boray, 2015 [68] United Kingdom 10
S15 Mezghani et al., 2017 [69] France 16
S16 Patel & Kannampallil, 2015 [70] United States 31
S17 Sedig et al., [71] Canada 3
S18 Klatt, 2018 [72] United States DNA
S19 Fox, 2017 [73] United Kingdom 10
S20 Savoy et al., 2018 [74] United States DNA
S21 Chang, 2016 [75] United States 9
S22 Gad et al., 2018 [76] Egypt DNA
S23 Franklin et al., 2017 [77] United States 11
S24 Devarakonda et al., 2017 [78] United States 4
S25 Lintern & Motavalli, 2018 [79] Australia 3
S26 Chen et al., 2016 [80] United States 138
S27 Marshall et al., 2017 [81] United States 3
S28 Murphy et al., 2018 [82] United Kingdom 2
S29 Chen et al., 2017 [83] China 50
S30 Williamson et al., 2018 [84] United Kingdom 2
 32

S31 Chen et al., 2017 [85] China 90

S32 Schnittker et al., 2017 [86] Australia 4

Legend: DNA: Data Not Available

 33

Table A2: Count of source of studies by year

Study Source and Study # (Referred from Table A1) 2015 2016 2017 2018 Total
IEEE Internet Of Things Journal (S1) 1 1
ACM Transactions on Internet Technology (S2) 1 1
IEEE Sensors Journal (S3) 1 1
Future Generation Computer Systems (S4, S22) 2 2
The Journal of Arthroplasty (S5) 1 1
International Journal of Engineering & Technology (S6) 1 1
Journal of e-Learning and Knowledge Society (S7, S8) 2 2
IEEE Pulse (S9) 1 1
Genetic Epidemiology (S10) 1 1
Journal of Cognitive Engineering and Decision Making (S11, S32) 1 1 2
International Journal of Pharmacy & Technology (S12) 1 1
OMICS: A Journal of Integrative Biology (S13) 1 1
Computers in Biology and Medicine (S14) 1 1
IEEE Transactions on Emerging Topics In Comp. Intel. (S15) 1 1
Journal of Biomedical Informatics (S16, S19, S20, S23) 1 2 1 4
International journal of evidence-based healthcare (S17) 1 1
Journal of pathology informatics (S18) 1 1
Progress in Pediatric Cardiology (S21) 1 1
International Journal of Medical Informatics (S24) 1 1
BMC Medical Informatics and Decision Making (S25) 1 1
Clinical Therapeutics (S26) 1 1
Health Information Science and Systems (S27) 1 1
BMC Health Services Research (S28) 1 1
MDPI (big data and cognitive computing) (S29) 1 1
Discourse: Studies in the Cultural Politics of Education (S30) 1 1
IEEE Access (S31) 1 1
Grand Total 3 6 10 13 32
 34

Table A3: Distribution of studies by year and digital database

Digital Database Study # Year No. of Papers
Scopus S1-S8 2018 8
S9, S31, S32 2017 3
S10-S13 2016 4
S14 2015 1
DBLP S15 2017 1
S16 2015 1
PubMed S18 2018 1
S17 2015 1
ScienceDirect S20, S22 2018 2
S19, S23, S24 2017 3
S21, S26 2016 2
Springer S25, S28 2018 2
S27 2017 1
Taylor & Francis S30 2017 1
MDPI S29 2017 1
 35

Table A4: Source by Country, Indexer and H Index (As of December end 2018)
Study Source and Study # (Referred from Table A1) Country H Index
IEEE Internet Of Things Journal (S1) United States 31
ACM Transactions on Internet Technology (S2) United States 47
IEEE Sensors Journal (S3) United States 89
Future Generation Computer Systems (S4, S22) Netherlands 85
The Journal of Arthroplasty (S5) DNA DNA
International Journal of Engineering & Technology (S6) DNA DNA
Journal of e-Learning and Knowledge Society (S7, S8) Italy 10
IEEE Pulse (S9) United States 76
Genetic Epidemiology (S10) United States 87
Journal of Cognitive Engineering and Decision Making (S11, S32) United States 20
International Journal of Pharmacy & Technology (S12) India 12
OMICS: A Journal of Integrative Biology (S13) United States 50
Computers in Biology and Medicine (S14) United Kingdom 68
IEEE Transactions on Emerging Topics In Comp. Intel. (S15) DNA DNA
Journal of Biomedical Informatics (S16, S19, S20, S23) United States 76
International journal of evidence-based healthcare (S17) United States 24
Journal of pathology informatics (S18) India 5
Progress in Pediatric Cardiology (S21) Netherlands 27
International Journal of Medical Informatics (S24) Netherlands 88
BMC Medical Informatics and Decision Making (S25) United Kingdom 56
Clinical Therapeutics (S26) United States 120
Health Information Science and Systems (S27) Portugal DNA
BMC Health Services Research (S28) United Kingdom 83
MDPI (big data and cognitive computing) (S29) DNA DNA
Discourse: Studies in the Cultural Politics of Education (S30) DNA DNA
IEEE Access (S31) United States 36

Legend: DNA: Data Not Available

(Source: compiled by author with SCIMAGO Institution Ranking. https://www.scimagojr.com/)
 36

Table A5: Source distribution by % of international collaboration (As of December end 2018)
Study Source and Study # (Referred from Table A1) 2015 2016 2017
IEEE Internet Of Things Journal (S1) 38.71% 41.09% 33.64%
ACM Transactions on Internet Technology (S2) 57.14% 65.38% 46.81%
IEEE Sensors Journal (S3) 21.32% 26.19% 23.16%
Future Generation Computer Systems (S4, S22) 43.61% 34.47% 32.33%
The Journal of Arthroplasty (S5) DNA DNA DNA
International Journal of Engineering & Technology (S6) DNA DNA DNA
Journal of e-Learning and Knowledge Society (S7, S8) 13.51% 10.64% 8.57%
IEEE Pulse (S9) 5.33% 3.8% 3.8%
Genetic Epidemiology (S10) 25.81% 37.5% 38.57%
Journal of Cognitive Engineering and Decision Making (S11, S32) 13.79% 20.59% 8.7%
International Journal of Pharmacy & Technology (S12) 2.39% 2.17% 0%
OMICS: A Journal of Integrative Biology (S13) 42.11% 51.81% 50.68%
Computers in Biology and Medicine (S14) 26.18% 31.88% 33.91%
IEEE Transactions on Emerging Topics In Comp. Intel. (S15) DNA DNA DNA
Journal of Biomedical Informatics (S16, S19, S20, S23) 22.03% 21.43% 18.14%
International journal of evidence-based healthcare (S17) 35% 7.41% 24%
Journal of pathology informatics (S18) 0% 21.57% 23.81%
Progress in Pediatric Cardiology (S21) 20.69% 12.9% 6.76%
International Journal of Medical Informatics (S24) 24.35% 28.26% 22.34%
BMC Medical Informatics and Decision Making (S25) 23.57% 21.08% 20.69%
Clinical Therapeutics (S26) 24.09% 23.44% 26.2%
Health Information Science and Systems (S27) DNA DNA DNA
BMC Health Services Research (S28) 33.78% 33.7% 30.66%
MDPI (big data and cognitive computing) (S29) DNA DNA DNA
Discourse: Studies in the Cultural Politics of Education (S30) DNA DNA DNA
IEEE Access (S31) 30.13% 41.82% 36.37%

Legend: DNA: Data Not Available

(Source: compiled by author with SCIMAGO Institution Ranking. https://www.scimagojr.com/)
 37

Table A6: Selected studies with quality score

Study # Reference Quality Evaluation
QE1 QE2 QE3 QE4 QE5 Score
S1 [55] 2 2 2 1 1 8
S2 [56] 1.5 2 2 1 0 6.5
S3 [57] 2 1 2 1.5 1 7.5
S4 [58] 2 1 2 2 1 8
S5 [59] 0 1 2 1.0 1 5
S6 [60] 0 2 2 0 1 5
S7 [61] 0 2 1 1 1 5
S8 [62] 0 2 2 1 1 6
S9 [63] 0 2 2 1.5 0 5.5
S10 [64] 0 2 2 2 0 6
S11 [65] 0 2 2 1 0 5
S12 [66] 0 2 2 1 0 5
S13 [67] 0 1 2 1.5 0 4.5
S14 [68] 2 2 1 1.5 0 6.5
S15 [69] 2 1 1 1 1 6
S16 [70] 2 2 2 2 1 9
S17 [71] 0 2 2 1 1 6
S18 [72] 0 2 1.5 1 0 4.5
S19 [73] 1.5 1 1 1.5 0 5
S20 [74] 2 2 2 1.5 0 7.5
S21 [75] 0 2 2 1 0 5
S22 [76] 2 2 2 1.5 1 8.5
S23 [77] 0 2 2 1.5 0 5.5
S24 [78] 2 2 2 1.5 0 7.5
S25 [79] 2 1 1 1.5 1 6.5
S26 [80] 1.5 2 2 2 0 7.5
S27 [81] 0 2 2 1 0 5
S28 [82] 1.5 2 2 1.5 1 8
S29 [83] 2 2 2 1.5 0 7.5
S30 [84] 0 2 1 1 1 5
S31 [85] 2 2 2 2 1 9
S32 [86] 1.5 1 1 1 0 4.5
 38

Appendix B

Table B1: Research Articles Categorization (Study # is referenced from Table A1 in Appendix A)

Study# & Setting Participants Data Collection Output

Reference Method
S1 [55] Simulated Patients Case studies Emotion-aware health system connected with a powerful module for emotion detection
S2 [56] Simulated Patients Case studies Group of design patterns to reduce the complexity of the system design
S3 [57] Simulated mHealth Documents and Robust approach to activity recognition for intelligent healthcare using deep
public dataset Records Convolutionary Neural Network and body sensors.
S4 [58] Simulated Users Documents and Functional prototype of an intelligent healthcare system based on ECC to monitor and
Records analyze physical health.
S5 [59] N/A N/A N/A Basic understanding of AI, cognitive computing and demystify technology to better
understand how and where to use it for practicing surgeons.
S6 [60] N/A N/A N/A National and international research status and problems concerning personalized
medicine and cognitive computing.
S7 [61] N/A N/A N/A Ontology-based eHealth system with a conventional communication strategy.
S8 [62] N/A N/A N/A Motivations to adopt computer-based solutions for healthcare and surveys application
of cognitive computing in healthcare.
S9 [63] N/A N/A N/A IBM‟s Cognitive Computing effort in healthcare.
S10 [64] N/A N/A N/A The current use, development and monitoring of future drug discovery in public and
personal healthcare and principled hindrances to untangle the full potentiality in
healthcare.
S11 [65] N/A N/A N/A An environmentally valid and cohesive IT system to reveal and support cognitive
function in a burn intensive care unit.
S12 [66] N/A N/A N/A Discussed "Cognitive IoT" by defining demand, services and open challenges in
healthcare.
S13 [67] N/A N/A N/A Introduced the concept of “super customer" versus a "customer" and refers to
healthcare in the 21st century. Proposed PPT-DAM strategic solution for providing
"super customers" and customers with quality health services.
S14 [68] Simulated Patients Random Extended Q-UEL, i.e. universal interoperability language to more traditional areas of
Sampling public health surveys.
 39

S15 [69] Simulated Patients Random Suggested a series of autonomous cognitive design patterns to mitigate the
Sampling complication of the design of intelligent IoT systems.
S16 [70] N/A N/A Digital database Presented current research areas, future research directions, current research challenges
records and challenges facing health and biomedical research in cognitive information
technology.
S17 [71] N/A N/A N/A The concept of evidence was examined, particularly in the illumination of human
cognition models leading to greater robustness.
S18 [72] N/A N/A N/A Discussed that people are the ultimate users of biomedical information and computer
scientists can recognize the limitations of human knowledge and use cognitive science
to inform the design and evaluation of solutions for technical information management
and the interface of the healthcare team.
S19 [73] Clinical Patients Documents and A framework for comprehending human expertise, and the design and deployment of
Records cognitive task-supporting systems.
S20 [74] Simulated Primary care Interviews and Implemented a design approach to cognitive system engineering (CSE) to build a
providers and Observations model supporting referring clinicians‟ cognitive needs and improving communication
specialists with referrals.
S21 [75] N/A N/A N/A Discussed Big Data and robust data management in conjunction with data analytics
and suggested that AI will lead to a new information and knowledge paradigm called
"medical intelligence".
S22 [76] Simulated IRIS databases Records Developed an iris-based recognition technology for handling authentication problems
with multi-biometric scenarios as a unimodal biometric.
S23 [77] Simulated Patients Questionnaires Discussed the challenges of visualizations design and implementation of emergency
departments by supporting decision-making in real time.
S24 [78] Simulated Physicians Random Described an automated method of generation of problem lists and reports on the
Sampling results from a pilot study
S25 [79] Clinical Patients Records No important healthcare processes and workflows were taken into account by health
systems. In contrast to this, a cognitively-focused design strategy shows how
technology can be introduced to support and improve patient care work strategies.
S26 [80] Simulated Full set of Documents and Watson's application to pilot studies in identification of drug targets and reuse of
MEDLINE Records drugs. The result suggested that Watson can speed up the identification of new drug
abstracts candidates and new drug targets by exploiting the potential of Big Data.
S27 [81] N/A N/A N/A Cognitive computing and eScience are the disruptive factors in the research
methodologies of health and life science. Presented AI-based research models.
 40

S28 [82] Simulated Patients Random The development process was described by cognitive interviews with primary care
Sampling patients to improve the quality of the item and to test the face validity of the Primary
Care Outcomes Questionnaire (PCOQ).
S29 [83] Simulated Users Random Proposed a new healthcare system for the physiological and psychological
Sampling management of the health status of the patient based on a 5G cognitive system (5G-
Csys).
S30 [84] N/A N/A N/A Analysis and development of brain-based research for the use of AI in education by
Pearson and IBM and, then, to develop cognitive learning systems to improve the
performance of education systems.
S31 [85] Simulated Real-life Random Experimented with modified prediction models to overcome the difficulty of
hospital data Sampling incomplete data and achieved 94.8 percent prediction accuracy at convergence rate.
S32 [86] Simulated Nurses Random Use of cognitive system engineering frameworks to investigate the decisions of
Sampling experienced providers of anesthesia and to identify the decision-making pathways used
in difficult situations in airways.
 41

Table B2: Current research in terms of method, algorithms, applications, results, strength and limitation
Study# & Method Algorithm Application Results Strength Limitation
Reference
S1 [55] IoT devices capture Feature Extraction, Children, Achieved 99.87% Significant contribution to 5G The proposed system's
patient's voice and image ML, DL, pattern elderly and accuracy in terms of personalized and time requirement
signals in an intelligent recognition, and people with seamless emotional health
home scenario and the AI algorithms mental illness services
scores from the signals are
combined to generate a
concluding score for
emotional decision
S2 [56] Discussed a series of design Recommendations Management Autonomic Automation of system Response time and
patterns and the design and algorithm, of patient Cognitive management, and coordination scalability management
development of smart IoT Blackboard comorbidity Management pattern of business processes in order during discovery
systems pattern based on is the highest level of to manage complex demands processes
wearable maturity and was
generic
S3 [57] Used different body sensors Deep Person with Superiority over Robustness for cognitive Inadequacy of the
for healthcare and then Convolutional sensors others assistance in intelligent implementation of a real
extracted features from the Neural Network installed healthcare systems based on smart home
sensors‟ data body sensors environment. Cloud-
based distributed
platform was utilized
S4 [58] Combined cognitive Supervised Patient Offers a better user A better user Quality of Users „emotions are not
calculation and edge learning algorithm emergency experience and Experience was offered recognized in the
calculation to monitor and situations improves the survival proposed system and
analyze users‟ health rates of patients in corresponding care is
sudden emergencies not carried out
significantly
 42

S5 [59] Use of AI as a tool by IBM's Watson Doctors Recognition of Hospitals and clinics will use The creation of AI-based
doctors to support human Health ML provide clinical front patterns analytics based on historical systems requires
cognitive functions algorithm, ANNs increasingly improves the data to optimize workflows enormous costs, since
complex accuracy of and allocation of resources the machines are very
patients with diagnostics based on complex. Restoration
care images, and AI and maintenance require
platforms support enormous costs as well
clinicians in
decision-making

S6 [60] Different studies have been Cognitive Personalized The analysis between Presentation of case study on PM requires massive
used to discuss the current computing Medicine traditional medicines how traditional medicine is infrastructure investment
status of personalized algorithm versus personalized different from personalized to collect, store and
medicine (PM) research medicine is presented medicine and prediction of share information
and how to use cognitive disease based on various
computing to address parameters
personalized medicine
S7 [61] Use semantic web Ontologies related Patient with Representation of The proposed IoT Fitness Didn't discuss about
technologies to describe the algorithms cardiovascular linguistic Ontology is a domain-specific handling of data silos,
importance of sensor data health compatibility ontology for lightweight data privacy and fake
explicitly problems between expansion news
heterogeneous IoT
fitness and log data

S8 [62] Surveys the adoption of DL, and IBM Common Cognitive Cognitive computing in Lack of proper
cognitive computing-based Watson problems in computing-based healthcare is an encouraging infrastructure settings
solutions in healthcare medicine solutions in topic and requirements of
adoption, and healthcare provide hardware to run on
visual pattern machines with systems effectively is
recognition, human-like still high
healing reasoning
capabilities for
successful findings
 43

S9 [63] Shows that IBM Watson NLP, hypothesis Personalized, Real-time analysis of Cognitive healthcare decisions Cognitive Computing in
platform enhances generation and its Integrated, health information to have the advantages of Healthcare has a limited
physicians and ensures evaluation, and transparent make informed improving patient health with risk analysis and
ability to innovate through dynamic learning and high- decision reduced cost requires a thorough
huge volume of health data quality care of training process
patients

S10 [64] Carried out cognitive Algorithms related Patient The functional Big Data analysis makes the Inaccessible data are not
studies and used results to to cognitively- electronic prototype of a tool data transparent, easy to handled in a
develop a cognitive based methods health recordintegrated with an access, enables stakeholders in standardized manner
framework for prototype (EHR) EHR improves the drug development and helps to without clear
design efficiency of a take evidence-informed international codes of
summary review of decisions practice
patient data
S11 [65] Used the naturalistic Used ethno- Intensive Care Produced a Key challenges and barriers to Needs to pay attention to
decision-making approach methodology to Unit descriptive BICU effective BICU clinical care efficiency, accuracy,
to examine the clinical investigate cognitive work have been identified reliability, salience,
decision-making of burn practical actions' model, developed a communication
intensive care units(BICU) rational properties series of prototypes effectiveness and trust in
for information the cognitive work
design

S12 [66] Used literature reviews to Algorithm related Patients In applications where The CIoT makes better use of Primary limitations are:
discuss "Cognitive IoT" to supervised and people cannot spend resources and, in the future, it energy conservation/
(CIoT) by clearly defining unsupervised their time and will provide intelligent harvesting, resource
the demand, services and learning attention, CIoT is methods for bio-medical discovering and routing,
the open challenges very useful applications and clinician and location proofs and
diagnosis security
S13 [67] Presented PPT-DAM No specific Patients Cognitive systems In-depth understanding and Lack of more in-depth
method to offer superior algorithms, but is enabled by and IBM Watson can effective communication with insights into emerging
quality service by based on unprecedented generally bring customers in the broader social and new intersections of
introducing super-customer Experiment– social transformational and ecosystems is surrounding Big the proposed approach
and customer Execute–Evaluate practices endurable healthcare Data and Big Data
implementations skills
 44

S14 [68] Uses the principles of Unsupervised data Clinical trials Interoperability and Q- UEL is the 10-year Q-UEL overlaps with
observation, interpretation, mining and inference in health roadmap for IT infrastructure quantum mechanics
evaluation and decision- supervised data for health problems with mental,
making to extend Q-UEL, mining statistical and zeta
the language of universal functions
exchange for
interoperability and
inference in healthcare and
biomedicine
S15 [69] Proposed a model-based Autonomic Patient health System requirements Efficient cognitive design Gaps in the performance
methodology that combines cognitive design based on must be better patterns measurement as
the principles of software patterns, and heterogeneous integrated by parameters other than
modeling and knowledge Semantic wearable modeling the response time and
engineering to facilitate Knowledge interactions between scalability were not
autonomous and cognitive Mediator management considered
IoT systems design and processes and
development knowledge sources
S16 [70] Extracted journal articles Exploratory review biomedicine Highlighted the In promotion and sustainment Cognitive studies with
from Journal of Biomedical and healthcare importance of research in cognitive computer smaller samples are
Informatics and conducted cognitive and science, the journal played an carried out. For future
the literature review on learning sciences and important role practitioners and
cognitive informatics their growing role researchers, critical
training and education
directions are required

S17 [71] To examine evidence- Evidence Primarily in Evidence is The conceptualization of Number of
based healthcare, the study hierarchies and diagnosis, and essentially evidence has a number of consequences for
used analytical and grading systems management information that consequences for evidence- evidence-based
synthetic methods of healthcare takes different forms based healthcare healthcare
activities
 45

S18 [72] Identifies a fundamental Deep Learning Workplace Informaticians to Informaticians can recognize Refinement of graphical
challenge for environment in become aware of the the limitations of human displays that provide the
informaticians, i.e. to healthcare constraints cognition and use cognitive healthcare team with
connect with the healthcare associated with science to inform the design information on care
team by acquiring, cognitive processing and evaluation of solutions for phases
retrieving and processing and workplace technical information
information within the factors management and the interface
cognitive abilities of the of the healthcare team
human brain
S19 [73] Proposed CREDO Clinical decisions, Applications Argumentation Results in a technology stack, Did not take the full
framework clinical for treatment theory for with the key element as an responsibility of the
recommendations, in clinical uncertainty agent specification language accuracy of information
machine learning specialties and reasoning, symbolic that proved to be a versatile
algorithm based on environments decision theory and tool for the design of care
Big Data domino agent point applications
S20 [74] Conducted interviews and Qualitative Electronic CSE improves Supports the implementation The degree of impact of
observations, identified thematic analysis, medical referral of an approach to CSE design existing EHR systems
cognitive requirements, Iterative consultation communication for electronic medical varies according to the
designed prototypes for consensus-based orders consultation orders usability
user interfaces and approach
compared to currently
implemented user
interfaces(CSE)
S21 [75] Presented the application of NLP, cognitive Clinical trials BD, Data analysis, Big Data and well-built data Effective handling of the
Big Data in healthcare computing, ML AI expertise should management in combination problem of "signal-to-
using a three-step "bottom- and DL be used by healthcare with data analysis and AI leads noise" which can lead to
up" data management organizations in to a new information and diagnosis and treatment
strategy order to collect knowledge exemplar of the disease
meaningful medical
data to improve
quality and results
S22 [76] A CIoT framework was Delta-mean and Biometric For general hardware Satisfactory performance with Lack of fuse scores in
proposed in this study multi-algorithm- recognition and real-time accuracy is over 99.4 percent identification mode
mean applications, the and resolved the loss of iris
CIoT framework is data and invariants of
suitable inconsistency
 46

S23 [77] Discussed the challenges ofAlgorithm related ED Developed an Support for decision-making in There needs to be more
design and implementation to Situational information-rich real time work to determine the
of Emergency Department Awareness suite to provide at-a- impact of performance
(ED) visualizations Global Assessment glance information management in real time
Technique
S24 [78] Evaluated the Watson NLP and ML- Patient- At least one The IBM Watson system Did not talk about
method with random based Watson centered important problem increases efficiency, improves modeling cognitive
selection of de-identified method models care failed by doctors support for clinical decisions patient-centered work
patient records which was identified and improves the quality of for high-value care
by Watson patient care
S25 [79] Health systems developed Algorithms for Patient Important problems Cognitive analysis and The costs associated
without cognitive analysis decision support evacuation, arise in the cognitive design are time and with an organized effort
and cognitive design does anesthesiology development of effort-related challenges for cognitive analysis
not take into consideration technological support and cognitive design in
important healthcare systems. A cognitive engineering is
processes comprehensive modest in relation to the
attempt is required to costs of technological
understand the development
cognitive abilities of
patient care
professionals
S26 [80] Showed the application of Advanced Life Sciences IBM Watson can In the identification and Time-consuming
cognitive computing to Big reasoning, research speed up the encoding of adverse event extensive training for
Data challenges using IBM predictive identification of new reports, cognitive computing IBM Watson
Watson tool modeling, and ML drug candidates and adds value
targets

S27 [81] Presents AI-based research Deep ML Health By increasing the The models support the The systems require a
models and identifies Big (obesity) and performance of processing of large data sets great deal of raw
Data as a catalyst for life science human tasks with AI, that enable statistical modeling material to gain insight,
innovation, presents a research cognitive computing to be relaxed and provides the learn and forecast. There
computer-based human and e-science, basis for the creation and are important legal and
problem solving framework research models are democratization of new privacy consequences
more efficient knowledge for the search of all
adaptive intervention these data
programs for obesity
 47

S28 [82] Used two formats of the Tourangeau‟s Primary care PCOQ-Status The cognitive interviews have The PCOQ-change was
Questionnaire on Primary model captured a subjective helped to develop the PCOQ misunderstood and
Care Outcomes (PCOQ): health outlook that successfully and the status was indicates a need for
PCOQ-Status and PCOQ- could not vary found to be well-understood more cognitive testing of
Change to improve the according to age or by patients transitional
quality of the item long-lasting questionnaires
conditions
S29 [83] For verification, 5G Algorithm related Speech 5G-Csys is a 5G The system ensures extremely Protection of sensitive
cognitive system (5G-Csys) to resource emotion cognitive voice low latency and ultra-high data and privacy
healthcare system and a cognitive engine recognition emotion recognition reliability problems for users and
5G- Csys prototype and data cognitive system that can the lack of a unified
platform are proposed engine recognize the voice development API
emotion of users
S30 [84] Provides analysis of brain- Neuro-scientific, Neuroscience Understanding, The emergence of AI and There‟s much unknown
based R&D for the use of neuro-anatomy, Education modeling, cognitive systems in education of current impacts,
AI in education brain-inspired simulating, and actively creates new types of potential impact of
computing, and integrating brain into educational spaces neurocomputational
artificial neural AI and cognitive governance in education
networks application with emerging
algorithms neuroeducation practices

S31 [85] Experimented on real-life New multimodal Chronic The predictive The proposed algorithm is Did not tackle efficient
hospital data with modified risk prediction and disease of accuracy of the faster than the unimodal cost management by
prediction models and used ML algorithm cerebral proposed algorithm disease risk prediction reducing wasteful
a latent factor model to based on infarction at convergence speed algorithm procedures
retrace the inexistence data convolutional is 94.8 percent
to overcome the difficulty neural network
of incomplete data
S32 [86] Investigated decisions of Categorization of Anesthesia Illustrated a number Anesthetic teams have to make The limitations are
experienced anesthesia cognitive pathways emergencies in of decisions taken by different decisions to ensure mainly linked to
providers and to identify airway the anesthetic team the patient's oxygenation voluntary participation,
the decision-making management to address possible during the surgical period self-reporting,
processes airway management experimental prejudice
variability and data research
 48

References

[1] Chen, M., Herrera, F., & Hwang, K. (2018). Cognitive computing: architecture, technologies and intelligent
applications. IEEE Access, 6, 19774-19783.

[2] Hwang, K., & Chen, M. (2017). Big-data analytics for cloud, IoT and cognitive computing. John Wiley &
Sons.

[3] Chen, M., Hao, Y., Hwang, K., Wang, L., & Wang, L. (2017). Disease prediction by machine learning over
big data from healthcare communities. IEEE Access, 5, 8869-8879.

[4] Chen, M., Yang, J., Hao, Y., Mao, S., & Hwang, K. (2017). A 5G cognitive system for healthcare. Big Data
and Cognitive Computing, 1(1), 2.

[5] Chen, Y., Argentinis, J. E., & Weber, G. (2016). IBM Watson: how cognitive computing can be applied to
big data challenges in life sciences research. Clinical therapeutics, 38(4), 688-701.

[6] Belle, A., Thiagarajan, R., Soroushmehr, S. M., Navidi, F., Beard, D. A., & Najarian, K. (2015). Big data
analytics in healthcare. BioMed research international, 2015.

[7] Asante-Korang, A., & Jacobs, J. P. (2016). Big Data and paediatric cardiovascular disease in the era of
transparency in healthcare. Cardiology in the Young, 26(8), 1597-1602.

[8] Goli-Malekabadi, Z., Sargolzaei-Javan, M., & Akbari, M. K. (2016). An effective model for store and
retrieve big health data in cloud computing. Computer methods and programs in biomedicine, 132, 75-82.

[9] Tranfield, D., Denyer, D., & Smart, P. (2003). Towards a methodology for developing evidence‐informed
management knowledge by means of systematic review. British journal of management, 14(3), 207-222.

[10] Kitchenham, B. A., & Charters, S. M. (2007). Guidelines for performing systematic literature reviews in
software engineering Version 2.3. EBSE Technical Report EBSE-2007-01. Keele University and Durham
University.

[11] IBM research stakes its future on cognitive computing. (2013).retrieved from
https://www.zdnet.com/article/ibm-research-stakes-its-future-on-cognitive-computing/

[12] Evans, R. S. (2016). Electronic health records: then, now, and in the future. Yearbook of medical
informatics, 25(S 01), S48-S61.

[13] Nishida, T., Nakazawa, A., Ohmoto, Y., & Mohammad, Y. (2014). History of Conversational System
Development. In Conversational Informatics (pp. 43-62). Springer, Tokyo.

[14] Coccoli, M., Maresca, P., &Stanganelli, L. (2016). Cognitive computing in education. Journal of E-learning
and Knowledge Society, 12(2).
 49

[15] Banavar, G. S. (2015). Watson and the era of cognitive computing. In Pervasive Computing and
Communications (PerCom), 2015 IEEE International Conference on (pp. 95-95). IEEE.

[16] Coccoli, M., Maresca, P., &Stanganelli, L. (2017). The role of big data and cognitive computing in the
learning process. Journal of Visual Languages & Computing, 38, 97-103.

[17] Noor, A. K. (2015). Potential of cognitive computing and cognitive systems. Open Engineering, 5(1).

[18] Hurwitz, J. S., Kaufman, M., & Bowles, A. (2015). Cognitive computing and big data analytics. John
Wiley & Sons.

[19] Applications of Cognitive Computing. (2018).retrieved from

https://developer.ibm.com/code/2018/01/21/the-age-of-cognitive-computing/

[20] From tabulating, programming to cognitive computing era. (2018).retrieved from https://www-
07.ibm.com/events/th/ibmthailand65/p/tabulating_programming_cognitive_era/

[21] Piai, S., & Claps, M. (2013). Bigger data for better healthcare. IDC Health Insights.

[22] Six barriers to adopting AI applications in healthcare. (2018).retrieved from

https://searchhealthit.techtarget.com/tip/Six-barriers-to-artificial-intelligence-adoption-in-healthcare

[23] Five top challenges affecting healthcare leaders in the future. (2018).retrieved from
https://www.beckershospitalreview.com/hospital-management-administration/five-top-challenges-affecting-
healthcare-leaders-in-the-future.html

[24] 5 Reasons Healthcare Data Is Unique and Difficult to Measure. (2018).retrieved from
https://www.healthcatalyst.com/a-new-way-to-look-at-healthcare-data-models

[25] How Big Data Is Changing Healthcare. (2015).retrieved from

https://www.forbes.com/sites/bernardmarr/2015/04/21/how-big-data-is-changing-healthcare/#2176a6ff2873

[26] Ogiela, M. R., You, I., &Yim, K. (2013). Cognitive and secure computing in Information Management. Int
J. Information Management, 33(2), 243-244.

[27] Gupta, S., Kar, A. K., Baabdullah, A., & Al-Khowaiter, W. A. (2018). Big data with cognitive computing:
A review for the future. International Journal of Information Management, 42, 78-89.

[28] Chen, Y., Argentinis, E., & Weber, G. (2016). IBM watson: How cognitive computing can Be applied to
Big data challenges in life sciences research. Clinical Therapeutics, 38(No. 4), 688–701.

[29] IBM Watson & Cognitive Computing - Tech in Asia. (2016). retrieved from
https://www.slideshare.net/ngito/ibm-watson-cognitive-computing-tech-in-asia-2016

[30] Hodach, R., Grundy, P., Jain, A., & Weiner, M. (2016). Provider-Led Population Health Management: Key
Strategies for Healthcare in the Cognitive Era. John Wiley & Sons.
 50

[31] Tarafdar, M., Beath, C., & Ross, J. (2017). Enterprise cognitive computing applications: opportunities and
challenges. IT Professional.

[32] Chen, Y., Argentinis, J. E., & Weber, G. (2016). IBM Watson: how cognitive computing can be applied to
big data challenges in life sciences research. Clinical therapeutics, 38(4), 688-701.

[33] Kelly III, J. E., & Hamm, S. (2013). Smart machines: IBM's Watson and the era of cognitive computing.
Columbia University Press.

[34] Murtaza, S. S., Lak, P., Bener, A., &Pischdotchian, A. (2016). How to effectively train IBM Watson:
Classroom experience. In System Sciences (HICSS), 2016 49th Hawaii International Conference on (pp. 1663-
1670). IEEE.

[35]Coccoli, M., Maresca, P., &Stanganelli, L. (2016). Cognitive computing in education. Journal of E-learning
and Knowledge Society, 12(2).

[36] Finch, G., Goehring, B., & Marshall, A. (2017). The enticing promise of cognitive computing: high-value
functional efficiencies and innovative enterprise capabilities. Strategy & Leadership, 45(6), 26-33.

[37] Savaglio, C., &Fortino, G. (2015). Autonomic and cognitive architectures for the Internet of Things. In
International Conference on Internet and Distributed Computing Systems (pp. 39-47). Springer, Cham.

[38] Savaglio, Claudio, Giancarlo Fortino, and Mengchu Zhou (2016). Towards interoperable, cognitive and
autonomic IoT systems: An agent-based approach. Internet of Things (WF-IoT), 2016 IEEE 3rdWorld Forum
on. IEEE, 2016.

[39] Savaglio, C., Fortino, G., Ganzha, M., Paprzycki, M., Bădică, C., &Ivanović, M. (2017). Agent-based
computing in the internet of things: a survey. In International Symposium on Intelligent and Distributed
Computing (pp. 307-320). Springer, Cham.

[40] De Lemos, R., Giese, H., Müller, H. A., Shaw, M., Andersson, J., Litoiu, M.,&Weyns, D. (2013). Software
engineering for self-adaptive systems: A second research roadmap. In Software Engineering for Self-Adaptive
Systems II (pp. 1-32). Springer, Berlin, Heidelberg.

[41] Foteinos, V., Kelaidonis, D., Poulios, G., Vlacheas, P., Stavroulaki, V., &Demestichas, P. (2013).
Cognitive management for the internet of things: A framework for enabling autonomous applications. IEEE
vehicular technology magazine, 8(4), 90-99.

[42]Islam, S. R., Kwak, D., Kabir, M. H., Hossain, M., & Kwak, K. S. (2015). The internet of things for health
care: a comprehensive survey. IEEE Access, 3, 678-708.

[43]Kim, J., & Lee, J. W. (2014, March). OpenIoT: An open service framework for the Internet of Things. In
Internet of Things (WF-IoT), 2014 IEEE World Forum on (pp. 89-93). IEEE.
 51

[44] Mendoza, J. A., Baker, K. S., Moreno, M. A., Whitlock, K., Abbey‐Lambertz, M., Waite, A., ... & Chow,
E. J. (2017). A Fitbit and Facebook mHealth intervention for promoting physical activity among adolescent and
young adult childhood cancer survivors: a pilot study. Pediatric blood & cancer, 64(12), e26660.

[45] Feldman, M., Juldaschewa, F., & Bernstein, A. (2017). Data Analytics on Online Labor Markets:
Opportunities and Challenges. arXiv preprint arXiv:1707.01790.

[46] Afrooz, S., &Navimipour, N. J. (2017). Memory designing using quantum-dot cellular automata:
systematic literature review, classification and current trends. Journal of Circuits, Systems and Computers,
26(12), 1730004.

[47] Milani, B. A., &Navimipour, N. J. (2017). A systematic literature review of the data replication techniques
in the cloud environments. Big Data Research, 10, 1-7.

[48] Luxton-Reilly, A. (2009). A systematic review of tools that support peer assessment. Computer Science
Education, 19(4), 209-232.

[49] Aznoli, F., &Navimipour, N. J. (2017). Deployment strategies in the wireless sensor networks: systematic
literature review, classification, and current trends. Wireless Personal Communications, 95(2), 819-846.

[50] Aznoli, F., &Navimipour, N. J. (2017). Cloud services recommendation: Reviewing the recent advances
and suggesting the future research directions. Journal of Network and Computer Applications, 77, 73-86.

[51] Soltani, Z., &Navimipour, N. J. (2016). Customer relationship management mechanisms: A systematic
review of the state of the art literature and recommendations for future research. Computers in Human Behavior,
61, 667-688.

[52] Ouhbi, S., Idri, A., Fernández-Alemán, J. L., &Toval, A. (2015). Requirements engineering education: a
systematic mapping study. Requirements Engineering, 20(2), 119-138.

[53] Idri, A., azzahraAmazal, F., &Abran, A. (2015). Analogy-based software development effort estimation: A
systematic mapping and review. Information and Software Technology, 58, 206-230.

[54] Zapata, B. C., Fernández-Alemán, J. L., Idri, A., &Toval, A. (2015). Empirical studies on usability of
mHealth apps: a systematic literature review. Journal of medical systems, 39(2), 1.

[55] Hossain, M. S., & Muhammad, G. (2018). Emotion-aware connected healthcare big data towards 5G. IEEE
Internet of Things Journal, 5(4), 2399-2406.

[56] Mezghani, E., Expósito, E., &Drira, K. (2018). An Autonomic Cognitive Pattern for Smart IoT-based
System Manageability: Application to Comorbidity Management. ACM Transactions on Internet Technology
(TOIT), 19(1), 8.
 52

[57] Uddin, M. Z., & Hassan, M. M. (2018). Activity Recognition for Cognitive Assistance Using Body Sensors
Data and Deep Convolutional Neural Network. IEEE Sensors Journal.

[58] Chen, M., Li, W., Hao, Y., Qian, Y., &Humar, I. (2018). Edge cognitive computing based smart healthcare
system. Future Generation Computer Systems.

[59] Bini, S. A. (2018). Artificial Intelligence, Machine Learning, Deep Learning, and Cognitive Computing:
What Do These Terms Mean and How Will They Impact Health Care?. The Journal of arthroplasty.

[60] Jayanthi, A., Ramu, G., Rani, P. (2018). A review on personalized medicine technique using cognitive
computing. International Journal of Engineering & Technology, 7 (4) (2018) 3423-3425.

[61] Carbonaro, A., Piccinini, F., & Reda, R. (2018). Integrating Heterogeneous Data of Healthcare Devices to
enable Domain Data Management. Journal of e-Learning and Knowledge Society, 14(1).

[62] Coccoli, M., &Maresca, P. (2018). Adopting Cognitive Computing Solutions in Healthcare. Journal of e-
Learning and Knowledge Society, 14(1).

[63] Ahmed, M. N., Toor, A. S., O'Neil, K., & Friedland, D. (2017). Cognitive computing and the future of
health care cognitive computing and the future of healthcare: the cognitive power of IBM Watson has the
potential to transform global personalized medicine. IEEE pulse, 8(3), 4-9.

[64] Leyens, L., Reumann, M., Malats, N., & Brand, A. (2017). Use of big data for drug development and for
public and personal health and care. Genetic epidemiology, 41(1), 51-60.

[65] Nemeth, C., Blomberg, J., Argenta, C., Serio-Melvin, M. L., Salinas, J., & Pamplin, J. (2016). Revealing
ICU cognitive work through naturalistic decision-making methods. Journal of Cognitive Engineering and
Decision Making, 10(4), 350-368.

[66] Amuthadevi, C., &Monicka, S. G. (2016).Emerging Bio-Medical Applications and Open Research
Challenges in Cognitive Internet of Things (CIoT).

[67] Kolker, E., Özdemir, V., &Kolker, E. (2016). How Healthcare can refocus on its Super-Customers
(Patients, n= 1) and Customers (Doctors and Nurses) by Leveraging Lessons from Amazon, Uber, and Watson.
Omics: a journal of integrative biology, 20(6), 329-333.

[68]Robson, B., &Boray, S. (2015). Implementation of a web based universal exchange and inference language
for medicine: Sparse data, probabilities and inference in data mining of clinical data repositories. Computers in
biology and medicine, 66, 82-102.
 53

[69] Mezghani, E., Exposito, E., &Drira, K. (2017). A model-driven methodology for the design of autonomic
and cognitive IoT-based systems: application to healthcare. IEEE Transactions on Emerging Topics in
Computational Intelligence, 1(3), 224-234.

[70] Patel, V. L., &Kannampallil, T. G. (2015). Cognitive informatics in biomedicine and healthcare. Journal of
biomedical informatics, 53, 3-14.

[71] Sedig, K., Parsons, P., Naimi, A., & Willoughby, K. (2015). Reconsidering healthcare evidence as dynamic
and distributed: the role of information and cognition. International journal of evidence-based healthcare, 13(2),
43-51.

[72] Klatt, E. C. (2018). The human interface of biomedical informatics. Journal of pathology informatics, 9.

[73] Fox, J. (2017). Cognitive systems at the point of care: The CREDO program. Journal of biomedical
informatics, 68, 83-95.

[74] Savoy, A., Militello, L. G., Patel, H., Flanagan, M. E., Russ, A. L., Daggy, J. K., ... & Saleem, J. J. (2018).
A cognitive systems engineering design approach to improve the usability of electronic order forms for medical
consultation. Journal of biomedical informatics, 85, 138-148.

[75] Chang, A. C. (2016). Big data in medicine: The upcoming artificial intelligence. Progress in Pediatric
Cardiology, 43, 91-94.

[76] Gad, R., Talha, M., El-Latif, A. A. A., Zorkany, M., Ayman, E. S., Nawal, E. F., & Muhammad, G. (2018).
Iris recognition using multi-algorithmic approaches for cognitive Internet of things (CIoT). Future Generation
Computer Systems.

[77] Franklin, A., Gantela, S., Shifarraw, S., Johnson, T. R., Robinson, D. J., King, B. R., ... & Rubio, A.
(2017). Dashboard visualizations: supporting real-time throughput decision-making. Journal of biomedical
informatics, 71, 211-221.

[78] Devarakonda, M. V., Mehta, N., Tsou, C. H., Liang, J. J., Nowacki, A. S., &Jelovsek, J. E. (2017).
Automated problem list generation and physicians perspective from a pilot study. International journal of
medical informatics, 105, 121-129.

[79] Lintern, G., &Motavalli, A. (2018). Healthcare information systems: the cognitive challenge. BMC medical
informatics and decision making, 18(1), 3.

[80] Chen, Y., Argentinis, J. E., & Weber, G. (2016). IBM Watson: how cognitive computing can be applied to
big data challenges in life sciences research. Clinical therapeutics, 38(4), 688-701.
 54

[81] Marshall, T., Champagne-Langabeer, T., Castelli, D., &Hoelscher, D. (2017). Cognitive computing and
eScience in health and life science research: artificial intelligence and obesity intervention programs. Health
information science and systems, 5(1), 13.

[82] Murphy, M., Hollinghurst, S., & Salisbury, C. (2018). Qualitative assessment of the primary care outcomes
questionnaire: a cognitive interview study. BMC health services research, 18(1), 79.

[83] Chen, M., Yang, J., Hao, Y., Mao, S., & Hwang, K. (2017). A 5G cognitive system for healthcare. Big Data
and Cognitive Computing, 1(1), 2.

[84] Williamson, B., Pykett, J., &Nemorin, S. (2018). Biosocial spaces and neurocomputational governance:
brain-based and brain-targeted technologies in education. Discourse: Studies in the Cultural Politics of
Education, 39(2), 258-275.

[85] Chen, M., Hao, Y., Hwang, K., Wang, L., & Wang, L. (2017). Disease prediction by machine learning over
big data from healthcare communities. IEEE Access, 5, 8869-8879.

[86] Schnittker, R., Marshall, S., Horberry, T., Young, K., &Lintern, G. (2017). Exploring decision pathways in
challenging airway management episodes. Journal of Cognitive Engineering and Decision Making, 11(4), 353-
370.

[87] Bhadani, A. K., &Jothimani, D. (2016). Big Data: Challenges, Opportunities, and Realities. In Effective
Big Data Management and Opportunities for Implementation (pp. 1-24). IGI Global.

[88] Yao, Y. (2009). Interpreting concept learning in cognitive informatics and granular computing. IEEE
Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 39(4), 855-866.

[89] Wang, Y. (2007). The theoretical framework of cognitive informatics. International Journal of Cognitive
Informatics and Natural Intelligence (IJCINI), 1(1), 1-27.

[90] Shivhare, R., Cherukuri, A. K., & Li, J. (2017). Establishment of cognitive relations based on cognitive
informatics. Cognitive Computation, 9(5), 721-729.

[91] Wang, Y. (2009). A cognitive informatics reference model of autonomous agent systems (AAS).
International Journal of Cognitive Informatics and Natural Intelligence (IJCINI), 3(1), 1-16.\

[92] Li, J., Mei, C., Xu, W., & Qian, Y. (2015). Concept learning via granular computing: a cognitive
viewpoint. Information Sciences, 298, 447-467.

[93] Huang, C., Li, J., Mei, C., & Wu, W. Z. (2017). Three-way concept learning based on cognitive operators:
an information fusion viewpoint. International Journal of Approximate Reasoning, 83, 218-242.
 55

[94] Li, J., Huang, C., Qi, J., Qian, Y., & Liu, W. (2017). Three-way cognitive concept learning via multi-
granularity. Information Sciences, 378, 244-263.

View publication stats