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

net/publication/351792114

Data Analytics in Healthcare Systems – Principles, Challenges, and

Applications

Chapter · May 2021

DOI: 10.1201/9781003185246-1

CITATIONS READS

0 306

4 authors:

s. Suganthi Vaishali Gupta

Galgotias University Galgotias University
6 PUBLICATIONS 1 CITATION 10 PUBLICATIONS 2 CITATIONS

SEE PROFILE SEE PROFILE

Varsha Sisaudia T. Poongodi

Delhi Technological University Galgotias University
10 PUBLICATIONS 37 CITATIONS 49 PUBLICATIONS 133 CITATIONS

SEE PROFILE SEE PROFILE

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

Data Analytics in Healthcare View project

Watermarking View project

All content following this page was uploaded by s. Suganthi on 19 September 2021.

The user has requested enhancement of the downloaded file.

 1 Data Analytics in
Healthcare Systems –
Principles, Challenges,
and Applications
S. Suganthi, Vaishali Gupta,
Varsha Sisaudia, and T. Poongodi

CONTENTS
1.1 Introduction....................................................................................................... 1
1.1.1 Data Analytics in Healthcare................................................................. 2
1.1.2 Characteristics of Big Data.................................................................... 3
1.2 Architectural Framework.................................................................................. 3
1.2.1 Data Aggregation...................................................................................4
1.2.2 Data Processing.....................................................................................5
1.2.3 Data Visualization.................................................................................6
1.3 Data Analytics Tools in Healthcare...................................................................7
1.3.1 Data Integration Tools...........................................................................7
1.3.2 Searching and Processing Tools............................................................8
1.3.3 Machine Learning Tools........................................................................9
1.3.4 Real-Time and Streaming Data Processing Tools.................................9
1.3.5 Visual Data Analytical Tools............................................................... 10
1.4 Data Analytics Techniques in Healthcare....................................................... 11
1.5 Applications of Data Analytics in Healthcare................................................. 12
1.6 Challenges Associated with Healthcare Data.................................................. 18
1.7 Conclusion....................................................................................................... 19
References................................................................................................................. 19

1.1 INTRODUCTION
The healthcare industry is multidimensional, with multiple data sources involving
healthcare systems, health insurers, clinical researchers, social media, and govern-
ment [1], generating different types and massive amounts of data. It is impossible to
handle this big data with traditional software and hardware and the existing storage
methods and tools. Data analytics is the process of the analysis of data to identify

1
2 ML and Analytics in Healthcare Systems

trends and patterns to gain valuable insights. The data generated in the health indus-
try are characterized by the four Vs of big data, namely volume, velocity, variety,
and veracity, which play crucial roles in health data analytics. Also, evidence-based
decision making has gained importance, which involves the sharing of data among
various data repositories. According to Deloitte Global Healthcare Outlook, it is
expected that global healthcare expenditure will continue to increase at an annual
rate of 5.4% between 2017 and 2022. This is due to the increased importance of
personalized medicine, the use of advanced technologies, the demand for new pay-
ment models, improvement and expansion of care delivery sites, and competition.
Various research attempts, based on big data, have provided strong evidence that
the efficiency of healthcare applications is dependent upon the basic architecture,
techniques, and tools used. Statistical data and reports can be generated with the use
of patient records, aiding in knowledge discovery, and thereby influencing value-
added services to the patients, improving healthcare quality, the making of timely
decisions, and minimizing the costs incurred. Hence, there is a need to incorporate
and integrate big data analytics into existing healthcare systems. Despite healthcare
analytics having massive potential for value-added change, there are many tech-
nological, social, organizational, economic, and policy barriers associated with its
application [2].

1.1.1 Data Analytics in Healthcare

Health industries employing data analytics can use big data for the early detection
of diseases and their treatment, clinical operations, genomic analysis, patient profile
analytics, and prevention of fraud. Data processing involves analytics being applied
to the transformed data to obtain meaningful insights from the healthcare data for
evidence-based decision making. There exist three levels of analytics with increas-
ing complexity and value, which are given below.

• Descriptive Analytics describes the current events or summarizes the

past events by generating reports with the help of statistical tools such
as tables and graphs. Thus, it helps medical practitioners to study and
understand the patient’s behavioral patterns in the past with the help of
the patient’s operational data, which can be used to solve problems in cur-
rent situations.
• Predictive Analytics enables the user to predict the future with the use of
descriptive data, using empirical methods, such as machine learning, mod-
eling, and data mining, to analyze the data.
• Prescriptive Analytics enables the user to choose the best solution from
several possible alternatives to the issues in question. The real-time data
are analyzed with the historical data by the use of artificial intelligence,
data mining, and machine learning. The practitioners can determine the
possible effects of their decisions by using simulation and optimization
techniques and can prepare themselves to handle any event in case of
failure or success.
Data Analytics in Healthcare Systems 3

1.1.2 Characteristics of Big Data

The main characteristics of big data are the four Vs, which are as follows.

• Volume: The healthcare industry generates an enormous amount of data

coming from various sources, such as EHRs (Electronic Health Records),
LIMS (Laboratory Information Management System), diagnostic or moni-
toring instruments, supply chains, insurance claims/billing, pharmacy,
real-time locating systems [3], and social media. The data that are collected
are used in continuous learning by employing various technologies and pro-
cesses to derive insights from the information, which will help improve the
quality of healthcare. In addition, the reduction in storage costs and the
development of advanced architectures have led to large volumes of data
being stored, processed, and managed, using the existing traditional sys-
tems adopted by the healthcare industry.
• Velocity: The speed with which medical data are generated is high and
requires specific processing requirements. The healthcare industry has not
adapted to technological advances, and hence various processing methods
have been adopted to cope with the speed with which the data are pro-
duced. Batch processing, stream processing, near-real-time processing and
real-time processing methods are used to handle data in the healthcare
industry [4].
• Variety: Data in healthcare come from several sources and are in various
forms. About 80% of the total healthcare data is unstructured (e.g., images,
signals, text), which does not fit into any predefined data format, data type,
or structure. The remaining 20% of the data is structured (e.g., tempera-
ture, blood pressure, patient demographics, etc.), possessing a predefined
data format. The structured data are easy to handle, store, process, and
manipulate. Hence, the unstructured data must be converted to a structured
form by efficient automatic transformation methods. When combined with
structured data, the unstructured data provide valuable information, which
can be harnessed to improve value-added services in healthcare by adopt-
ing more-efficient methodologies.
• Veracity: This refers to the accuracy of the data collected, which is directly
proportional to the value of the insights which can be obtained from them.
The results derived from data analytics are error free if the data obtained
are trustworthy, accountable, authenticate, and available.

1.2 ARCHITECTURAL FRAMEWORK
Hadoop/MapReduce is an open-source platform for big data analytics used in
healthcare, which performs parallel processing in a distributed environment, involv-
ing multiple nodes in the network. The use of Hadoop and MapReduce technologies
has been found to be fruitful in many healthcare applications, by improving the
performance of, for example, image processing, neural signal processing, protein
4 ML and Analytics in Healthcare Systems

FIGURE 1.1 Conceptual framework of big data architecture in healthcare.

structure alignments, signal detection algorithms, and lung texture classification [5].
The architectural framework of big data in healthcare is composed of three major
components, namely data aggregation, data processing, and data visualization [6].
Figure 1.1 illustrates the conceptual framework of big data architecture in healthcare.

1.2.1 Data Aggregation
Data aggregation in healthcare involves the process of collecting and integrating raw
data from various modalities and multiple systems and converting them into a single
standard format suitable for analysis, processing, and storage in a data warehouse.
The functionalities involved in the process include data extraction, data transforma-
tion, and data loading.

• Data Extraction
Healthcare data occupy large volumes and come from heterogeneous
sources. The primary sources of medical data include medical records,
health surveys, claims data, disease registries, vital records, surveil-
lance data, peer-reviewed literature, clinical trial data, and administrative
Data Analytics in Healthcare Systems 5

records. The data from these sources can be structured or unstructured.

The structured data, which contain numeric, categorical, and nominal data
types, have a predefined format and are easy to handle. Unstructured medi-
cal data, such as image data from imaging devices, text data (e.g., doctor
notes, shorthand notations etc.), and signal data (e.g., biosignals from wear-
able devices), do not have a predefined format. They must be converted into
a standard format for further processing. The medical data can be in any
format, which includes EHRs, images, biomedical data, genomic data, sen-
sor data, and clinical text data. These data are in the form of text/ASCII,
XML, JSON, or images in DICOM formats. Usually, medical image
data from local workstations are stored in PACS (Picture Archiving and
Communication Systems) and are transferred and communicated to other
workstations following DICOM standards.
• Data Transformation
The acquired raw data are transformed in order to apply particular business
logic by passing through the phases of data filtering, which is the process
of removing unwanted data (“data cleaning”), which includes the processes
of normalization, noise reduction, and management of missing values, and
data manipulation. The data converted in this way are in a standard format
that is consistent and suitable for further analysis.
• Data Loading
Customized relational databases are usually designed for each particular
healthcare system, with its own defined data models and schemes for the
storage of medical data. The transformed medical data, which are suitable
for further analytics, are loaded into the target database or a data ware-
house, such as HDFS (Hadoop Distributed File System), SQL relational
databases, or NoSQL databases, or combinations of these.

1.2.2 Data Processing
The data processing used in healthcare includes batch processing and stream pro-
cessing methods [7]. Batch processing is the method of analyzing data in batches,
which are collected and stored over a period and in which response time is not
considered. On the other hand, stream processing is the method of analyzing huge
volumes of data, to which a real-time response is required. Some applications in
healthcare require real-time processing of data and they are characterized by noisy
data with missing or redundant values, continuous changes in data, and the need for
a rapid response. Stream processing overcomes these difficulties with simple and
rapid information extraction by using data-mining methods, such as clustering, clas-
sification, and frequent pattern mining [7]. Apache Hadoop MapReduce is a popular
framework used for batch processing, whereas Storm and S4 are frameworks used
for stream processing, with Apache Spark and Apache Flink being frameworks used
for both batch and stream processing.
The Hadoop platform is most widely used for batch processing in which paral-
lel processing of huge volumes of data are carried out in a distributed manner. It is
6 ML and Analytics in Healthcare Systems

a framework in which the process of big data analytics is conducted through a col-
lection of various tools, methodologies, and libraries. It consists of two main compo-
nents, the HDFS and Hadoop MapReduce.

• Hadoop Distributed File System (HDFS)

The HDFS is a distributed file system used for storing and retrieving
extremely large volumes of data at great speed. The data are split into
several blocks with uniform block sizes of 64 MB or 128 MB, which are
distributed across many server nodes, enhancing parallel processing. The
HDFS consists of two nodes, namely a NameNode and a multiple num-
ber of DataNodes. The DataNodes contain the application data and the
NameNodes contain metadata related to the storage and retrieval of data
from the DataNodes. The data are replicated and the copies are distributed
to many nodes, making the system fault tolerant in the case of any node
failure.
• MapReduce
MapReduce is a framework and programming model for distributed pro-
cessing of huge datasets, involving multiple nodes. The processing is broken
down across several individual nodes and carried out in parallel because the
data are vast. The process consists of two phases, the map phase and the
reduce phase, conducted with the mapper function and the reducer func-
tion, respectively.
1. Map Phase: the data stored in the HDFS are split into smaller, fixed-
sized elements and passed to a mapper function. The mapper function
produces structured output data tuples of each component in the form
of key/value pairs, which are written into an intermediate file.
2. Reduce Phase: the output from the map phase is passed to the reduce
phase, before which it is shuffled to consolidate and combine the rel-
evant data tuples into smaller ones. The reducer aggregates the output
from shuffling by merging the same keys to form a smaller set of tuples,
which is written to a single output file.

Data storage can be carried out with tools other than HDFS, such as HBase, Hive,
Cassandra, Pig, Apache Flume, Apache Squoop, and other relational databases.
Whereas Apache Oozie is used in the case of large numbers of interconnected sys-
tems, Apache Zookeeper is used to maintain application reliability, and Mahout is
used for machine- learning purposes [7].

1.2.3 Data Visualization
Visualization is the graphical representation of data, which helps the practitioner
gain more insights from the data. The analytical tool cleanses and evaluates the data
with the help of data-mining algorithms, evaluation, and software tools before the
data are visualized. The main applications of data analytics to healthcare are que-
ries, reports, online analytical processing (OLAP), and data mining. They are used
Data Analytics in Healthcare Systems 7

for displaying predictive reports, proactive messages, visualization of patient health

records, real-time alerts, and dashboards for monitoring the daily health status of
patients.

1.3 DATA ANALYTICS TOOLS IN HEALTHCARE

One of the critical challenges in healthcare systems is utilizing the massive amount
of data generated daily in an efficient and cost-effective manner. Hence, healthcare
systems require effective and efficient tools to ensure the appropriate use of the data.
Figure 1.2 depicts the data analytics tools used in healthcare.

1.3.1 Data Integration Tools

Data integration in healthcare applications refers to the act of combining health data
from a myriad of sources into a unified set of accumulated data that provide action-
able business intelligence. Integration of multiple medical databases can be useful
in identifying different methods of disease prevention, providing more sophisticated
and personalized care, and reducing costs by avoiding overuse of resources.
As health data come from multiple, disparate sources like medical devices, wear-
able devices, etc., healthcare professionals face a major challenge in dealing with
these unstructured data. Collecting and consolidating health data from various
sources is extremely beneficial but still poses a major challenge to the health sector.
The process of integrating huge volumes of medical data is complicated and presents
significant challenges. Some of these challenges include:

• Lack of standard data formats

• Data privacy and confidentiality regulations
• Data format inconsistency among various healthcare applications
• Need for greater integration processing power
• Low end-user adoption

However, data integration tools can be used to integrate health data from multi-
ple sources to generate meaningful insights from the data. Data integration tools
include software and platforms that can aggregate data from disparate sources. The

FIGURE 1.2 Data analytics tools in healthcare.

8 ML and Analytics in Healthcare Systems

following are some data integration solutions that can be considered in healthcare
organizations to make better and more efficient use of healthcare data.

• Attunity is an integration tool that can aggregate disparate data and files
across all major databases, including cloud platforms, data warehouses, and
Hadoop. It also supports the Health Level 7 (HL7) messaging standard,
which is a healthcare standard. It can integrate and connect with web appli-
cations in real time.
• Informatica is an advanced, multi-cloud and hybrid data integration tool
that can integrate data from multiple, disparate datasets, such as data ware-
houses, Hadoop, enterprise applications, message applications, and mid-
range systems. It also provides data management tools for companies in
the healthcare field to facilitate patient services with improved outcomes
and reduced costs. Informatica’s cloud integration allows administrators to
integrate data with in-house applications, claims processing, etc., in health
organization environments.
• Information Builder is a data integration and business intelligence tool
that can measure and aggregate very large healthcare data collected
throughout the patient lifecycle. These tools ensure the availability of data
in real time across the healthcare environment and enhance the quality of
health services.
• Jitterbit [8] is a single, secure, cloud-based data integration platform that
aggregates structured and unstructured health data or clinical data retrieved
from sources such as EHR. It enables more-efficient operations and pro-
vides complete access to health data in a format that can be used with other
systems.
• Magic is a data integration tool that connects disparate systems for health-
care organizations. It ensures the best possible care of patients by keeping
all health-related records up to date and available to all healthcare provid-
ers. Magic’s integration platform combines diverse systems into a single
interface via the graphical user interface.

1.3.2 Searching and Processing Tools

Since healthcare organizations deal each day with a large volume of unstructured
data, there is a great need for data indexing, searching, and processing tools to opti-
mize the efficient use of clinical data . These tools are employed for effective man-
agement of data that is stored in HDFS, which allows multiple files to be stored and
retrieved at the same time in a big data environment. The following are some of the
searching and processing tools used in healthcare.

• Lucene [8] is a scalable tool for indexing large blocks of unstructured text
that provides advanced, full-text search capabilities. It can integrate easily
with Hadoop to facilitate distributed text management.
Data Analytics in Healthcare Systems 9

• Google Dremel [9] is a distributed system that uses multi-level execution

trees for interactive query processing of large datasets.
• Cloudera Impala is an MPP (Massively Parallel Processing) SQL query
tool that supports the accessing of a massive volume of data stored in
Hadoop clusters. It is a scalable and flexible parallel database technology
that enables users to directly query data stored in HDFS file formats with-
out requiring data format transformation.
• Apache Hive is a database query tool which facilitates querying, writing,
and managing large datasets stored in distributed storage using a HiveQL
language, which is a SQL-like query language. It allows structure to be
projected onto data already in storage.

1.3.3 Machine Learning Tools

In a healthcare environment, machine learning tools are used to convert comprehen-
sive health data into actionable knowledge that supports effective decision making to
perform informed clinical activities [8].

• Apache Mahout [10] is an open-source, powerful, scalable machine learn-

ing library that runs on top of Hadoop MapReduce. The Mahout library
facilitates the execution of distributed or scalable machine learning algo-
rithms. The Mahout algorithms mainly focus on classification, clustering,
and collaborative filtering techniques.
• Skytree is a big data machine learning tool that offers more accurate
and faster predictive analytics models that are easy to use. These models
enhance the processing of massive datasets in a precise manner without
down sampling.
• Apache SAMOA (Scalable Advanced Massive Online Analysis) is an
open-source-distributed, streaming machine learning framework that
enables users to create distributed, streaming machine learning algorithms
and to execute them on multiple DSPEs (Distributed Stream Processing
Engines) [11].
• BigML [8] is a scalable, open-source machine learning tool that provides
a framework to perform sophisticated machine learning workflows such as
classification, regression, cluster analysis, anomaly detection, and associa-
tion discovery. It includes a cloud infrastructure that can readily be inte-
grated with machine learning features to build cost-effective, scalable,
flexible, and reliable applications.

1.3.4 Real-Time and Streaming Data Processing Tools

With the rapid growth of massive Internet of Things (IoT), including wearables
and medical devices in smart healthcare systems, there is a need for real-time and
streaming data processing tools to carry out real-time processing of health data.
10 ML and Analytics in Healthcare Systems

• Apache Storm [12] is an open-source, real-time data processing platform

to process streaming data in real time. It also provides a fault tolerant, real-
time computational framework. The applications of Apache Storm include
real-time analytics, log processing, ETL jobs, continuous computation, dis-
tributed real-time processing, and machine learning.
• SQL Stream Blaze [12] is a streaming analytics platform that supports
a variety of available streaming sources in all formats and at all speeds.
Applications of SQL streaming include high throughput, data discovery,
data wrangling, real-time threat detection, and analytics.
• Apache Flink [12] is a distributed processing engine, and it provides a
framework for stateful computations over bounded and unbounded data
streams. It is capable of both batch and stream processing and offers effi-
cient, fast, accurate, and fault-tolerant handling of massive streams of
data.
• Apache Kafka is an open-source, distributed event-streaming platform
that provides a framework for building high-performance real-time stream-
ing data pipelines, streaming analytics, and applications. There are five
core application programming interfaces (APIs) in Kafka, namely Producer
API, Consumer API, Streams API, Connector API, and Admin API, to
facilitate message passing, storage, and stream processing services.

1.3.5 Visual Data Analytical Tools

Data visualization is a pictorial or graphical representation of data that enables
users in any organization to analyze and understand the trends or patterns of data.
In healthcare, data visualization tools help in creating and monitoring the dynamic
data of a patient, presenting clinical records, identifying patterns and trends, and
carrying out time-series analysis to improve healthcare services and public health
policy.

• SAS Visual Analytics [13] is a web-based analytical tool that allows multi-
ple users to access a massive amount of real-time data simultaneously from
a LASR analytical server. It allows parallel networking by transferring data
from one machine to another machine to access secure data quickly.
• Tableau [13] is a business intelligence visualization tool that transforms
raw and large datasets into a defined format to provide real-time structured
data to support decision making. One of the uses of Tableau has been to
quickly diagnose genetic diseases and to help health practitioners in provid-
ing rapid treatment to the patients.
• QlikView [13] is a visualization tool that transmits related data from differ-
ent sources into electronic medical records. It provides in-memory analysis
and reduces the risk of medical error by tracking safety metrics and lowers
the cost of delivering services to the patients. It ensures that all regulatory
compliance in the healthcare system is delivered and maintained in a timely
manner.
Data Analytics in Healthcare Systems 11

1.4 DATA ANALYTICS TECHNIQUES IN HEALTHCARE

Healthcare big data refers to multidimensional health data amassed from various
sources, including medical imaging (X-ray, magnetic resonance imaging (MRI), and
computed tomography (CT) scan images), structured data EHRs, biomedical signals
(ECG, EEG, etc.), handwritten prescriptions, and data from wearables and medical
devices. Since health data are dynamic and complex, they are difficult to manage
and analyze using traditional techniques and technologies. There is a great demand
for effective data analytics techniques to study these diverse data and to facilitate
decision making. Figure 1.3 depicts the various big data analytics techniques used in
healthcare, some of which are described below [5].

• Data Mining is useful for discovering patterns and for extracting mean-
ingful information from large databases. With the rapid growth in mas-
sive health data, data-mining techniques have helped to search for new
and valuable information (knowledge) from large complex databases in

FIGURE 1.3 Data analytics techniques in healthcare.

12 ML and Analytics in Healthcare Systems

healthcare systems, that facilitate the decision-making process. The main

application areas of data mining include prediction and determination of
various diseases, biosignal monitoring of patients, assistance in diagnosis
and treatments, and exploratory data analysis in healthcare.
• Machine Learning (ML) in healthcare helps to analyze data and suggest
outcomes. Applications of ML include prediction of diseases and diagnosis,
drug discovery and manufacture, hospital performance assessment, smart
health records, personalized patient care, etc.
• Cluster Analysis helps to discover hidden structures and clusters found
in massive databases. In healthcare, cluster analysis helps to identify sub-
groups in the patient population defined by the patient’s characteristics,
disease severity, and treatment responses. Cluster analysis can be used, for
example, to determine obesity clusters, to identify high-risk patient groups,
and to identify population clusters with specific disease determinants in
order to optimize treatment.
• Graph Analytics techniques analyze data in the form of a graph, where
individual data entities are represented as nodes and the relationships that
connect those entities are represented as edges. Graph analytics are used
in healthcare to estimate healthcare fraud risk and hospital performance
analysis based on quality measures [14].
• Natural Language Processing (NLP) is the technique used to make com-
puters understand human speech and text. In healthcare, NLP has great poten-
tial to search, analyze, and interpret large volumes of patient-related datasets.
Applications of NLP include the provision of efficient patient care, control of
health service cost, extraction of meaningful health-related data from clinical
notes [15], provision of training, consultations, and treatments, etc.
• Neural Networks are highly useful in analyzing, modeling, and interpret-
ing healthcare data. Applications of artificial neural networks (ANN) to
healthcare include clinical diagnosis, prediction of cancer, medical analy-
sis, and drug development.
• Pattern Recognition methods help to enhance the clinical decision support
system by focusing mainly on the patient’s condition, based on symptoms
and demographic information. They also help to improve public health
screening in healthcare systems [16].
• Spatial Analysis plays a vital role in the effective use of geographic infor-
mation systems (GIS) in healthcare, to facilitate health data exploration,
data modeling, and hypothesis testing. Spatial analysis applications include
direct patient care, epidemic disease prevention and intervention, assistance
with strategic planning, etc.

1.5 APPLICATIONS OF DATA ANALYTICS IN HEALTHCARE

Big data and analytics have various applications in healthcare. With the digitization
of patient records, medical history, X-rays, etc., there is a tremendous opportunity for
data analytics to uncover useful information. Along with the volume of data is the
Data Analytics in Healthcare Systems 13

sheer variety of data generated from different sources at different time points. The
volume of and variation in data generated by this sector are what makes it a topic of
great interest for data analysts. Conventional computing mechanisms and systems
fail to provide real-time monitoring and preventive plans for patients as well as for
the doctor. Hence, there is a need for smart strategies that decipher the incoming
data to uncover trends and anomalies and which give recommendations for patients,
helping doctors in their practice.
Some applications of data analytics in healthcare are as follows.

• Image-Based Analytics
Image-based datasets are a common source of information and are primar-
ily used by doctors for internal imaging. X-rays, mammography, CT, posi-
tron emission tomography–CT (PET–CT), ultrasound, and MRI are some
of the imaging technologies commonly used for diagnostics [5, 17]. Many
organizations and medical institutions release open datasets in a hope to
foster research activities. Neuro-images and MRI of the brain are widely
used for detecting tumors, an anomaly where the cells are enlarged and
form solid neoplasms. Early-stage detection is necessary for effective treat-
ment and recovery, and for decreasing the risk of mortality [18–20]. X-rays
are carried out for detecting fractures, pneumonia, cancer, etc. A large
amount of research has also been carried out for analyzing X-rays and CT
scans to promote early detection of the novel coronavirus COVID-19, with
minimal human intervention and interaction [21, 22].
With the growing number of medical records, the reliance on computer-
aided diagnosis and analysis is increasing [8]. High-performance comput-
ing and advanced analytical methods, like ML and optimization techniques,
are aiming to minimize predictive errors. Countries with low doctor-to-
patient ratios can benefit greatly from such machine-aided diagnostics.
A significant challenge associated with image-based analytics is the
amount of data generated. Images are space intensive. A single X-ray can
take up several megabytes (MBs) of storage space. The quality of an image
plays a crucial role in correct diagnosis and hence must not be compro-
mised. What adds to the processing challenges is that the data are highly
unstructured. Image processing techniques like segmentation, denoising
(noise reduction), and enhancement form the pre-requisites before use-
ful features, like color, contour, shape, pixel intensity, edges, etc., can be
extracted to train models for classification and diagnosis.
• Signal-Based Analytics
In a world full of wearable sensors, time-based signals are being generated
at a frequency greater than that at which they can be processed. Wearable
devices, like smart watches, smart rings, and fitness trackers, continuously
track heart rate, blood pressure, sleep patterns, calories burned, etc. Apart
from personal devices and gadgets, time-stream data are being generated
by electrocardiograms (ECG), ventilators, electroneurograms (ENG), elec-
troencephalograms (EEG), phonocardiograms (PCG), etc. [23]. Analysis of
14 ML and Analytics in Healthcare Systems

these signals plays a significant role in deciding and prescribing medica-

tion, care regime, routine check-ups, etc. Readings of such signals can pro-
vide useful information concerning the current status of the patient.
With lifestyle changes over the past decade, heart diseases have become
more common [24]. ECG signals, providing first-hand information on the
well-being of the heart, can be sensed through carbon nanotubes and sent
directly to doctors for real-time analysis, while the patient may still be at
home [25]. Such a set-up reduces the need for patients to physically be pres-
ent at the hospital or clinic to obtain a consultation. Furthermore, a live feed
can also be sent to intelligent devices which have been trained to use ML
models, to identify any anomaly in the signals [26]. On detection of any
abnormal signs, the doctor can be contacted.
The mental state of a driver can be monitored using physiological signal
analysis. This can prevent accidents which occur due to negligence or lack
of attention by the driver [27]. In stressful times like today, when people
are restricted by lockdown to their homes because of a global pandemic,
mental health monitoring becomes more crucial than ever. Thus, detection
and monitoring of neurological and mental health markers in patients can
be a game changer [28].
Signals from a wide range of sensors can be generated continuously.
These signals may be analyzed periodically for a routine check-up, or con-
tinuously, if the condition of a patient is critical. Nevertheless, all data still
need to be saved in the patient’s history, for future reference. Imagine the
amount of data being generated by an individual wearing a smart watch
24/7. As with image-based analytics, such data are enormous in volume
and rich in variety. Every sensor generates a different class of data. Such
time-stream data needs to be placed in context in order to derive mean-
ingful results regarding the current status of the patient. Furthermore, it
adds another layer of difficulty to the analysis of signal-based data. The
complexity of the problem is increased by the need for real-time stream
diagnostics and analytics.
• Clinical Diagnosis and Research
The study of signs, symptoms, and medical history of a patient to assess
underlying conditions or disease is known as a clinical diagnosis. It usually
does not involve any laboratory testing but uses the prior records of the
patient, along with current symptoms and characteristics, to reach a pre-
liminary level of diagnosis. Subsequently, additional tests can be ordered by
the doctor to pinpoint the exact extent of the disease or to uncover anoma-
lies and false symptoms. While a doctor uses years of experience to reach
this first level of diagnosis, it is a time-consuming process to go through the
medical history of the patient and past treatments to which the patient has
been subjected.
With the digitization of data, the entire medical history of every patient
is stored as an EHR. Information like clinical notes, prescriptions, admin-
istrative data, laboratory test results, medical imaging data, the patient’s
Data Analytics in Healthcare Systems 15

personal data, etc., can all be easily retrieved and stored permanently
because of ever-decreasing hardware cost [17, 29].
Data analytics can be helpful not only in the detection of disease at the
earliest stage, but can also attain high levels of accuracy, predicting the
timeline and disease development trajectory. Furthermore, data analytics
can bring to the doctor’s notice whether there is any change in the vital
signs, indicating a deviation from a healthy state. It can provide transpar-
ency to patients and provide them with a more personalized experience
of the entire healthcare management system [8, 30]. Healthcare organiza-
tions also benefit from data analytics as they help to provide cost-effective
care and personalized predictions for each patient. Clinical data can also
be useful for research purposes, such as the demographics of patients with
a particular condition, predicting the sales of drugs and their profitability,
identifying drug competitors, usage patterns of drugs, and effective drug
design, uncovering inter-drug associations, etc. [8].
Clinical decision support systems (CDSS) [5, 23, 31] may therefore be
an all-round solution for automated clinical diagnostics and research. But
achieving systems with high levels of accuracy and efficiency is a big chal-
lenge. The data gathered from different sources in multiple formats over
time adds volume, variety, and velocity to the data. Systems with high com-
putational power to deal with structured, semi-structured, and unstructured
data need to be implemented. Furthermore, handwritten clinical notes, pre-
scriptions, and medical journals need advanced ML algorithms with con-
cepts of NLP. Thus, the heterogeneity of clinical data remains currently the
most significant challenge and an open research area.
• Disease Transmission and Prevention
Some diseases are infectious and can be spread via direct or indirect con-
tact with the infected person or carrier. To prevent the outbreak of such con-
ditions, it becomes critical to study the means of transmission and to predict
the spread of the disease to develop better mitigation plans and improved
disease management strategies [32]. With the help of data analytics, math-
ematical and stochastic models can be generated to predict the outreach of
the disease and to estimate its impact.
Many researchers have studied the transmission of the novel coronavirus
that emerged in Wuhan in late 2019 and which has spread throughout the
world. It has been declared by the World Health Organization (WHO) to be
the worst epidemic in the past two decades [33]. Based on early available
data, symptoms, numbers of positive cases, and international travel history,
a predictive model was developed by researchers to identify the extent of
transmission and the risk it posed to human life [34]. Many government
organizations have also funded projects to research the prevention and
preparedness of individual countries. Massive amounts of data gathered
globally have been used to develop preventive measures to stop or mini-
mize further spread of the virus. Many countries opted for complete lock-
down, halting businesses, closing schools and universities, banning travel,
16 ML and Analytics in Healthcare Systems

etc. Medical equipment, ventilators, masks, personal protective equipment

(PPE), and sanitizers have been mass produced to deal with the prevail-
ing situation. Predictive analyses for other diseases, like HIV/AIDS, have
previously been carried out to enable early detection and treatment [35].
Transmission of all chronic illnesses could be prevented if the necessary
measures are undertaken, although this would be costly [36].
Data play an essential role in understanding the transmission model
of communicable diseases. The success of any disease prevention model
depends largely on the timely identification of the pathogen. Hence, accu-
rate classification, clustering, and associative models need to be developed,
which can help physicians to prescribe medication and other treatments [36].
• Precision Medicine
With an increase in the volume of data being generated on a daily basis
from sensors, implants, EHR, clinical practices, etc., the potential to fos-
ter healthcare and medical functionalities has increased dramatically.
Population health management, prevention of disease transmission, and
CDSS, have all become possible. Ever-increasing sizes of EHR, and their
dimensionality and variety, along with the incorporation of behavioral,
social, and omics data, have given data analysts the power to propose and
develop models for personalized patient care and precision medicine. The
layered architecture of such models, working on different aspects of health-
care, has facilitated the development of such comprehensive and detailed
healthcare solutions. By incorporating data from varied sources, and iden-
tifying relationships and patterns of interest, the entire healthcare solution
framework has migrated from a disease-centric view to a patient-centric
view. Diverse formats of data, the generation of bulk volumes of data, and
the inherent uncertainty associated with sources of big data complicate the
task of data curation. Transformation of raw data into useful facts and infor-
mation is vital for these healthcare framework solutions to achieve their
intended goals [8].
Precision medicine promises better healthcare delivery by improving
prognosis, diagnosis, and treatments being given to patients. Improving the
quality of clinical practices translates directly to personalized healthcare
routines for each individual, optimized for their direct benefit [37].
Some challenges identified for precision medicine involve upgrading of
ongoing clinical practices to incorporate newer, disruptive technologies in
the field of big data, the ability to handle large volumes of data generated
from different data sources and their analysis to achieve meaningful results
[38]. High-scale cognitive computing, using advanced ML models, can
drive data-driven analysis of biomedical big data [39].
• Health Insurance
Insurance agents and companies today enjoy an extensive database of cus-
tomers belonging to different demographics. The application of data ana-
lytics to this large dataset can help the agents to identify patterns, clusters,
and typical human behavior. With insurance models and schemes moving
Data Analytics in Healthcare Systems 17

online, consumers expect transparency and an accurate breakdown of the

money they are spending on health insurance. Thus, the agencies cannot
add some hidden costs to make additional profits. Competition among agen-
cies is stiff and, hence, attracting customers with lucrative benefits is how
these business agencies function [40].
Insurers today need to focus on individual customizable plans for indi-
viduals, instead of targeting groups of people. Predictive analysis is an
essential branch of analytics, which is widely used in such scenarios. It
identifies the appropriate coverage for the needs of people and further cus-
tomizes plans based on individual inputs. Smart models deployed online
first take the necessary information from users before developing proposals
for each individual. This personal experience and the tailor-made environ-
ment are valuable for customers.
The insurance costs can be intelligently calculated based on the current
health and medical history of an applicant. For instance, with the growing
obesity rate, the insurers may calculate the potential risk associated with an
individual applicant and modify the premium for health insurance accord-
ingly. Furthermore, as obesity increases the risk of heart failure, prelimi-
nary tests may be advised before the policy can be issued [41].
Thus, health insurance companies use big data analytics to deliver a per-
sonalized experience to each individual, to predict the occurrence of fraud
before it is realized, and to fast-track claims by predictive analytics.
• Service Delivery Systems
Today, healthcare providers across the globe are facing competition and
understand the need to deliver the highest quality of services to their cus-
tomers. Companies work on low profit margins and hence need to optimize
their service delivery systems to minimize the cost of providing the service.
The easiest way to do this is by automating the healthcare services and
reducing human intervention with respect to suggesting remedial actions,
preventive measures, prescriptions, diagnostics, etc. The majority of the
costs incurred by such organizations are on high-quality medical equip-
ment that needs to be upgraded over time, to keep pace with global techno-
logical advances. The research and development cost for new medicines is
also subject to extensive clinical trials before they can be rolled out to the
public for consumption, leading to high costs [42].
Furthermore, medical devices, equipment, and sensors continuously
generate large volumes of data which can be used for forecasting and deci-
sion making. Hence, data analytics can serve as a powerful tool for improv-
ing healthcare delivery systems.
A service delivery network usually consists of multiple organizations
working together collaboratively to provide an overall healthcare service
delivery system. These organizations are closely connected with one another
and share data among themselves to make smart, informed decisions. Players
involved in such networks may include sales representatives, doctors, physi-
cians, insurance personnel, laboratory technicians, hospital staff, equipment
18 ML and Analytics in Healthcare Systems

vendors, etc. They continuously explore ways to understand data being gener-
ated from different sources and to understand the impact such data have on
their policies and services. This co-development and deployment environ-
ment aims to reduce the cost of healthcare services [43, 44].

1.6 CHALLENGES ASSOCIATED WITH HEALTHCARE DATA

With the use of big data analytics, the healthcare industry has improved in many
aspects, including operational efficiency in healthcare management, reduction in
healthcare costs, improved drug discovery, higher-quality healthcare services, per-
sonalized patient care, effective treatments, and improved clinical outcomes. Despite
the benefits, however, big data introduces several challenges.

• One of the foremost challenges facing the healthcare industry nowadays

involves capturing health-related data. Due to the extensive use of IoT
devices and wearables in the smart healthcare system, which constantly
generate massive volumes of streaming data, the capture and processing of
data has become a major challenge. Lack of efficient governance practices
for data sources is another challenge for the capture of data from heteroge-
neous data sources, a problem which can lead to inaccurate data.
• Due to the exponential volume growth of streaming health data, data stor-
age has become a primary challenge for the healthcare industry. Most
health organizations prefer in-house data storage so that they can have con-
trol over data security and data access, but, with the rapid growth of health
data, in-house data storage infrastructures become difficult to scale up as
maintenance and scale-up costs are high.
• The privacy and confidentiality of medical data, in the form of a patient’s
health data, are of utmost importance in healthcare. Data sharing among
the various practitioners in a healthcare system escalates the need for pri-
vacy, while informed consent issues are another challenge faced by data
analytics in healthcare.
• Since healthcare data are generated from multiple sources, such as medical
images, wearable sensors, and EHRs, there are no fixed unified standards
for these data, leading to difficulties in consolidating and processing of the
data. Therefore, the lack of data protocols and standards are one of the gov-
ernance challenges facing healthcare data analytics.
• As digital healthcare has no geopolitical boundaries, health services are
available across international and national borders. It is difficult to form
uniform legislation because medical licenses, privacy of patient data, and
the advertisement and marketing of healthcare services may vary between
countries, representing major challenges for the healthcare industry.
• Data analytics in healthcare also pose new ethical and legal challenges,
including personal autonomy, risk of compromising the patient’s privacy,
need for informed consent, trust and transparency when using biomedical
big data.
Data Analytics in Healthcare Systems 19

1.7 CONCLUSION
Data analytics with big data in healthcare is still at the developing stage and advances
in tools and techniques will improve and their applications will expand. In addition,
establishing proper standards and governance of data, ensuring data privacy and
security, and updating the healthcare systems continuously are some of the chal-
lenges faced by the healthcare industry. Improving communication and data sharing
among related sectors in healthcare would increase the overall efficiency by provid-
ing value-added services, with minimal additional costs incurred.

REFERENCES
1. A. Belle, R. Thiagarajan, S. M. R. Soroushmehr, F. Navidi, D. A. Beard, K. Najarian.
Big data analytics in healthcare. Biomedical Research International, 2015, 16 pages,
2015. http://dx​.doi​.org​/10​.1155​/2015​/370194.
2. A. Kankanhalli, J. Hahn, S. Tan, G. Gao. Big data and analytics in healthcare:
Introduction to the special section. Information Systems Frontiers, 18, 233–235, 2016.
3. M. J. Ward, K. A. Marsolo, C. M. Froehle. Applications of business analytics in health-
care. Business Horizons, 57, 571–582, 2014.
4. S. Kumar, M. Singh. Big data analytics for healthcare industry: Impact, applications,
and tools. IEEE, 2 (1), 48–57, 2019.
5. N. Mehta, A. Pandit. Concurrence of big data analytics and healthcare: A systematic
review. International Journal of Medical Informatics, 114, 57–65, 2018.
6. Y. Wang, N. Hajli. Exploring the path to big data analytics success in healthcare.
Journal of Business Research, 70, 287–299, 2017.
7. N. El aboudi, L. Benhlima. Big data management for healthcare systems: Architecture,
requirements, and implementation, Hindwai. Advances in Bioinformatics, 2018. https://
doi​.org​/10​.1155​/2018​/4059018.
8. V. Palanisamy, R. Thirunavukarasu. Implications of big data analytics in developing
healthcare frameworks–A review. Journal of King Saud University-Computer and
Information Sciences, 31(4), 415–425, 2019.
9. S. Melnik, A. Gubarev, J. J. Long, G. Romer, S. Shivakumar, M. Tolton, T. Vassilakis.
Dremel: Interactive analysis of web-scale datasets. In: 36th International Conference,
2010.
10. C. E. Seminario, D. C. Wilson. Case study evaluation of Mahout as a recommender
platform. In: 6th ACM Conference on Recommender Engines (RecSys 2012), pp. 45–50,
2012.
11. https://www​.softwaretestinghelp​.com​/ big​-data​-tools/.
12. https://www​.pre​dict​ivea​naly​ticstoday​.com ​/top​-open​-source​-commercial​-stream​-ana-
lytics​-platforms/.
13. https://www​.yourtechdiet​.com​/ blogs​/impact​-data​-visualization​-healthcare/.
14. N. Downing, A. Cloninger, A. Venkatesh, A. Hsieh, E. Drye, R. Coifman, et al.
Describing the performance of U.S. hospitals by applying big data analytics. PLoS One
12(6), e0179603, 2017.
15. A. Khalifa, S. Meystre. Adapting existing natural language processing resources for
cardiovascular risk factors identification in clinical notes. Journal of Biomedical
Informatics, 58, S128–S132, 2015.
16. D. D. Luxton, J. D. June, A. Sano, T. Bickmore. Intelligent mobile, wearable, and ambi-
ent technologies for behavioral health care. Artificial Intelligence in Behavioral and
Mental Health Care, Elsevier, 137, 2015.
20 ML and Analytics in Healthcare Systems

17. B. Ristevski, M. Chen. Big data analytics in medicine and healthcare. Journal of
Integrative Bioinformatics, 15 (3), 1–5, 2018.
18. A. R. Kavitha, C. Chellamuthu. Brain tumour detection using self-adaptive learn-
ing PSO-based feature selection algorithm in MRI images. International Journal of
Business Intelligence and Data Mining, 15 (1), 2019.
19. S. Tchoketch Kebir, S. Mekaoui, M. Bouhedda. A fully automatic methodology for
MRI brain tumour detection and segmentation. The Imaging Science Journal, 67(1),
42–62, 2019.
20. T. V. N. Rao, H. Katukam, D. Guvva. Early brain tumour detection in MRI using
enhanced segmentation approach. image, 8, 9, 2019.
21. L. Brunese, F. Mercaldo, A. Reginelli, A. Santone. Explainable deep learning for pul-
monary disease and coronavirus COVID-19 detection from X-rays. Computer Methods
and Programs in Biomedicine, 196, 105608, 2020.
22. A. Jacobi, M. Chung, A. Bernheim, C. Eber. Portable chest X-ray in coronavirus dis-
ease-19 (COVID-19): A pictorial review. Clinical Imaging, 64 (April), 35–42, 2020.
23. C. K. Reddy, C. C. Aggarwal. An introduction to healthcare data analytics. In
Healthcare Data Analytics, 2015, pp. 1–18.
24. S. Dalal, V. P. Vishwakarma. GA-based KELM optimization for ECG classification.
Procedia Computer Science, 167, (2019), 580–588, 2020.
25. M. Bansal, B. Gandhi. IoT & Big Data in Smart Healthcare (ECG Monitoring). In
2019 International Conference on Machine Learning, Big Data, Cloud and Parallel
Computing (COMITCon), 390–396, IEEE, 2019, February.
26. S. Dalal, V. P. Vishwakarma, V. Sisaudia. ECG classification using Kernel extreme
learning machine. In 2nd IEEE International Conference on Power Electronics,
Intelligent Control and Energy systems (ICPEICES-2018), pp. 988–992, 2018.
27. Barua, S., Ahmed, M. U., & Begum, S. Distributed multivariate physiological signal
analytics for drivers’ mental state monitoring. In International Conference on IoT
Technologies for HealthCare, pp. 26–33, Springer, Cham, 2017, October.
28. M. Neumann, O. Roesler, D. Suendermann-oeft, V. Ramanarayanan. On the utility
of audiovisual dialog technologies and signal analytics for real-time remote monitor-
ing of depression biomarkers. In Proceedings of First Workshop on Natural Language
Processing for Medical Conversations, pp. 47–52, 2020.
29. Belle, A., Thiagarajan, R., Soroushmehr, S. M., Navidi, F., Beard, D. A., & Najarian, K.
Big data analytics in healthcare. BioMed Research International, 2015, 2015.
30. Wang, Y., & Hajli, N. Exploring the path to big data analytics success in healthcare.
Journal of Business Research, 70, 287–299, 2017.
31. Shafqat, S., Kishwer, S., Rasool, R. U., Qadir, J., Amjad, T., & Ahmad, H. F. Big data
analytics enhanced healthcare systems: A review. The Journal of Supercomputing,
76(3), 1754–1799, 2020.
32. Wong, Z. S., Zhou, J., & Zhang, Q. Artificial intelligence for infectious disease big data
analytics. Infection, Disease & Health, 24(1), 44–48, 2019.
33. Koubâa, A. Understanding the covid19 outbreak: A comparative data analytics and
study. arXiv preprint arXiv:2003.14150, 2020.
34. Kucharski, A. J., Russell, T. W., Diamond, C., Liu, Y., Edmunds, J., Funk, S., ... &
Flasche, S. Early dynamics of transmission and control of COVID-19: A mathematical
modelling study. The Lancet Infectious Diseases, 20(5), 553–558, 2020.
35. Das, N., Das, L., Rautaray, S. S., & Pandey, M. Detection and prevention of HIV
aids using big data tool. In 2018 3rd International Conference for Convergence in
Technology (I2CT), pp. 1–5. IEEE, 2018, April.
36. Razzak, M. I., Imran, M., & Xu, G. Big data analytics for preventive medicine. Neural
Computing and Applications, 32(9), 4417–4451, 2020.
Data Analytics in Healthcare Systems 21

37. Panayides, A. S., Pattichis, M. S., Leandrou, S., Pitris, C., Constantinidou, A., &
Pattichis, C. S. Radiogenomics for precision medicine with a big data analytics per-
spective. IEEE Journal of Biomedical and Health Informatics, 23(5), 2063–2079, 2018.
38. Hulsen, T., Jamuar, S. S., Moody, A. R., Karnes, J. H., Varga, O., Hedensted, S., ... &
McKinney, E. F. From big data to precision medicine. Frontiers in Medicine, 6, 34,
2019.
39. D. Cirillo, A. Valencia. Big data analytics for personalized medicine. Current Opinion
in Biotechnology, 58, 161–167, 2019.
40. Gupta, S., & Tripathi, P. An emerging trend of big data analytics with health insurance
in India. In 2016 International Conference on Innovation and Challenges in Cyber
Security (ICICCS-INBUSH), pp. 64–69, IEEE, 2016, February.
41. Revels, S., Kumar, S. A., & Ben-Assuli, O. Predicting obesity rate and obesity-related
healthcare costs using data analytics. Health Policy and Technology, 6(2), 198–207,
2017.
42. Alotaibi, S., Mehmood, R., & Katib, I. The role of big data and twitter data analytics
in healthcare supply chain management. In Smart Infrastructure and Applications, pp.
267–279, Springer, Cham, 2020.
43. M. A. Pikkarainen. Data as a driver for shaping the practices of a preventive healthcare
service delivery network. Journal of Innovation Management, 1, 55–79, 2018.
44. M. Usak, M. Kubiatko, M. Salman. Health care service delivery based on the
Internet of things: A systematic and comprehensive study. International Journal of
Communication Systems, 33, 1–17, 2019.
View publication stats