You are on page 1of 7

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

net/publication/259759844

Business Intelligence Solutions in Healthcare A Case Study: Transforming OLTP


system to BI Solution

Conference Paper · June 2013


DOI: 10.1109/ICCITechnology.2013.6579551

CITATIONS READS
11 3,730

3 authors:

Osama Ali-Ozkan Ali Bou Nassif


The University of Western Ontario University of Sharjah
6 PUBLICATIONS   21 CITATIONS    87 PUBLICATIONS   1,106 CITATIONS   

SEE PROFILE SEE PROFILE

Luiz Fernando Capretz


The University of Western Ontario
253 PUBLICATIONS   3,217 CITATIONS   

SEE PROFILE

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

Capturing Emirati-Accented Speech Corpora for Applications of Speech Signal Processing View project

MBTI Personality types in Software Engineering View project

All content following this page was uploaded by Ali Bou Nassif on 27 March 2014.

The user has requested enhancement of the downloaded file.


ICCIT-2013: Special Session-Computational Intelligence Applications in Software Engineering (CIASE), Beirut

Business Intelligence Solutions in Healthcare


A Case Study: Transforming OLTP system to BI Solution
Osama T. Ali, Ali Bou Nassif and Luiz Fernando Capretz
Electrical and Computer Engineering
The University of Western Ontario
London, Ontario, Canada, N6A 5B9
{oali8,abounas,lcapretz}@uwo.ca

Abstract – Healthcare environment is growing to include not to make the right decision whether at enterprise or
only the traditional information systems, but also a business departmental level, and this service requires many other
intelligence platform. For executive leaders, consultants, and important components, such as clinical/operational data
analysts, there is no longer a need to spend hours in design and stores (ODS), data warehouse (DW), OLAP, query and
develop of typical reports or charts, the entire solution can be
reporting, data mining (DM), etc.
completed through using Business Intelligence “BI” software.
This paper discusses current state-of-the-art B.I components
(tools) and outlines hospitals advances in their businesses by II. WHAT IS BUSINESS INTELLIGENCE (BI)
using B.I solutions through focusing on inter-relationship of BI is the one of hottest buzzword for the last 4-5 years in
business needs and the IT technologies. We also present a case the business administration and information management
study that illustrates of transforming a traditional online fields. There are a lot of definitions for BI as follows:
transactional processing (OLTP) system towards building an
online analytical processing (OLAP) solution. A “BI is a strategic initiative by which organizations
Keywords- business intelligence; healthcare informatics; data measure and drive the effectiveness of their competitive
warehouse; data mining; data analytics; decision support; strategy” [1]. In order to achieve this grand goal, there is
need for analysis, software, resources, technical leadership,
I. INTRODUCTION process specialists, executive leaders and much more.
Health Informatics is one of the rapidly growing fields that Gartner [2] also defines a BI platform as a “software
focus on applying “Computer Science” and “Information platform that delivers the 14 capabilities within three main
Technology” to medical and health data. categories of functionality” as follows:
Every day hospitals are collecting huge volumes of data, 1) Integration: BI infrastructure, Development tools,
and the challenge they now have is how to make the raw data Metadata management, Collaboration.
inside the relational databases meaningful information for 2) Information Delivery: Reporting, Ad-hoc query,
executive leaders in the organization to make effective Dashboards, Data integration, Search-based BI, Mobile BI.
decisions based on what is going on now and to predict what 3) Analysis: Online analytical processing (OLAP),
will happen in the near future. Interactive visualization, Data mining and Predictive
The implementation of BI solutions within healthcare modelling, Scorecards (Key Performance Indicators -KPIs,
settings is still one of the controversial issues among Performance Management Methodology) [2].
healthcare providers, analysts and executive managements’
levels, which executive leaders like to see high level of III. BUSINESS INTELLIGENCE ARCHITECTURE
aggregated data to make strategic decisions. On the other Fig. 2 represents the essence of BI with the proper
side, analysts need to work with the transactional data to workflow of the interdependent components as follows:
build daily basis operational and monitoring reports.
The hierarchy of different organizational solutions based A. The Multidimensional Data Warehouse
on their complexity and strategic insight levels can be Data Warehouse (DW) is the core of any solid BI solution.
depicted in the Fig. 1. Data Warehouse can be defined as a “repository for keeping
High data in a subject oriented, integrated, time variant and non-
volatile manner that facilitates decision support” [3][4].
Strategic Insight

Basically it is a big database containing all the data needed


H.W & S.W
Complexity

for performance management, decision making, and


prediction. The multi-dimensional modeling techniques is
using facts and dimensions within relational or multi-
dimensional databases and it is typically used for the design
Low of corporate data warehouses and departmental data marts.
Fig. 1: Organizational solutions according to strategy and complexity Such a model can adopt one of the data mart’s building
schema whether a star, snowflake, or fact constellation
Eventually, the destination of all healthcare administration schema [5]. The task of designing and construction of a Data
levels is to use analysis service tools in a very efficient way warehouse is very complex, it involves many technical issues
related to a number of fields and subfields [6].

978-1-4673-5307-6/13/$31.00 ©2013 IEEE 209


ICCIT-2013: Special Session-Computational Intelligence Applications in Software Engineering (CIASE), Beirut

work to clean up and harmonize the data coming from


different systems, that’s why most of BI professionals
describe ETL work as 70% of the IT side of a BI project.
Through working on the development of ETL component, a
separate database in the data warehouse is reserved as storage
space for intermediate results of the required transformations.
This area is called staging area or work area. Once the
transformation work is done, the prepared data can be loaded
into the multi-dimensional model.

D. Operational Data Store (ODS):


Operational data store is an off-line copy of one or more
production source systems which has some characteristics as
follows:
1) Has Entity Relational (ER) data model.
2) Has additional functionality of storing historical
versions of the data.
3) Potentially has some or full integration between the ER
models of different applications
In any case the ODS is built for supporting the operational
reporting that related to the functional scope of a specific
application and it is not to support strategic decision making.

E. Online Analytical Processing (OLAP) cubes:


Fig. 2: The main components of BI architecture [7]
OLAP is analysis techniques including variety of
functionalities such as aggregation, summarization, and
A Clinical Data Warehouse (CDW), as defined by Gray consolidation as well as the ability to view information from
[8], is a “Place where healthcare providers gain access to
different angles [5]. OLAP offers high performance in
clinical data gathered in the patient care process”. Some
analysis and loading of the data. OLAP cubes have had very
researches consider the extent of the essential background
high success rates for business environments where the BI
knowledge of the architect is slightly different between DW
solution is used for what-if analysis, financial simulations,
and CDW which is the CDW is immensely complex to build,
budgeting and target setting, etc. Fig. 3 shows the logical
and maintain when compared to other business. representation of multi-dimensional cube.

B. The source systems:


The source systems are not really considered as a part of
the BI environment, but they feed the BI solution and so they
are at the basis of whole architecture and should be totally
understood by the developers. One of the important things in
the set-up of BI environment is that to consider all the types
of data that may need to be included in the analysis process.
The BI architect should take in his/her consideration to
include all information in databases, external data feeds from
other stakeholders, e.g. through XML; or connecting to
Enterprise Resource Planning (ERP) systems on application
server level; or uploading Excel sheets and flat data files
(always needed); and many more candidate data sources.

C. ETL: Extract, transform and load:


It’s a process of pulling the required data from different Fig. 3: A 3-D data cube representation of the data, according to the
data sources and populates the multi-dimensional data dimensions time, age groups, and location. The measure displayed is number
warehouse within BI environment, the potential connectivity of patients [9]
issue is a keyword here. After extraction task, the data needs
to be transformed. The transformation process can mean a lot F. Semantic layer for reporting:
of things which includes all tasks to make the data match and The objective of a BI solution is to offer a tool that enables
fit the multi-dimensional model. end users easily access to information for analytical
Given the significant difference between entity relational purposes. Unfortunately, databases are often not user-
and multi-dimensional models, the transformation process friendly, therefore most vendors of reporting solutions started
may become quite complex specially when it includes extra to create a layer between the database and the reporting

210
ICCIT-2013: Special Session-Computational Intelligence Applications in Software Engineering (CIASE), Beirut

which made data access easier for end users. In this layer the 3) Online Analytical Processing (OLAP): Support analysts
database fields can be translated into Objects, each object has with the capability of perform multi-dimensional analysis of
a business definition. These objects can be easily dragged data (i.e rollup, drill down, slice-and-dice, ‘what if’ analysis).
and dropped onto a report, making report creation easy. 4) Query and Reporting Services: Provide quick and easy
access to the data with predefined report design capabilities.
G. BI portal: 5) Data Mining (Predictive Model): Examines data to
When the amount of different reports begins to grow, the discover hidden facts in databases using different techniques
best solution is to create a single point of access to (i.e, statistical analysis, machine learning, frequent
information within the organization. Usually the efforts of pattern/relation finding, infer predictive and descriptive
creating a single point of access is results into building some information). Examples of successful predictive models
kind of intuitive portal solution that contains different reports include [12][13][14].
with clear descriptions of the scope of each report, as well as 6) Operational Data Services: Collect data from end users,
indication who is the business owner of the report. organizing data, establishing solid data structures and store
them in different databases, retrieve data from multiple
H. Data mining: databases.
Data mining is “the automated process of discovering 7) Integration Services: Design and implement of process
previously unknown useful patterns within structured data.” flow of data extracting, transforming, and loading to the data
The data warehouse is a perfect environment to conduct data warehouse.
mining exercises along with online analytical processing
OLAP in which users can slice and dice, pivot, sort, filter B. Business Solutions: Business focused analytical
data to discover patterns often using visual data mining form. applications, as follows:
Without applying data mining techniques, it is difficult to 1) Patient Analysis: Focuses on analysis of patients’
realize the potential of data collected within healthcare demographic and satisfaction processes.
organization as data under analysis is massive, multi- 2) Electronic Health Record Analysis: Focuses on analysis
dimensional, distributed and uncertain [10]. of the quality of clinical data (illness, diagnosis, medication,
etc.)
IV. BI APPLICATIONS FOR HEALTHCARE 3) Performance Analysis: Streamline and optimize the way
BI applications in healthcare can be categorized in two that a business uses its resources (budget, human,
major sets of solutions as shown in Fig. 4 [11]: equipment).
4) Fund Channel Analysis: Devise, implement, and
Business Solutions Technology Solutions evaluate fund strategies, then use the corporate metrics to
continuously monitor and enhance the fund process
Patients (Demographic and 5) Productivity Analysis: Focuses on building business
Satisfaction) Analysis
Decision Support Systems
metrics for activities such as quality improvement, risk
Electronic Health Record mitigation, asset management, capacity planning, etc.
Analysis Executive Information Systems 6) Behavioural Analysis: Understanding and predicting
Performance Analysis trends and patterns that provides business advantage.
(Staffing & Scheduling, Case Costing) Online Analytical Processing 7) Supply Chain Analysis: Monitor, benchmark, and
Fund Channel Analysis (OLAP) improve supply chain activities from materials ordering
Query & Reporting Services through service delivery.
Productivity Analysis 8) Wait Time Analysis: Focuses on the factors that are
Data Mining (Predictive Model) associated with longer waiting times and the effects of delays
Behavioural Analysis
in scheduling and operation.
Operational Data Services
Supply Chain Analysis
Integration Services (ETL) There is a big competition among BI vendors based on
Wait Time Analysis many factors, such as performance, customization, risk
tolerance, and business improvement.
Fig. 4: BI business and technology focused applications [11]
V. CASE STUDY: TRANSFORMING OLTP
A. Technology Solutions: It’s Data & Information Tools and RELATIONAL DB TO B.I SOLUTION
Services, as follows: A. Corporate Project Purpose
London Health Sciences Centre (LHSC) continues to have
1) Decision Support Systems (DSS): Support managerial significant challenges related to infection rates, reporting
decision making, usually day-to-day tactical. rates well above the provincial average and often showing
2) Executive Information Systems: Support decision making LHSC as one of the poorest performing teaching centres in
at the senior management level which provide and Ontario. The corresponding costs to the organization (time,
consolidate metrics-based performance information. energy, supplies and equipment) are extreme. Reducing
infections has become a strategic priority for LHSC.

211
ICCIT-2013: Special Session-Computational Intelligence Applications in Software Engineering (CIASE), Beirut

The infection safety solution will consist of multiple • Creating a solid data infrastructure which is made up by a
projects and strategies to improve infection rates of antibiotic collection of heterogeneous systems
resistant organisms (AROs) across LHSC. Having a solid • Automating the process of extraction, transformation, and
framework for collecting, managing and reporting data and loading in order to integrate all the different data sources
information will be a key success factor in these projects and in central repository.
overall solution. • Building multi-dimensional cubes to provide with all the
The intent of this project is to define and recommend an required information to different level of users (managers,
appropriate data management and reporting structure to specialists, consultants, analysts, etc.).
support the Infection Safety corporate strategy. This will
include (but may not yet be inclusive) the following 4) Infection Prevention and Control (IPaC) Reporting
deliverables: Tool/Dashboard
1) Hand Hygiene Audit Technology • Review existing Internal and External reporting
2) Corporate Data Warehouse with Data Integration and processes, tools and reports for ARO and Safer healthcare
Analysis Tools Now (SHN) indicators and evaluate effectiveness and
3) Infection Prevention and Control (IPaC) Reporting Tool efficiency of data analysis and review.
/ Dashboard • Obtain feedback from IPC and identify internal
departmental reporting needs.
B. Project Scope: • Solicit clinical director and manager level feedback re
This project will cover the approaches of analysis and current ARO reports, formats, delivery.
implementation of B.I solution for traditional OLTP • Consider dashboard/ business intelligence approach;
relational database in order to expand the classical use of benchmark with other healthcare organizations regarding
data from different perspectives through showing how SQL software tools, review potential vendors in collaboration
Server 2008 Integration and Analysis Services solves many with ITS and make recommendations
business problems in a simple and cost-effective manner via
providing a simple, integrated view of data, data mart and D. Technology used for design and implementation:
specialized application consolidation, intelligent views of • For Data Mart Building: Microsoft SQL Server 2008 R2
data, and real-time business intelligence data. Management Studio
The most challenging part of this project was the analysis • For Data Transformation (ETL): SSIS packages
and design of data warehouse model by using the Star (Business Intelligence Development Studio with Visual
schema as well as building the cube and KPI. Studio 2008
• For Reporting: SSRS or MS Excel 2010
C. Project Outcomes: E. Traditional System’s Diagram:
The expected outcomes from the project include the Fig. 5 illustrates the traditional solution architecture and
following: workflow.
1) All Data Management and Reporting
• Understanding of our current state regarding how data is
collected, analysed, and disseminated including what
infrastructure, tools and support exist.
• Defining standards and processes regarding future
collection, analysis and reporting of data related to
infection safety.
• Recommending infrastructure, appropriate tools and
technology required to support quality data management
and reporting.

2) Hand Hygiene System


• Clear definition of roles and responsibilities regarding the
auditing and data entry process.
• To have a short-term interim solution that will allow us to
deliver monthly audits as soon as possible until a long-
term solution has been developed and implemented.
• Inform requirements and support ITS as they consider an
appropriate long term solution for the auditing.
Fig. 5: Diagram of data systems and infrastructure of the existing traditional
3) Data Warehouse Building systems [15]
• Creating a standard data dictionary for all heterogeneous
information

212
ICCIT-2013: Special Session-Computational Intelligence Applications in Software Engineering (CIASE), Beirut

F. Project Methodology:
In this project as a mid-sized solution, the BI developer
was applying the following development life cycle phases:
1) Design the data warehouse first through using SQL
Server Management Studio. In particular, design the tables
(dimensions and fact) that are needed as part of the DW,
ignoring any staging tables as shown in the Fig. 6.

Fig. 6: Star schemal data mart design (dimensions and fact tables)

2) Design the ETL by using SQL Server Integration


Services project which contains a lot of data integration
components as shown in the Fig. 7, also sometimes the stored
procedures have been used within the production databases:
• Its fine if any staging tables are required as part of the
ETL, but at the same time they should get cleaned up. A
staging table used only as part of a single series of ETL
steps should be truncated after those steps are completed.
• SSIS packages refer to the OLTP database at least to pull
data into the staging tables. Depending on the situation,
they may process the OLTP tables directly into the data
warehouse.
• Documenting and make it clear what inputs are used by
each package, where the output goes, and the criteria by
which the input are selected (i.e. last 24 hours? Since last
success? New identity values? All rows?)
Fig. 7: Building the ETL framework by using SQL server integration
3) Designing Analysis Service project by using VS 2008 services project
SSAS as follows:
• Create Data Source. 4) Build required reports by using SSRS or Excel 2010
pivot table.
• Create Data Source View.
5) Data mining component
• Create associated dimensions plus DateTime Dim.
• Create multi-dimensional cube. VI. CONCLUSIONS
• Create KPI’s (Key Performance Indicators) to display In this paper, the general architectural approaches for BI
monthly metrics. solution have been outlined and its major development

213
ICCIT-2013: Special Session-Computational Intelligence Applications in Software Engineering (CIASE), Beirut

components have been introduced in order to give the reader [7] R. Guro. Components of business intelligence. The Business Intelligence
Guy [Online]. 2011. Available: http://www.the-business-intelligence-
a high level picture with some important details regarding
guy.com/components-of-business-intelligence-bi/.
healthcare informatics. [8] G. W. Gray, "Challenges of Building Clinical Data Analysis Solutions,"
Many healthcare organizations struggle with the lack of Journal of Critical Care, vol. 19, pp. 264-270, 2004.
utilization of data collected through non-integrated OLTP [9] Information technology for students and professionals. United States
Information Source [Online]. Available: http://www.info-source.us/.
system which have been used for decision making and data
[10] H. Kaur and S. K. Wasan, "Empirical Study on Applications of Data
mining. For successful healthcare organization it is important Mining Techniques in Healthcare," Journal of Computer Science, vol. 2, pp.
to empower the staff and management with data warehousing 194-200, 2006.
based on critical thinking and knowledge management tools [11] M. Peco, "TDWI Business Intelligence Fundamentals: From Data
Warehousing to Business Impact," The Data Warehouse Institute, 2011.
for strategic decision making.
[12] A. B. Nassif, D. Ho and L. F. Capretz, "Towards an Early Software
Decision support tools such as data mart, OLAP and data Estimation Using Log-linear Regression and a Multilayer Perceptron
mining techniques can support on building a solid foundation Model," Journal of Systems and Software, vol. 86, pp. 144-160, 1, 2013.
for clinical data warehouse. [13] A. B. Nassif, L. F. Capretz and D. Ho, "Estimating software effort
based on use case point model using sugeno fuzzy inference system," in
A data mart is a subset of data warehouse. It focuses on 23rd IEEE International Conference on Tools with Artificial Intelligence,
selected business subjects. OLAP solutions provide a multi- Florida, USA, 2011, pp. 393-398.
dimensional view of the data found in relational databases. [14] A. B. Nassif, L. F. Capretz and D. Ho, "Software effort estimation in
Storing and presenting data in three dimensional format as a the early stages of the software life cycle using a cascade correlation neural
network model," in 13th ACIS International Conference on Software
OLAP cube makes it possible to analyze potentially large
Engineering, Artificial Intelligence, Networking and Parallel & Distributed
amount of data with very fast response times and provides Computing (SNPD), Kyoto, Japan, 2012, pp. 589-594.
the ability for users to go through the data and drill down, [15] A project charter, "Data Management System for Infection Prevention
roll up, or slice-and-dice through various dimensions as and Control," IPAC Project Team, London Health Sciences Centre, 2011.
defined by the data structure.
The primary motivations of the case study in this paper
have been to expand the horizon of a software engineer to
mix the business acumen with the technical expertise in order
to maximize the strategic thinking through moving from
classical information systems towards enterprise BI
solutions.

VII. FUTURE WORK


Within big healthcare organizations there is vast potential
for implementation of BI solutions (Clinical Data
Warehouse, Information Visualization, and Data Mining
Applications), such as, operational performance, quality and
risk management, case costing, patient safety factors, wait
time management, staffing and scheduling, evaluation of
effective treatment and best practices, predictive medicine,
image and pattern recognition, etc.
Through data mining and predictive analytics, historical
data can reveal patterns that are used to predict trends.
Historical data analysis and predictive analytics, together
with expert knowledge will effectively assist in the diagnosis
and treatments of numerous diseases.

REFERENCES
[1] Business intelligence – an endless story. MAIA Intelligence [Online]. pp.
1-14. 2011. Available: http://www.maia-intelligence.com/pdf/BI-An-
endless-story-wp.pdf.
[2] J. Hagerty, R. L. Sallam and J. Richardson. Magic quadrant for business
intelligence platforms. Gartner [Online]. 2012. Available:
http://www.microstrategy.com/download/files/whitepapers/open/gartner-
magic-quadrant-for-bi-platforms-2012.pdf.
[3] W. H. Inmon. What is a data warehouse? Prism Solutions, Inc [Online].
1995. Available: www.cait.wustl.edu/cait/papers/prism/vol1_no1/.
[4] Y. Naddaf. Data mining in health informatics. [Online]. Available:
http://yavar.naddaf.name/downloads/Data%20Mining%20in%20Health%20I
nformatics.pdf.
[5] J. Han and M. Kamber, Data Mining: Concepts and Techniques.
Kaufmann, 2006.
[6] A. Sen and A. P. Sinha, "A Comparison of Data Warehousing
Methodologies," Communications of the ACM, vol. 48, pp. 79-84, 2005.

View publication stats


214

You might also like