e-Health – For Continuity of Care 657

C. Lovis et al. (Eds.)

© 2014 European Federation for Medical Informatics and IOS Press.
This article is published online with Open Access by IOS Press and distributed under the terms
of the Creative Commons Attribution Non-Commercial License.
doi:10.3233/978-1-61499-432-9-657

User Needs and Requirements Analysis for

Big Data Healthcare Applications
Sonja ZILLNERa,b,1, Nelia LASIERRAc, Werner FAIXb, and Sabrina NEURURERc,d
a
Corporate Technology, Siemens AG, Munich, Germany
b
School of International Business and Entrepreneurship,
Steinbeis University, Berlin, Germany
c
Semantic Technology Institute Innsbruck, University of Innsbruck, Austria
d
Department of Medical Statistics, Informatics and Health Economics,
Innsbruck Medical University, Austria

Abstract. The realization of big data applications that allow improving the quality
and efficiency of healthcare care delivery is challenging. In order to take
advantage of the promising opportunities of big data technologies, a clear
understanding of user needs and requirements of the various stakeholders of
healthcare, such as patients, clinicians and physicians, healthcare provider, payors,
pharmaceutical industry, medical product suppliers and government, is needed.
Our study is based on internet, literature and market study research as well as on
semi-structured interviews with major stakeholder groups of healthcare delivery
settings. The analysis shows that big data technologies could be used to align the
opposing user needs of improved quality with improved efficiency of care.
However, this requires the integrated view of various heterogeneous data sources,
legal frameworks for data sharing and incentives that foster collaboration.

Keywords. Big data, user needs, requirements analysis

Introduction

Recent publications [1-4] highlight the potential impact of big data applications and
health data analytics for the healthcare sector. However, as of today, big data
technology is rarely used within health care settings. The aim of this paper is to present
and discuss the results of a big data related user needs and requirements analysis that
was accomplished in the course of a big data technology road map development for the
healthcare domain. Big data technologies establish means to aggregate and analyze
large sets of heterogeneous data sources, and allow generating new insights and
knowledge. Thus, one essential step for developing a technology road map for big data
in the healthcare domain is to determine the particular information needs of the
involved user groups. The context of our work is the Big Data Public Private Forum2
Project, which aims towards developing a technology road map for big data
technologies for the European healthcare market3.

1
Corresponding Author: [email protected]
2
http://www.big-project.eu/
3 Beside the healthcare sector, several other industrial sectors, such as Energy, Transport, Finance,
Manufacturing, Retail and the Public Sector are addressed within the project but not focus of this publication.
658 S. Zillner et al. / User Needs and Requirements Analysis for Big Data Healthcare Applications

The goal of our study was to identify the most relevant information needs of the
various stakeholder groups in the healthcare delivery domain. They could be addressed
by aggregating and analyzing large and heterogeneous data sets. Particular
requirements that need to be addressed in order to foster the implementation of big data
applications in the healthcare domain were identified as well.

1. Methods

As depicted in Figure 1, our study consisted of two different stages.

Figure 1. Two-stage approach of the presented study.

The goal of the first stage was to identify both stakeholders and use case
applications of big data. Therefore, we conducted a literature review including
scientific reviews, market studies and other internet sources. This knowledge allowed
us to identify and select potential interviews partners, and to develop the questionnaire
for the domain expert interviews. The interview guide consisted of 12 introductory
questions with open response options which were clustered into three parts: a) direct
inquiry of specific user needs b) indirect evaluation of user needs by discussing the
relevance of the use cases identified at Stage 1 as well as any other big data
applications they were aware of c) reviewing constraints that need to be addressed in
order to foster the implementation of big data applications in healthcare. At Stage 2, we
conducted 13 semi-structured interviews with an average length of 75 minutes. At least
one expert of each stakeholder group identified in Stage 1, such as patients, clinicians,
hospital operators, pharmaceutical industry, research and development (R&D), payors,
and medical product providers was interviewed. To derive the user needs from the
collected material respectively interview transcripts, we aggregated the most relevant
and frequently mentioned use cases into high level application scenarios. Our data
collection and analysis strategy was inspired by the triangulation approach [5].
Reviewing and quantitatively assessing the high-level application scenarios (see [6]),
we derived a reliable analysis of user needs (see Section 0), and by examining likely
constraints of big data applications the relevant requirements that need to be addressed
were identified (see Section 2.2).

2. Results

In order to derive user needs and requirements from the collected data, we
aggregated the most relevant and frequently mentioned use cases (53 of the 67 use
cases that we had discussed with our interviewees) into six high-level application
scenarios: 1) Comparative effectiveness research aiming to compare clinical and
financial effectiveness of clinical care services 2) Clinical Decision Support assisting
 S. Zillner et al. / User Needs and Requirements Analysis for Big Data Healthcare Applications 659

the decision making process of clinicians 3) Clinical operation intelligence aiming to

optimize clinical processes 4) Secondary Usage of health data aiming to discover new
knowledge by means of data analytics 5) Public health analytics relying on
comprehensive disease management of chronic and severe disease and 6) Patient
engagement platforms that foster the active engagement of patients in the care process.
The six high-level application scenarios established the basis for our user needs and
requirements analysis.

2.1. User Needs

User needs in the context of big data can be captured by identifying the most
relevant information needs of the various user groups. Within all application scenarios,
any information unit that helped the involved users to improve the quality of care
without increasing the costs was of great value. For instance, individualized treatment
paths cannot be standardized and thus are likely to become very labor and cost-
intensive without means of big data-based analytics. Thus, any information that could
help to improve the quality AND the efficiency of care at the same time was indicated
as most relevant and useful for the user groups. In general, those high impact insights
can only be realized if the data analytics is accomplished on heterogeneous data sets
encompassing data from the clinical, administrative and financial domain.

2.2. Requirements

Within our study, we identified several requirements that need to be addressed in

order to foster the implementation of big data healthcare applications. We distinguished
requirements a) that are business-related (BR), b) technical-related (TR) and c)
requirement that are business and technical-related (BTR):
High investment needed (BR): The majority of big data applications in the healthcare
sector rely on the availability of large-scale, high-quality and longitudinal health
care data. The collecting and maintaining of such comprehensive data sets require
not only high investments, it usually takes several years until the data sets are
comprehensive enough for producing insightful analytics results. In general, such
expensive and long-term based investments can rarely be covered by one single
party, such that the conjoint engagement of multiple stakeholders, often including
the government, is needed.
Value-based system incentives needed (BR): Current system incentives enforce “high
number” instead of “high quality” of treatments. Although, it is very obvious that
nobody wants to pay for treatments that are ineffective, this is still the case in
many medical systems. In order to avoid low-quality reimbursements, the
incentives of the medical systems need to be aligned with outcomes and thus
foster cooperation between stakeholders.
Business cases with several partners needed (BR): Business cases for big data-bases
solutions are difficult to identify. Several partners with diverging points of
interests need to cooperate. Often, the one who is benefiting from the solution, is
not the one who is in the position to drive the solution or able to pay for the
complete solution. For instance, the implementation of data analytics solutions
using clinical data requires high investments and resources to collect and store
patient data [1]. Although, it seems to be quite obvious how the involved
660 S. Zillner et al. / User Needs and Requirements Analysis for Big Data Healthcare Applications

stakeholder could benefit from the aggregated data sets, it remains unclear
whether the stakeholder would be willing to pay or drive such an implementation.
Data security and privacy (BTR): As of today, legal frameworks defining data access,
security and privacy issues and strategies are missing, and the seamless sharing
and exchange of data is hindered. Simply because the involved parties lack
procedures for sharing and communicating relevant findings, important data and
information often remains within one department, group or organization.
Data quality (BTR): Although there exist many big data applications that are mainly
looking for patterns in data, and thus do not need clean data, this is not the case
for the healthcare domain. In order to derive reliable insights for health-related
decision, high data quality standards need to be fulfilled. For instance, the
features and parameter list used for describing the patient health status need to be
standardized in order to enable the reliable comparison of patient or population
data sets.
Data digitalization (TR): Still today, a high percentage of health related data is
documented in paper-based form. However, to derive the maximum benefit from
health-related data analytics, the data needs to be available in digital format. And,
it needs to be complete and in good quality. In order to fulfill this requirement,
the demand for standardization as well as technologies supporting the
documentation process (e.g. context sensitive information extraction) is pointed
out.
Semantic annotation (TR): Health data consist of very heterogeneous data, such as lab
reports, medical images, clinical reports, sensor data or gene test results. Only a
small percentage of this data is documented in a structured or standardized
manner (e.g. International Classification of Diseases4 (ICD) codes for diagnoses,
laboratory data). It is estimated that in the upcoming years, 90% of health data
will be provided in of unstructured format (e.g. medical reports) [7]. Semantic
annotation facilitates automated content processing of unstructured data. This
establishes the data foundation for a holistic analysis of the patient's status or the
processing of complex research questions. For semantic annotation it is necessary
to rely on standardized and commonly used vocabularies, terminologies or
ontologies or coding schemes.
Data sharing (TR): Today a large amount of health-related data is stored in data silos.
The efficient and automated data sharing is hardly possible as it faces multiple
media disruptions. Although first approaches facilitating data interoperability (e.g.
Health Level 75, OpenEHR6, Integrating the Healthcare Enterprise7) are available,
additional work and research is still required. At present, the health data exchange
is mainly based on individualized solutions. This problem could be resolved by
standardized clinical data models, commonly agreed terminologies and coding
systems. Although, different coding systems are available, they are mainly used
in country-specific adaptations (e.g. ICD-10) or lack usability (e.g. SNOMED
Clinical Terms8).

4
http://www.who.int/classifications/icd/en/
5 https://www.hl7.org/
6 http://www.openehr.org/
7 http://www.ihe.net/
8
http://www.ihtsdo.org/snomed-ct/
 S. Zillner et al. / User Needs and Requirements Analysis for Big Data Healthcare Applications 661

3. Discussion

Our investigations showed that big health data applications indicate a high potential for
improving the efficiency and quality of care delivery. However, we could identify only
a limited number of already implemented big data application scenarios. Although non-
advanced healthcare analytics applications - such as analytics for improved accounting
or quality control - are available in a wide-spread manner [1], those do not yet make
use of the potential of big data technologies. To use this potential, it is necessary that
the various dimensions of health data, such as a) the clinical data describing the
patient's history and health status b) the administrative data c) the knowledge about
diseases as well as related population data d) the knowledge about changes in time are
incorporated in the automated health data analysis. Hence, the highest clinical impact
of big data approaches for the healthcare domain can be achieved when aggregating,
comparing and relating data from the four dimensions. Therefore, big data technologies
could help to produce new insights enabling personalized treatments [6].
This leads to one major problem: Health data cannot be easily accessed. High
investment and efforts would be needed. As a consequence, convincing business cases
are difficult to identify as the burden of initial required investments strongly reduces
profit expectations [6]. In other words, one of the biggest challenges for the realization
of big health data applications is the need for high investments, standards and
frameworks as well as new supporting technologies to make health data available for
subsequent big data analytics applications [8]. Technical and non-technical challenges
need to be addressed to foster the seamless access to health data. Within our future
work, we plan to elaborate these requirements in further detail.

Acknowledgment

This research has been supported in part by the Big Data Public Private Forum, a
project that is co-funded by the European Commission within the 7th Framework
Programme under the grant number 318062. The responsibility lies with the authors.

References

[1] Frost and Sullivan. U.S. Hospital Health Data Analytics Market [Internet]. Frost and Sullivan; 2012
[updated 2012 Aug 7; cited 2014 May 6]. Available from: http://tinyurl.com/myb2f3c.
[2] McKinsey and Company. Big data: The next frontier for innovation, competition, and productivity
[Internet] McKinsey and Company; 2011 [updated May 2011; cited 2014 May 6]. Available from:
http://tinyurl.com/cplxu6p.
[3] Kayyali B, Knott D, Van Kuiken S. The ‘big data’ revolution in healthcare [Internet]. McKinsey &
Company; 2013 [updated 2013 April, cited 2014 May 6]. Available from: http://tinyurl.com/pq4dzrn.
[4] Porter M, Teisberg OE. Redefining Health Care: Creating Value-Based Competition on Results.
Boston: Harvard Business Review Press; 2006.
[5] Flick U. Triangulation: Eine Einführung, Wiesbaden: VS Verlag,; 2011.
[6] Zillner S, Rusitschka S, Munne R, Lippell H, Vilela FL, Hussain K, et al. D2.3.1 First Draft of Sector’s
Requisites. Public Deliverable of the EU-Project BIG; 2013.
[7] Lünendonk Gmbh. Lünendonk Trendpapier: Big Data bei Krankenversicherungen: Bewältigung der
Datenmengen in einem veränderten Gesundheitswesen [Internet]. 2013 [cited 2014 May 6]. Available
from: http://tinyurl.com/ph6bwmh.
[8] Lobillo F. et al. D2.4.1.First Draft of Sector’s Roadmap. Public Deliverable of the EU-Project BIG;
2013.