=Paper=
{{Paper
|id=Vol-2336/MMHS2018_paper_7
|storemode=property
|title=
Development of an E-mental Health Infrastructure for Supporting Interoperability and
Data Analysis
|pdfUrl=https://ceur-ws.org/Vol-2336/MMHS2018_paper_7.pdf
|volume=Vol-2336
|authors=Fazle Rabbi,Yngve Lamo
}}
==
Development of an E-mental Health Infrastructure for Supporting Interoperability and
Data Analysis==
https://ceur-ws.org/Vol-2336/MMHS2018_paper_7.pdf
Development of an E-mental Health
Infrastructure for Supporting Interoperability
and Data Analysis
Fazle Rabbi, Yngve Lamo
Western Norway University of Applied Sciences, Bergen, Norway
{Fazle.Rabbi@hvl.no, Yngve.Lamo@hvl.no}
Abstract. Digital technology plays an increasingly important role in
addressing the challenges faced by health and care services such as ris-
ing costs, changing demographics, shortage of healthcare professionals.
eHealth is the use of information and communication technologies (ICT)
for healthcare systems which helps patients and healthcare providers
work together to ensure faster, safer and better care. eHealth strengthen
the use of ICT in health development through a range of services or
systems including electronic health record, clinical decision support sys-
tem, health informatics, self-monitoring healthcare devices, personalized
medicine. This paper presents an eHealth infrastructure for E-mental
health which is under development. The infrastructure is being designed
to provide internet based interventions and support for interoperability
and data analysis.
Keywords: healthcare systems, internet of things, process mining, ma-
chine learning, HL7 FHIR
1 Introduction
Today’s vast amount of medical data need to be integrated and accessed intel-
ligently to support better healthcare delivery. Interoperability in healthcare can
bring together partners and facilitate knowledge sharing which can potentially
create new networks of knowledge. Delivering context relevant clinical informa-
tion enables decision making through healthcare data analysis. By measuring
and monitoring processes digitally, we can compare data more easily. Such in-
sight facilitates streamlined workflows, greater efficiency and improved patient
care. Systematic analysis of healthcare data can help to detect patterns so that
healthcare providers can optimize their resource allocation and clinicians can
conduct treatments to individuals and project better health outcomes. Accord-
ing to WHO a mental health information system should enable managers and
service providers to make well-informed decisions that improve the quality of
care [7]. To improve the effectiveness and efficiency of mental health services,
mental health information systems go through the following essential stages:
– Collection : Data collection from source
2 Rabbi and Lamo
– Processing : movement of data from the source
– Analysis : examination and study of the raw data
– Dissemination : communication of the results of the analysis
– Use : utilization of the results of the analysis for service improvement, plan-
ning, development and evaluation.
In this paper, we give an overview of an E-mental health infrastructure that
facilitates the development of mental health information systems. In many de-
veloped countries, majority of their citizens use public healthcare services. To
support a variety of healthcare service providers, these healthcare systems of-
ten use a large number of software applications. For collecting and processing
healthcare data from various sources we require healthcare interoperability. The
term ‘interoperability’ refers to the ability of different information systems to
exchange information and understand the semantic of information. Healthcare
interoperability is very essential to reduce the processing time that is required
for the conversions of healthcare information originated by different healthcare
providers and/or systems. It is considered as one of the biggest challenge in
today’s healthcare systems due to the fact that healthcare systems are inher-
ently complex and there are many players involved in the healthcare industry.
There has been a lot of initiatives to address healthcare interoperability over
the past decade. HL7 [8] is an international community who is involved in the
development of a set of international standards, guidelines and methodologies
for sharing healthcare information among healthcare providers. These standards
are referred to as HL7 standards. HL7 FHIR (Fast healthcare Interoperability
Resources) [6] is the latest standard developed by HL7 for exchanging healthcare
information with a main focus on implementation. It provides a number of re-
source types which are the building blocks for exchanging healthcare data. Any
healthcare information that needs to be exchanged among organizations should
be specified as FHIR resources. The FHIR standard is suitable to integrate
healthcare applications across organizations, medical devices, and also mobile
healthcare applications. FHIR resources utilize standard terminologies from on-
tologies which provide semantic interoperability. Currently we are developing
an interoperability healthcare platform based on HL7 FHIR in collaboration
with Helse Vest ICT [2], a large IT service provider in Western Norway. The
infrastructure development is partly supported by the ‘Intromat’ project [3].
There is a great need for doing research in health service improvement to
provide the best care possible to the patients with limited resources. Research
related to health service improvement needs to undertake many complex tasks
such as root cause analysis, capture information from previous steps into a simple
document and study the variability of a large number of patient population. It
is challenging to accomplish these tasks and therefore we require techniques and
tool support. For the examination and study of healthcare information we use
data mining techniques. Data mining techniques provide deeper insights into
patients health by analyzing historical healthcare information including patients
diet, appointments, exercise, lab results, vital signs, prescriptions, treatments,
allergies, etc. Data mining techniques such as process mining in healthcare brings
Development of a Software Infrastructure for E-mental Health 3
the opportunity to learn from patients healthcare information including children,
women, elderly, patients with co-morbidity and the results can be utilized for
optimizing healthcare resources and the improvement of health service delivery.
We need to disseminate the analysis results to a diverse group of people in the
healthcare system. Healthcare managers, analysts and clinicians need to visual-
ize healthcare processes across disciplines to investigate the common pathways
of patients. Identifying common pathways for patients flow in healthcare systems
is complex as we need to deal with a variety of patients group. While analyzing
common pathways for patients, different context need to be setup to focus on dif-
ferent group of patients and visualize their careflows. For instance, the manager
of the pediatrics department in a hospital would be interested to look at the flow
of patients’ admission at the children clinics and other departments to make a
better planning of resource distribution; a clinician would be interested to inves-
tigate the common pathways of patients with mental and behavioral disorders
to extract knowledge about concurrent common mental disorders; an analyst
may be interested to investigate the efficiency of a new planning strategy. Re-
sources in healthcare may include time, money, facilities, equipment, people and
competences. Proper resource planning needs to be carried out in healthcare to
ensure that healthcare providers are not overloaded with work, patients are not
waiting too long to get services, and the overall cost of healthcare is optimized.
In our approach, we disseminate the results of careflow analysis to healthcare
professionals for service quality improvement.
2 System architecture
There are a variety of health service providers in a healthcare system and the
healthcare data are often siloed away from other data. Data silos is not only the
problem in a healthcare setting but different standards are being followed by
different health facilities to code diagnosis, lab test results, medical procedures
and drugs. Variety of data models of different service providers are making data
analysis challenging. To address this issue, it is therefore important to create
an infrastructure for ad-hoc exploration of large collections of data. Such an in-
frastructure needs to be flexible and scalable yet supporting suitable format for
decision making. We envision a healthcare information system that provides ac-
cess to information from various healthcare providers as well as patients personal
healthcare devices and/or applications. Availability of information from patients
personal healthcare devices can potentially be used to detect complicated prob-
lems correctly in their early stage and monitor the effects of treatment. For
instance, bipolar disorder can be difficult to diagnose and according to a study
published in Psychiatry, around 69 percent of bipolar disorder cases are misdi-
agnosed [13]. Analyzing patients personal healthcare information can be used
to identify periods of mania and depression. Correct identification of depressive
and one manic or hypomanic episodes are important factors for the diagnosis of
bipolar disorder. However, the integration of personal healthcare devices in the
mainstream treatment process will require the use of healthcare standards.
4 Rabbi and Lamo
We chose to use HL7 FHIR as it allows us to integrate healthcare infor-
mation collected from several sources. Although HL7 FHIR provides a suitable
way to harmonize healthcare information, it does not provide any sophisticated
visualization technique to get an overview of patients health or administrative
information. In our approach, we apply process mining techniques for extracting
an overall picture of healthcare information from various contextual view and
from different level of abstraction and utilize machine learning techniques to
constantly monitor patients condition and raise alarms.
Process mining
output
devices and applications
personal healthcare
Data from patients
Processed
Data from healthcare event logs
service providers
Mobile and VR apps
Machine learning
FHIR
FHIR Server Database Careflow
Wellness and fitness
Identity and
Analyzer
devices
Access
Management
FHIR Interceptor
Remote monitoring
devices
Fig. 1. A system architecture for E-mental health
Figure 1 illustrates an E-mental health system architecture where a FHIR
database is used to store the data captured from healthcare service providers as
well as patients personal healthcare devices and applications. A careflow analy-
sis tool is used to perform data analysis including process mining and machine
learning. The careflow analyzer prepares the event logs by querying the FHIR
database. Security and privacy are major concerns for healthcare systems. Dif-
ferent types of users may be involved in the process mining related work. The
users must have proper authorization to access patients healthcare information.
It might be possible to give partial access to the healthcare information stored
in the FHIR database. A FHIR interceptor is incorporated in the system archi-
tecture to handle users authorization to access FHIR resources. The interceptor
will intercept all the FHIR queries and consults with an identity and access
management module and returns data that the user is authorized to access.
We have developed applications to provide digital interventions for some
clinical cases such as managing depression, monitoring bipolar patients status,
treatment for social anxiety disorder [3]. Currently patients need to sign up
to become a part of a clinical study program and they use their BankID for
authentication which is a personal electronic ID used to identify and sign online.
BankID is a Public Key Infrastructure (PKI) solution offered by Finance Norway.
Development of a Software Infrastructure for E-mental Health 5
The solution supports both authentication and signing. Our future plan is to
incorporate blockchain technology using hyperledger fabric [5] where the patients
would be able to receive mental healthcare services anonymously. Public stigma
is a barrier to mental healthcare and many people either do not seek treatment
or dropout from mental health treatment. We look forward to a solution where
patients identity will be hidden but they will be able to get help from a support
group of professionals. The patient will own the data and they will be able to
decide if they want to share their records with their therapist. We will study the
applicability of blockchain technology in developing E-mental health solutions
by exploring the potential of using blockchain technology to incorporate security,
privacy and integrity of medical records.
3 Data analysis
Typically, a process model describes the activities needed to be performed within
a given process by different actors such as physicians, nurses, and lab technicians.
Therefore, mining a process would in general extract a process model represent-
ing the activities being performed in a healthcare system. The primary focus of
process mining in healthcare is to provide evidence-based process analysis tech-
niques for effective process management [11]. It is used to discover trends and
patterns of process executions by analyzing the trace of activities (a.k.a event
logs) performed in a system. Due to the multidisciplinary nature of healthcare,
the event logs need to be harmonized before they can be processed. Getting the
right setup for data preparation is important to get the best understanding out
of the data as effectively as possible [12]. In a healthcare setting, the data prepa-
ration task is complex due to the involvement of various healthcare systems and
variations of data formats. We propose to employ data warehouse techniques to
pre-process vast amount of information.
Existing process mining tools or techniques [1, 14, 15] have limited support to
provide abstraction from different perspective, and healthcare analysts currently
need to perform a lot of manual investigation to find out the pattern of patients
treatment flow. Given a large number of patients records, this is not an efficient
process as they need to change context from one patient group to another and,
need to look into the data for a specific time range. To overcome this limita-
tion, we propose to develop a diagrammatic approach that will allow analyst to
specify process mining requirements diagrammatically such as the context and
abstraction level.
3.1 Role of ontologies in process mining
Ontologies are often used to standardize terminologies in healthcare. For exam-
ple, the ICD-10 (International Classification of Diseases) ontology is designed
to provide diagnostic codes for classifying diseases, including wide variety of
signs, symptoms, abnormal findings, etc. The SNOMED CT [4] ontology pro-
vides a comprehensive terminology for clinical health. It has been well accepted
6 Rabbi and Lamo
HL7 FHIR resources
Paediatrics Orthopedic
EpisodeOfCare Observation
Gynocology
Dermatology
Psychiatry
Episode
Of Care Procedure CarePlan
Clinical
Care Plan finding
Clinical findings
Healthcare
regarding mental and
behavioral disorders
Condition
Procedure
Observation
Fig. 2. Use of dimensional model for specifying process mining requirements
by healthcare professionals worldwide and its use has improved the quality of
medical health records by providing consistency in using medical terms. Ontolo-
gies can be used to define suitable level of abstraction for selecting a particular
patient group and for visualizing care-flow from a high level of abstraction. We
intend to provide a customizable framework where domain ontologies such as
ontologies for care-plans, symptoms can be easily constructed and attached to
the data source.
3.2 Dimensional modeling
The concept of dimensional modeling originated from data warehousing and busi-
ness intelligence (DW/BI) [9]. Dimensional modeling has emerged as the leading
architecture for building integrated DW/BI systems. Dimensional models pack-
age the data in a format that allows simplicity for displaying understandable in-
formation to business users and also supports developing efficient data analytic
tools in terms of query performance. We propose to use dimensional modeling
for specifying process mining requirements. The dimensional models are used
for both filtering and selecting the level of abstraction for visualizing the process
mining output. For instance, an analyst may be interested to investigate the
admission flow of patients who have issues related with mental and behavioral
disorders. He needs to know which other departments the patients also visit. He
does not need to know the details about the clinics in the departments where
Development of a Software Infrastructure for E-mental Health 7
the patients visit. We assume that the department hierarchy of the hospital is
used for the ‘Admission’ dimension. We illustrate the situation in Figure 2 to
visualize how the dimensional model and the hierarchical representation of data
can be utilized to specify such requirements. The purpose of this dimensional
model is to provide an easy to use visualization for its user to investigate careflow
from different context. We have used ontological hierarchies to provide hierar-
chical representation of healthcare information along each dimensional model.
In this figure, ‘F00-F99’ is the ICD-10 code for ‘mental and behavioral disor-
der’ diseases. Selecting ‘F00-F99’ for filtering essentially means to filter based
on the sub-diseases under ‘F00-F99’ which are depicted as small orange circles
in the figure. Performing this filter over the FHIR resources extracts the ‘con-
dition’ FHIR resources where patients condition has been identified as one of
the sub-disease code of ‘F00-F99’. We use this filtered patients identifications to
extract their admission resources. Patients admission resources contain informa-
tion about patients visit to different clinics. Since we need to display patients
admission to the departments, we use departments hierarchical information to
manipulate the results displaying department names instead of clinics name.
This example illustrates the department hierarchy of the Haukeland University
Hospital for the ‘Admission’ dimension. We use this filtered patients’ identifica-
tions to extract their admission resources. Patients admission resources contain
information about patients visit to different clinics.
4 Conclusion and Future work
The establishment of any large IT infrastructure for healthcare on a regional,
national or international level governs by political influences. The Norwegian
Center for E-health Research recently published a report [10] on reviewing the
focus on machine learning, natural language processing, data mining and process
mining methods: their usefulness, use cases, tools and relatedness to Norwegian
healthcare. The report emphasized on doing more research in machine learning,
data and process mining and natural language processing. Our effort on devel-
oping a software infrastructure for E-mental health is aligned with the focus and
interest published by the report. Interoperability and healthcare analytics are
two major topics in healthcare related research. In this paper, we proposed an IT
infrastructure for E-mental health to achieve interoperability and data analysis
with cutting-edge technologies.
Acknowledgement
This work is partially supported by The Research Council of Norway as a part of
the INTROducing Mental health through Adaptive Technology (INTROMAT)
project under grant agreement 259293.
References
1. Fluxicon Disco. https://fluxicon.com/disco/.
8 Rabbi and Lamo
2. Helse Vest IKT. https://helse-vest-ikt.no/seksjon-engelsk.
3. INTROMAT: INTROducing Mental health through Adaptive Technology. www.
intromat.no.
4. SNOMED CT. www.snomed.org/snomed-ct.
5. E. Androulaki, A. Barger, V. Bortnikov, C. Cachin, K. Christidis, A. De Caro,
D. Enyeart, C. Ferris, G. Laventman, Y. Manevich, S. Muralidharan, C. Murthy,
B. Nguyen, M. Sethi, G. Singh, K. Smith, A. Sorniotti, C. Stathakopoulou,
M. Vukolić, S. W. Cocco, and J. Yellick. Hyperledger fabric: A distributed operat-
ing system for permissioned blockchains. In Proceedings of the Thirteenth EuroSys
Conference, EuroSys ’18, pages 30:1–30:15, New York, NY, USA, 2018. ACM.
6. FHIR. Fast healthcare interoperability resources. www.hl7.org/fhir/.
7. M. Funk and W. H. Organization. Mental health information systems / World
Health Organization. World Health Organization Geneva, 2005.
8. HL7. Health level seven international. www.hl7.org.
9. R. Kimball and M. Ross. The Data Warehouse Toolkit: The Definitive Guide to
Dimensional Modeling. Wiley Publishing, 3rd edition, 2013.
10. A. Makhlysheva, A. Budrionis, T. Chomutare, A. T. Nordsletta, P. A. Bakkevoll,
T. D. Henriksen, J. S. Hurley, J. G. Bellika, H. Blixgrd, F. Godtliebsen, S. O.
Skrvseth, T. Solvoll, and L. H. Linstad. Health analytics. Norwegian Center for
E-health Research, 2018.
11. A. Partington, M. Wynn, S. Suriadi, C. Ouyang, and J. Karnon. Process mining
for clinical processes: A comparative analysis of four australian hospitals. ACM
Trans. Manage. Inf. Syst., 5(4):19:1–19:18, Jan. 2015.
12. D. Pyle. Data Preparation for Data Mining. Morgan Kaufmann Publishers Inc.,
San Francisco, CA, USA, 1st edition, 1999.
13. T. Singh and M. Rajput. Misdiagnosis of bipolar disorder. In Psychiatry (Edg-
mont), volume 3, pages 57–63. 2006.
14. W. M. P. van der Aalst. Process Mining: Discovery, Conformance and Enhance-
ment of Business Processes. Springer Publishing Company, Incorporated, 1st edi-
tion, 2011.
15. B. F. van Dongen, A. K. A. de Medeiros, H. M. W. Verbeek, A. J. M. M. Weijters,
and W. M. P. van der Aalst. The prom framework: A new era in process mining
tool support. In Proceedings of the 26th International Conference on Applications
and Theory of Petri Nets, ICATPN’05, pages 444–454, Berlin, Heidelberg, 2005.
Springer-Verlag.