REUSABILITY IN PATIENT REGISTRIES
Implementation of a Generic Extensible Web-based Patient Registry System
Eric Tröger, Elena Prokofyeva, Robert Wilke
Biomedical Engineering Lab, Institute for Ophthalmic Research, Centre for Ophthalmology, University of Tübingen
Paul-Ehrlich-Strasse 17, Tübingen, Germany
Eberhart Zrenner
Institute for Ophthalmic Research, Centre for Ophthalmology, University of Tübingen
Schleichstrasse 12-16, Tübingen, Germany
Keywords: Patient registry, patient database, generic database, Operational Data Model, ODM, XSLT, ophthalmology.
Abstract: Usually patient registry systems are developed in very specific scientific contexts where the reusability is
considered a topic of minor concern. To improve reusability as well as to achieve better separation between
the data definition step and the technical implementation of the patient registry, a generic approach has been
chosen to automatically generate application components based on the data definition. This is done by using
the ODM standard format as meta-model and applying XSL transformations to create and configure system
components. Components required among several patient registries are implemented once and reused.
1 INTRODUCTION
To allow scientific analyses of large amounts of
patient datasets as well as to assemble populations of
patient datasets for recruitment for upcoming
clinical trials it is of great importance to have a
robust and easily accessible patient registry system.
Usually such a system is built focusing on a very
specific scientific purpose - and therefore has limited
usefulness for other applications – or patient data is
gathered from paper files or operative systems on
demand. Problems regarding technological as well
as organisational aspects are usually appearing
frequently among such systems.
In order to enhance reusability, we are pursuing a
generic approach for our patient registry system,
enabling free specifications of data definitions by
using a standardised meta-model. This approach not
only allows to automatically create user interfaces
like Web-based data entry forms, it also enables
interconnectivity among multiple systems generated
by this technique by matching the models used for
the generation-processes
. It also ensures the
reusability of a highly robust technological
framework.
2 CONSIDERATIONS
In medical contexts scientific expertise usually has a
crucial influence on data structure design. Although
data definition is one of the most important steps
while developing a patient registry, it can also
become very time consuming and may unacceptably
delay further development steps.
Especially within larger consortiums where
differing scientific interests are present, it can be
difficult to reach a point of mutual consent. This e.g.
held true for the common ophthalmic patient registry
system developed within the scope of the European
Integrated Project EVI-GENORET. The system
covers about 400 data fields per patient, and while
the overall system has already been serviceable from
a technical point of view, field definition changes
needed to be made even after patient data already
had been entered due to continuously developing
scientific requirements.
As a result our current approach in building
patient registry systems aims at a strict separation of
the data definition process and the technical
implementation of the system. While core
components are taken from a common set of
438
Tröger E., Prokofyeva E., Wilke R. and Zrenner E. (2009).
REUSABILITY IN PATIENT REGISTRIES - Implementation of a Generic Extensible Web-based Patient Registry System.
In Proceedings of the International Conference on Health Informatics, pages 438-441
DOI: 10.5220/0001538504380441
Copyright
c
SciTePress
program modules, data definition specific parts are
mainly generated during compile time.
3 IMPLEMENTATION
3.1 Technological Infrastructure
Since our previously developed patient registry
systems are J2EE Web-applications, the generic
approach also targets this kind of application.
Following common best practices, the framework is
structured as a 3-tier application utilising widely
used libraries for implementing the different tiers.
Hibernate 3 is used as object relational mapping
layer to allow for database abstraction. Struts 2 is
used as Web-application framework realising the
MVC pattern, whereas Freemarker is used as
template engine implementing the View to render
HTML-pages. Furthermore certain already
implemented components from previous systems are
reused, e.g. the auditing component logging user
activity on the database (Section 3.5.2).
3.2 Generic Approach
It is essential to point out that the target system is
clearly defined as patient registry system and
therefore allows for restrictions on what components
are to be generated and to which extent. Our efforts
were not directed towards generating arbitrary
configurable database systems, but rather focused on
this special kind of database system. This also leads
to implications e.g. regarding data security and
protection components.
While our approach cannot directly be defined as
such, it still can be considered as being related to
Model-Driven Software Development (MDSD) in
terms of building a software system based on a
model, whereas the model itself first and foremost
contains data structure definitions.
3.3 Meta-Model
A common meta-model was needed as basis for the
generic platform. After searching for an appropriate
meta-model as well as considering defining a new
meta-model, the CDISC (Clinical Data Interchange
Standards Consortium) Operational Data Model
(ODM) has been chosen.
ODM is a common standard for interchange of
clinical data, especially in clinical trials systems.
ODM is an XML-dialect, which was being
supplemented by Vendor Extensions to allow for the
usage as platform-specific model. The ODM-file not
only contains definitions of data structures, but also
validation rules, user interface configuration and
access control rules that allow to configure database
access rights as will be described in the following
sections
3.4 Compiling an Application
After defining an ODM file a patient registry system
can be built automatically. Since about half of the
target application consists of reusable components
(Section 3.5) this step can also be thought of as
configuring the target system.
3.4.1 Principle
The system is built by Maven 2. The ODM file is
being transformed by multiple XSL files using an
XSLT Maven plug-in, whereas different components
are generated. While some XSL files only produce
one target file, others produce many at once,
depending on the type of the transformation. There
is a difference if only one configuration file is to be
generated, or multiple classes representing domain
entities. Internationalisation is completely supported
throughout all generated and non-generated
components and within the ODM file.
3.4.2 Interfaces for the Datasets
The ODM file is structured into StudyEventDef-
Elements (XML tags), whereas each of them can
reference a FormDef-Element. The StudyEventDef-
Elements are use to model specific groups of
datasets belonging together. In our sample patient
registry the StudyEventDef “Monogentic Retinal
Dystrophy” encapsulates FormDefs representing sub
datasets for this kind of disease like “Clinical
Examination” or “Technical Examination”.
The FormDef-Elements are used to generate Java
interfaces representing the corresponding datasets,
e.g. for “Clinical Examination”. For FormDef-
Elements references to ItemDef-Elements are
modelled, and for each ItemDef an identifier and the
data type are declared. This information is taken to
generate corresponding getter and setter methods for
the Java interfaces representing the datasets.
3.4.3 Object Relational Mapping
To configure the database abstraction layer,
Hibernate 3 specific domain classes are generated
that implement the interfaces described in Section
REUSABILITY IN PATIENT REGISTRIES - Implementation of a Generic Extensible Web-based Patient Registry System
439
3.4.2. Those classes are automatically extended with
Hibernate Annotations containing database mapping
instructions of which most are dependent on the
ODM content.
Furthermore service classes are generated that
allow for database operations as database access
objects (DAOs).
3.4.4 Web-Forms
The target patient registry system is Web-based and
therefore requires HTML-forms to show and enter
data. Since Freemarker is used to generate HTML
output, Freemarker templates (FTL) are generated at
compile time. One FTL represents one dataset
defined as FormDef in the ODM file, and therefore
will be associated with one specific Hibernate
domain object (Section 3.4.3) at runtime.
The ODM file has been extended by Vendor
Extensions to accommodate additional tags
describing a basic user interface layout for each
dataset. This was done as intended by CDISC by
adding custom XML Schemata (XSDs) referenced
by the ODM file. This layout information is not only
used to structure Web-forms, but also for reports
(Section 3.4.7). It enables visual grouping of
elements.
3.4.5 User Input Validation
To facilitate the definition of validation rules within
the ODM file Vendor Extension tags have been
added to allow for data entry restrictions like regular
expressions. Range checks are also possible for
numerical data types.
Apart from manually defined validation rules,
data type dependent rules are added at compile time
to e.g. automatically check against date-format
patterns when data is entered. All validation rules
are converted into an application specific format and
applied at client side (JavaScript), as well as at
server side (Struts 2 Interceptor).
3.4.6 Restrictions on Data Collection
The ODM standard provides Collection-Exception-
Conditions that can be set for each item referenced
within FormDef-Elements. Those conditions were
used to set user input restrictions by defining special
XPath expressions.
An example expression may look like:
../ItemDef[@OID='CLINICALEXAMINATION
.ANTERIORSEGMENT.IRIS.ABNORMAL.OS'][
@Value='OTHER']
Each item referencing this condition as
Collection-Exception-Condition is only to be
collected if the Item with the given ODM identifier
has the value “OTHER”. This ODM compatible
expression is converted into an application-specific
rule at compile time. These rules are not only used to
show/fade-out user input elements on Web-pages
while entering data depending on previously entered
data (JavaScript), but also to validate the data before
saving it in the database (Struts 2 Interceptor).
3.4.7 Reports
For each dataset a printable report can be generated
when using the target application. The reports are
compiled and filled by JasperReports, which needs
JRXML files as templates. At compile time JRXML
files are generated using XSTL, whereas the
structure of the reports is mainly defined by the user
interface layout defined in the ODM file.
After the JRXML files have been generated, they
are compiled by the Maven Jasper plug-in into
Jasper files to enhance performance at runtime.
3.4.8 Option-Lists for Data Fields
Many data field values are restricted to specific
options which can be defined in the ODM file as
CodeLists. The values defined in the CodeLists are
parsed and used at runtime to fill select-boxes on
Web-pages or to validate entered data before saving
it.
Special CodeLists can be marked as persisted
within the database, which is appropriate for large
sets of available options. This was true for long
custom diagnosis lists or ICD10 lists in our sample
application. As additional feature options stored
within the database can be structured as a tree. A
special AJAX-based user interface component has
been developed to handle tree-structured option lists.
3.4.9 Additional Generated Components
The ODM file allows the configuration of the target
patient registry in much more depth than has already
been described.
For example file upload components can be
specified, causing special tables holding binary
content to be generated. Additionally AJAX-based
file upload user interface components are then
integrated into the Web-forms.
Web-based search masks are pre-configured
according to ODM content allowing restrictions for
the searchable values of the AJAX-based query
builder and backend search engine. So if only a
HEALTHINF 2009 - International Conference on Health Informatics
440
predefined range of values was allowed for a
numerical field or a specific set of options is defined
for a field, this will be respected when dynamically
building the user interface.
Further generated components like an additional
importer module to integrate data from external
databases can be compiled on demand. The importer
also includes all validation checks that are
performed when manually entering data. Further
data integrity checks are performed by validating the
data to be imported (XML format) against an
automatically generated XML Schema.
3.5 Reusable Components
Apart from the generated portion of the target
application, many components are shared among all
patient registries. Some of them are being addressed
in the following sections.
3.5.1 Application Framework
The basic infrastructure of the application covering
HTTP-request handling and HTTP-response filling
(Web-pages, reports, binary streams), as well as
database connectivity, logging, HTML-rendering
fragments, search routines or the validation
framework are common for all generated patient
registries.
The framework can be augmented and extended
separately, providing features for all patient registry
systems to be developed in the future.
3.5.2 Data Protection and Security
To provide regulation compliant data protection and
security mechanisms that are required for medical
databases, the authentication system is constructed
as non-generic component. ACEGI is used as
extensible security framework.
Security rules and roles can be defined in the
ODM file, both on dataset and attribute level.
Typical rights include read, write, edit, search,
export and are evaluated throughout different layers
of the application.
An auditing mechanism logging all changes
made to patient datasets on attribute level guarantees
complete traceability. This also covers non-manual
database access like initiated by the importer.
3.5.3 User Management
User data is stored within the database at the
moment, whereas different user management
techniques may be implemented in the future. Users
can register themselves using a registry form.
Afterwards an appointed administrator can approve
or reject pending registrations for his clinical centre.
The user registration is handled via email.
Configurable templates (Freemarker) are used to
personalise the email content.
4 CONCLUSIONS
The generic system was successfully used in
Tuebingen to construct a patient registry named
Ophthabase. Although created for ophthalmology,
the generic concept allows the creation of almost
any kind of patient registry, either as standalone
system or as basis for a distributed patient registry
within a scientific consortium.
The generated patient registry can easily be
modified or extended by adjusting the ODM-file,
which saves much time especially in situations
where the data structure definition is still under
review. The generated system is less error-prone
during system development, since data modelling is
strongly separated from technological concerns.
The generic concept will be extended in the
future to further augment the patient registry
systems, e.g. by adding more sophisticated search
routines, analytic functionality, miscellaneous export
modules, and optional pedigrees.
Multi-centre data interchange will be possible
based on Web Services using a specialised mediator-
server able to match ODM-models of different
patient registries. Patient registry systems will be
able to communicate directly after mediation.
By removing the Vendor Extensions from the
ODM-file it can also be used to configure clinical
trial systems. To amend the definition of the ODM-
file a specialised visual editor may be developed. It
is also planned to provide a tight integration into
isTUmas, the study management system developed
at the University of Tuebingen (Strasser el al.,
2008).
REFERENCES
Clinical Data Interchange Standards Consortium,
http://cdisc.org, accessed 15/07/2008
European Vision Institute – GENORET, http://www.evi-
genoret.org, accessed 15/07/2008
Strasser et al., 2008. An integrated System for Workflow
and Data Management in Clinical Trials, ARVO 2008
Annual Meeting. Fort Lauderdale, April 27, 2008 –
May 01, 2008.
REUSABILITY IN PATIENT REGISTRIES - Implementation of a Generic Extensible Web-based Patient Registry System
441