The
SPREAD CARE is a project founded by the Italian Ministry
of Health and required to provide about 27 Neurological
Departments with a decision support system, based on
“The Italian guidelines for stroke prevention and
management”, delivered by the
SPREAD
association.
Our decision support system is based on workflow
technology. The workflow logic is based on the rules
provided by the SPREAD guidelines (GLs) for stroke
management. The already existing clinical chart has
been transformed into an evidence-based, real-time
decision support system, meanwhile maintaining the same
look the users were familiar with. Since the final aim
of the Project was to improve evidence-based behavior
and detect possible organizational bottlenecks,
non-compliance to the clinical prac-tice guidelines,
before and after the system introduction, have been
analyzed, as well as the accuracy of the clinical chart
compilation, some care process variables, and system
usability. Results show that the system enhances the
clinical practice without boring users.
Evidence-based clinical
practice guidelines have been widely promoted as a
way of improving health outcomes. In current healthcare
systems, however, scientific knowledge about best care
is not applied systematically or expeditiously to
clinical practice. GLs usually capture both
literature-based and practice-based evidence into a
textual format, which can be easily diffused but
uneasily used in daily work. Thus there is a great
effort to disseminate them in computer-interpretable
representations, more suitable for individual clinical
decision support. Contemporary, there is increasing need
of smooth integration of GLs into the existing hospital
information systems.
The current workflow
technology seems to offer a convenient solution to
build a cooperative system in which the activities of a
care providers’ team can be coordinated within a process
properly designed on the basis of available best medical
knowledge. This projectb presents an approach to the
design, implementation and evaluation of an
evidence-based Careflow management system (CfMS).
As borrowed from the Workflow Management Coalition, a
CfMS is a system that defines, create and manages the
execution of careflows (Cfs) through the use of
software, running on one or more Cfs engines, which are
able to interpret the care process definitions, interact
with Cfs participants and invoke the use of ICT tools
and applications. Careflow indicates the automation of a
care process, in whole or in part, during which
information, documents or tasks are passed from one
participant to another for action, according to a
process definition. Cfs are case-based, i.e.,
every piece of work is executed for a specific patient.
One can think of a patient care process as a Cf
instance. A Cf process definition specifies which
medical tasks needs to be executed and in what order. A
CfMS may also contribure to solve the communication
problem within the health care organizations since it is
able to manage automatically a great amount of
communication acts among organizational agents involved
in patient care.
The Neurological
Departments, involved in the Project, were already
equipped with a Computerized Clinical Chart (CCC)
implemented with the commercial tool WINCARE®
(by
TSD Projects). To add the
decision-support funcionalities reengineering of both
structure and interface of the electronic patient record
(EPR) was performed. A deep analysis of the SPREAD GL,
the existing EPR, and its interface, allowed to
determine the minimum data set required for implementing
all the GL reccomendations, to check where these data
were or not already managed by WINCARE®,
and whether they were in the oppurtune format. Thus, the
data model has been updated both by increasing
information and by changing the nature of the existing
information, mainly shifting from free text to encoding.
To meet physicans' needs to quickly produce printed
reports in natural language we developed a module for
Generation of Inferable Free Text (GIFT). The
idea was to create a generic module able to produce
textual reports starting from a set of encoded data in
whichever WINCARE® form.
After the data model analysis, and its consequent update,
we implemented the GL recommendations through CfMS using
Oracle Workflow™. The
Cf model
is described on the basis of the SPREAD GL, with some
site-specifications decided by the involved neurologists.
All the GL recommendations were implemented, regardless
of their scientific evidence level (SPREAD uses four
levels, from A to D). The CfMS needs patient data in
order to feed the workflow engine and to interpret GL
rules, then it must be able to communicate the
patient-specific recommendations to the users, or Cf
participants, through messages and to-do-lists. Our
choice was to integrate all the needed functionalities
within the existing end-user application, making it more
“dynamic”, according to the CfMS execution, without
creating a new specific interface. A
middleware layer
has been developed to keep the systems independent,
while granting communication. It is composed by a
supporting database (Oracle DB) and an Interpreter
written in PL/SQL.
In
implementing and integrating the decision support system
attention has been put in maintaining the end-user
interface as much as possible unchanged, so that users
perceive the new system just as an update of the
clinical chart, with some new functionalities, that are:
-
the sections of the clinical chart waiting for new
data are listed runtime in an “intelligent”,
patient-specific, dynamic way, i.e. accounting
for the most useful data for that patient in that
moment ;
-
data relevant for interpretation of the guideline
rules appear as
yellow-coloured fields, in order to facilitate a
complete record filling ;
-
according to the urgency of the recommendation, the
related message is shown directly on the screen or
made accessible through a
communication box;
-
at the patient discharge a module, called
RoMA (Reasoning
on Medical Actions) is actived and the list of
non-compliances is shown and physicians may provide
motivations.
The
test-bed for the proposed methodologica and
techonological solutions is the Stroke Unit in Pavia.
The decision support system has been installed in April
2006, but the data model of the EPR is the same from
January 1st 2005. Since that time to mid January 2007,
about 400 ischemic stroke patients have been admitted to
the SU. In a recent work we compared data collected in
the two periods April /December 2005 and April/December
2006 in order to check the impact of the CfMS on the
clinical routine. Preliminary results are encouraging:
the system has been accepted by healthcare personnel and
entered the daily practice; completeness of encoded
data input is increasing for the most part of the data
forms useful for guideline interpretation and also
compliance is in general improved.
