TRANSFORM: Clinical prediction rules as a basis for decision support
Corrigan D¹, Dimitrov BD¹, Fahey T¹
¹HRB Centre for Primary Care Research, Department of General Practice, Royal College of Surgeons in Ireland, Republic of Ireland.
How will TRANSFoRm work?
Clinical knowledge changes rapidly. Information Systems have huge potential to support
clinical practice by enabling staff to access the latest evidence-based care. This support
can be provided in the form of access to the very latest clinical guidelines or
recommendations relating to a clinical problem. A fundamental challenge in doing this is
how to represent known clinical facts or knowledge in a format that can easily be used
and shared by different computer based systems.
Current paper-based guidelines are not easily translated to a machine interpretable
format that computer-based systems can reason with. In addition these types of
guidelines cannot be implemented and refined by Information Systems themselves
based on the newly available epidemiological data.
There is therefore a requirement for development of more formal definitions of what
constitutes ‘clinical knowledge’. One such representation that could potentially be used
is the clinical prediction rule (CPR). The CPR provides a more rigid and formal structure
with which to reason as part of a Clinical Decision Support System (CDSS). The CPR
definition can be saved in a computer friendly format in the form of:
•A list of diagnostic cues indicative of a clinical outcome
•A list of criteria to be applied to those diagnostic cues indicative of grades of a
particular clinical outcome
•A scoring system to evaluate which criteria apply to a particular clinical case
•A clinical interpretation of diagnostic severity based on the scoring system
•A clinical assessment/decision based on the score interpretation.
TRANSFORM will implemented using an iterative process implemented in 5 main
information system components:
1. A CPR repository of evidence in the form of electronic CPRs which will be
implemented by the HRB Centre for Primary Care Research as part of TRANSFORM
Work Package 4.
2. A dynamic interface that uses this evidence incorporated into electronic health
records (EHRs) for diagnostic decision support for primary care GPs
3. A research study designer to recruit patients from their electronic health records to
generate new clinical evidence for refinement/updating of CPRs
4. Tools to manage research studies and their associated data
5. A data mining tool to extract research data and feed it back to update the CPR
repository of evidence resulting in generation of new or updated CPRs.
Example - The CRB 65 clinical prediction rule for predicting Pneumonia mortality
Respiratory Rate >= 30/min,
BP : SBP<90 mm Hg or DSP <= 60
Age >= 65 years
1 OR 2
CPR Analysis &
5 CPR Data
3 OR 4
Work to Date
Work to date has primarily focussed on the design phase by gathering clinical evidence
from available literature to support three main clinical scenarios; acute chest pain,
chronic abdominal pain and dyspnoea. Diagnostic criteria in the form of diagnostic cues,
risk factors and existing clinical prediction rules have been identified by review of the
available clinical literature and will be used as the basis for the Clinical Prediction Rules
Repository. An initial generic data model has been developed which will describe the
available clinical evidence in a format that can then be used to provide decision support
for the defined clinical diagnostic scenarios.
The TRANSFORM Project
The “Translational Research and Patient Safety in Europe” (TRANSFORM) project is a 5
year EU funded research project that has been running since March 2010. There are 17
participant bodies including the HRB Centre for Primary Care Research working on 19
work packages. The HRB Centre for Primary Care Research is leading work package 4 –
“Decision Rules and Evidence”. The primary aim of TRANSFORM is to use clinical
evidence in the form of CPRs to bridge clinical research and primary care practice.
Future work will focus on building the evidence repository and making it available as a
web-based service from which clinical decision support can be provided in the GP
Electronic Health Record. A focus will be on investigating potential use of currently
available clinical decision support service frameworks to do this.
Conflict of Interest Statement: The authors declare no conflict of interest.
Funding: This research is funded through the EU FP7 TRANSFoRm grant.
Corresponding author: Derek Corrigan, HRB Centre for Primary Care Research, RCSI
Ireland, [email protected]