Background: EHR systems are widely used in hospitals and primary care centres but it is usually

difficult to share information and to collect patient data for clinical research. This is partly due to

the different proprietary information models and inconsistent data quality. Our objective was to

provide a more flexible solution enabling the clinicians to define which data to be recorded and

shared for both routine documentation and clinical studies. The data should be possible to reuse

through a common set of variable definitions providing a consistent nomenclature and validation of

data. Another objective was that the templates used for the data entry and presentation should be

possible to use in combination with the existing EHR systems.

Methods: We have designed and developed a template based system (called Julius) that was

integrated with existing EHR systems. The system is driven by the medical domain knowledge

defined by clinicians in the form of templates and variable definitions stored in a common data

repository. The system architecture consists of three layers. The presentation layer is purely webbased,

which facilitates integration with existing EHR products. The domain layer consists of the

template design system, a variable/clinical concept definition system, the transformation and

validation logic all implemented in Java. The data source layer utilizes an object relational mapping

tool and a relational database.

Results: The Julius system has been implemented, tested and deployed to three health care units

in Stockholm, Sweden. The initial responses from the pilot users were positive. The template

system facilitates patient data collection in many ways. The experience of using the template system

suggests that enabling the clinicians to be in control of the system, is a good way to add

supplementary functionality to the present EHR systems.

Conclusion: The approach of the template system in combination with various local EHR systems

can facilitate the sharing and reuse of validated clinical information from different health care units.

However, future system developments for these purposes should consider using the openEHR/

CEN models with shareable archetypes.