Providing Likeable and Understandable Guidelines using GRADE in the EHR with Direct links to INdividual patient data (PLUGGED-IN)
PLUGGED-IN was a part of the research program Evicare, supported by the Norwegian Research Council. Main contributing partners from Evicare are Sykehuset Innlandet and DIPS ASA. Linn Brandt is the phd-candidate in this project, Per Olav Vandvik is the main Supervisor. Evicare ended in 2014, but the PLUGGED-IN research continued and through implementation of its strategies into MAGICapp (the MAGIC authoring and publication platform for GRADE guidelines), and real integration with the DIPS ASAs new platform Arena, we will further pursue guideline-integration into the EHR
What are the the issues with current CDSS strategies, and what does PLUGGED IN try to solve
- All advice is not equal
Traditional Clinical decision support systems (CDSS) in Electronic Medical Records (EMR) provide clinicians with recommendations based on algorithms/rules using patient-specific information as inclusion/exclusion criteria. Improved systems (GRADE) for developinge evidence-based guidelines result in the majority of recommendations being weak and warrant balanced clinical judgments without clear inclusion/exclusion criteria, while traditional decision support systems are still typically developing manual rules and algorithms for when to show recommendations based on the assumption that “all advice are equal”.
Solution: If CDSS where to use structured GRADE recommendations, they would get the strength as an inherent part of the recommendation content.
- CDSS is time consuming to create and maintain
CDSS is based on guidelines, but to create CDSS from unstructured text is both difficult and time consuming. You have to code the different parts you are using from the guidelines, and cut them into manageable parts. When made, the content needs to be updated. If the content was taken from a guideline, the CDSS creators need to know when the guideline is updated.
Solution: If CDSS where to use structured GRADE recommendations they could be used directly for decision support, bypassing the need for recoding. There will also be possible to track when the original recommendation has been updated. The ontology-coded information in the structured recommendations can be reused to create CDSS.
- Algorithms have their place, but they do not always work well
Recommendations are given for populations, and many times that population has no clear inclusion/exclusion criteria. The recommendations will cover as specific populations as the evidence lets it do. Guideline authors should not specify populations more specific than their evidence cover. Implementors however, might want to be more specific as a part of a policy decision, and add inclusion/exclusion criteria for when advice should trigger or show to clinicians. These policy decisions might differ from organization to organization, although the recommendation and its evidence stays the same.
Solution: Not all recommendations fit well in a algorithm of strict inclusion and exclusion criteria, so we suggest using a supplementary approach where the recommendations is contextualized by patient specific information. See more.
Implementors that would like to create algorithms, will get easy access to more details with structured multilayered recommendations, as they can dig down into the level of detail they want, all the way to the primary studies.