April 1, 2016: Leveraging EHR Data to Evaluate Sepsis Guidelines

April 1, 2016: Leveraging EHR Data to Evaluate Sepsis Guidelines


Leveraging EHR Data to Evaluate Sepsis Guidelines


Bonnie Westra, PhD, RN, FAAN, FACMI, Associate Professor and Director of the Center for Nursing Informatics at the University of Minnesota


Electronic health records; EHR; Evidence-based guidelines; Flowsheet data; Common data model; CDM; i2b2

Key Points

For EHR data to be useful for research, it needs a common data model, standardized coding, and standardized queries to compare data across healthcare systems. PCORnet is an example of a common data model whose domains include condition, demographic, enrollment, encounters, patient-reported outcomes, and more.

EHR data are documented in flowsheets containing templates, groups of measures, observations, value sets, and data points. But there are challenges with using data flowsheets, including the sheer volume of data; the design of data flowsheets, which varies within an EHR system; and multiple measures for the same concept.

  • In most EHR systems, there is no overarching information model that harmonizes the different ways that data are captured in flowsheets. Developing an information model requires many steps:
  • Identify the clinical topic important to researchers/operations
  • Develop a list of concepts from research questions, clinical guidelines, and literature
  • Search for concepts in templates/groups/measures
  • Search associated groups for additional concepts
  • Add matched concepts to a running list
  • Categorize into assessment and interventions
  • Organize into a hierarchy
  • Combine similar concepts that have similar value sets
  • Validate via a second researcher

Many CTSAs use i2b2 (Informatics for Integrating Biology and the Bedside), a self-service cohort discovery tool that allows researchers to determine if there is a cohort of patients in a clinical data repository that meets their criteria of interest.

In a recent study from researchers at the University of Minnesota, deidentified EHR data was mapped to evidence-based sepsis guidelines in order to estimate compliance and effectiveness of individual recommendations for the prevention of in-hospital mortality and sepsis-related complications.

Discussion Themes

How scalable is the approach used to map flowsheet data to a more structured model?

When using EHR data in research, what about unmeasured characteristics that can’t be controlled for?

You need a multidisciplinary team when developing information models: domain experts, computer scientists, informaticists, healthcare system data experts, and researchers. It's important to work closely with the IT group when doing data extraction because of the large volume of data in a clinical repository.

For More Information

Read more about i2b2 here: https://www.i2b2.org/.

#ElectronicHealthRecords; #CommonDataModel; #i2b2; #pctGR
@PCTGrandRounds, @Collaboratory1, @PCORnetwork