Data & Technology Infrastructure

Learning enabled by data & technology – A learning health system approach requires robust data and technology to implement, disseminate, and scale evidence-based discovery and innovation

We promote FAIR principles for technology and data management and stewardship, which seek to improve the Findability, Accessibility, Interoperability, and Reuse of digital assets. Our projects utilize approaches that are transparent, replicable, and able to provide computational and technology support for real-world data and problem-solving to improve health. We support projects across the Center for Learning Health System Sciences’ units and provide infrastructure that benefits researchers broadly in collaboration with our partners. This includes a collaborative data platform, customer-centric design capabilities, a multi-institutional federated learning partnership, and industry partnerships supporting interoperable clinical decision support (CDS). 

LHS Data Platform

Our LHS data platform includes real world data to support research and operations. Data elements are defined for reuse and clinical insights.  We collaborate with the Clinical and Translational Science Institute, the Center for Clinical Quality & Outcomes Discovery and Evaluation, the Department of Surgery, investigators from the Academic Investment Clinical Program (AICP), and Fairview Health Services. Key aspects of the data platform include:

  • Compliant data infrastructure leveraging the Academic Health Center-Information Exchange Infrastructure. A collaboration with the Center for Clinical Quality & Outcomes Discovery and Evaluation to enable access.

  • Datamarts focused on surgical and acute care.

  • Novel data sources including rich text format clinical EHR notes, logfile and other EHR metadata

  • Interoperable data management, including data mappings to the Observational Medical Outcomes Partnership (OMOP) common data model. We engage with Standards Development Organizations (SDOs) to improve data elements for reuse. 

  • In collaboration with Fairview Health Services, supporting data technology which includes a FHIR (Fast Healthcare Interoperability Resources) façade that consumes data in heterogenous formats and provides standardized FHIR-ized data outputs. FHIR façade capabilities enable downstream solutions such as application solutions, predictive analytics, and CDS that are lightweight and interoperable.

Federated Machine Learning Collaborative

Federated Machine Learning approaches allow for artificial intelligence (AI) / machine learning (ML) models to be trained and optimized on multiple cohorts across organizations without directly sharing sensitive data including protected health information (PHI) which is often part of clinical and other real-world data. 

Together with M Health Fairview and other partners, CLHSS leads a Federated Computer Vision in Healthcare Collaborative with other partner institutions. In addition to Computer vision, other AI/ML use cases utilize federated learning including use cases leveraging Natural Language Processing (NLP). The Collaborative includes experts with a wide range of expertise across the University of Minnesota including the Medical School, School of Public Health, Institute for Health Informatics, and the College of Science and Engineering. 

Current federated learning partners include Emory University, University of Florida, Indiana University, University of North Carolina, and University of Virginia. Current use cases include bowel injury, ultra-massive transfusion, pediatric surgery, entropy in clinical notes, surgical postoperative risk, and rib fracture. In addition, UMN also serves in a leadership role advancing NLP methods for federated learning through the N3C (National COVID Cohort Collaborative).

Interoperable Clinical Decision Support

With support from the Agency for Healthcare Research and Quality, the SCALED (Scaling AcceptabLE cDs) Venous Thomboembolism Prevention project features a multi-site collaboration with University of California Davis, Geisinger Health, Johns Hopkins, and Regenestrief Institute/Indiana University and a 3rd party industry partner, Stanson Health, building interoperable CDS solutions. In addition to standard FHIR resources, we are able to improve and support largescale rapid integration of data through a FHIR façade to provide optimal evidence-based care.  

Customer-Centric Design of Digital Solutions

In addition to expertise of the CLHSS Digital Technology Innovation Program in health IT usability, design, and workflow, CLHSS collaborates on digital solutions with the M Health Fairview GPHR Customer Design studio. Based in Fairview Information and Digital Services, GPHR builds innovative digital solutions for M Health Fairview to engage patients and care teams in their health.