Lecture Introduction: Dr. Lindsay Knaik

• Role and Affiliation: Visiting from the University of Iowa
  • Professor of Metrics
  • Associate CMIO at Stead Family Children's Hospital
• Background
  • Biomedical engineering and medical school in Iowa
  • Residency at Baylor in pediatrics
  • Fellowship in neonatology at Vanderbilt
  • Informatics background

Overview of Epic Cosmos

• What is Cosmos?
  • Research-related sharing network with Epic
  • Contains de-identified clinical EHR data from Epic institutions
  • 203 million patients, billions of encounters
• Benefits
  • Large database of clinical information
  • Fast query abilities, similar demographic spin to U.S. Census
  • Contains longitudinal patient charts and diverse datasets

Using Epic Cosmos

• Access Requirements
  • Belong to Epic Institution
  • Epic classes needed for access
• Data Available
  • Slicer dicer version: limited aggregate data
  • Data science side: line level detailed data

Example Study: Neonatal Hypertension

• Research Setup
  • Started with slicer dicer for cohort creation
  • Validation using clinical knowledge
• Findings
  • 1.7% diagnosis rate, consistent with literature
  • Medication patterns observed: ACE inhibitors, calcium channel blockers
  • Showed the power of Cosmos for large-scale analysis

Challenges and Feedback

• Data Quality and Limitations
  • Examples of mapping issues in care everywhere
  • Importance of validating datasets
• Sidekick Tool
  • Large language model in slicer dicer
  • Tested to improve query building
  • Needs clinical validation for accuracy

Cosmos Collaboration and Research

• Collaborations
  • Early adopters provide feedback
  • Multidisciplinary teams for validated cohorts
• Epic Research
  • Epic publishes findings to showcase Cosmos
  • Caution in definitions used by Epic teams

Case Study: Tiny Baby Program

• Objective: Analyze outcomes for extremely preterm infants
• Methodology
  • Comparison with historical data
  • Study of survival rates, medication use
• Results
  • Survival improvement in 22-weekers
  • Validated use of Cosmos for neonatal health trends

Future Directions

• Potential Improvements
  • Enhanced data quality dashboards
  • More structured note data
• Access and Training
  • Legal and mapping issues
  • Training available online and virtually

Conclusion

• Cosmos is a powerful, evolving tool
• Important for large-scale and multi-center research
• Continued feedback necessary to improve data quality and usability
• Encouragement for institutions to join and contribute to Cosmos