Lecture Notes: Introduction to Epic Cosmos and its Applications

Speaker: Dr. Lindsay Knaik

• Background
  • Visiting from University of Iowa
  • Professor of Metrics and Associate CMIO at Stead Family Children's Hospital
  • Education: Biomedical Engineering, Medical School in Iowa, Residency at Baylor, Fellowship at Vanderbilt
  • Focus: Neonatology and Informatics

Introduction to Cosmos

• What is Cosmos?

  • A large, de-identified clinical EHR database
  • Over 203 million patients and billions of encounters
  • Research-related sharing network
  • Fast querying capabilities

• Benefits of Cosmos

  • Largest database of clinical EHR information
  • Diverse data set: Labs, Meds, Social Determinants of Health
  • Longitudinal charts of patients
  • Similar demographic spin to U.S. Census

Accessing Cosmos

1. Belong to an Epic Institution
2. Legal approval within the institution
3. Epic training classes for users

• Cosmos Dashboard
  • Virtual machine access
  • Tools: Slicer Dicer, Data Science, R, Python
  • Data dictionary and publication checklist

Examples of Cosmos Use Cases

• Neonatal Hypertension Study

  • Slicer-dicer model to study neonatal hypertension
  • Importance of validating cohorts
  • Challenges in obtaining accurate data mappings

• Tiny Baby Program

  • Replication of neonatal research database results
  • Comparison of survival rates in Cosmos vs. curated databases
  • Medication usage analysis

Challenges and Considerations

• Data Quality

  • Issues with mapping and ICD coding
  • Need for valid and reliable data

• Research and Publication

  • Epic research team's publications
  • Collaboration and competition with researchers

• Legal and Ethical Considerations

  • De-identification and data protection
  • Epic's rules against institution identification

Future of Cosmos

• Continued Development

  • Improved data inclusion (e.g., state-level data)
  • Ongoing feedback and enhancements

• Large Language Models (Sidekick)

  • Beta testing for enhanced slicer-dicer querying
  • Potential for aiding non-expert users

Conclusion

• Cosmos provides a powerful tool for large-scale healthcare research
• Importance of domain knowledge and data validation
• Potential for growth and refinement of the system

Final Notes

• Users can collaborate across institutions through Cosmos
• Legal issues and Epic's involvement in training and data validation
• Plans to enhance participation and data quality across sites

End of Lecture. Questions and discussions continued beyond this outline.