Lecture Notes: Introduction to Dr. Lindsay Knaik and Epic Cosmos

Speaker Introduction

• Dr. Lindsay Knaik
  • Professor of metrics at the University of Iowa
  • Associate CMIO at Stead Family Children's Hospital
  • Background in biomedical engineering, medical school in Iowa, residency in pediatrics at Baylor, and neonatology fellowship at Vanderbilt
  • Focus on neonatology and informatics

Epic Cosmos Overview

• What is Epic Cosmos?

  • Large, de-identified database of EHR information from multiple institutions
  • Over 203 million patients, billions of encounters
  • Aims to aggregate clinical data for research purposes

• Benefits of Cosmos

  • One of the largest EHR databases
  • Fast querying capabilities (minutes per query)
  • Represents diverse demographics similar to the U.S. Census
  • Contains data on longitudinal patient charts, labs, meds, social determinants of health

Challenges and Considerations

• Data Integrity

  • Data quality varies; need to ensure questions asked are reliable and data is present
  • Unique identifiers attempt to track patients across institutions

• Accessing Cosmos

  • Requires belonging to an Epic institution
  • Legal agreements needed to contribute data
  • Training required for users (Epic classes)

Practical Applications

• Research Examples
  • Neonatal Hypertension Study
    • Use of Slicer-Dicer tools to analyze neonatal hypertension across demographics
    • Importance of sanity checks: ensuring data is plausible and matches known clinical trends
  • Tiny Baby Survival Rates
    • Comparison of survival rates of premature infants in Cosmos with manually curated databases
    • Validation of Cosmos data against existing literature

Data Science in Cosmos

• Tools and Limitations

  • Use of SQL, R, and Python in a secure virtual environment
  • Limitations on state and zip code data in de-identified settings

• Cosmos Sidekick

  • New tool using large language models to assist in data queries
  • Potential to accelerate data exploration but requires careful verification

Collaborations and Improvements

• Working with Epic Cosmos Team
  • Feedback processes in place to improve data representation and query tools
  • Examples of improved data entry options based on user feedback
  • EPIC's own research publications using Cosmos data

Future Directions

• Expansion and Evolution
  • Potential for expanded access to state-level data on data science side
  • Ongoing efforts to improve data quality and representation
  • Need for multidisciplinary collaboration to maximize the use of Cosmos

Funding and Use Cases

• Considerations for Research Funding
  • Epic's 10% fee on funded project grants using Cosmos
  • Potential impacts on research choices

Conclusion

• Summary
  • Epic Cosmos provides a powerful tool for large-scale EHR data research
  • Ongoing improvements and adaptations required to optimize data use
  • Encouragement for continued exploration of Cosmos capabilities by researchers

These notes summarize the key points from Dr. Lindsay Knaik's lecture on Epic Cosmos, including the database's potential for research, practical applications, and considerations for data quality and access.