Epic Launches County-Level Health Alerts

Epic Health Alerts for Communicable Diseases

Epic Research has introduced county-level monitoring for communicable health conditions. The research division of Epic Systems will publish Health Alerts indicating higher than typical rates identified by statistical models utilizing information from real-world medical records.

Health Alerts are available across the U.S. concerning conditions such as

• Measles
• Acute bronchitis
• Acute tonsillitis
• Strep throat
• Viral gastroenteritis

Epic issues alerts based on its Cosmos database, which pools data from a community of health systems that rely on the vendor’s electronic health record system. The dataset combines information from 2,067 hospitals, more than 47,000 clinics and 300 million patients to answer clinical questions and support better care delivery.

How Epic Identifies and Validates Rising Health Trends

Because Health Alerts stem from organizations that participate in Cosmos, they don’t represent complete case counts but do indicate population-level trends. The detection process uses diagnosis rates based on ICD-10-CM codes.

The methodology section of the tool specifies that Epic flags a condition for review if it meets these requirements:

• There is a year-over-year increase aside from typical seasonal variations, such as rising flu cases during the winter.
• The proportion of a particular diagnosis compared to all patients seen is growing at an accelerated rate. In other words, the share of the condition grew faster between this month and last month than it did between last month and the month before that.
• The increase is statistically significant and not explained by random variation. If a condition meets the first two requirements, the Epic team then assesses it utilizing a Farrington improved algorithm, which establishes expected baseline rates according to three years of historical data and further adjusts for long-term trends and seasonality. This ensures that alerts identify conditions rapidly on the rise rather than health concerns gradually increasing in prevalence.

The Epic Research team — consisting of both clinicians and data scientists — reviews each alert prior to publication. While there is no fixed publication schedule, the team usually reviews statistical model output on weekdays. Health Alerts are available online, and interested parties can also subscribe to receive alerts via email for specific states.