Born and raised in Arlington, Virginia, just outside Washington, D.C., Leduc recalls an early fondness for math. “As a kid,” Leduc says, “you’re taught that one plus one equals two.” There’s an answer to a math problem, in other words, and Leduc’s quantitative bent has served her well in her role as data detective, where she found math to be more complicated—and interesting—in the real world than it was at school.
If “data detective” sounds cheesy or overblown, consider that Leduc’s role is to help triage anomalies on the Surescripts network, which carried 20.4 billion transactions in 2021 alone. Anomalies are outliers that can be found in just about any large data set, but given the sensitive nature of health information, they must be taken seriously in order to continuously protect and improve the access, security and performance of the Surescripts network and the data it carries.
To do this, Leduc applies computer logic, machine learning and good, old-fashioned human ingenuity.
“My role in healthcare is to keep patient data secure, to make sure the data is used how it was intended to be used.”
When Leduc joined Surescripts in July 2021, following a five-year stint as a data scientist, she quickly got to work on conceiving and developing a method for prioritizing anomalous outliers.
The result—Leduc’s “severity score logic”—uses rules and a machine learning algorithm to automate triage, so Surescripts network access investigators like Eric Engelken can hunt anomalies from highest to lowest priority in search of answers: Was it data misuse? Was it a misconfiguration? Was it nothing?
As the name suggests, severity score logic applies a score to each anomaly, to help our investigators focus their efforts.
As satisfying as math was for Leduc in grade school, with its clear-cut answers, she later found a quantitative “artform” at Columbia University in New York City, where she earned a master’s degree in statistics. Her study included advanced data analysis, which she puts to use at Surescripts.
Once Engelken and the other investigators complete their latest round of inquiries, and there is enough data to conduct a meaningful statistical analysis, Leduc takes several measurements:
- Did the result of the investigation point to data misuse?
- What was the source of the anomaly referral, whether human or automated?
- How long did it take to close the investigation or move it to remediation, if necessary?
- How many investigations were conducted in the given timeframe?
- Were any “repeat offenders” a cause for concern?
These efforts result in a continuous, positive feedback loop, which Leduc and the Network Access team use to find the proverbial needle in the haystack of more than 20 billion transactions.