InsideTraCS — with Loretta Fearrington

Loretta Fearrington is a Research Informatics Specialist with the NC TraCS Informatics and Data Science (IDSci) service, which supports investigator studies with innovative technology and advanced analytics. She oversees the Electronic Health Record (EHR) Data Service, supervising the analysts who query the Carolina Data Warehouse for Health (CDW-H) and use other tools to provide data to researchers.

Marla Broadfoot, NC TraCS science writer, recently spoke with Fearrington about what drew her to informatics, the challenge of creating common data models, handling large amounts of PHI data, and helping researchers succeed.

How did you become involved in research informatics?

I honestly didn't know what I wanted to do when I graduated from college. I ended up working as an administrative assistant to senior executives at Roche Biomedical (now LabCorp). I loved the job, but craved learning something new. One day, the Senior Vice President of Operations told me, "There's this program called QMF (IBM Db2 Query Management Facility). If you could learn it, it would be really helpful."

When I had it installed on my computer and opened it, I was staring at a blank screen. I quickly realized that it was not "a program" per se—I had to learn to program! At first, I struggled; but I liked the challenge and continued to get better and better, eventually getting promoted into an analyst position, and even receiving the Presidential Achievement Award in the World of Health Services, along with two of my colleagues.

I worked in various analyst roles there for 19 years (21 years there in total), further developing my programming skills. I was hooked on data and analytics. Fast forward—I remarried after a divorce and over a decade as a single mom and moved to Chapel Hill to live with my husband. Growing tired of my commute, I applied for and was offered a job at NC TraCS. That was 2011, and I've been here ever since.

What changes to informatics, specifically the application of informatics to research, have you noticed since you came to NC TraCS?

There are so many stories to tell behind the data and creative ways to use it to solve problems and answer questions.

 Fearrington

When I first started in NC TraCS over a decade ago, Emily Pfaff (now Dr. Pfaff) and I were the only two analysts. We were still using a system called Business Objects, which used "drag and drop" functions to query the data. We eventually were able to query the data by writing our own Structured Query Language, a programming language used to communicate with and manipulate databases. That opened so many doors. We could provide more robust data to our researchers, and we've been able to work closely with the Information Service Division (ISD) of UNC Health and even help with development of the Carolina Data Warehouse for Health (CDW-H). There are so many stories to tell behind the data and creative ways to use it to solve problems and answer questions.

One of the biggest changes I've seen in informatics in the last decade is the creation of common data models that allow us to "normalize our data" so that, for example, a lab test for COVID-19 antibodies at UNC has the same "test number" as it does at Duke or Vanderbilt. When we combine our data, there's more statistical significance in our findings.

I've also seen the development of some really cool tools that can "make sense of the data," especially in the world of analyzing notes. At first, I thought that "free text" data (like in clinical notes) would evolve and clinicians would enter more structured or "discrete" data, but I've seen the opposite occur. While some of that did happen, I've watched as informatics has advanced and tools were created like natural language processing and machine learning that transform free text into meaningful discrete data.

Part of your role at NC TraCS entails overseeing the analysts who query the Carolina Data Warehouse for Health (CDW-H). What is this resource, and how do you think it has helped to transform clinical and translational research and UNC?

To understand the CDW-H data, imagine all the data that's generated from UNC hospitals and their associated clinics while patients are being scheduled and treated (recording vitals, ordering medications and procedures, test results, etc.) and the billing data generated from all these transactions. There are over 5,000,000 patients and their associated data in the CDW-H. Being able to query this much information and narrow the scope to only relevant data for a particular research project is a game changer.

Even the most advanced informatics techniques, like machine learning, require data. It's at the core of it all.

 Fearrington

In the past, so much data was obtained by reviewing charts and manually recording it. The ability to capture variables straight from the system enables us to have more accurate data, and to get it faster. It also gives researchers the ability to "search the data" in a way they never could. For example, when a researcher needs to find patients with a certain diagnosis, positive test result, and who are NOT taking certain medications for their clinical trial it can be like finding a needle in a haystack when reviewing charts. That's when the power of programmatically querying the data can be game-changing. Even the most advanced informatics techniques, like machine learning, require data. It's at the core of it all.

While big data repositories like CDW-H can promote clinical research efficiency, they can also raise concerns about protecting personal health data. Could you talk about how the concept of the "honest broker" could address those concerns?

Large amounts of data related to protected health information (PHI) does raise concerns, but we're well equipped to handle them. Although the regulatory aspects of obtaining data for research can seem daunting to investigators, we have services in NC TraCS that assist in "walking them through" our regulatory processes.

We're responsible for assuring that the data we provide is constrained to what's permitted by the IRB and approved through our regulatory process (for example, if the IRB indicates that the study doesn't include children, we constrain to patients 18 and older). We also make sure to remove any identifiers that are not permitted, and we provision the data securely and only to those permitted to see it. We are completely unbiased regarding the research and realize that it's not OUR study.

What is your favorite part of your job?

My favorite part of my job is feeling like my team is making a difference that will ultimately benefit the patient.

 Fearrington

My favorite part of my job is feeling like my team is making a difference that will ultimately benefit the patient. We're doing something good, and we're good at it. We play an important role in research. I love the variety of projects we get. I love the excitement behind this work, the potential it carries, learning every day, and seeing how incredibly smart our researchers are!

I've often quoted my very wise mother by saying "You know, if everyone in this world were good at the same thing; we'd be in trouble." I've had clinical researchers comment that they don't understand "all this technical stuff" that we do; to which I respond, "I am SO glad that you are good at what YOU do" (and there's no way I could begin to understand their work!). We join forces (with other important roles, like biostatistics and other services) to make the whole thing happen, everyone contributing what they're best at. How cool is that?