Harnessing data science in laboratory medicine

In January, the Association for Diagnostics & Laboratory Medicine (ADLM) launched a Data Science in Laboratory Medicine Certificate Program designed specifically for clinical laboratorians and pathologists. The new program teaches participants how to set up effective data governance systems; request and process lab data appropriately; ensure data integrity, accuracy, and structure; and evaluate when and how to use machine learning. In short, it puts laboratory professionals at the forefront of some of the most transformative technologies in healthcare.

To gain more insight, Clinical Laboratory News spoke with Anthony Killeen, MD, BCh, PhD, DABCC, FADLM, a faculty member for the program and ADLM past president. Killeen talked about the state of data science in the clinical laboratory and the need for this exciting educational offering.

How did data science become one of your areas of expertise?

I got into data science later in my career. The University of Minnesota, where I’m a professor of laboratory medicine and pathology with an interest in clinical chemistry, serves as a central laboratory for a large number of clinical trials, many of which the National Institutes of Health funds.

These trials generated many data points, which we then passed off to coordinating centers that had statisticians to analyze them. I thought, “gosh, I’d like us to analyze some of that data ourselves.” Then I earned a master’s degree in evidence-based healthcare and medical statistics at Oxford University.

How does your lab use data science to improve testing and patient care?

We use a lot of data science, mostly statistical analysis, to assess quality control in our clinical trials lab. Among other metrics, we investigate the long-term stability of analytes in stored specimens, such as frozen blood samples containing glucose. We want to know how long such samples are stable. Is it 3 months? Three years? Ten?

We’re also looking at implementing real-time monitoring of patient results as a statistical quality control. For instance, if a patient’s glucose assay begins to drift away, we can gauge that something’s wrong before the actual quality controls flag it.

Additionally, we use data science in computational pathology. Although I’m not personally involved in this work, other lab leaders analyze the information gleaned from high throughput genetic sequencing, for example.

As data science becomes more sophisticated, what do you hope labs will be able to accomplish with it?

One of the most exciting developments is the integration of laboratory medicine with higher-end artificial intelligence (AI) and machine learning capabilities.

Clinical laboratories have always generated huge amounts of data on a daily basis. But hardware and software were not readily available for doing higher-end analytics. AI is changing that.

To share an illustrative example from everyday life: For a long time, labs were operating similarly to ATM machines. We’d receive a sample — like a card in the machine — and spit a number back out, similar to an ATM’s cash output. While the ATM might give you a receipt with your balance on it, it doesn’t usually provide any additional information, such as whether there’s a sale at Nordstrom or that the gas you just paid for is 50 cents cheaper across the street.

Now imagine if we could take a lab result and generate a report filled with similarly useful, interpretive medical information. For instance, we might learn whether a lab value has risen by 10% within the last 24 hours, prompting a possible flag, or if lab results don’t align with imaging findings from the same patient. If you could put those pieces together, you could get a powerful, predictive prognostic tool. That’s the goal: to extract a wealth of valuable information out of the data we already generate.

Are there any hopes for data science that you think are unrealistic?

I’m optimistic that a lot of good things will happen. That said, as with any other tool, we need to be very careful about ensuring the accuracy of the information we glean through data science. There needs to be appropriate caution before you implement all new technologies. You must validate that they work as intended by doing reality checks along the way.

What hurdles face data science in lab medicine over the next 5 to 10 years?

One of the limitations is laboratorians’ lack of familiarity with the technology. We will need to train people in how to use these tools, become comfortable with them, and gain proficiency with them. Clinical laboratory professionals at multiple levels must understand the basic principles of data science and acquire the skills to apply them effectively. Not everybody will learn coding languages like R or Python, but lab professionals should grasp the goals of data science projects in their labs and know how to evaluate data outputs.

It’s also important to understand the costs. Although a software package may be free to run, the information generated in a project may have financial implications, which should be considered on the front end as much as possible.

Why is it important for laboratory medicine professionals to learn about data science?

We’re at the beginning of a big wave of data science applications in lab medicine. If you have the right tools, you can take millions of numbers and extract a story out of them. It’s really exciting.

More broadly, laboratorians need to be cognizant of the larger technical and scientific landscape in which they operate. To stay up to date on how key trends shape your own field, you must not only stay abreast of the literature, but also be able to critically evaluate it.

What will ADLM’s data science certificate program cover?

It’s an introductory course for people who either work in laboratory medicine or who are at least familiar with it and want to expand their skillset. Participants will engage with 12 hours of educational content from prerecorded lectures and take a test to assess what they learned.

What distinguishes this program from other data science educational programs?

The program was developed to meet the specific needs of professionals working in the context of lab medicine. While other professional development offerings might target engineers or social scientists, this one has educational materials and examples taken from the clinical lab. It’s very focused.

Once people complete this program, what should their next steps be?

They should get engaged in a couple of projects in their own institution so they can apply the skills they learned. Although this is not a full-year course or degree program, it does provide a solid framework that should empower people already familiar with laboratory medicine to use data science in their work.

I hope participants attend the ADLM Annual Meeting and present abstracts based on their projects, write papers on what they’re doing, and otherwise share their efforts.

I also want to add that, although the certificate program is open to both ADLM members and non-members alike, ADLM members in particular can go on to join the association’s Data Science and Informatics Division. The community has an online forum where people can discuss what they’re doing and any problems they encounter.

What advice would you give to lab professionals looking to get started in data science?

Jump in and do it! If you work in lab medicine, now is the time to start your journey in data science.

Jen A. Miller is a freelance journalist who lives in Audubon, New Jersey. +Bluesky: @byjenamiller.bsky.social