CLN - Focus on Data Science
Supported by
The ability to sort, collect, and analyze data was a key factor in how some healthcare systems were able to mobilize effectively during the worst of the COVID-19 pandemic. The University of Washington was one of the fastest in the country to expand testing for COVID-19, partially because of the amount of informatics resources they had prior to the pandemic.
Although COVID-19 has faded, it’s still circulating — and H5N1 bird flu looms on the horizon. We asked Patrick Mathias, MD, PhD, vice chair of clinical operations and associate medical director of the informatics division at the University of Washington School of Medicine, about how his institution was able to mount such a quick response, and what they learned for the next time a pandemic comes knocking.
How can informatics help laboratory medicine prepare for the next pandemic?
COVID taught us that we need broader awareness of different threats to our public health. We need a more robust infrastructure for collecting data and monitoring infectious diseases and potential pandemics. We also need more connected systems to make sure that our different data sources are plugged into the larger public health infrastructure.
This is all key in continuing to work with our public health colleagues, as we did during the COVID-19 pandemic, to ensure we’re monitoring new pathogens, like H5N1 and other variants of influenza.
During the pandemic, my institution pulled together a multidisciplinary group across departments and worked closely with county and city officials to think about how we could make our processes more efficient. We asked, how can we scale up our testing? How can we understand the various information systems in play? And how can we connect them by writing software to pull different tools together to get patient test results efficiently and to get those results to our public health authority?
We learned that informatics expertise is critical. We had to do a lot in the laboratory to get data to flow from point A to point B to point C as efficiently as possible. We also worked on continuous improvement methods, which aren’t explicitly informatics but do tie into analytic infrastructure.
We were able to have the laboratory handle tens to hundreds of times its normal volume of testing. A lot of that relied on having the right information systems and the right infrastructure to build tools on the fly and connect systems together using the data that we were collecting, and then also sharing that data with our colleagues.
How can labs use data science to improve readiness for events like this?
There are three main ways. The first is knowing our existing information systems, including how they work and communicate with one another. We need to have complete understanding of those pieces so we know what we could change relatively quickly. Informatics is not necessarily something where you can hire someone from the outside to fix your problems. You need to know where you’re starting from, have comfort with the current state of things, have a team of experts, and link them together.
The second is having data so we can react, and using it to understand how we can optimize operations. For example, there were times during the worst of the pandemic when we were low on supplies, like reagents and pipette tips. We had to scale up and track these things. We had what my chair, Geoffrey Baird, MD, PhD, called a “Noah’s ark strategy” for our instrumentation where we bought two of everything, and we had to shift on the fly, from day to day, based on what we had. We developed more rigorous inventory management techniques so we could know, at any given time, what supply we had on hand, and what our burn rate was. We developed dashboards so we could have day-to-day awareness of supply and know where we could make changes so we wouldn’t completely run out.
The third way is that, over time, we developed some in-house capabilities to write software. Many of the problems we had to solve were not overly complicated, like needing to get data from this instrument and reformat it to put it into that instrument or through that interface. We were able to utilize cloud capabilities and write software to connect everything to all the places it needed to be connected to.
What are some lessons your group learned that would make it easier to deal with the next emergent disease? What worked well? What were the barriers you had to overcome?
During the worst of the pandemic, a lot of our historical barriers shifted significantly. We were really freed up and given license to do what we needed to scale up our operation. We had lots of groups from different places all over the university – in addition to outside entities - pitching in to help move things along. Having a collective goal helped us align.
But since then, there’s been some falling back. Large disruptions can shake people loose a little bit to consider how to operate in a better, more efficient way. But that is harder when the institution is no longer in that mode of responding to an emergency.
In general, how would you describe the opportunities for lab medicine professionals to use informatics more in their institutions?
Often, people are attracted to working in laboratories because we have very predicable work. You need to make sure you’re following the same script every time.
Understanding information systems, workflow, and how staff operates — all those things can really help make improvements within the laboratory, not just being more efficient but in developing new ways to ensure patient safety and think about the value that we provide to our colleagues.
That’s true even with what seem like small ideas, like mapping a process or tackling turnaround times and how many samples are making their way from point A to point B during a given time. By asking questions about these basic processes, and by fostering collaboration between laboratory directors and staff, we can make those processes better.
Outside the laboratory, we have opportunities to improve the value of the results we provide. I’m particularly passionate about data analytics and data science and how we can apply data to improve patient care at a population scale.
For instance, how do we modify our electronic health records so we can nudge our colleagues to use the best laboratory test for a particular scenario? We also can improve patient safety by flagging things that might be less apparent to a healthcare provider, such as identifying and calling out common laboratory test interferences. Individual incidents of those kinds of cases might seem rare, but when you add them up, it happens to quite a few patients.
How do you recommend people get started with informatics and data science?
There are certifications and courses for those who are interested in informatics but don’t want to become board certified in it. These include courses from the American Medical Informatics Association and Association of Pathology Informatics, as well as educational resources from the Association for Diagnostics & Laboratory Medicine (ADLM, formerly AACC). The ADLM Data Analytics Steering Committee is also preparing a certificate program.
What are the most exciting developments around informatics in lab medicine for the near and medium-term future?
You can’t go too far in this space without bumping into something around artificial intelligence (AI) or machine learning (ML). While there’s been some hype around AI or ML for the last 10 or so years, there are now actual applications for those techniques in the laboratory in addition to broad applications around medicine.
I’m excited about stretching and pushing that boundary of what humans can do a little further by using some of these techniques. That way, we can provide better outcomes for patients in the laboratory. We can also identify trends and problems that we may never have recognized before.
This interview has been edited for clarity and length.
Jen A. Miller is a freelance journalist who lives in Audubon, New Jersey. +X: @byJenAMiller
Advertisement
Advertisement