Patricia Flatley Brennan, director of the National Library of Medicine (NLM), was the 2022 recipient of the Miles Conrad Award from NISO (National Information Standards Organization), and she delivered the annual Miles Conrad Lecture at this year’s NISO Plus Conference. Brennan holds an M.S. in nursing and a Ph.D. in industrial engineering and has led NLM since 2016. She is a past president of the American Medical Informatics Association, a member of the National Academy of Medicine, and a fellow of the American Academy of Nursing, the American College of Medical Informatics, the New York Academy of Medicine, and the American Institute for Medical and Biological Engineering.
Brennan’s lecture, “The Role of a Library in a World of Unstructured Data,” was delivered live via Zoom on Feb. 16, and NISO has made the recording available at niso.plus/miles-conrad-lecture-2022-dr-patricia-flatley-brennan. We spoke shortly after her lecture, and this is an edited and abridged version of our conversation.
Dave Shumaker: Dr. Brennan, I’m intrigued by your educational background. I imagine there aren’t many people with degrees in nursing and industrial engineering working in information management. Can you share a bit about your educational and professional journey?
Patricia Flatley Brennan: Sure. While working as a nurse, I got interested in how we could support patients in self-management. I also got interested in computer technology, and I wanted to do graduate study. I was unable to find a program in nursing that would address my interests in decision-making, computer systems, and patient self-management. Some nurse colleagues introduced me to a wonderful, visionary industrial engineering professor at the University of Wisconsin–Madison, Dave Gustafson, who was interested in mathematical models for decision-making. I ended up studying with him. Industrial engineering introduced me to systems thinking. Also, it has something in common with nursing: They both operate from the idea that human function is enhanced by technology, but shouldn’t be overwhelmed by it.
Shumaker: In your lecture, you touched on NLM’s responses to the COVID-19 pandemic. What were some of NLM’s key initiatives and changes?
Brennan: The first and most important was that we accelerated what we already do best, which is make sure that the needed information is available in a timely fashion. The two critical pieces that were different were helping to open the literature more, working with journals to make the research literature about COVID available without a paywall, and second, to get our genomic databanks—which are essential for understanding the virus and its genome—responsive in a faster turnaround.
Shumaker: Have the challenges of the pandemic led to any long-term changes at NLM? We’re all talking about the “next normal” or the “new normal.” How will NLM’s future change as a result?
Brennan: One of the things we’ve learned for certain is that our staff have been able to operate at a distance. We didn’t have any gap in services. So, I believe that in the future, we’ll have a hybrid work environment. We’ll continue to have the opportunity to work with the maximum flexibility available to our staff.
Shumaker: One of the key trends you’ve mentioned in your lecture and elsewhere is the importance of data science and informatics. As an employer of data scientists, is the educational system meeting your needs for graduates with the skill sets you need? Are you getting them from library science or other disciplines?
Brennan: We’re seeing a whole new group of librarians—data and open science librarians. So, we’re seeing a great role for librarians in this area. One of the critical changes is to move from providing access to the literature to providing access to data. Skills like curation and the development and use of formal terminologies are needed in data science—they transfer very nicely. The other skill that’s needed is new ways of implementing search and discovery. So, we look to library science to bring knowledge to improve data science. Also, we’re actively partnering with educational institutions. For example, we’re starting a program this summer to bring young people from historically Black colleges and other minority-serving institutions to do summer internships with us. The development of a diverse workforce will come when people can see the exciting things going on in data science, so almost all the data science and artificial intelligence programs we’re working on will have a strong training component to bring the workforce in. So, we’re not passively relying on the educational system. We can help young people see that there are exciting challenges for them to work on.
Shumaker: Another aspect of the development of data science is the accessibility of software code. Code, data, and published literature are all artifacts of research. How do NLM’s initiatives address code as well as data and literature?
Brennan: There are really three kinds of code. The first is the code that generates the data. The second is the code that curates the data—assigns terms and meanings to it. The third is the code that’s used to analyze the data. In the spirit of transparency, we need to not only know what the code was, but also to be able to reuse it. NLM doesn’t hold analytical code, but we do have the ability in PubMed Central for the investigator to attach supplemental files, so they can include executable files as well as data files. We’re also encouraging investigators to use generalist repositories. We know that as the National Institutes of Health moves toward requiring them to have data management and sharing plans, they need to have a place to put the data. Some of the repositories only accept data, but other services do accept executable code.
There are also questions emerging about the licensing of code and algorithms. Even if there’s a desire to share the code freely, sometimes licensing helps to preserve the integrity of the code set. This is an interesting shift—we’ve had 20 years of open source code, and now we’re seeing some pulling back—not to less sharing, but to share code in a way that’s trustable and persistent, so the code is preserved intact as the original developers expected. Ultimately, we see the future of digital objects becoming more interconnected. The ability to trace the relationships of the different artifacts of a research study will improve understanding and the reproducibility of research.
Shumaker: Returning to the pandemic, it’s been said that we’re really experiencing two pandemics: the COVID pandemic and the “infodemic,” or pandemic of misinformation and disinformation. How do you see NLM’s role in dealing with misinformation?
Brennan: I’d begin by saying that science is self-correcting. As we get a better understanding of the phenomena we’re dealing with, we change our interpretation. Remember, at the start of the pandemic, we were wiping down our groceries. Now we understand that this is a solely respiratory-transmitted virus, so we don’t do that anymore. As science self-corrects, new information comes out. But it’s difficult to help people understand that, when they are looking for consistency, and they believe that science should provide one answer and never change it. Having said that, people need to know what sources to trust. Understanding how people draw meaning from information is our best skill set at NLM. Within Medline Plus, our consumer-oriented resource, we have guides, tutorials, and other tools for teaching and helping people evaluate information. Also, we have our national network of medical libraries and partnerships with public libraries, and we find that having that human link with people who understand local context is really valuable. Even years ago, we found that in some communities, getting health information into barbershops and hairdressers worked well, because that’s where people went for trusted information.
Shumaker: Let’s follow up on that issue of trust. Could you say more about it?
Brennan: I believe strongly that it’s a responsibility of the federal government to provide individuals with unfettered access to information that has been properly vetted. We promote the individual’s access to information through PubMed Central, which houses articles that have been vetted by experts in the field. We encourage people to rely on that expertise. We at NLM aren’t the adjudicators of what is good science—the experts are.
A second aspect of trust is permanence, so that a question will be answered in a similar way over time. That relates to the earlier point about science being self-correcting. It’s important to help people realize that even though our understanding of health problems does change over time, we still preserve the older resources. For example, we still have the articles in PubMed that say stress causes ulcers. So, we retain the content, but it takes on a different meaning over time.
That’s where unstructured data and the idea of sense-making in the moment come in, and why I believe we need to develop meaning that is trustable. Going back to your question about storing code with data, we need to develop tools for people to ask questions and interpret meaning. So, it isn’t just a matter of trusting your doctor, or NLM, but developing the skill to understand the information that’s being presented and its relevance in my life. For example, you might be encouraged to drink eight glasses of water a day—but if you have a kidney problem, that might not be a good idea for you. So, helping people understand how to modify information to make more sense in their own lives is a very important part of building trust. Building trust is really about building capacity.
Shumaker: Let’s shift to a management question. You’ve written in your blog about your philosophy of team-based management. How would you describe your approach to management?
Brennan: In one of my blog posts, I wrote that I learned the most about management by playing squash. Squash is all about timing and angles. You have to know when to intervene and in what direction. So that’s one of my management secrets.
My team-based management approach originated with the idea that NLM is a group of interdependent centers of excellence. If we think that it’s only library operations, or only the Center for Biomedical Information, we under-power the interdependencies, and we make decisions that might not be optimized for the entire enterprise. To be able to balance local and enterprisewide optimization, members of the leadership team have to know and trust one another. Sometimes they have to have hard conversations. But that comes from a firm foundation of respect.
Also, I’ve learned not to involve people in things they don’t need to be involved in. They’re busy people. I cannot tie them up just because it’s fun to have a meeting. When I first came to NLM, I held leadership team meetings once a week. After 6 weeks, members said that was too often. So, we changed to every other week, and that seems to work well. And everyone can count on that meeting—that sacred hour. Sometimes, they have to miss a meeting, but they don’t send a deputy. They can count on it being the members of the leadership team. I can tell the operation is working better, because I hear stories about someone working with people from other units, so we’re leveraging our internal strengths in very positive ways.
Shumaker: I’ve never heard the connection between squash and management before!
Brennan: Of all the racquet sports, squash is the most precise, and it’s played in a very small area. You have to wait until the ball comes to you—you can’t rush the ball. It’s a lot of fun, and I’ve learned the most from it.
Shumaker: Here’s my last question: Is there anything else you’d like to add, before we wrap up?
Brennan: I want to return to the idea of partnerships. I came to NLM as a nurse and industrial engineer. I knew almost nothing about library science, about what it took to put our products out, and how important standards are to the work we do every day. I came in, I hope, with a sense of humility, to learn. One of the things our staff have taught me is that we don’t control standards. Standards arise through collaboration; they’re taken up by a community. We don’t make edicts—we foster and facilitate. If we don’t work together to build and implement a standard, it will never happen. We’ll never get the real value, which is to enable information from one context to be understood in a different context. As we’re coming into a time when we’re desperately trying to understand the diversity of the American healthcare public, the terms we use have to be enlightening about the Other, not melding everyone into the Same. This challenges the standards community to help us understand where phenomena are different: the experience of pain; the sense of self-efficacy. The words in different contexts have to be different words in order to make sense, but they can’t be willy-nilly. We need to find a way to make standards carry meaning about the Other. It’s through partnerships that we can build the overlapping ontologies, so that the meaning transfers.
Shumaker: That’s a great place to wrap up. Thank you.
Brennan: Thanks so much!