View  Transcript

Dr. Jody Platt. She is an Associate Professor of Learning Health Sciences at the University of Michigan Medical School. Trained in medical sociology and health policy, her research focuses on issues at the intersection of informatics and ethics. She is interested in understanding what makes data-driven health trusted, and the pathways for earning, achieving, and sustaining trust across stakeholders. Dr. Platt is the Academy Health and ABIM Foundation Senior Scholar in Residence advising on issues of measuring and building trust.

Learn more about Dr. Platt, and you can follow Dr. Platt @jeplatt and U-M Learning Health Sciences @umichDLHS  on X.

Resources

• Find out more about the research stories mentioned at the top of the show at the links below:
  • Clinicians could be fooled by biased AI, despite explanations
    U-M study shows accurate AI models improved diagnostic decisions, but biased models led to serious declines
  • Can AI help hospitals spot patients in need of extra non-medical assistance?
    Test of machine learning techniques shows promise for scouring medical record notes for social determinants of health
• You can learn more about The Fundamentals on our website.
• Learn more about other exciting research happening at the University of Michigan, by checking out Health Lab, Michigan Medicine's daily online publication featuring news and stories about the future of healthcare.

Transcript

Kelly Malcolm:

Welcome to The Fundamentals, a podcast focused on the incredible research and researchers at Michigan Medicine. I'm your host Kelly Malcolm.

Jordan Goebig:

And I'm Jordan Goebig. And in this week's episode, our guest scientist is in the studio to discuss a topic that really impacts the lives of everyone, patient health data.

Kelly Malcolm:

But before we jump into the interview, let's share one U of M research story we read this week. This particular story illustrates the challenges facing clinicians using AI within healthcare. Basically in this study, the researchers set up a scenario where they trained AI to review imaging tests and come up with a diagnosis based on biased information including an explanation for why the AI came to that conclusion. And the unfortunate outcome was that physicians reviewing the AI decision were less accurate in their own diagnosis and treatment plan.

Jordan Goebig:

In more positive AI news, a recent University of Michigan study found a unique and promising role that machine-learning techniques may play in using patient medical record notes to determine non-medical care like transportation, housing, and food that a patient may need support with. We'll provide links to the full articles and info about our featured guest in the show notes. Now let's get onto our guest.

Kelly Malcolm:

In the studio with us today we have Dr. Jody Platt. She is an Associate Professor of Learning Health Sciences at the University of Michigan Medical School. Trained in medical sociology and health policy, her research focuses on issues at the intersection of informatics and ethics. She is interested in understanding what makes data-driven health trusted, and the pathways for earning, achieving, and sustaining trust across stakeholders. Dr. Platt is the AcademyHealth and ABIM Foundation Senior Scholar in Residence advising on issues of measuring and building trust. Hello, Dr. Platt.

Dr. Jody Platt:

Hi. Thank you for having me.

Kelly Malcolm:

So great to have you here.

Jordan Goebig:

Yeah, thank you so much for coming in today. Before we jump into our more formal questions, I would actually love to just learn a little bit about you for a brief moment. How did you get to Michigan? What's your path look like?

Dr. Jody Platt:

I've been in Michigan a long time. I came to Michigan to get a master's in public health a long time ago, and then I just never left. I was very lucky to be able to be here from that first educational endeavor. So I got my master's in public health here at the School of Public Health. I worked at the School of Public Health for a number of years with a genetic epidemiologist, Sharon Kardia and Toby Citrin who passed away in 2019 right before the pandemic started. Maybe I'll get to talk about them a little later. So I worked with the university and then I decided I really want to get a PhD and I want to keep going. So I got my PhD here at Michigan and was hired onto the faculty. So very lucky to just be able to come and stay.

Jordan Goebig:

Yeah. Well, we're happy to have you here today and for the years that you've been with us. So I'll jump into my first question. So you just really, really briefly touched on what you do, but I'd love to hear more about what are the examples of the types of research data that you collect and then that you use for your research?

Dr. Jody Platt:

So I study people and their perspectives on how we use health information and how we should be using health information, and then the technologies that come out of that. So increasingly with that is AI. I do a lot of survey research. I collaborated over the years with NORC, which is one of these national survey firms that collects data from the people who live in the United States. And so I do public surveys and then I've also done a number of public deliberations with community-based organizations across the state of Michigan. And that's where we're able to bring people together for almost an entire day, sometimes a couple of days, to try to do both an educational piece and to learn about how health information is collected and used and shared. And then to begin to identify what kind of policies do we want in place, what do we want Michigan medicine to do with that health information? What do we want policymakers to do to be able to protect consumer interests?

Kelly Malcolm:

I think the concept of protecting consumers has been in the media a lot recently with the use of data for everything on social media and healthcare. But I am really interested because I know you've been studying this for a long time. How did you get involved in exploring the use of data especially in the healthcare setting?

Dr. Jody Platt:

So for me it started with public health. When I was working with Toby Citron and Sharon Kardia many years ago, they were working with the health department and specifically the newborn screening program. And in the state of Michigan we have a repository of over 4 million blood spots that, again, this was probably over 10 years ago now. But they had a very practical issue of a building was being demolished and they needed to move the blood spots from point A to point B. And in the process it was like, wait a minute, this is actually a really potentially valuable resource that we could be doing research. We have blood, we have information about babies, their moms, where they were born. It's blood spots so they can potentially get DNA information from that. So a really rich. And all babies born in a state go through this program. It's one of the few sort of universal healthcare programs that we have in the United States.

So again, it was seen as this boon for research, but it wasn't what people were expecting to happen with their blood spots. And so it raised this really fundamental ethical question, should we use this for research and if we should, under what circumstances can we do so ethically? That's how I got started. We're still working on how do we inform parents about the research uses of blood spots. Many of the same questions are occurring in healthcare today. We collect a lot of data about people and patients the minute they walk through the door and we can use that for a lot of things. It's not just for patient care anymore.

Jordan Goebig:

I think this is a good segue into, I always try to ask a fundamental question and something that when I was researching about you, when I was doing my research, collecting my data, the term ELSI research, if I'm pronouncing that correctly? Is it LC or is it ELSI, research kept popping up and I had never heard that term before. So I'd love if you could just define that for us.

Dr. Jody Platt:

Yeah. So ELSI is an acronym. It stands for Ethical, Legal and Social Issues or implications of whatever it is that you're trying to do. ELSI started as an issue for genomics. So when the human genome was sequenced, the NIH in particular said sequencing the human genome is a really big deal and we need to be able to attend to both the science of this, but we also need to attend to the ethical, legal, and social implications of what it is that we're doing. NHGRI, which is the National Human Genome Research Institute, and again, the federal side of the Human Genome Project did a very innovative thing at the time and said, you know what? We really need to be sure that we dedicate 10% of our budget to these issues. And so it became in essence a field at the same time that we were sequencing the human genome. So that's where it came from.

Jordan Goebig:

Really interesting.

Kelly Malcolm:

Yeah, had no idea.

Know all about that project, had no idea. So what have you found? I mean, I know my gut instinct is that as someone who is interested and dedicated to science, I'm okay with my data being used, but I know that's probably not universal feelings. What have you found and has the attitude about data use changed over time?

Dr. Jody Platt:

Yeah, that's a great question. So when we were doing our research with the Biotrust which is the newborn screening program, the sort of biobank that emerged from that program, we did a project where we went out and we talked to young adults. So those were the people that were in the biobank, but they were adults at the time. And so it was really there as sort of a key stakeholder group. So we went out to campuses across the state and asked them, what do you think? Uniformly people had some kind of shock response initially. It was like, "Oh, wow, that's happening?" And then people from there went into I think three main groups, and one of them was what you were describing. It's like, "Oh, that's kind of cool. It's research. That's an important piece of what we do in society."

And then other people were like, "Okay, that was a little bit weird that you told me that and I don't really care." And then the third group was actually pretty upset about it. It's sort of like, "Well, you didn't tell me. I'm not sure I trusted you before you told me. I still don't trust you now that you have told me." And the people that really wanted to stay engaged over time to see what's going to come of this and are you actually going to benefit me and my community.

Kelly Malcolm:

So for that last group, is there any way to opt out of these things? And I guess as a researcher, how do you deal with knowing that there's going to be a subset of folks who are really upset about you using their data in this way?

Dr. Jody Platt:

Yeah, it's a really important question. And the second part of what you just asked is how it's changed over time.

Kelly Malcolm:

Um-hum.

Dr. Jody Platt:

That particular issue I think has changed a lot over time. Because the sort of traditional research paradigm is, if I don't want to be a part of your study, I can opt out and then it's over. And now there's so much data flowing across so many different organizations and types of organizations and being used for so many different things, that to say I want to opt out is not a real option anymore. You can maybe turn it off on one sort of node, but it's going to be flowing somewhere else. So I think what we're beginning to see, I hope we're seeing, is looking for ways that we can really protect consumers about the harms that come from data sharing as opposed to the fact that it's being collected and shared in the first place because the data, it's out there.

Jordan Goebig:

Are there any actual policies in place that is around the use of data and research at all?

Dr. Jody Platt:

Yeah. And in many ways, research has a lot more rules around it than a lot of other data uses. So we have informed consent for research. We have rules and regulations. We have IRBs that oversee, make sure at least we're compliant with the current policies and are not doing things outside of the scope of what's allowable. So research has a fair amount of oversight. There are other data uses like for quality improvement for example, that is a little bit less clear because it's not research. We need it to make our health systems better, but there's a very fine line between quality improvement and research and one that consumers don't necessarily pick up on or they don't see it as differently as we would inside the healthcare system.

Jordan Goebig:

I know that you are also wading into the whole chat GPT, artificial intelligence arena, and that's sort of new.

Dr. Jody Platt:

Yes.

Jordan Goebig:

But what do you think those tools can be used for as far as maybe making healthcare better or I don't know. What are the positives, I guess, and then we can talk about some of the potential pitfalls.

Dr. Jody Platt:

Sure. I mean, they're very powerful tools, and if we can again, sort of harness all of this data to help with the administrative burdens of healthcare, to help patients and doctors communicate more effectively, to identify where the problem areas are that we don't even know exists. That's important work.

Kelly Malcolm:

And then some of the potential harms?

Dr. Jody Platt:

Yeah, I mean the harms are I think a couple of things. I think one is that we may be over-relying on technology to a certain extent. So the dream always has been with health information technologies that we can hopefully use this technology so that we have more time to be humans. And so that's again the dream. But sometimes when you have more time, it just means that you can do more work or see more patients. I think ideally if we can find ways to create that space so that we can have those better, more human connections, then that would be amazing. And if AI can help us get there and help us create the space for us to think and do better work, let's do it.

Jordan Goebig:

Okay. So pardon my ignorance on this topic, but when we're talking about ChatGPT, is it about uploading, I guess for lack of a better word, or maybe I'm using the right word. I don't work in the AI space, but is it about uploading all of that research data into artificial intelligence systems?

Dr. Jody Platt:

Again, I'm not a technology expert, but I think ChatGPT and AI in general, it benefits from a lot of data and different types of data. So in many ways, talking about electronic health records which may or may not be research in the traditional way. It's just the data that we happen to have about people and what can that tell us about health.

Jordan Goebig:

Yeah, that totally makes sense. Thank you.

Kelly Malcolm:

I'm not harping on the harms of AI. Trust me. I'm one of those people. I'm a skeptic by nature, and I know there's a lot of hype around this. I know that a lot of the data that these systems I've been trained on comes from a certain viewpoint of the world and has things built into it that we might not want to perpetuate. So do you think that the use of AI tools can exacerbate health disparities?

Dr. Jody Platt:

Absolutely. There's no doubt about it. And I think there are a few ways that it can do that, right? So part of it is exactly what you were talking about, where the data itself is biased. And if you're looking sort of in a closed healthcare system, that health system is going to have data about the people who walk in their door, and we already know that there are huge disparities around who has access to care and who doesn't. So we're starting, to the extent it's using that data set, that data set is biased. It doesn't have the information it needs to be not biased. ChatGPT in particular is basically making predictions about what should be coming next. And so if we live in a society that is biased and discriminates, it will perpetuate that. So if we've had racism in the past, it's going to predict racism in the future. I don't know how else to sort say that, but...

Kelly Malcolm:

Yeah. I mean, it's kind of why I am always taking a pause when I'm reading about this, all the different ways that it's going to make life incredible. But I know that Learning Health Sciences is about collecting that data but then feeding it back into the system to improve it.

Dr. Jody Platt:

That's right.

Kelly Malcolm:

Can you think of a way to maybe mitigate against some of those effects?

Dr. Jody Platt:

Yeah. I think if we check... If you feed the data in, it gives a prediction and then you can then assess those predictions for that racism or bias or whoever it's discriminating against and then try to basically change your practice, then that's one way of mitigating some of those harms. We can also, I think, ask better questions around what are the outcomes that we care about? And so if we're looking at the appropriate outcomes, then we can hopefully be also looking at the appropriate inputs. So the third one has to do with what kinds of data are we collecting period. And do we really have a real representation of people's life in the information we have? We don't right now, but...

Jordan Goebig:

You're working on it.

Dr. Jody Platt:

We're working on it.

Jordan Goebig:

You're working on it.

Dr. Jody Platt:

I'm working on it specifically.

Jordan Goebig:

It felt like, as I was reading your bio, I was like, it feels like you're working on this.

Dr. Jody Platt:

I'm trying.

Jordan Goebig:

Are there any specific projects you would like to talk about that you are working on right now? I know you had mentioned when we did a quick brief phone call that you did have some upcoming things coming down the pipeline. I'd love to hear more about those.

Dr. Jody Platt:

Yeah, I mean, so again, we've done a lot of work on data and data sharing and that part of the health information infrastructure, and we now have a couple of federally funded grants to look at public perspectives on the use of AI, how they want to see AI labeled and how do we communicate with people about how AI is used. And we have some exciting sort public information about even some of the ChatGPT and AI and what are the policies we'd like to see in place to make that a safe technology and not one that...

Jordan Goebig:

Is freaking me out? You better fill your survey out.

Kelly Malcolm:

No, but I appreciate that we have people like you thinking about these things, and we've done a lot with ethics and data use at Michigan for a really long time, so it does make me feel better about it.

Dr. Jody Platt:

The skepticism is justified.

Kelly Malcolm:

So are there any collaborations you want to mention or note or ones that you're hoping to make just in case they're listening to this?

Dr. Jody Platt:

Well, okay, I love working with people, so feel free to contact me. I'd love to talk to you. I think Sharon Kardia has been a mentor for many, many years and continues to be so and the School of Public Health in general, I think is a good partnership with the medical school. I've got Paige Nong who is at the University of Minnesota who does really excellent work looking at... She's an anti-racist scholar looking at, again, health information technologies and I don't know. This is hard. This feels like the Oscars. My mom.

Jordan Goebig:

That's fair. It seems like when I was, again, kind of going back to you, and me being a creep and looking you up on the internet and Googling you. I was telling them before we started this that I was telling my husband, I'm like, "She just is doing so many things" like in a very positive way.

Dr. Jody Platt:

That's so nice.

Jordan Goebig:

I don't know exactly what I'm going to ask her because she's really been involved in so many different types of quantitative data collection, which makes sense because an expert in that area. So folks are asking you to collaborate and probably you're seeking them out as well, so it makes sense that you have an academy award list of folks you could mention.

Dr. Jody Platt:

Yeah, it's also the nature of interdisciplinary work. You just need a lot of people to make it happen and I'm grateful to all of them.

Kelly Malcolm:

Well, thank you so much, Dr. Platt. This has been truly fascinating and I can't wait to hear what you learn about people's thoughts about AI. I mean, it's here and I think it's here to stay. So yeah, it's really curious. I'm really curious about it, and I hope that it'll be something that will make healthcare better for all.

Dr. Jody Platt:

Me too.

Jordan Goebig:

Yeah. Thank you so much. I'm looking forward to continuing to read about your work and some of the things that are up and coming. I really appreciate you taking the time today.

Dr. Jody Platt:

Thank you. Thanks for having me.

Kelly Malcolm:

The Fundamentals is produced by the Michigan Medicine Department of Communication in partnership with the University of Michigan Medical School. Find us and subscribe wherever you listen to podcasts.