Leaders in Public Health: Dr. Dena Bravata
Jan 9, 2025
Hi, everybody.
It's good to see you all.
Welcome to our November Leaders in Public Health series.
I'm absolutely delighted to welcome Dr. Dina Bravata
to our series today.
For those who don't know her,
Dr. Bravata is a healthcare entrepreneur,
executive, and scientist.
She's the co-founder of Lyra Health,
where she also served as chief medical officer.
She was the founding chief medical officer
and head of products at Cast Light Health,
twice, from 2008 to 14, and again, 2019 to 22.
And most recently, she served as co-founder
of Solara Health, a company that's using AI
for mental health and palliative care.
Dr. Bravata was a practicing internist for 16 years,
as well as an attending physician at Stanford University
and the Palo Alto VA.
And she's a nationally recognized
health services researcher.
Dr. Bravata has published widely
on the clinical effectiveness of a variety
of key health technology, clinical,
and public health topics,
and has served as an inspiring mentor
and force behind many of the health services researchers
that some of us recognize
as amongst the most influential in the field today.
I met her through her position
on the Aspen Health Strategy Group,
for which she's a founding member.
They meet on a yearly basis to take on big topics
of importance to the health of folks
across the United States, as well as across the globe.
She's been incredibly influential
in the way that they think about that work,
as well as in their yearly reports.
She's also served as an assistant public health officer
for Santa Clara County,
and currently serves on the board
of several health services, technology,
and services companies.
I will say that I frequently like
to highlight folks' connections,
and Dr. Bravata was a Yale undergrad.
So we are quite happy to welcome her back
to her old stomping grounds.
Her MD is from Columbia, so we share that in common,
and her MS is in health policy from Stanford.
She also has a twin sister
who directed the RWJ Clinical Scholars Program
for many, many years.
So has a lot of deep ties,
both to our community and to public health in general.
So with that, Dr. Bravata, Dina, welcome,
and truly thank you for coming back and joining us.
It's an absolute delight to get to spend time with you.
Thank you so much.
It has been so fun, even just before the conversation,
so spending the day here has been great so far.
Thank you.
Sure.
Well, I thought that we'd get started
with a question that I ask many
of our leaders in public health,
which is to share a bit about your journey, right?
So to situate people, some may have read up on your bio,
some may not have, but I think thinking about
kind of how you went from being a baby physician
and public health professional to your current world,
where you're working in the startup world,
health tech, health innovation.
Can you share a little bit
about kind of how you got from there to here?
Yeah, thank you.
So first, people often ask me like,
well, did you plan this?
And let me demystify that out of the gate.
So my journey has been one purely of great good fortune
and serendipity more than any planfulness.
So I think like many clinicians,
so I did the standard kind of academic gig,
you shared where I had my training.
And I loved being at Stanford.
So I was there for 16 years.
I was in the health policy, health economics group.
I did health services research.
And because I was more of a methodologist
as opposed to having like one clinical area,
I got to work on lots of different projects.
And I did a variety of technology assessments.
And you talked a little bit
about public health work that I did.
And I think that I just,
I am a primary care doc by training.
I'm interested in a lot of different clinical domains
and mental health.
We'll talk more about mental health later.
But like I care about cancer as much as I do
about mental health in lots of ways.
And when I was a junior faculty member,
I had the chance to kind of dabble
and do little side projects here and there.
And I didn't even know what it meant
to be a product manager for a technology company.
But I was asked to help develop a product feature
for an early innovation really that enabled patients
to provide information to their clinicians.
And so I helped them do that in a structured way
so that clinicians could make triage decisions
about that information.
And I kind of liked doing that.
And one of my mentors was Alan Garber who-
Now president of Harvard.
So, you know, she chose fairly well, yeah.
And Alan was friendly with a set of healthcare entrepreneurs
and they were starting Caslight at a time
when I was kind of at an inflection point
in my academic career.
And so I left the clinical practice of medicine,
which I love, and I left my academic world
and all of my grants and mentees
to be the founding chief medical officer for Caslight.
And so I haven't seen patients in a long time,
but that then kind of launched the subsequent journey
in healthcare entrepreneurship.
But still being very much rooted in academics
and in kind of thinking the way
that academics tend to think about innovation.
So it's inspiring and I think really helpful
to maybe some of our students who are like,
well, what's the next thing that I need to do?
And a takeaway often is the best thing that you need to do
is to be really good at your methods and at your work.
And as you shine, you'll find opportunities
that you may never have even imagined.
I mean, when you started your career,
who would have thought-
By no means.
Yeah.
We didn't have smartphones when our-
So a lot of your work has been in the mental health space
and you just alluded to this.
Lira Health in particular has been an incredible leader.
I wondered if you would mind speaking a little bit
about kind of what drew you to that work
and also share with the community
who may not be familiar a little bit
about Lira Brightline Caslight.
Yeah, so I think like most Americans statistically,
I have family members who have suffered
from serious mental illness.
And so I think that that has always been
in the background for me.
As a primary care doc,
you cannot avoid taking care of people's depression,
anxiety, sleep disorders, substance use disorders.
And so it was something that throughout my life
as a clinician was important to me.
I was in attending at the VA and I think there,
obviously there's a lot of very effective
mental health related treatment.
So I saw some best practices in play there.
When I was at Caslight,
so Caslight for those of you who don't know
was a really early digital health intervention.
It was launched commercially in 2010
and provided for, so it's a technology that was provided
through commercial health benefits.
So employees had access to this tool
through their health benefits
and it enabled them to see what services would cost them,
what services were made available to them
through their health benefits.
And I was the chief medical officer there
and head of products.
I also ran the analytics team.
And one of the things that I saw in our work
doing analytics was the unrecognized economic
and mortality burden of mental illness
for these employee populations.
So when I would talk to benefit leaders
of these innovative companies around the United States
that were providing Caslight to their populations,
I could see that there are populations with diabetes
who also had comorbid depression
or costing them four or five fold
what risk adjusted that same diabetic population
would be costing them.
And I think that that was not really
that well understood at the time.
And so I was eager as I was thinking about
what I would do after Caslight.
Caslight was a technology company at that point
without a services offering.
So I had two things that I wanted to do.
One was to do something that was at the intersection
of technology plus services.
And I was really motivated by this problem
that at that point was not really that well understood
for mental health,
especially for commercially insured populations.
And so that led to the founding of Lyra.
My co-founder, David Ebersman is a wonderful man
and he was on the board of Caslight.
And so we started Lyra in January of 2015.
And so we're coming up on the 10 year anniversary.
And in full disclosure, I have not been involved
in the day-to-day of Lyra for many years.
It's in his very capable hands.
And Lyra serves the same population.
So commercially insured populations,
it's a benefit provided
through their employee sponsored benefits.
And it provides a full range of mental health services
from digital interventions.
So digital cognitive behavioral psychotherapy
and other related tools, telehealth related.
So people can get psychotherapy
or see a provider virtually
or they can get matched to a provider in person.
I love it.
As you talk about this kind of this journey,
one of the things that strikes me
is kind of the way that we often think about innovation
and care delivery within academia
is a little different perhaps
from how you've learned to think about it
in the startup and now no longer startup,
but actually quite stable company space.
How has your thinking about kind of innovation
and care delivery shifted over this past decade plus
of working outside of academia in actual service provision?
Yeah.
So I think that one of the things
that is most notable about sort of both of these environments
is that fundamentally patients and providers, right?
Like that dyad is where the magic happens.
And I don't think it really matters that much
if that's happening in a large academic hospital
or if it's happening virtually.
Like ultimately when it's services
between patients and providers,
I think that it is relatively similar.
That said, I think that in a tech enabled world, right?
We have the opportunity to implement services,
to use technology, to really evaluate those
in much, much more rapid fire.
I think that we have the opportunity to recruit clinicians
who want to be part of innovating their clinical domain.
And so there's this spirit of let's do our best work,
let's evaluate that, let's iterate on that
in a way that I think is certainly like the mindset
in large academic medical centers,
but just may not move it quite the same.
NIH grants don't always facilitate that type of pace.
Or that type of real flexibility.
And may lack the fundamental infrastructure
for doing that ongoing quality assessment.
Well, and I think that's actually
one of the really neat things
about the work that you've done.
So I will say in full disclosure,
I hired one of Dina's former PhD students at Brown
as one of my last hires before I left,
not knowing that he had been her PhD student
until we were chatting.
But I think that one of the neat things
about what you've done is that you've kind of merged
that all the strengths of industry
with also some of the rigor
and evaluative capacity of academia.
Maybe you can talk a little bit about that,
particularly for some of the students
and or faculty in the audience
who may be interested in those types of partnerships.
Yes, especially as we're looking at a world
where NIH funding may be a little more uncertain
and we're all thinking about kind of
how do we continue to do great impactful work?
I think that you've set up some really neat examples
of doing impactful work through partnership.
Yeah, so the gentleman to whom Megan is referring
is Chris Whaley.
He is a PhD health economist now at Brown.
And Chris, I think, is a perfect example of this.
So at Caslight, we had been around for about three years
and so we had developed a sufficiently large,
in this case, claims repository
to really be able to do some innovative analytic work.
Chris was getting his PhD at Berkeley at the time
and he used that data set to really begin to understand
if you have a benefit design that is there
to incentivize particular health utilization behaviors
and that is coupled with a technology
that enables people to understand the choices
that are available to them.
He asked questions about how do those two worlds interact
and what are the health economic outcomes
of having benefit design plus transparency tools.
Chris then went on, so he got a PhD on that.
He then went on to RAND and we, by being friends,
continued to collaborate as we have
for the last decade plus.
And so I think that our relationship
is just this perfect example of people in the private sector
have access to amazing data sets about, be they claims
or what are the kinds of health services and technologies
that patients are interacting with?
What ones are they not interacting with?
How do we think about the effects on health
or outcomes or mortality or whatever
the outcome of interest may be?
And many commercial entities don't have
like a dorky academic co-founder
and whack somebody with the methodologic experience
to be able to rigorously evaluate
those products and services.
And the buyers of these technologies and services
are becoming increasingly sophisticated, right?
And so, can they tell the difference
between a propensity match cohort
that somebody did a nice job on
versus not, probably not, right?
I don't know that I could these days anymore either,
but there's a greater and greater obligation
on the part of private sector companies
that are delivering health services
to demonstrate in peer-reviewed
or other methodologically rigorous ways,
does their product work as they say?
For whom does it work?
For whom doesn't it?
And part of that is really then accretive to their brand
by saying, look, we are thought leaders
in whatever the clinical domain may be.
If we, for example, had faculty or students
who were interested in that type of partnership,
how does one go about it?
Yeah, I think like this is a perfect forum for that.
One of the reasons that we have people like you here,
but it's, yeah.
Yeah, so I think, so there are research,
so if someone is interested
in dipping their toe in the water,
but they don't really have a bunch of pals
who are in the startup world,
I would say that some of the clinical research organizations
whose job it is to actually do this work for hire
are great places to do that.
But to the extent that you do have this network
of, in this case, Yale alums or so
who are out in the real world,
I think just having these kinds of conversations,
so I work with many companies,
and one of the things that I do with them
is to help them design their evaluations.
And sometimes those evaluations
are not for public consumption.
They're for them to understand
how well are their tools and services working?
Are they living up to,
are they fulfilling their promise just to themselves?
And then, obviously, there's the marketing benefit
that can accrue as well.
And I'll say there, for those in the audience
or on Zoom who are interested,
if this is something that you're interested in,
certainly reach out to me.
We have a whole office through Yale Ventures
that helps with these corporate partnerships
to structure them in a way
that you know when you're going into it,
whether or not it's for pay or for experience,
to know whether or not you are allowed
to say a priori, yes, I'm going to publish, no, I'm not.
Is it that you're doing kind of,
you have a contract to do some stuff
on behalf of the company that's gonna stay private?
Or is it a real intellectual partnership
and that you're co-creating together?
Which I think is, to set out those terms
from the get-go is really helpful.
That's right, and there are SBIR grants.
There are other, there are like granting mechanisms
that can be, that are a little bit more
academic-y leaning, right?
So where as students, as faculty members,
you are co-investigators with the product folks
on the other side.
And I have worked on some of those grants
that have just gone like beautifully,
and some that did not go quite as well.
And I think the thing that characterizes the best of these
is a very cleared understanding of
what exactly are we gonna evaluate,
and depending upon if the outcomes are good,
bad, or indifferent with respect to showing
the product or service offering in the best light,
what will we do with respect to publication?
Because if you have a junior faculty member,
they need those publications for their tenure track
and the like, and yet, for marketing purposes,
that could be devastating, especially for a young company.
I think also the timelines need to be very well understood
so that if you're getting a grant,
and it's over a year or two years or whatever,
and the company in the meantime has like pivoted
and gone off in some totally different direction,
that can be like really not so great.
Yeah, that can be, and the timelines just between,
having been part of a couple of SBIRs and STTRs,
the timeline of the startup, right,
is often very different from my timeline.
I wanna have the perfect study design,
and they're like, I have nine months, right,
like the phase one grants are super, super short,
and so thinking about how we can
maybe do approximate truth
rather than approaching perfection,
which I know is a big, hot debate in the world right now
about kind of how perfect do you have to be
in the data that you're collecting
and in the estimates that you're creating.
So that leads me into a different,
there are kind of two other things,
big things that I wanna get to
before I open it up to the group,
and I'm gonna set them both out.
One is talking about quantitative science today.
The other is thinking about the future of public health,
and I know you sit on a bunch of boards
and have some, including some Buddhist medicine work,
and so have some interesting thoughts there.
So let me take the first one,
but I wanna just highlight for those in the audience
that I will get to the second as well.
So I'm gonna kind of disclose vulnerability here.
So I had lunch with Dina a little over a year ago,
and I had not yet started this job,
and she said to me, you know,
one, as I'm talking about kind of what schools
of public health should be focusing on,
she's like, you should not be graduating anyone
who doesn't have a strong,
quantitative data science skillset.
And if she could, you know, anything,
as you're looking at kind of school of public health
graduates, that is something that is distinguishing,
and as a top school, we should have that.
And I kind of took that into my brain
and thought I really need to talk to her more
about what she means, but I don't have a plan yet,
and I don't know the school well enough yet
to be able to come back.
So it took me a year to go back to you,
but that comment stuck with me.
And as we've been starting to do our work in our school
around our strategic plan, right,
we have our priority around kind of data science
and AI and data equity.
It's something that I would love now
that I'm in a slightly different space
for you to share some of your thoughts
about what you meant by that
and where you see the future
of public health data science going
before I get to that topical future of public health,
but kind of within the data science world,
kind of what you're seeing as the rapid changing landscape,
but also what you meant by a rigorous quantity,
like what would you want our students
and or some of us as faculty
who may wish we could go back and take some courses,
what do you see as kind of those areas
that we need to lean into?
Well, first of all, I mean, like,
if anybody has done an awesome job in a year, right,
like, I mean, seriously,
I could not be more impressed and delighted.
And like, proud as an alum to, you know,
be affiliated with all that work that you've done.
Thank you, Dina.
What I meant by that at the time was that I had,
so I did my postdoctoral training at Stanford, as you heard,
and I was in a very methodologically rigorous environment,
right, so I'm not a health economist,
but I was surrounded by them.
The health-
You had Jay Bhattacharya down the hall, yeah.
And a lot of the work that I did was in meta-analysis,
systematic reviews, regression-based prediction modeling,
and that kind of thing.
And I had a number of colleagues,
very, very smart colleagues,
thoughtful people who were public health graduates,
not of this institution.
And what I noticed about the distinction
between our training was that they were great
at thinking about populations
and setting policies in nuanced ways
that would be appropriate for, you know,
this or that clinical situation.
But what I came out of my training with
was like very pragmatic, useful skills
that I could then apply in a set of situations
to then go on to inform,
be it a policy or a product design or that kind of thing.
And I saw, like at least for me,
how very useful being able to do a regression
or do some data science was for my own career,
and so that was a lot of the motivator.
I think also, you know, when I looked at hiring people,
right, so the very first people that I hired
in my first role outside of academic
were people who were trained like me.
And, you know, and I think that that's sort of like
a human foible, maybe, anyway.
But, you know, what I found was that clinicians
who were health services researchers,
somehow, like they had this data-driven approach
that enabled them to work better with the engineers
than other clinicians who were a little,
perhaps, more risk-averse and couldn't use the data
to help shape their intuition about product design
in quite as facile a way, somehow.
And so, again, I saw that there,
and, you know, that's, I think,
part of the answer to your question,
like in terms of what I look for when I'm hiring people.
I think, by and large,
especially on like the clinical product strategy side,
we have data, we should use it
to inform those kinds of decisions, yeah.
And then you and I, relatedly,
have had a lot of discussions about AI
and the challenges of teaching in kind of this world
where AI is both advancing faster
than we can create a syllabus,
but also changing the way to the students in the audience,
right, the way that you study, the way that you learn,
the way that you do projects.
And we've had some, you have a colleague
who recently went back to get an MS in AI and ethics?
Yes.
Yeah, so, and from your side, from the business side,
I'm sure there's no way you're not having discussion.
So I'd love your thoughts
about kind of what we need to be learning
about AI or teaching and learning.
Those of us that are faculty are learning sometimes
along with the students.
Yeah, so I would say, you know,
outside of the academic environment,
so among innovators, among entrepreneurs,
I think among venture capitals,
private equity firms that are making investments
in technology, the intersection of AI and healthcare
is absolutely the greatest area of buzz.
I'd love to talk maybe a little bit about GLP-1s
and obesity care.
Oh, sure, we can do that too.
As well, but like, I would say that those are probably
like the two sexiest, most interesting topics right now.
Along with palliative care.
Yeah.
But yes, go ahead.
And so you mentioned Solara,
which is a startup that's working in the space
of using AI to help deliver protocolized,
so highly evidence-based psychotherapy for populations
who don't have access to it today.
And I think it's pretty easy for us to imagine, right?
Computers, we've had digital versions of CBT
for two decades.
Right, so all right, let's add a little flavor of AI
to that, that can make it a little more personalized,
perhaps a little more empathic.
But I think the magic actually comes beyond that.
So the magic actually can come in and actually,
like now, having the technology summarize clinical notes
so that when you see this patient for the next time
or you're a new provider who's never seen this patient
before, you would never go through
and read all those therapy notes.
And here is a nice summary of where they are
and what kinds of very personalized bits of information
might be most motivating for that patient.
I think some other uses for it that are very compelling
are in training and recruiting and training.
So let's say you're the chair of the Department of Psych
and you want to hire somebody
who's good at doing psychotherapy, right?
Like, let's see it do it.
And not with a vulnerable patient, but with an AI patient.
And let's see if they say anything
that might raise a red flag,
or perhaps they're good at it, but not that efficient.
It took them like a really long time to get to it.
So maybe you can have assessments
along whatever criteria are important to you.
From a quality point of view,
and we talked about this a little bit
with some folks this morning.
So as somebody who has been responsible for clinical teams,
oh, whoops.
It's a little foam.
Yeah, it's on your, here we go.
Okay, whoops.
Yeah, it's fine.
The thing's still on.
We're good.
From a quality assurance point of view.
So at many times in my past life,
I have had to listen to interactions
between nurses or doctors and patients,
or had to do chart review or that kind of thing.
And then usually, you know,
it's the sort of exhausting, cringe-worthy experience
where you then have to think about
a way to thoughtfully provide feedback to your colleague
in a way that's going to resonate with them
so that they can do a better job next time.
And, you know, and even if you do it every week,
like you're supposed to, you know,
maybe you can do one or 2% of all patient encounters,
you know, which always just leaves me like sleepless
at the thought of like, well, what am I not seeing?
What am I not hearing about?
And now these tools can do quality assurance
on 100% of patient interactions.
They can do so with objective criteria
and provide feedback to the teams as a team in a useful way
and to individuals in a useful way
with very specific learning for those individuals.
I really think that the quality lift
that we should expect that of these technologies.
Any of us that develop interventions
and do intervention trials
can imagine that fidelity component of the intervention
just gets transformed.
And then from an implementation science perspective too,
you think about, right, like how many studies do we have?
And I'll look at Susan Bush here,
but like how many studies do we have
showing that provision of care
does not follow quality guidelines?
What if you could use this?
But then there are equity concerns on the other side, right?
So like the training database is not representative.
What if those objective criteria are deeply biased?
And you've been doing some thinking about that as well.
Yeah, I mean, and I look to your colleagues here
as folks who will advance this.
I think that I at least applaud the efforts
on the part of many machine learning experts
to recognize that if their training sets are biased,
so too will their outcomes be.
And I think this is really important
when we think about this concept
of looking for digital twins.
I don't, so just, so for some of these AI-based tools
that are really trying to predict future events,
so to predict future high-cost events
or preventable ED visits or hospitalizations,
one of the methods by which they do that
is to look at a large collection of related patients,
pick a digital twin of the patient
that we're trying to understand what their future,
what the future might hold for them.
And if that large collection,
if the training set doesn't include vulnerable populations,
low-income populations of color and so on,
we're never going to be able to do an accurate job
in our predictions.
And so what, so there are then, I think,
very interesting and creative computer science techniques
for improving that, which are well beyond my pay grade.
But which I'm grateful that we have some faculty here
that are working on, which is super cool.
Okay, so you mentioned GLP-1s,
and then we've also talked about,
so there's this real tension,
and we talked about this frequently, right,
between GLP-1s, which are magical in many ways,
but then also the structural factors.
And particularly, we were talking last night at dinner
about kind of looking forward,
having some concerns about states
that are eliminating free lunches for kids
during the summer, for example.
So kind of how are you thinking about GLP-1s?
You said you think they're magical,
but then how do you balance that
with kind of that attention to the structural
or kind of more classic population
or public health side of things?
Yeah, so for me, I think that GLP-1s
are one of the most important health innovations
of my career, and they are also incredibly costly
and have low availability to some
of the most vulnerable populations who need them.
I think critically, I have spent most
of my kind of post-academic career
thinking about commercially insured populations.
And so here, every employer in America covers GLP-1s.
And for those of you, right, just to be clear,
these are the Ozempic-like drugs
that have been FDA approved for diabetes
for a very long time.
So every company covers them for that, by and large.
And so the question now is,
now that there have been expansions
for these meds for weight loss,
and if you estimate that 40% of the adult population
in the United States is overweight or obese
and may be a candidate for them,
this could just wipe out the expenditures
on health benefits, like no more musculoskeletal care,
no more cancer care, right?
We're just gonna spend all of our dollars on this,
which some people sort of think is okay
because if we can eradicate the obesity epidemic,
that then cancer rates will fall as will many other issues.
I'm not so sure I would go that far.
But nevertheless, I do think that it is very important
to think about ways that we can increase access
to these really important services.
And so we are at a crossroads
when it comes to commercial entities thinking about this.
So among small employers,
so those are employers with less than 1,000 employees,
about 5% cover GLP-1s for weight loss in this calendar year.
I think that number is estimated to go up.
I see estimations between sort of 12 and 25%.
So we'll see.
I don't think it'll be anywhere near 25%.
And there are also concerns that if folks,
if employers choose to cover compounded versions,
so not the branded versions of these drugs,
that they may get into trouble,
that the rebates that they get from drug companies
may go away.
So I think there is a lot of complexity to this.
These compounded drugs,
especially when they're made thoughtfully
in compounding pharmacies
that are compounding chemotherapeutic agents,
and these are excellent facilities,
that offers an affordable option to large populations
that otherwise would not have access to these drugs.
So the reason I think it's so interesting
is because of the powerful effectiveness of these.
Just the huge size of the populations, no pun intended,
who might benefit from them,
and the enormous cost-effective kind of concerns.
And the short-term versus long-term cost-effectiveness,
and then all these other things that now they're saying,
well, GLP-1s may also be effective for treating addiction.
So some really interesting questions,
and as we go back to that mental health space.
But for my final question
before I open it up to the audience,
as you think kind of about the future
of places like this school,
are there things that you see on the horizon?
You sit on a bunch of boards,
in addition to being part-time often,
well, you live in Wyoming,
but so then part-time in the Bay Area.
As you kind of work in this mix of people
that are thinking about the future,
what is the kind of space of the future
that you would say to us is five years ahead,
or things that others aren't thinking about yet
that you would want to see us lean into?
Well, I think as I reviewed the strategic vision
that you set for the school,
I think you have done a beautiful job.
I don't really like that.
I didn't plant that, by the way, at all.
So that was not, yeah, thank you.
I mean that with all sincerity.
And I really don't like this term future-proofing.
I think that there is no such thing.
But you have been thoughtful
about training the future generation
to think independently,
to use the best tools that are available,
recognizing that those tools are advancing at a pace
that outstrips our historical ways
of even creating a curriculum.
I would be so daunted if I sat in the shoes
of people who are teaching today
and thought about, okay,
what do I want to teach 12 weeks from now?
It's like, oh my God,
the large language models will have evolved exponentially
by then, so maybe I'll just leave
that part of the syllabus open.
Our education committee will love that.
But I do think that it is about these ideas
of ethics, considerations around diversity
and vulnerable populations,
keeping our eye out for those.
There's been a technological divide forever,
and I can't imagine that these newer technologies
are not just gonna exacerbate that.
Is that just gonna be one more public health problem,
or how can we all make sure
that that is less likely to happen?
Thank you.
As I look around the audience,
and I know some of you that are online,
I know that we have many folks
that are stepping into that space,
and I'm excited to report back to you
as we continue to make progress.
All right, I will open it up for questions
both in person and on Zoom.
As always, we will pay attention to the chat
in addition to welcoming questions
from the in-person audience.
And if you don't mind just introducing yourself
and your role in the school
to give Dina a sense of who she's speaking with.
We'll do a trade-off.
Technology, our best friend and our worst enemy.
Does this work?
Yeah.
My name is Shelby.
I'm one of the data scientists under Dr. McCurgey.
We're a new team doing data science and data equity.
So while I was doing my master's at Johns Hopkins,
I did a consulting project
with med track students in neurobiology.
And we did such a project on mental health.
Like it was a company that was seeking
to do extra clinical mental health diagnoses.
And in premise, the idea of being able to like create
an AI that was able to do that
was technology that's been around for decades,
really the ability to do such things.
But where we ran into difficulties
was with the ethics of it.
And you mentioned doing like mental health
and it sounds like it's within the offices of like a hospital
and under like direct supervision of a physician.
But I was wondering how you navigate those challenges,
especially as people outside of the clinic
are looking to do some of these things.
Yeah.
So Shelby, like that is I think the fundamental question
that I bring to my thinking about Solara in this case.
And so we were talking last night at dinner.
You know, you hear about in the early days of CHAT-GPT,
for example, someone expressing suicidal ideation
and may pose a question along the lines of like,
I'm thinking of hanging myself, you know,
what do I do?
And actually it provides guidance, right?
The opposite of what you would want it to do, right?
You know, perhaps like direct someone
to a national cancer suicide hotline
or something like that.
Instead it like gives you, you know,
you can even provide diagrams for how best to do that.
So that is, right, so obviously in a modern approach,
right, it's all about setting up a set of constraints
so that that could almost like you try to have
the probability of that ever happening go to zero.
And I think that that is around this novel science
of prompt engineering and really around thinking about
what is the corpus of information
upon which these large language models are being trained.
But I think that anybody who is in the world
of trying to develop these technologies, right,
like that's the worst thing.
And you mentioned last night that was it 9 million people
are using GPT right now or?
No, we were talking about a different.
Was it Replica?
That was Replica.
Okay.
Yeah, and so Replica for those of you
who may not be familiar is a personal companion
where people have personal companions for many reasons
and sometimes for emotional comfort.
And so this is a company that has that technology.
And I think that they, I was surprised
as I began to learn about this world
about the prevalence of the use of just
sort of non-healthcare, non-academic types of interventions
that are being extremely widely used
by people who have loneliness, low grade,
depression, anxiety, and so on.
And I think that's part of the ethics question.
It's also, right, is this better,
is this more harmful than nothing, right?
And so how do you start to like tease out
those risks versus benefits?
And I think I'll say risks versus benefits
is a core question for us as a field in everything
as we saw with COVID response, right?
Like thinking about and actually quantifying
what are the potential harms
as well as what's the potential benefit
and how do you put your finger on the scale?
And there's costs, there's personal,
there's individual outcomes, there's societal outcomes,
there may be outcomes that are not health related as well.
And I think it's a deep area.
For those who are interested,
we do have a digital ethics center here at Yale.
One of our grads is actually serving
as one of their inaugural staff.
And so it's a good place to work with
and obviously also have bioethicists on our staff,
on our faculty, so.
We have a question in the back.
Yeah, so to continue in the vein
of what you were just discussing
and maybe getting you to speculate a little.
Introduce yourself, please, Jason.
I'm Jason Hockenberry, professor of health policy.
And I do work in behavioral health policy.
How far away do you think we are from CBT,
so cognitive behavioral therapy,
like well-validated therapy being delivered by AI?
Oh, it's, we are there.
It's here, yeah.
So I mean, I think the, yeah, I mean,
I think the, you know, maybe the,
from a regulatory perspective, right,
are there digital CBT, right?
So there are digital CBT tools
that are FDA-approved for standalone psychotherapy.
And so, and they've been around for a long time.
I think like perhaps sort of the second level question
that you're asking is, well, how about today, right?
Will we let these AI-based tools loose on populations
without any human in the loop?
And so the tool, I would love for you to play
with the Solara tool.
There are others, right?
They're extremely good.
They are true to the protocol in empathic ways.
Patients can interact with them in the middle of the night
or during the day.
They can have, you know, like long, long sessions
and short sessions, things that would never be feasible,
even with some of the like best asynchronous models.
So, but the question will be,
when will they be FDA-approved for delivery
and for reimbursement without a human in the loop?
And I would say that we're extremely close to that.
And there, I will say, I'll just put in a plug.
We have Michelle Tarver,
who's the director of the FDA Center
on Digital Health and Diagnostics,
who's gonna be joining us in February, I think,
for one of these lectures.
So ask the question, yeah.
We have some good Zoom questions.
So I'm gonna jump into a Zoom one.
Before I do Zoom, and then we'll,
excuse me, I saw your hand, so.
This is from Christine, who's a visiting PhD student.
She asks, how do you look at the balance
between using more resources on giving people access
to weight-reducing medication
versus using the resources on public health interventions
which prevent the development of overweight
and obese conditions?
There's some large ethical issues
about giving medication to up to 50% of the population.
Yeah, I mean, so I think that this question
about prevention versus treatment
is fundamental to our entire healthcare system.
So I would say we don't have infinite resources.
We need to do both.
But there are people who are having the deleterious effects
of being overweight or obese today,
and it's incumbent upon us to do something for them
every bit as much as we should do prevention.
And you work on some, you sit on,
you work on food insecurity as well,
which I just pointed, like as we think about obesity,
right, but it's about food insecurity
as much as it is about overweight.
It's people who are food insecure
are more likely to eat ultra-processed foods
because they are cheaper, yeah.
Yeah, I work with a Connecticut-based company, Nourished Rx,
which is in the food is medicine space
and is really trying to deliver.
So they take care of patients
who are at the intersection of food insecurity
and some medical condition
that requires them to be on a healthy diet.
So diabetes, hypertension, high-risk pregnancy, obesity,
and I think it's very interesting to see the advances
in the payment models for these kinds of interventions.
And I think it's a little bit of a misnomer.
One of the things that I've really learned
from the Nourished Rx team is that,
you know, giving people food is a great thing.
I like to feed my friends and family,
but that's really just kind of a hook
to getting this population,
this vulnerable population to engage,
to learn about healthier foods,
to learn about where to access them,
to learn more about nutrition science,
what's appropriate for them
and culturally relevant for their families.
And so this isn't just about, like,
giving people food for the rest of their life.
And on my list as a post-follow-up is to connect you.
We have a team in our Office of Public Health Practice
that does a fair amount of work
in the state of Connecticut around food is medicine.
It'd be fun to connect you guys.
So I'll do that after.
Yes, Susan.
Hi, I'm Susan Bush,
and I teach in the Department of Health Policy and Management
and I've done a lot of work on behavioral health.
That's one thing I'm very interested in.
So one thing I'm really interested in
is workforce shortages in mental health.
As you know, this is a huge problem,
and I've really been engaged in thinking,
like, how can we address it?
It's a really, really tough problem.
So I've attended a lot of webinars,
say, in the past six months,
National Academies, National Council,
just to hear different people's ideas.
And there really aren't that many really, like,
ideas that are really gonna take us very far
on this website.
So I was on these webinars.
I was on a webinar just yesterday,
and the big idea being floated,
I think this was National Academies, actually,
was let's require Medicare and Medicaid
to pay for psychotherapy at cost.
And that's a proposal, apparently, right now.
To me, that's like a non-starter.
You know, if you talk about that,
I just don't think that's gonna happen,
and it's strictly not gonna happen,
you know, with the new administration.
That's, in my opinion, a good thing.
And I noticed that there were at least two things
that you mentioned that could really address
the workforce shortage,
and I'll just call everything you just said about AI,
like improving the quality of psychotherapy.
Potentially, this could be, you know,
shorter duration for people when they come to psychotherapy,
and that could potentially relieve some pressure
from the workforce.
Also, with LERA, I did take the opportunity
to go on the website.
There's a really nice study that suggests
if you have an EAP that works really well,
very robust EAP, LERA,
you can reduce your mental health cost,
and I assume a lot of that is through labor,
so that that would also potentially
relieve some of the pressure,
even though I understand the EAP does have
some mental health workforce within that intervention.
So, why, can you talk a little bit about
what this disconnect is?
I'm sitting here so excited about what you're saying,
and I go on the webinars, and I'm not excited.
Like, they're not talking to each other,
and I will say, if you can talk to industry folks,
and like I said, I remember I sat on a call
with ABHW, you know, the sort of lobbying group
for the behavioral health insurers.
You know, have you thought about,
in terms of provider networks,
we're talking about mental health provider networks,
and how you can have a more adequate network,
have you thought about maybe EAP,
like, really developing your EAP as a way to do this?
And, like, people look at me like I was crazy,
like, they had never even,
it's almost like this had never occurred to them.
If you could speak to that, I'd appreciate it.
Yeah, you're asking just an amazing question,
and I think that the answer is sort of yes,
all of the above, right?
So, I think, you know, I am a primary care doc.
Primary care docs have a real role to play
in increasing access to mental health services, right?
I was not at all trained, I was never trained
in motivational interviewing, or CBT, or anything like that.
My poor patients suffered as a result,
and, you know, it would have been good
if I had gotten a little bit of that along the way.
Maybe a tool to help primary care docs like me
begin to be able to deliver a bit more
psychotherapeutic interventions inside of primary care,
I think could be useful,
but that has to come with reimbursement as well.
I think that AI, right, like, this is the reason for Solara,
right, and for other companies like that, right?
We have got to address this access problem.
Can technology do it?
I think in some cases it can.
Can technology support, like, lower levels
of trained individuals to deliver this
in safe and effective ways?
I absolutely think that it can as well,
but, you know, EAPs, and, like, I think of those
as different ways of reimbursing, right,
like classical, you know, sort of delivery mechanisms,
and so we need to have innovation
in who is paying for what.
Collaborative care is an intervention
that some of you may not be familiar with,
which is basically where you have
expert mental health providers
over doing sort of population level mental health,
and then providing guidance to often master's level
social workers, psychotherapists,
in some cases primary care providers,
to actually deliver the intervention,
and it has been studied in, you know,
well over 100 randomized controlled trials
published in very fancy journals
that demonstrated effectiveness
for lots and lots of populations,
and yet it's almost always only been deployed in,
you know, kind of- Low income.
Yeah, federally qualified health centers,
or in, you know, European countries
that have socialized healthcare systems,
and so, and when you ask why,
often the answer is because of reimbursement, right,
because we have not incentivized,
in a fee-for-service environment,
those providers to actually take that guidance,
you know, and especially, right,
like if, you know, back to me as a primary care doc,
if someone said, well, you know, Mr. Smith,
for whom you take, you know, enormous care
about his diabetes and hypertension, right,
like what he really needs from you is a little Seroquel,
or some other fancy drug that, you know,
an atypical antipsychotic that like, you know,
like I really don't feel comfortable writing for that,
if you feel so comfortable writing for that,
why don't you feel free to do that,
but that's, of course, not how that kind of model works,
and so, so I think, like I would love to see
regulatory and financial mechanisms
that would incentivize more of that kind of program
that would increase access, I think enormously.
And I think, I know we're getting,
we have a couple of hands, I just,
just to kind of add on to that,
that I think that this is an incredible opportunity for us
as a school, as we think through,
kind of as we're thinking through
our new federal administration,
and where there may be opportunities
to do good for populations,
I think there is going to be a willingness
to look at models of payment and of care delivery
that are different.
And so, thinking about kind of what sort of alliances
we can build with people like Dina,
with kind of behavioral health, right,
how can we create those alliances now
so that we're ready to propose things
that have data behind them,
and could be transformative for the American public, so.
We had a question back there,
and Mike and Joan, and that'll probably be our last,
just for the sake of time.
We'll do the lightning round now,
for the last couple minutes.
All right, no pressure.
Mike Hansberger, I'm the Director of Academic Affairs,
and I spend a lot of time thinking about
how we can make our curriculum more authentic,
preparing our students for the workforce
that they're gonna be very soon thrust into.
And so, you mentioned quantitative skills
as something we should really kind of hone in on,
I heard you.
I'm assuming just learning the mechanics
of a regression is not sufficient.
So, what are the missing pieces
between learning the mechanics of the quantitative skills,
and using them effectively in the workforce
to kind of be the foundation
of data-driven decision-making?
Yeah, it's a great question,
and I think that it depends on the person,
and the role that they're going to fulfill.
So, shockingly, I still do, you know,
analytic work from time to time.
Really?
Yeah, oddly enough. That's awesome.
And partially because I love it,
and that kind of thing,
but, and partially because I'll sometimes work
with a company that, like, you know,
my rusty skills are better than, like,
whatever else might be available in-house.
But by and large, for somebody like me,
like, I should not be, you know,
like, doing that kind of work,
but I should be able to understand it well enough
to ask the analyst, right, like, you know,
why did you choose those sets of variables,
or did you consider this, or that kind of thing?
And I think that for someone else
who's actually the data scientist, right,
like, the very best of those folks with whom I work,
usually, right, like, I'll ask for, like,
oh, okay, let's look at, let's try to predict this,
or understand the relationship
between these variables, or whatever.
Then they'll do whatever I suggested,
but then they'll come back, usually,
and say, like, well, yeah, yeah, yeah, so I did that,
but I did this, too, this thing that you didn't ask for,
which is obviously, like, vastly clever,
and leads to the most interesting and relevant outcome.
And so it's for that person,
they really have to understand the method,
but then also the clinical domain
in which they are working.
And so I think the answer is really
very much a role-dependent one.
But I think you have to have done it enough, right,
like, to have the ability to ask questions,
because otherwise, these technologies are moving so quickly
that it's easy to get intimidated and be like,
oh, okay, well, there are, like,
a lot of numbers on this page,
and, you know, I'm not so sure
that I'm gonna question you.
You seem like a smart person.
Well, with that, and I know that I promised Joan
the last question, but I am conscious of the time,
and many people have to get to meetings and class.
I would love to give a big round of applause
to Dr. Bravada, I am so grateful.
Thank you all.
And Yale is proud of you, you know?
And in good Yale fashion, we do have swag,
because, you know, you can't come here without it.
She'll be sticking around for a little bit,
feel free to ask questions,
and look forward to partnerships for years to come.
So thank you, Dina.
Thank you all.
Thank you.

