EPISODES

image of Dr. Charles P. Mouton and Dr. Peter McCaffrey

Exploring AI’s role in health care with Dr. Peter McCaffrey

MP3

  • This episode of The Pulse with Dr. McCaffrey highlights UTMB’s pioneering role in integrating artificial intelligence into health care and research. Dr. Charles P. Mouton, executive vice president, provost and dean of the John Sealy School of Medicine, introduces Dr. McCaffrey, UTMB’s Chief AI Officer, who elaborates on the new AI Council’s mission to harmonize AI efforts across UTMB. The council ensures each department's needs are met while focusing on the rapid, safe and aligned deployment of AI initiatives.

    Dr. McCaffrey explains how AI is transforming patient care, particularly in emergency radiology and pathology. Using AI, UTMB’s care teams can identify life-threatening conditions faster, streamline workflows and enhance diagnostic accuracy. The integration of AI extends to screening processes that detect underlying conditions, adding a crucial layer of preventative care. These advancements enrich clinician-patient interactions by reducing administrative tasks, enabling clinicians to focus on direct patient care.

    In research, AI facilitates patient-trial matching by structuring vast amounts of unstructured data within medical records. This innovation allows for quicker, more precise recruitment of eligible participants in studies, strengthening UTMB’s research capabilities.

    Addressing potential concerns, Dr. McCaffrey reassures listeners that AI at UTMB is a tool that supports, rather than replaces, clinician decision-making. Safety, transparency and ethical considerations remain at the forefront, affirming UTMB’s commitment to high-quality, patient-centered care.

  • 1
    00:00:00,840 --> 00:00:09,600
    Welcome to The Pulse: UTMB Health's podcast, your gateway to the latest insights, ideas and innovations shaping our community and beyond.

    2
    00:00:10,230 --> 00:00:16,530
    Stay tuned for engaging discussions, compelling stories, and the information you need to stay informed and connected.

    3
    00:00:16,830 --> 00:00:22,470
    Right here, on The Pulse. Welcome.

    4
    00:00:23,410 --> 00:00:31,210
    This time on The Pulse, I'm Dr. Charles Mouton, executive vice president, provost and dean of the John Sealy School of Medicine.

    5
    00:00:32,200 --> 00:00:38,530
    As we continue to bring you compelling conversations about education, research, health care, and innovation,

    6
    00:00:38,890 --> 00:00:46,090
    you'll occasionally hear from various hosts, each bringing their own unique perspective and expertise to the conversation.

    7
    00:00:46,870 --> 00:00:56,920
    Together, we'll dive into topics discussing the future of medicine and how that will shape the impact here at UTMB, both locally and globally.

    8
    00:00:57,670 --> 00:01:05,620
    Today, I'm joined by Dr. Peter McCaffrey, Chief Artificial Intelligence Officer at UTMB.

    9
    00:01:06,520 --> 00:01:09,970
    Dr. McCaffrey, welcome to The Pulse. Thank you for having me.

    10
    00:01:11,020 --> 00:01:20,140
    Can you spend a few minutes talking about the new artificial intelligence, aka AI Council, and what it will do at UTMB,

    11
    00:01:20,170 --> 00:01:27,340
    and a little bit more about how you see your role as the new and our first Chief AI Officer.

    12
    00:01:27,460 --> 00:01:29,710
    Yeah, absolutely. Thank you. Uh, so really,

    13
    00:01:29,710 --> 00:01:38,680
    the AI Council exists to let us be efficient and aligned while not ignoring the needs that people have in certain areas of the enterprise,

    14
    00:01:38,860 --> 00:01:42,700
    and being mindful of what our workforce wants and our various members want.

    15
    00:01:42,730 --> 00:01:47,470
    So really, it's our representation across the enterprise of stakeholders, really in leadership,

    16
    00:01:47,980 --> 00:01:54,010
    uh, stretching from academic to operational to clinical areas, knowing what they want out of AI,

    17
    00:01:54,010 --> 00:01:57,730
    what they feel like they need in their departments and their areas out of AI,

    18
    00:01:58,390 --> 00:02:05,080
    so that when we pick initiatives to dive into on an annual and quarterly basis, we all know what we're collectively trying to do.

    19
    00:02:05,080 --> 00:02:08,620
    I think we recognize we want to be fast. We want to be efficient.

    20
    00:02:08,650 --> 00:02:14,379
    We want to be up-to-date, but we really don't want to be irresponsibly so.

    21
    00:02:14,380 --> 00:02:18,490
    So, it's the way to do it with safety and with awareness. That's great.

    22
    00:02:19,120 --> 00:02:25,720
    Can you discuss how UTMB is using artificial intelligence in terms of providing health care to our patients?

    23
    00:02:25,900 --> 00:02:31,629
    Yeah, absolutely. So UTMB is quite innovative in its use of AI, uh, stretching back many years actually.

    24
    00:02:31,630 --> 00:02:34,630
    Uh, obviously more so now than maybe in the past.

    25
    00:02:34,630 --> 00:02:39,430
    But, we use AI in several areas in clinical care. So radiology is a great example of this,

    26
    00:02:39,430 --> 00:02:46,150
    where if you come into UTMB and you have symptoms of a stroke or symptoms of a pulmonary embolism, both are really emergent.

    27
    00:02:46,510 --> 00:02:53,469
    Uh, time sensitive, intense clinical scenarios. Uh, an AI will probably be the first thing that really sees that.

    28
    00:02:53,470 --> 00:02:59,950
    You have that on imaging. It will, uh, annotate the image, put a circle around where it sees that pathology,

    29
    00:02:59,950 --> 00:03:04,329
    and notify the care team as this one very kind of tangible thing where we've seen this,

    30
    00:03:04,330 --> 00:03:08,620
    I saved people's lives, for example, by getting the care pathway activated quicker.

    31
    00:03:09,310 --> 00:03:13,360
    We also use AI a lot in what we'd call, uh, screening settings.

    32
    00:03:13,360 --> 00:03:19,960
    Really. So how do these, uh, chest radiographs that we take for sometimes general health care purposes.

    33
    00:03:20,230 --> 00:03:23,500
    How might they indicate other diseases that perhaps no one was quite aware of?

    34
    00:03:23,770 --> 00:03:28,630
    Things like heart disease or kidney disease or liver disease or skeletal osteoporosis, for example.

    35
    00:03:28,990 --> 00:03:33,310
    So yeah, really helps kind of covering these bases and preventing things from slipping through the cracks.

    36
    00:03:33,700 --> 00:03:39,430
    We use about 40 modules in radiology for AI. We use it in pathology too, especially with things like prostate.

    37
    00:03:39,460 --> 00:03:41,800
    Make sure that our pathologists... I'm a pathologist for example.

    38
    00:03:42,010 --> 00:03:47,920
    Don't miss things on those prostate cords a little piece of cancer here... or invasion of a nerve there...

    39
    00:03:47,950 --> 00:03:51,580
    The AI really helps us cover those bases and ensure that nothing falls through the cracks.

    40
    00:03:52,300 --> 00:03:56,680
    And lastly, AI helps enrich this experience of doing care for patients.

    41
    00:03:56,680 --> 00:03:59,950
    So, um, depending on who you are as a clinician or patient,

    42
    00:03:59,950 --> 00:04:03,159
    you may have benefited from these tools where you go to talk to your doctor and you get to

    43
    00:04:03,160 --> 00:04:07,180
    have a face to face conversation and not the side of someone's head while they type in Epic.

    44
    00:04:07,480 --> 00:04:10,570
    And AI takes care of that scribing activity for you. Really,

    45
    00:04:10,570 --> 00:04:17,260
    just to make it a more personal and direct interaction and to to remove the the bookkeeping so much from

    46
    00:04:17,260 --> 00:04:22,089
    the clinical workflow, and all those things that I mentioned are live in practice today and more will come,

    47
    00:04:22,090 --> 00:04:30,190
    but we use those even now. While I'm sure that all the clinicians here on UTMB thank you and the AI Council for making their lives better.

    48
    00:04:30,190 --> 00:04:36,520
    And our patients certainly thank you for creating an environment that brings higher quality to the health care,

    49
    00:04:36,850 --> 00:04:44,740
    Uh, that we deliver and puts UTMB on the map in terms of this cutting edge technology and utilizing it for the betterment of our patients.

    50
    00:04:45,340 --> 00:04:50,590
    Can you talk a little bit about how AI is helping us do better at conducting research?

    51
    00:04:51,070 --> 00:04:53,740
    Yeah, so I can give a very tangible use case for this.

    52
    00:04:53,740 --> 00:05:00,160
    So... many enterprises, us included, when we look over patient data to do something like match a patient to a trial,

    53
    00:05:00,160 --> 00:05:07,630
    for example, or conduct a research study that needs to have a certain population, there's a lot of filters you would apply.

    54
    00:05:07,660 --> 00:05:10,780
    Okay, well, people need to have this diagnosis, but not that one.

    55
    00:05:10,780 --> 00:05:17,890
    Or maybe they have this genetic test, but not that one. A lot of this information in any EMR, anywhere you go is not really structured.

    56
    00:05:18,370 --> 00:05:22,900
    It's kind of narrative. It's in notes. And historically the only way we could access that

    57
    00:05:22,940 --> 00:05:28,730
    was workforce allocation for someone to read all those notes and make like a spreadsheet out of it.

    58
    00:05:29,060 --> 00:05:32,629
    And, that is expensive, time consuming, and not always correct either.

    59
    00:05:32,630 --> 00:05:39,320
    You know there's challenges there. A really great thing I can do, like GPT, for example, which is good at speaking in notes,

    60
    00:05:39,320 --> 00:05:45,650
    understanding and comprehending them and turning them into structured things... is create this structure where we lacked it before.

    61
    00:05:46,280 --> 00:05:51,919
    And for us, that means things like being able to more efficiently say... What patients have this,

    62
    00:05:51,920 --> 00:05:54,379
    this and this feature, but not that, that and that feature.

    63
    00:05:54,380 --> 00:05:59,780
    So, I can enroll them in a study, give them access to clinical research easily and more efficiently.

    64
    00:05:59,780 --> 00:06:07,250
    I think this is a tremendous value add for AI, basically bringing structure on to the medical record where there wasn't any historically.

    65
    00:06:08,090 --> 00:06:16,820
    Well, I think you and I are believers in AI, but what are some of the concerns that people who may not be so ready to adapt AI,

    66
    00:06:17,180 --> 00:06:22,070
    particularly as it pertains to health care, and particularly as it pertains to our patients?

    67
    00:06:22,280 --> 00:06:25,999
    Yeah, there are very valid concerns that people can have about AI in health care.

    68
    00:06:26,000 --> 00:06:29,270
    And, um, I'm sure patients have them and their right to have them.

    69
    00:06:29,270 --> 00:06:34,849
    We share them as clinicians and as an enterprise. And... and essentially they boil down to a couple of key areas.

    70
    00:06:34,850 --> 00:06:39,860
    So, I think the main one is will this replace my doctors ability to reason about my care?

    71
    00:06:40,250 --> 00:06:44,990
    I don't want, you know, uh, a non-human or a robot deciding what happens to me.

    72
    00:06:45,260 --> 00:06:52,700
    And, I think we would say we fully agree with that, that where AI is helpful today is not in replacing the doctor's reasoning,

    73
    00:06:53,000 --> 00:07:01,190
    but letting the doctor get to do the reasoning more often and more efficiently, and that... it should be comforting, I think.

    74
    00:07:01,190 --> 00:07:07,700
    So, I helping to collect and surface the info a doctor needs to think about, but not thinking on their behalf.

    75
    00:07:07,730 --> 00:07:13,760
    So, that's a valid concern we share. Another one is, you know, how will these tools be from a safety standpoint in general?

    76
    00:07:14,510 --> 00:07:19,460
    Sometimes I that looks over an image for example, is it perfect and will it miss things?

    77
    00:07:19,820 --> 00:07:25,040
    Do we know how often it misses things? How do we deal with that? And I would say this is also something that of course we share.

    78
    00:07:25,340 --> 00:07:32,209
    We're not new to this. You know, I might be a new modality, but we've done this for decades in other kinds of things diagnostic test, therapeutics,

    79
    00:07:32,210 --> 00:07:36,640
    etc. and we apply that same kind of logic, which is evidence base.

    80
    00:07:36,650 --> 00:07:43,850
    What works and what doesn't. Doctor is central to the judicious use of these tools, and we're always vigilant how they perform after the fact.

    81
    00:07:43,850 --> 00:07:46,309
    And I don't think we'll... we ever veer from that.

    82
    00:07:46,310 --> 00:07:53,150
    Uh, keeping to our core ethical principles of doctors guiding for the benefit of the patient every aspect of their care.

    83
    00:07:54,200 --> 00:08:01,160
    That's great. Uh, what are some of the other opportunities that you think AI presents for the future to improve health care,

    84
    00:08:01,550 --> 00:08:07,790
    not only here at UTMB, but in society in general? Yeah, I think there's a tremendous opportunity in access.

    85
    00:08:08,300 --> 00:08:13,520
    You know, bias and imperfect algorithms are always a risk, but they're there risks...

    86
    00:08:13,520 --> 00:08:17,180
    We... we kind of know that we can overcome because we've done it in other areas.

    87
    00:08:17,510 --> 00:08:22,459
    But, lack of access has been this macro thing in health care for a very long time.

    88
    00:08:22,460 --> 00:08:27,290
    You know, in our state, almost every county has some critical physician shortage, some have no physicians.

    89
    00:08:27,710 --> 00:08:34,160
    What people want at a health care is a level of access that I don't think we can really globally give them without AI.

    90
    00:08:34,310 --> 00:08:40,670
    But, the opportunity it lets us have is to give them this. So to say you're in, you know, a rural setting and you have a symptom,

    91
    00:08:40,940 --> 00:08:46,549
    can you get an answer right now about that symptom and being able to kind of scale that intelligence,

    92
    00:08:46,550 --> 00:08:51,650
    that clinical reasoning to everyone, everywhere is a really inherent social good.

    93
    00:08:51,650 --> 00:08:56,990
    I think, I we've known we wanted it in a long time for health care, but I think I finally let's us have this,

    94
    00:08:57,140 --> 00:09:01,130
    uh, not not too quick, you know, but eventually we will get there.

    95
    00:09:02,000 --> 00:09:10,160
    Uh, I think that's great. We know how I can really help us in doing the access piece, the health care piece, even a research piece.

    96
    00:09:10,640 --> 00:09:17,690
    Can you talk a little bit about how you see AI intersecting with our new fourth pillar here at UTMB?

    97
    00:09:17,690 --> 00:09:25,300
    That is... the innovation mission. Yes. So one of the great things about... I'll take language models as an example...

    98
    00:09:25,310 --> 00:09:30,950
    We've all heard of ChatGPT. This type of AI is kind of a new to come out and be so useful.

    99
    00:09:31,310 --> 00:09:38,090
    What's lovely about this than different from historical AI tools, is that if you have a subject matter expertise,

    100
    00:09:38,090 --> 00:09:42,170
    like a workflow you work with, you experience a problem and you can write English.

    101
    00:09:42,470 --> 00:09:45,470
    You are kind of a developer, kind of an innovator.

    102
    00:09:45,830 --> 00:09:50,270
    And that wasn't the way things were in the past. You had to have a computer scientists make a special model do this and that.

    103
    00:09:50,600 --> 00:09:54,739
    What we see now is that the ability for people to be innovative and to make real

    104
    00:09:54,740 --> 00:09:59,960
    solutions in their workforce is really democratized in a way that it never was before.

    105
    00:10:00,260 --> 00:10:05,930
    And by giving people the education, the tools to do that, everyone can drive innovation.

    106
    00:10:05,930 --> 00:10:08,389
    And it wasn't it just wasn't quite that way before.

    107
    00:10:08,390 --> 00:10:14,270
    But I think we're going to see this emerge out of these sort of copilot paradigm that we see coming out now.

    108
    00:10:14,270 --> 00:10:17,300
    And as an institution where we want innovation everywhere.

    109
    00:10:17,570 --> 00:10:22,520
    Uh, we we can have some level of that, and it is to the benefit of everyone, but it lets all the

    110
    00:10:22,720 --> 00:10:29,710
    brain power and collective expertise we have here at UTMB manifest in tools that people can actually build and create and bring to life.

    111
    00:10:30,760 --> 00:10:39,639
    That's great. Well, you know, I've seen articles in different places talking about how AI is going to eliminate the workforce.

    112
    00:10:39,640 --> 00:10:47,260
    That it's going to eliminate COVID, for instance. Uh, some people say that will radiologists won't be needed... will just be needed.

    113
    00:10:48,130 --> 00:10:52,420
    Do you think I will replace our help to traditional health care workforce?

    114
    00:10:52,810 --> 00:10:55,810
    I think it'll I think it'll help. I think some of our roles will evolve.

    115
    00:10:55,960 --> 00:11:04,360
    Um, but I think the comforting thing to clinicians should be that we're a far way from having a discussion about needing less doctors.

    116
    00:11:04,750 --> 00:11:07,479
    I think that's the only discussion we've really had over many years,

    117
    00:11:07,480 --> 00:11:12,940
    is we need more and more and more and more doctors, and we can never seem to get ahead of the curve of having enough.

    118
    00:11:13,390 --> 00:11:17,710
    And, I think the reality is that we're going at we're going to need AI to even have enough,

    119
    00:11:17,890 --> 00:11:24,790
    much less have too many, because what people want of them is much more access than they've ever had before.

    120
    00:11:25,150 --> 00:11:28,180
    And, so even as a pathologist, as a radiologist, what you're going to see, I think,

    121
    00:11:28,180 --> 00:11:35,620
    is a higher volume of diagnostic testing, more data available to people, more information to reason about.

    122
    00:11:35,830 --> 00:11:39,910
    And you will require AI to kind of keep abreast of that volume.

    123
    00:11:40,270 --> 00:11:44,290
    And ultimately, that's what patients expect when they say, I want to ask a question.

    124
    00:11:44,290 --> 00:11:48,429
    Now I need to get an answer now. Um, we'll only be able to get there with AI.

    125
    00:11:48,430 --> 00:11:55,570
    I really don't think we're going to face contraction just because the demand is almost limitless for what doctors provide.

    126
    00:11:56,640 --> 00:11:58,770
    Well, thank you, Dr. McCaffrey, for being here.

    127
    00:11:59,100 --> 00:12:06,450
    I know that you're going to really allow UTMB to be on the cutting edge of what AI can do for our patient care mission,

    128
    00:12:06,750 --> 00:12:11,220
    our research mission, our education mission, and our innovation mission.

    129
    00:12:11,460 --> 00:12:14,220
    And, we look forward to seeing all the great things that you're going to be

    130
    00:12:14,220 --> 00:12:19,140
    able to do and allow UTMB to be the be the leader globally and locally on how,

    131
    00:12:19,140 --> 00:12:23,130
    AI gets deployed in an academic medical center. Thank you for being here.

    132
    00:12:23,160 --> 00:12:26,910
    Thank you. Appreciate it. Thank you for tuning in to The Pulse.

    133
    00:12:27,390 --> 00:12:32,640
    We hope you've gained valuable insights, discover new perspectives, and feel inspired by the stories shared today.

    134
    00:12:33,900 --> 00:12:38,730
    Please join us for future episodes to stay informed and connected right here on The Pulse.

More Information

Dr. Peter McCaffrey, MD, MS, FCAP, serves as UTMB Health’s Chief Artificial Intelligence (AI) Officer, where he leads AI strategy and implementation, advancing patient care and operational efficiency across UTMB. Board-certified in Clinical Pathology and Clinical Informatics, Dr. McCaffrey is also co-chair of the University of Texas System AI Collaborative, a statewide initiative guiding AI innovation and strategy for Texas’ health institutions.

Dr. McCaffrey brings extensive experience from a unique blend of academic and industry backgrounds. He completed his medical training at Johns Hopkins, with residency and chief residency at Massachusetts General Hospital/Harvard Medical School, and went on to Stanford University for graduate training in AI and Biomedical Data Science.

Before UTMB, he was a Biodesign fellow at the Texas Medical Center Innovation Factory, where he collaborated with venture incubators. He later co-founded Pragma Biosciences, an AI-focused biotechnology company, which has raised nearly $20 million in venture funding and partners with industry leaders like GSK and Merck on AI-driven drug discovery.

At UTMB since 2018, Dr. McCaffrey has pioneered projects like AI-enhanced automation of COVID-19 diagnostics, nationally recognized by America's Essential Hospitals, and developed Texas’ most advanced laboratory for Single Cell and Pharmacogenomics. His vision and expertise are helping UTMB expand its leadership and innovation in medical AI.