1 00:00:00,320 --> 00:00:03,679 Welcome to the Becker's podcast. I'm Naomi Diaz, 2 00:00:03,679 --> 00:00:06,319 health IT reporter here at Becker's. And today, 3 00:00:06,319 --> 00:00:09,679 we're joined by doctor Mirage Gaine, director of 4 00:00:09,679 --> 00:00:10,179 responsible 5 00:00:10,559 --> 00:00:13,059 AI in health at the Coalition for Health 6 00:00:13,439 --> 00:00:15,139 AI, also known as CHI. 7 00:00:15,515 --> 00:00:18,074 In this episode, we'll discuss what hospital leaders 8 00:00:18,074 --> 00:00:20,234 need to know about AI today, how to 9 00:00:20,234 --> 00:00:22,554 prepare for the next few years, and where 10 00:00:22,554 --> 00:00:24,875 AI is really headed in health care. Doctor 11 00:00:24,875 --> 00:00:26,714 Gain, thank you so much for joining us 12 00:00:26,714 --> 00:00:29,059 today. I appreciate it. I wanna start off 13 00:00:29,059 --> 00:00:30,980 with our first question here. Just a simple 14 00:00:30,980 --> 00:00:33,539 one of just introducing yourself and tell us 15 00:00:33,539 --> 00:00:35,159 a little bit about your background. 16 00:00:36,899 --> 00:00:38,759 Great to be here. Thanks so much, Naomi. 17 00:00:40,020 --> 00:00:40,679 I am. 18 00:00:41,460 --> 00:00:44,234 I have a PhD in clinical psychology, 19 00:00:44,615 --> 00:00:45,115 actually, 20 00:00:45,975 --> 00:00:48,695 and I have a background in computational and 21 00:00:48,695 --> 00:00:49,674 clinical neuroscience, 22 00:00:50,454 --> 00:00:53,015 behavioral science, and I used a lot of 23 00:00:53,015 --> 00:00:54,315 machine learning and AI 24 00:00:54,774 --> 00:00:57,255 in my research back in the academic days. 25 00:00:57,255 --> 00:00:57,755 But, 26 00:00:58,320 --> 00:00:58,820 I'm 27 00:00:59,600 --> 00:01:02,399 far from those, not super far, but pretty 28 00:01:02,399 --> 00:01:04,260 far from the academic days. 29 00:01:05,439 --> 00:01:05,939 Since 30 00:01:07,200 --> 00:01:09,680 finishing my postdoc, I've been working 31 00:01:10,375 --> 00:01:12,454 I I started working as a principal behavioral 32 00:01:12,454 --> 00:01:15,015 designer at ideas forty two, which is a 33 00:01:15,015 --> 00:01:16,314 mission driven nonprofit, 34 00:01:17,174 --> 00:01:19,114 and I focus my work in, 35 00:01:19,575 --> 00:01:21,435 AI and machine learning and health, 36 00:01:22,534 --> 00:01:25,629 sort of things around, like, identifying bias and 37 00:01:25,629 --> 00:01:27,090 health AI solutions, 38 00:01:28,269 --> 00:01:30,909 for organizations and helping organizations do that in 39 00:01:30,909 --> 00:01:32,609 a behaviorally informed way. 40 00:01:33,390 --> 00:01:36,369 I actually started working with CHI then. 41 00:01:36,895 --> 00:01:39,295 So before CHI was ever a nonprofit, it 42 00:01:39,295 --> 00:01:40,834 was a coalition for the willing, 43 00:01:41,775 --> 00:01:42,995 coalition of the willing. 44 00:01:44,415 --> 00:01:46,975 And I led their fairness and bias work 45 00:01:46,975 --> 00:01:47,475 group, 46 00:01:48,495 --> 00:01:50,995 in developing sort of responsible AI guidance, 47 00:01:52,310 --> 00:01:52,969 and then 48 00:01:53,270 --> 00:01:56,150 I transitioned to this role. So here at 49 00:01:56,150 --> 00:01:59,109 CHI, I work across many sectors. I bring 50 00:01:59,109 --> 00:01:59,609 together 51 00:02:00,950 --> 00:02:01,450 any 52 00:02:02,150 --> 00:02:04,230 sector you can think of in health care. 53 00:02:04,230 --> 00:02:07,605 So anybody on the developer side, the clinical 54 00:02:07,665 --> 00:02:08,165 side, 55 00:02:09,824 --> 00:02:10,884 sometimes policymakers. 56 00:02:13,425 --> 00:02:15,925 Who else do we have? Patient advocacy groups, 57 00:02:15,985 --> 00:02:17,205 community health centers, 58 00:02:17,585 --> 00:02:20,490 all of them into one space, and we 59 00:02:20,490 --> 00:02:22,750 talk about how to build shared frameworks, 60 00:02:23,849 --> 00:02:24,349 tools, 61 00:02:24,810 --> 00:02:25,310 processes, 62 00:02:28,169 --> 00:02:30,430 best practice guidance to make AI, 63 00:02:31,050 --> 00:02:32,270 safer, more transparent, 64 00:02:32,730 --> 00:02:33,469 and responsible 65 00:02:33,849 --> 00:02:34,830 in health care. 66 00:02:36,014 --> 00:02:38,094 Amazing. Great to hear about your background. And, 67 00:02:38,094 --> 00:02:39,395 again, just responsible, 68 00:02:40,014 --> 00:02:43,055 safety of AI implementations is just really what 69 00:02:43,055 --> 00:02:44,814 we need right now. Would love if you 70 00:02:44,814 --> 00:02:46,415 could just dive in and just tell us 71 00:02:46,415 --> 00:02:49,135 about your work with consensus driven frameworks for 72 00:02:49,135 --> 00:02:50,574 AI in health care. What have you learned 73 00:02:50,574 --> 00:02:52,340 over these past years, and what do you 74 00:02:52,340 --> 00:02:53,800 feel like is most important today? 75 00:02:55,219 --> 00:02:58,760 Great question. Consensus work is hard and traditionally, 76 00:02:59,620 --> 00:03:02,599 very slow, but we don't actually have that 77 00:03:02,900 --> 00:03:03,400 privilege, 78 00:03:04,340 --> 00:03:06,979 really in the AI space of being slow 79 00:03:06,979 --> 00:03:09,594 because things are moving so so fast. And 80 00:03:09,594 --> 00:03:10,094 so, 81 00:03:11,354 --> 00:03:11,854 we 82 00:03:12,634 --> 00:03:15,775 at CHI use sort of modified Delphi methods 83 00:03:16,235 --> 00:03:18,655 to build these consensus driven frameworks, 84 00:03:19,849 --> 00:03:22,669 And we convene, like I said, many stakeholders 85 00:03:23,049 --> 00:03:23,549 across, 86 00:03:24,090 --> 00:03:27,129 large academic medical centers like the Mayo's, like 87 00:03:27,129 --> 00:03:28,509 the Dukes, UT Health, 88 00:03:30,090 --> 00:03:33,229 as well as non academic medical centers, small 89 00:03:33,609 --> 00:03:34,829 community health centers, 90 00:03:35,395 --> 00:03:37,634 and, like I said, the developers of these 91 00:03:37,634 --> 00:03:40,134 solutions as well to kind of think about, 92 00:03:42,435 --> 00:03:45,314 how can we take what it currently exists 93 00:03:45,314 --> 00:03:48,534 in many different silos with many different incentive 94 00:03:48,594 --> 00:03:49,094 structures, 95 00:03:50,009 --> 00:03:51,789 and bring them together to find 96 00:03:52,090 --> 00:03:52,750 the shared 97 00:03:53,209 --> 00:03:53,709 motivations. 98 00:03:54,810 --> 00:03:57,289 How can we help each other improve the 99 00:03:57,289 --> 00:03:58,669 health care system, 100 00:03:59,610 --> 00:04:03,770 without sacrificing innovation, without sacrificing patient safety and 101 00:04:03,770 --> 00:04:04,270 privacy, 102 00:04:05,465 --> 00:04:07,644 and keep things moving at a pace that 103 00:04:08,185 --> 00:04:10,764 that this this whole field really needs. 104 00:04:11,944 --> 00:04:14,104 I think one of the biggest things that 105 00:04:14,104 --> 00:04:16,824 I've learned is that everyone is trying to 106 00:04:16,824 --> 00:04:17,884 do the right thing, 107 00:04:19,384 --> 00:04:20,285 with AI. 108 00:04:20,610 --> 00:04:21,110 We 109 00:04:21,569 --> 00:04:23,430 most people. We have, 110 00:04:23,889 --> 00:04:26,449 I think different lenses depending on where we're 111 00:04:26,449 --> 00:04:28,550 coming at this problem from. 112 00:04:29,410 --> 00:04:31,410 But and there is no map that says 113 00:04:31,410 --> 00:04:33,729 here is how all these different lenses should 114 00:04:33,729 --> 00:04:35,750 connect, and that's where we come in, 115 00:04:36,314 --> 00:04:38,254 to sort of co create 116 00:04:39,274 --> 00:04:40,495 practical practical, 117 00:04:41,274 --> 00:04:42,814 sort of flexible guardrails, 118 00:04:43,914 --> 00:04:45,055 tools, resources 119 00:04:45,435 --> 00:04:46,175 to help 120 00:04:46,555 --> 00:04:49,774 everyone in this landscape make better safer decisions. 121 00:04:50,810 --> 00:04:52,730 And I think it's more important than ever 122 00:04:52,730 --> 00:04:54,589 now because the consequences 123 00:04:55,129 --> 00:04:56,350 of AI misuse 124 00:04:57,050 --> 00:04:59,069 in health care in particular, 125 00:05:00,730 --> 00:05:02,029 is huge, and 126 00:05:02,644 --> 00:05:04,404 we have a health care system that is 127 00:05:04,404 --> 00:05:05,384 already struggling, 128 00:05:06,644 --> 00:05:07,944 both as a business 129 00:05:08,485 --> 00:05:10,105 and for patients. 130 00:05:10,805 --> 00:05:12,725 And so how can we really take this 131 00:05:12,725 --> 00:05:14,644 opportunity, and I see this as a big 132 00:05:14,644 --> 00:05:15,144 opportunity 133 00:05:15,605 --> 00:05:16,425 to do better, 134 00:05:17,605 --> 00:05:18,105 without 135 00:05:19,389 --> 00:05:21,089 hurting the most vulnerable people. 136 00:05:21,790 --> 00:05:22,290 Mhmm. 137 00:05:22,830 --> 00:05:24,830 Mhmm. I I love what you touched on 138 00:05:24,830 --> 00:05:26,910 there too is not stifling that innovation, but 139 00:05:26,910 --> 00:05:29,389 also ensuring that, you know, you have those 140 00:05:29,389 --> 00:05:32,129 safety guardrails for a technology that's continue 141 00:05:32,595 --> 00:05:35,574 continuously evolving. I think that's super important there. 142 00:05:35,795 --> 00:05:36,935 I think for our audience, 143 00:05:37,314 --> 00:05:39,574 obviously, and our listeners of health care execs, 144 00:05:39,875 --> 00:05:41,235 I would love to ask you as as 145 00:05:41,235 --> 00:05:43,154 a follow-up here as well. What do you 146 00:05:43,154 --> 00:05:45,475 feel like hospital executive teams really need to 147 00:05:45,475 --> 00:05:48,079 know about health care AI today in this 148 00:05:48,079 --> 00:05:50,639 such rapidly changing landscape, and what do they 149 00:05:50,639 --> 00:05:52,399 need to do to prepare their teams for 150 00:05:52,399 --> 00:05:53,779 the next two to three years? 151 00:05:54,719 --> 00:05:55,699 It's a great question. 152 00:05:56,079 --> 00:05:57,759 And they're we're doing a lot of work 153 00:05:57,759 --> 00:05:59,355 in this space, but I think the first 154 00:05:59,355 --> 00:06:01,754 thing to realize is AI is already being 155 00:06:01,754 --> 00:06:02,254 used. 156 00:06:03,834 --> 00:06:04,574 And and, 157 00:06:05,274 --> 00:06:05,935 you know, 158 00:06:06,314 --> 00:06:07,935 sometimes it's chat GPT 159 00:06:08,314 --> 00:06:09,694 on somebody's phone. 160 00:06:10,475 --> 00:06:12,495 And and it's also being used 161 00:06:13,480 --> 00:06:15,740 for real use cases and documentation 162 00:06:16,199 --> 00:06:17,740 and triage and 163 00:06:18,120 --> 00:06:18,620 diagnostics. 164 00:06:19,959 --> 00:06:22,699 Radiology is a big area and and administrative 165 00:06:22,920 --> 00:06:25,000 tasks. So people think about AI and health 166 00:06:25,000 --> 00:06:26,600 care like, oh, it must be a clinical 167 00:06:26,600 --> 00:06:27,100 use, 168 00:06:27,415 --> 00:06:30,074 but there are so many operational and administrative 169 00:06:30,774 --> 00:06:31,274 processes 170 00:06:31,574 --> 00:06:33,355 in health care that are struggling, 171 00:06:34,375 --> 00:06:35,595 and leading to workforce 172 00:06:36,615 --> 00:06:37,115 issues, 173 00:06:37,814 --> 00:06:40,295 and and cost of having to retrain people 174 00:06:40,295 --> 00:06:42,314 because people are so burnt out. 175 00:06:42,730 --> 00:06:43,230 So, 176 00:06:44,569 --> 00:06:45,310 just recognizing 177 00:06:45,610 --> 00:06:47,770 that it is already being used, and there 178 00:06:47,770 --> 00:06:50,250 are lots of spaces that you can use 179 00:06:50,250 --> 00:06:51,870 it within the health care 180 00:06:52,569 --> 00:06:53,069 landscape. 181 00:06:54,250 --> 00:06:56,805 I think the second thing is that folks 182 00:06:56,805 --> 00:06:57,785 need a plan, 183 00:06:58,564 --> 00:07:01,465 and that means establishing internal policies, 184 00:07:01,845 --> 00:07:02,985 governance structures, 185 00:07:03,925 --> 00:07:04,425 processes, 186 00:07:06,324 --> 00:07:08,965 working together with your teams and and within 187 00:07:08,965 --> 00:07:11,189 your network. So some organizations 188 00:07:11,490 --> 00:07:14,529 are much larger and have very robust internal 189 00:07:14,529 --> 00:07:15,029 teams. 190 00:07:15,810 --> 00:07:19,029 Some organizations like safety net hospitals, FQHCs, 191 00:07:19,490 --> 00:07:20,629 community health centers, 192 00:07:21,089 --> 00:07:22,470 they don't have that 193 00:07:22,805 --> 00:07:25,044 luxury, so to speak, but they do have 194 00:07:25,044 --> 00:07:26,425 access to things like 195 00:07:26,805 --> 00:07:27,305 their 196 00:07:28,564 --> 00:07:31,464 health center control network, their primary care association 197 00:07:31,604 --> 00:07:33,604 network. And so how can we get really 198 00:07:33,604 --> 00:07:37,100 creative about what it means to establish internal 199 00:07:37,160 --> 00:07:38,780 policies, governance structures, 200 00:07:39,160 --> 00:07:41,879 and not, you know, burden those that don't 201 00:07:41,879 --> 00:07:44,040 have resources and and be able to develop 202 00:07:44,040 --> 00:07:45,259 that? So we're doing that, 203 00:07:46,600 --> 00:07:47,259 in partnership 204 00:07:47,560 --> 00:07:47,954 with 205 00:07:48,834 --> 00:07:51,814 NAC, the National Association for Community Health Centers, 206 00:07:53,314 --> 00:07:56,194 but also with just our member organizations that 207 00:07:56,194 --> 00:07:57,794 have come up and said we really care 208 00:07:57,794 --> 00:07:59,875 about this. Even the larger organizations will come 209 00:07:59,875 --> 00:08:01,875 up and say, if we're proposing this as 210 00:08:01,875 --> 00:08:02,534 a process, 211 00:08:03,160 --> 00:08:04,620 how will the smaller 212 00:08:07,000 --> 00:08:09,639 entities, the less resource entities, actually make this 213 00:08:09,639 --> 00:08:12,220 happen? So we're really thinking about this carefully. 214 00:08:13,400 --> 00:08:15,240 So as part of that plan, folks need 215 00:08:15,240 --> 00:08:17,875 to understand the risks. They need to determine 216 00:08:17,935 --> 00:08:20,035 what are some of the responsible use cases. 217 00:08:20,095 --> 00:08:21,235 Where can they start, 218 00:08:22,254 --> 00:08:25,795 in this AI journey for their particular organization? 219 00:08:25,935 --> 00:08:28,514 And, again, context will matters a lot. 220 00:08:29,500 --> 00:08:31,819 And then thinking about the processes that you're 221 00:08:31,819 --> 00:08:35,100 developing, the governance structures, the internal policies should 222 00:08:35,100 --> 00:08:36,639 address things like bias 223 00:08:37,419 --> 00:08:40,639 or risk of bias, security and privacy issues, 224 00:08:41,684 --> 00:08:44,404 the usability and usefulness of that solution. You're 225 00:08:44,404 --> 00:08:47,044 paying so much money not just to procure 226 00:08:47,044 --> 00:08:49,125 this, but to use it. And if it's 227 00:08:49,125 --> 00:08:51,365 not helpful, if it's just gonna make your 228 00:08:51,365 --> 00:08:52,745 staff angry, like, 229 00:08:53,605 --> 00:08:54,105 shouldn't 230 00:08:54,450 --> 00:08:56,610 shouldn't be the you know, even if it's 231 00:08:56,610 --> 00:08:58,149 bright and shiny and new, 232 00:08:58,690 --> 00:09:00,690 there's a lot of change management that needs 233 00:09:00,690 --> 00:09:01,429 to go in 234 00:09:01,730 --> 00:09:05,009 to to bringing new, processes in and making 235 00:09:05,009 --> 00:09:07,490 sure that they're helpful. And then safety and 236 00:09:07,490 --> 00:09:09,945 transparency. So those are, I think, areas that 237 00:09:10,024 --> 00:09:12,845 folks really need to start developing plans around. 238 00:09:14,105 --> 00:09:16,845 And then finally I think investing in capacity 239 00:09:17,225 --> 00:09:18,445 and in education 240 00:09:19,465 --> 00:09:22,605 for staff, so upskilling existing staff, 241 00:09:24,029 --> 00:09:27,070 training new staff to sort of have that 242 00:09:27,070 --> 00:09:29,950 come in with the mindset of, like, where 243 00:09:29,950 --> 00:09:32,029 AI is used, when it's helpful, how it's 244 00:09:32,029 --> 00:09:32,529 helpful, 245 00:09:33,389 --> 00:09:35,789 and what is it, for for many folks. 246 00:09:35,789 --> 00:09:37,009 That's the first question. 247 00:09:38,205 --> 00:09:40,605 So investing in that capacity now and investing 248 00:09:40,605 --> 00:09:42,144 in a road map for the future, 249 00:09:43,004 --> 00:09:43,904 data infrastructures, 250 00:09:44,445 --> 00:09:46,865 assets, all of that making sure that 251 00:09:48,524 --> 00:09:50,044 you have a sense of, like, where your 252 00:09:50,044 --> 00:09:52,149 data's at, how it could be used, how 253 00:09:52,149 --> 00:09:53,529 it could be leveraged 254 00:09:53,830 --> 00:09:57,049 to actually improve the business of your organization, 255 00:09:58,950 --> 00:10:00,490 in this sort of 256 00:10:01,990 --> 00:10:04,309 in these times where data data is king 257 00:10:04,309 --> 00:10:05,129 and queen, 258 00:10:05,589 --> 00:10:06,570 all of the above. 259 00:10:08,205 --> 00:10:10,605 I I really loved your takeaways, doctor Gaine. 260 00:10:10,605 --> 00:10:11,965 Really hit the nail on the head when 261 00:10:11,965 --> 00:10:13,725 it comes to just things I'm even hearing 262 00:10:13,725 --> 00:10:15,884 from our audience here on just the the 263 00:10:15,884 --> 00:10:18,144 obstacles they're going through in their own organizations 264 00:10:18,365 --> 00:10:21,089 there. Really appreciate you diving deep. I wanna 265 00:10:21,089 --> 00:10:23,169 stay on that track of what you mentioned 266 00:10:23,169 --> 00:10:25,809 of, you know, those governance, those guardrails, and 267 00:10:25,809 --> 00:10:28,309 ask you, you know, how should hospitals and 268 00:10:28,450 --> 00:10:32,070 really health systems just address regular regulatory concerns 269 00:10:32,129 --> 00:10:34,144 with AI and health care? And I know 270 00:10:34,144 --> 00:10:35,345 you touched on this a bit in your 271 00:10:35,345 --> 00:10:36,945 answer as well, but how do you really 272 00:10:36,945 --> 00:10:39,445 create ethical standards within the organization 273 00:10:39,904 --> 00:10:43,125 to deploy this safety and ethically as well? 274 00:10:44,625 --> 00:10:45,605 Yeah. I think 275 00:10:46,409 --> 00:10:49,149 I think we're in this sort of regulatory 276 00:10:49,610 --> 00:10:52,329 purgatory, I would call it. And and I 277 00:10:52,329 --> 00:10:54,009 don't think I think at the end of 278 00:10:54,009 --> 00:10:57,049 the day, we shouldn't have to wait. Nobody 279 00:10:57,049 --> 00:10:59,529 should have to wait for full regulation to 280 00:10:59,529 --> 00:11:00,350 act responsibly. 281 00:11:01,209 --> 00:11:03,285 If we're gonna only do the right thing 282 00:11:03,285 --> 00:11:04,884 when somebody tells us to do the right 283 00:11:04,884 --> 00:11:06,424 thing, I think, you 284 00:11:06,804 --> 00:11:08,825 know, maybe it loses its value a little. 285 00:11:09,524 --> 00:11:10,024 So 286 00:11:10,804 --> 00:11:13,044 for me, ethical standards has to be they 287 00:11:13,044 --> 00:11:15,044 have to be built into the AI life 288 00:11:15,044 --> 00:11:15,544 cycle. 289 00:11:15,845 --> 00:11:17,544 So all the way from procurement 290 00:11:18,470 --> 00:11:20,710 to development, if that's something you're doing internally 291 00:11:20,710 --> 00:11:22,710 in your organization or co development with, 292 00:11:24,470 --> 00:11:26,170 with a vendor or purchasing, 293 00:11:27,430 --> 00:11:30,090 all the way to monitoring and oversight. So 294 00:11:30,389 --> 00:11:31,529 when we think about 295 00:11:33,464 --> 00:11:35,625 ethics in health care, how can we then 296 00:11:35,625 --> 00:11:37,164 extend that to 297 00:11:37,544 --> 00:11:38,764 ethics in AI? 298 00:11:39,304 --> 00:11:40,125 And I think 299 00:11:41,304 --> 00:11:42,204 there is this 300 00:11:43,625 --> 00:11:46,105 sense that sometimes people try to reinvent the 301 00:11:46,105 --> 00:11:48,105 wheel. They're like, oh, AI is so different. 302 00:11:48,105 --> 00:11:49,139 Everything about it needs 303 00:11:51,139 --> 00:11:51,460 to be different, but safety still matters. Security 304 00:11:51,460 --> 00:11:54,420 privacy still matters, and these things existed. You 305 00:11:54,420 --> 00:11:55,160 know, transparency, 306 00:11:55,540 --> 00:11:58,259 consent, like, these things have existed in health 307 00:11:58,259 --> 00:11:58,759 care 308 00:11:59,139 --> 00:12:01,779 even pre AI. Now the question is, how 309 00:12:01,779 --> 00:12:04,360 do we identify where AI uniquely, 310 00:12:05,894 --> 00:12:08,534 needs ethics in these spaces, and how can 311 00:12:08,534 --> 00:12:10,455 we just add that to the processes that 312 00:12:10,455 --> 00:12:11,914 already exist without 313 00:12:12,615 --> 00:12:14,315 fully reinventing the wheel? 314 00:12:16,054 --> 00:12:18,370 And I think at CHI, we try to 315 00:12:18,370 --> 00:12:20,870 align our work with some of the regulatory 316 00:12:21,089 --> 00:12:22,610 momentum that's out there. 317 00:12:23,329 --> 00:12:25,409 You know, FDA has action plans out there. 318 00:12:25,409 --> 00:12:27,829 ONC has their HTI one rule, 319 00:12:28,769 --> 00:12:29,269 but 320 00:12:29,649 --> 00:12:31,730 I think what we hear from our community, 321 00:12:31,730 --> 00:12:33,875 and that's we are a community led nonprofit, 322 00:12:34,575 --> 00:12:36,115 is they want practical 323 00:12:36,495 --> 00:12:39,554 playbooks. They want practical tools, resources 324 00:12:40,735 --> 00:12:43,615 to make these things possible. So one of 325 00:12:43,615 --> 00:12:44,754 the things we're doing, 326 00:12:45,579 --> 00:12:47,740 in partnership with the joint commission is we're 327 00:12:47,740 --> 00:12:50,639 developing a set of governance playbooks for organizations 328 00:12:50,779 --> 00:12:51,759 of all sizes. 329 00:12:53,100 --> 00:12:55,339 And hopefully down the line, I mean, in 330 00:12:55,339 --> 00:12:58,779 the near future, accompanying tools, templates, resources that 331 00:12:58,779 --> 00:12:59,839 folks can leverage. 332 00:13:00,774 --> 00:13:02,855 And as part of that playbook, I think 333 00:13:02,855 --> 00:13:04,534 one of the biggest things that we wanna 334 00:13:04,534 --> 00:13:05,034 prioritize, 335 00:13:05,495 --> 00:13:07,735 not only getting feedback from our community, which 336 00:13:07,735 --> 00:13:09,014 we are gonna do. We have a series 337 00:13:09,014 --> 00:13:10,394 of workshops coming up, 338 00:13:11,894 --> 00:13:13,355 but also illustrating 339 00:13:13,735 --> 00:13:17,570 how differently sized organizations are aligning with some 340 00:13:17,570 --> 00:13:18,549 of these ethical 341 00:13:19,409 --> 00:13:21,329 guidelines that we're putting out there or some 342 00:13:21,329 --> 00:13:23,570 of these governance controls that we're putting out 343 00:13:23,570 --> 00:13:26,370 there in the playbook. So if we say 344 00:13:26,370 --> 00:13:29,009 everybody should have AI policy, what does that 345 00:13:29,009 --> 00:13:30,709 look like? How are different organizations 346 00:13:31,089 --> 00:13:31,909 doing this? 347 00:13:32,695 --> 00:13:34,695 So that folks have kind of a road 348 00:13:34,695 --> 00:13:37,095 map or template to look to. Say, oh, 349 00:13:37,095 --> 00:13:39,174 that worked for this size organization, which is 350 00:13:39,174 --> 00:13:41,095 similar to me, so maybe I could start 351 00:13:41,095 --> 00:13:41,595 there. 352 00:13:43,414 --> 00:13:45,735 And also being very clear about what the 353 00:13:45,735 --> 00:13:46,235 challenges 354 00:13:46,789 --> 00:13:49,509 are that people are facing in aligning with 355 00:13:49,509 --> 00:13:51,029 some of these ethical guidelines, 356 00:13:52,149 --> 00:13:52,649 and, 357 00:13:54,230 --> 00:13:56,809 you know, recommended governance controls 358 00:13:57,509 --> 00:13:59,269 and being able to say, like, okay. Are 359 00:13:59,269 --> 00:14:01,829 there tools, are there templates, resources that we 360 00:14:01,829 --> 00:14:03,845 can provide to help reduce some of that 361 00:14:03,845 --> 00:14:04,345 barrier 362 00:14:04,884 --> 00:14:07,445 and making sure that the ones who are 363 00:14:07,445 --> 00:14:09,524 resourced, who have it, and I can develop 364 00:14:09,524 --> 00:14:12,184 those things, great. For the ones who can't, 365 00:14:13,125 --> 00:14:14,804 how can we help them get a step 366 00:14:14,804 --> 00:14:15,304 up 367 00:14:16,600 --> 00:14:19,559 in that process and not widen that digital 368 00:14:19,559 --> 00:14:20,059 divide? 369 00:14:22,360 --> 00:14:24,379 And I think ethical AI 370 00:14:25,399 --> 00:14:28,200 isn't shouldn't just be within the health care 371 00:14:28,200 --> 00:14:30,860 system. I think a big part of 372 00:14:31,575 --> 00:14:34,054 what is now happening if we really want 373 00:14:34,054 --> 00:14:36,875 to create person centered or human centered 374 00:14:37,495 --> 00:14:37,995 solutions, 375 00:14:40,054 --> 00:14:41,434 it is a lot about 376 00:14:41,735 --> 00:14:45,254 the developers also understanding what ethics means in 377 00:14:45,254 --> 00:14:47,115 health care and being able to 378 00:14:47,600 --> 00:14:49,759 align their products with that so that when 379 00:14:49,759 --> 00:14:50,980 they go to their customers, 380 00:14:51,680 --> 00:14:53,779 the health care systems, and they say, 381 00:14:54,480 --> 00:14:56,399 I have this really cool thing that they're 382 00:14:56,399 --> 00:14:58,639 already speaking the same language. And I think 383 00:14:58,639 --> 00:15:00,160 that's a lot of what we're trying to 384 00:15:00,160 --> 00:15:02,340 do is get folks to hear each other's 385 00:15:03,654 --> 00:15:06,294 different languages and at least understand, like, do 386 00:15:06,294 --> 00:15:07,995 a little bit of translation there. 387 00:15:10,535 --> 00:15:12,075 And and it's about the relationships. 388 00:15:12,535 --> 00:15:15,414 So it's not it's also about how are 389 00:15:15,414 --> 00:15:17,894 we supporting each other, how are developers supporting 390 00:15:17,894 --> 00:15:19,019 their customers 391 00:15:19,480 --> 00:15:21,259 in in aligning, and how 392 00:15:21,639 --> 00:15:22,379 are the 393 00:15:22,759 --> 00:15:23,579 the health systems 394 00:15:23,879 --> 00:15:26,600 supporting the developers and improving the tools to 395 00:15:26,600 --> 00:15:27,899 better fit their needs. 396 00:15:28,360 --> 00:15:30,360 And so I think that's another big piece 397 00:15:30,360 --> 00:15:32,299 of ethical AI is that relationship 398 00:15:32,654 --> 00:15:33,154 between 399 00:15:34,175 --> 00:15:34,675 the 400 00:15:35,695 --> 00:15:37,075 the developer and the implementer. 401 00:15:39,134 --> 00:15:41,955 Wow. I you just mentioned so many important 402 00:15:42,014 --> 00:15:43,774 topics here. I think one of just an 403 00:15:43,774 --> 00:15:45,759 incredible quote you just, you know, said in 404 00:15:45,759 --> 00:15:47,519 the beginning as well, just nobody should wait 405 00:15:47,519 --> 00:15:49,279 for regulations to do the right thing. I 406 00:15:49,279 --> 00:15:52,559 think in essence, that was just amazingly said, 407 00:15:52,639 --> 00:15:54,740 doctor Gaines. So appreciate that. 408 00:15:55,120 --> 00:15:57,360 I wanna I I wanna look as as 409 00:15:57,360 --> 00:15:59,679 well to the future a little bit here 410 00:15:59,679 --> 00:16:01,845 for for our last question and and ask 411 00:16:01,845 --> 00:16:04,485 you just where do you really see AI 412 00:16:04,485 --> 00:16:06,325 headed in health care over the next two 413 00:16:06,325 --> 00:16:08,325 years, and what do you expect will be 414 00:16:08,325 --> 00:16:10,105 a little bit different because of AI? 415 00:16:12,565 --> 00:16:13,065 Yeah. 416 00:16:14,004 --> 00:16:15,065 So there's a lot 417 00:16:15,590 --> 00:16:16,970 happening in the space. 418 00:16:17,910 --> 00:16:21,050 I think we'll see definitely some expanded 419 00:16:21,750 --> 00:16:24,490 AI use in areas like ambient documentation, 420 00:16:25,110 --> 00:16:25,929 prior authorization, 421 00:16:27,110 --> 00:16:29,129 both on the provider and the payer side, 422 00:16:30,014 --> 00:16:31,634 and patient risk stratification. 423 00:16:32,334 --> 00:16:33,875 I think that's gonna continue, 424 00:16:34,254 --> 00:16:35,154 sort of predictive 425 00:16:35,534 --> 00:16:38,575 prescriptive analytics types pieces aren't it's not gonna 426 00:16:38,575 --> 00:16:40,735 go away even though GenAI has has come 427 00:16:40,735 --> 00:16:43,054 up. I think GenAI will help improve some 428 00:16:43,054 --> 00:16:43,954 of those processes. 429 00:16:46,480 --> 00:16:47,139 I think 430 00:16:48,720 --> 00:16:49,860 we have started 431 00:16:50,160 --> 00:16:52,500 focusing on a lot of back office tools 432 00:16:52,960 --> 00:16:53,460 as 433 00:16:54,000 --> 00:16:56,399 a first pass for AI and health care. 434 00:16:56,399 --> 00:16:58,179 Like, that's been the cautious approach. 435 00:16:58,634 --> 00:17:01,835 I think this will expand to frontline clinical 436 00:17:01,835 --> 00:17:02,815 decision support, 437 00:17:03,355 --> 00:17:05,535 and I think it will also 438 00:17:06,075 --> 00:17:07,055 move towards 439 00:17:08,875 --> 00:17:12,160 patients taking agency over their health care differently 440 00:17:12,460 --> 00:17:15,660 and interacting differently with their providers. Right? So 441 00:17:15,660 --> 00:17:16,160 with 442 00:17:17,819 --> 00:17:19,039 tools like ChatGPT, 443 00:17:19,500 --> 00:17:22,539 people are inevitably going to ask their health 444 00:17:22,539 --> 00:17:23,039 questions, 445 00:17:23,339 --> 00:17:23,839 bring 446 00:17:24,634 --> 00:17:26,975 a different kind of knowledge to their appointments, 447 00:17:27,115 --> 00:17:28,875 and so I think that will really shift 448 00:17:28,875 --> 00:17:30,575 the landscape quite a bit. 449 00:17:32,394 --> 00:17:34,634 I think there will be greater demand for 450 00:17:34,634 --> 00:17:35,134 transparency 451 00:17:35,755 --> 00:17:36,494 and evaluation 452 00:17:37,595 --> 00:17:39,914 over time, and I think we at CHI 453 00:17:39,914 --> 00:17:40,414 are 454 00:17:40,740 --> 00:17:42,200 developing sort of recertification 455 00:17:42,900 --> 00:17:44,119 programs for 456 00:17:45,380 --> 00:17:47,000 assurance resource providers, 457 00:17:47,299 --> 00:17:49,539 which might be folks that are validated, like 458 00:17:49,539 --> 00:17:51,799 third party validation of AI solutions, 459 00:17:52,914 --> 00:17:56,195 individuals that enable sort of local validation or 460 00:17:56,195 --> 00:17:57,494 training in a secure, 461 00:17:57,795 --> 00:17:59,414 privacy enhanced way, 462 00:18:00,195 --> 00:18:02,835 really making sure that people can actually access 463 00:18:02,835 --> 00:18:05,555 these tools in a way that is is 464 00:18:05,555 --> 00:18:06,055 safe. 465 00:18:07,700 --> 00:18:09,779 And then I think we'll see a lot 466 00:18:09,779 --> 00:18:10,279 more 467 00:18:10,820 --> 00:18:11,320 involvement 468 00:18:11,940 --> 00:18:15,299 from nursing staff, from community health centers, from 469 00:18:15,299 --> 00:18:16,519 under resourced providers, 470 00:18:16,980 --> 00:18:18,660 but we need to equip them with the 471 00:18:18,660 --> 00:18:21,140 tools and the training to really participate in 472 00:18:21,140 --> 00:18:21,880 this space. 473 00:18:22,715 --> 00:18:24,315 So one of the things that we are 474 00:18:24,315 --> 00:18:25,994 doing, I think there is quite a bit 475 00:18:25,994 --> 00:18:27,755 of curricula out there for, 476 00:18:28,715 --> 00:18:31,035 for physicians, and we are working on sort 477 00:18:31,035 --> 00:18:34,234 of expanding that into specialty society type work, 478 00:18:34,234 --> 00:18:36,700 but we also are partnering with Florida State 479 00:18:36,700 --> 00:18:39,740 University to bring a responsible AI education to 480 00:18:39,740 --> 00:18:40,240 nurses. 481 00:18:41,500 --> 00:18:42,320 And I think 482 00:18:42,940 --> 00:18:43,840 that will 483 00:18:45,420 --> 00:18:46,559 be pretty important. 484 00:18:47,100 --> 00:18:49,259 We have cross functional teams, and if only 485 00:18:49,259 --> 00:18:51,600 some parts of the team understand what's happening, 486 00:18:52,615 --> 00:18:55,174 and other parts don't, it can continue to 487 00:18:55,174 --> 00:18:57,335 increase the tensions, the silos in a space 488 00:18:57,335 --> 00:18:58,634 where we want actually 489 00:18:59,255 --> 00:19:01,835 better care continuity and better collaboration. 490 00:19:03,255 --> 00:19:05,174 So I think that's that's some of it. 491 00:19:05,174 --> 00:19:06,154 And then agentic 492 00:19:08,269 --> 00:19:09,089 things, hopefully, 493 00:19:09,390 --> 00:19:12,109 maybe will come out in health care. Maybe 494 00:19:12,109 --> 00:19:13,650 not in the next two years, 495 00:19:14,029 --> 00:19:16,670 but definitely down the line, especially for some 496 00:19:16,670 --> 00:19:18,990 of that back office stuff. I think for 497 00:19:18,990 --> 00:19:19,730 the administrative 498 00:19:20,029 --> 00:19:21,250 and operational work, 499 00:19:21,734 --> 00:19:23,595 it it makes a little bit more sense. 500 00:19:24,375 --> 00:19:24,875 But 501 00:19:25,654 --> 00:19:27,255 there are no solutions that I know of 502 00:19:27,255 --> 00:19:28,954 currently that are being deployed, 503 00:19:29,654 --> 00:19:31,255 in the health care system there, but it 504 00:19:31,255 --> 00:19:31,755 will 505 00:19:32,055 --> 00:19:32,795 come up, 506 00:19:33,494 --> 00:19:35,115 down the line for sure. 507 00:19:37,390 --> 00:19:39,869 Lots to look forward to. Lots evolving so 508 00:19:39,869 --> 00:19:42,109 quickly. I think, doctor Gain, you just had 509 00:19:42,109 --> 00:19:44,269 such great light for our audience on what 510 00:19:44,269 --> 00:19:46,029 this means for them and just the great 511 00:19:46,029 --> 00:19:48,109 work Chai is doing as well. So it 512 00:19:48,109 --> 00:19:50,195 doesn't become a place of AI have and 513 00:19:50,195 --> 00:19:51,955 have nots as I like to describe it 514 00:19:51,955 --> 00:19:54,835 where, you know, smaller system versus larger systems 515 00:19:54,835 --> 00:19:55,654 can't get that 516 00:19:55,955 --> 00:19:58,674 access to really, deploying this technology, like you 517 00:19:58,674 --> 00:20:01,234 said, ethically and safety. So I wanna just, 518 00:20:01,394 --> 00:20:03,234 give you a a a huge thank you 519 00:20:03,234 --> 00:20:05,730 for shedding light on there, sharing your insights 520 00:20:05,789 --> 00:20:07,309 on this, and as well as to our 521 00:20:07,309 --> 00:20:07,809 listeners, 522 00:20:08,109 --> 00:20:10,349 for taking the time out, listening to the 523 00:20:10,349 --> 00:20:13,309 Becker's podcast today. That's it for our episode 524 00:20:13,309 --> 00:20:15,309 today. Doctor Gaine, any last words or anything 525 00:20:15,309 --> 00:20:16,750 you would like to share with our audience 526 00:20:16,750 --> 00:20:18,849 before, we end today's episode? 527 00:20:19,654 --> 00:20:20,534 No. I think, 528 00:20:21,095 --> 00:20:22,634 join our newsletter list. 529 00:20:23,335 --> 00:20:25,095 Keep up to date even if you're not 530 00:20:25,095 --> 00:20:26,615 a member. If you want to be a 531 00:20:26,615 --> 00:20:28,694 member, feel free to reach out to us. 532 00:20:28,934 --> 00:20:30,694 We would love to chat and see how 533 00:20:30,694 --> 00:20:31,755 we can be helpful. 534 00:20:33,095 --> 00:20:33,595 And, 535 00:20:34,849 --> 00:20:36,609 yeah, I'm just great I'm grateful for the 536 00:20:36,609 --> 00:20:39,170 community. Without the community, like, the the things 537 00:20:39,170 --> 00:20:41,250 we put out, the understanding we have of 538 00:20:41,250 --> 00:20:43,089 the field just, like, really would not be 539 00:20:43,089 --> 00:20:43,570 possible. 540 00:20:43,890 --> 00:20:44,630 And so 541 00:20:45,089 --> 00:20:47,170 I'm grateful for that. We we, 542 00:20:47,570 --> 00:20:48,950 can only do this together. 543 00:20:50,154 --> 00:20:52,875 Actualizing that learning health system and learning community. 544 00:20:52,875 --> 00:20:54,714 Well, thank you again, doctor Gaines, and I 545 00:20:54,714 --> 00:20:57,195 appreciate your time today. Appreciate you. Thanks so 546 00:20:57,195 --> 00:20:58,015 much, Naomi.