1 00:00:00,000 --> 00:00:02,259 Hi, everyone. This is Laura Dierda with Becker's 2 00:00:02,279 --> 00:00:04,000 Healthcare. Thank you for tuning in to the 3 00:00:04,000 --> 00:00:06,719 Becker's Healthcare podcast series. We are thrilled today 4 00:00:06,719 --> 00:00:08,639 to be joined by doctor Holly Urban, vice 5 00:00:08,639 --> 00:00:11,699 president of business development strategy and clinical effectiveness 6 00:00:12,080 --> 00:00:14,160 at Walter's Kloer Health. Doctor Urban, it's a 7 00:00:14,160 --> 00:00:15,855 pleasure to have you on the podcast today, 8 00:00:15,934 --> 00:00:17,535 and thank you so much for joining us. 9 00:00:17,535 --> 00:00:19,214 We are excited to dive into your journey 10 00:00:19,214 --> 00:00:19,875 and experience. 11 00:00:21,054 --> 00:00:23,134 Thanks so much, Laura. I really appreciate the 12 00:00:23,134 --> 00:00:23,634 opportunity 13 00:00:23,934 --> 00:00:25,554 to, speak with you all. 14 00:00:25,934 --> 00:00:28,890 Fantastic. Well, let's dive right in. Doctor Urban, 15 00:00:28,890 --> 00:00:31,130 could you introduce yourself and your work in 16 00:00:31,130 --> 00:00:32,590 the health care to our listeners? 17 00:00:33,130 --> 00:00:35,289 Yeah. Absolutely. I'm Holly Urban. I am a 18 00:00:35,289 --> 00:00:36,670 pediatrician by training. 19 00:00:37,130 --> 00:00:40,250 I practiced in primary care pediatrics for several 20 00:00:40,250 --> 00:00:43,015 years before starting a second career in health 21 00:00:43,015 --> 00:00:44,695 care IT, and I've been in health care 22 00:00:44,695 --> 00:00:46,234 IT for about 15 years. 23 00:00:47,015 --> 00:00:50,074 In multiple leadership roles, I was vice president 24 00:00:50,215 --> 00:00:52,954 of product management at McKesson Horizon Clinicals, 25 00:00:53,460 --> 00:00:56,500 and then recently was CMIO at Oracle Cerner 26 00:00:56,500 --> 00:00:57,719 for their federal business. 27 00:00:58,100 --> 00:01:00,740 And I started working at Wolters Kluwer this 28 00:01:00,740 --> 00:01:01,539 year in, 29 00:01:02,020 --> 00:01:04,120 and my role in business development 30 00:01:04,579 --> 00:01:07,939 is really focused on innovation opportunities and how 31 00:01:07,939 --> 00:01:08,840 we do innovation, 32 00:01:09,594 --> 00:01:12,635 with some of Walter Scott's most influential and 33 00:01:12,635 --> 00:01:14,495 important products, which are our, 34 00:01:14,875 --> 00:01:16,015 up to date product, 35 00:01:16,954 --> 00:01:19,354 our up to date patient engagement, and, up 36 00:01:19,354 --> 00:01:21,674 to date Lexi drug content, formerly known as 37 00:01:21,674 --> 00:01:22,575 Lexi Comp. 38 00:01:22,909 --> 00:01:25,069 People who don't know the UpToDate product, it 39 00:01:25,069 --> 00:01:27,630 is probably the most used clinical decision support 40 00:01:27,630 --> 00:01:30,030 product used by physicians and other clinical users 41 00:01:30,030 --> 00:01:32,609 across the US. We have about 2,400,000, 42 00:01:34,030 --> 00:01:35,569 clinicians who use UpToDate, 43 00:01:36,270 --> 00:01:39,594 and clinicians really trust the content. They know 44 00:01:39,594 --> 00:01:41,134 that our 74100 45 00:01:41,834 --> 00:01:42,814 medical experts 46 00:01:43,114 --> 00:01:44,655 who are leaders in the field, 47 00:01:45,194 --> 00:01:47,614 they curate the content. They author the content 48 00:01:48,155 --> 00:01:49,694 as well as it's being updated 49 00:01:50,079 --> 00:01:52,719 every business day. During COVID, it was every 50 00:01:52,719 --> 00:01:53,219 day. 51 00:01:53,920 --> 00:01:56,400 So it's constantly being updated with the most 52 00:01:56,400 --> 00:01:58,479 important, most recent evidence. And so the our 53 00:01:58,479 --> 00:02:00,319 customers who use the content have a lot 54 00:02:00,319 --> 00:02:01,540 of trust in the content. 55 00:02:02,255 --> 00:02:03,615 So we did a big study of of 56 00:02:03,615 --> 00:02:04,115 providers 57 00:02:04,495 --> 00:02:06,655 this year that said, how do you think 58 00:02:06,655 --> 00:02:09,215 about generative AI tools? Since I'm in innovation 59 00:02:09,215 --> 00:02:11,215 with UpToDate, this is something I think about 60 00:02:11,215 --> 00:02:11,875 a lot. 61 00:02:12,254 --> 00:02:14,415 And, you know, we did learn that over 62 00:02:14,415 --> 00:02:17,170 40% of these users are are ready to 63 00:02:17,170 --> 00:02:19,569 use generative AI. And that's been a change 64 00:02:19,569 --> 00:02:21,409 over the past year. So that's been increased 65 00:02:21,409 --> 00:02:22,469 quite a bit, 68% 66 00:02:22,849 --> 00:02:24,849 over they're they have changed their views on 67 00:02:24,849 --> 00:02:26,550 generative AI since last year. 68 00:02:26,930 --> 00:02:27,750 But 91% 69 00:02:28,210 --> 00:02:30,764 said that any generative AI tool, they would 70 00:02:30,764 --> 00:02:33,405 wanna know where the materials were sourced and 71 00:02:33,405 --> 00:02:35,645 and know that they were created by physicians 72 00:02:35,645 --> 00:02:36,865 and medical experts, 73 00:02:37,165 --> 00:02:39,205 and they'd also want their vendors of j 74 00:02:39,324 --> 00:02:42,044 Gen AI tools to be transparent where the 75 00:02:42,044 --> 00:02:43,425 information came from. 76 00:02:43,849 --> 00:02:44,349 So 77 00:02:44,650 --> 00:02:47,050 as we think on up to date clinical 78 00:02:47,050 --> 00:02:48,030 decision support, 79 00:02:48,650 --> 00:02:50,650 we we know that there's some limitations. And 80 00:02:50,650 --> 00:02:52,730 and I'm and I'm not talking about, 81 00:02:53,530 --> 00:02:55,370 some of the really amazing tools out there 82 00:02:55,370 --> 00:02:57,870 that are all about physician efficiency. So, like, 83 00:02:58,364 --> 00:03:00,044 you know, some of the ambient listening or 84 00:03:00,044 --> 00:03:02,604 chart summarization tools. What I'm really talking about 85 00:03:02,604 --> 00:03:03,104 is 86 00:03:03,485 --> 00:03:06,444 AI tools that are focused on, clinical decision 87 00:03:06,444 --> 00:03:07,584 support, like diagnostics 88 00:03:08,044 --> 00:03:11,159 decision support or therapeutic decision support. And, you 89 00:03:11,159 --> 00:03:12,919 know, to be totally frank, we see a 90 00:03:12,919 --> 00:03:13,739 lot of limitations. 91 00:03:14,840 --> 00:03:17,019 You know, hallucinations are 1. I mean, 92 00:03:17,319 --> 00:03:20,359 even today, it continues to hallucinate. And when 93 00:03:20,359 --> 00:03:22,840 you're talking about making care decisions for a 94 00:03:22,840 --> 00:03:23,340 patient, 95 00:03:23,735 --> 00:03:25,735 that level of risk is just unacceptable for 96 00:03:25,735 --> 00:03:28,314 most clinicians. Again, it comes back to trust. 97 00:03:28,534 --> 00:03:30,854 So if if you're if you're concerned that 98 00:03:30,854 --> 00:03:32,854 there may be a hallucination, how can you 99 00:03:32,854 --> 00:03:35,334 possibly trust what the what the LM is 100 00:03:35,334 --> 00:03:37,514 telling you? The other is the inconsistency 101 00:03:37,814 --> 00:03:40,080 of the response. You ask the same question 102 00:03:40,080 --> 00:03:41,439 3 days in a row, you could get 103 00:03:41,439 --> 00:03:43,700 different answers from an LLM, which again, 104 00:03:44,719 --> 00:03:46,800 I think would impact trust. Right? So it 105 00:03:47,120 --> 00:03:48,800 how can you trust what they're really telling 106 00:03:48,800 --> 00:03:51,060 you if you're not getting a consistent response? 107 00:03:51,485 --> 00:03:53,004 And then the final, and this is something 108 00:03:53,004 --> 00:03:54,865 that I think about a lot is, 109 00:03:55,884 --> 00:03:57,664 is patient context. So 110 00:03:58,044 --> 00:04:00,284 one example is so let's say you, 111 00:04:00,685 --> 00:04:01,905 put it into an LLM, 112 00:04:03,004 --> 00:04:04,925 there's a patient who has a urinary tract 113 00:04:04,925 --> 00:04:06,224 infection, a bladder infection, 114 00:04:06,569 --> 00:04:08,890 and they're allergic to penicillin, so what antibiotic 115 00:04:08,890 --> 00:04:09,710 should I use? 116 00:04:10,170 --> 00:04:12,090 And you'll get a response, it's probably a 117 00:04:12,090 --> 00:04:12,590 fluoroquinolone. 118 00:04:13,849 --> 00:04:15,530 Great. I've got my answer. I'm gonna go 119 00:04:15,530 --> 00:04:16,990 write a prescription for a fluoroquinolone 120 00:04:17,449 --> 00:04:18,350 for my patient. 121 00:04:18,735 --> 00:04:20,254 Well, the challenge there is you don't have 122 00:04:20,334 --> 00:04:21,694 the L and M didn't have all the 123 00:04:21,694 --> 00:04:24,194 patient context. What if it's a pregnant patient? 124 00:04:24,735 --> 00:04:26,414 Patient is pregnant. You do not wanna be 125 00:04:26,414 --> 00:04:27,634 prescribing a fluoroquinolone. 126 00:04:28,334 --> 00:04:29,555 What if it's a child? 127 00:04:29,935 --> 00:04:32,354 Again, you do not wanna be prescribing fluoroquinolone. 128 00:04:32,735 --> 00:04:34,720 What if it's a recurrent bladder infection? What 129 00:04:34,720 --> 00:04:37,120 if it's a complicated bladder infection? There's just 130 00:04:37,120 --> 00:04:39,600 a lot of patient context that's there that 131 00:04:39,600 --> 00:04:41,379 you don't get in that LLM response, 132 00:04:42,000 --> 00:04:44,639 and that makes it really challenging again to 133 00:04:44,639 --> 00:04:46,959 trust what the response you're getting. And that's 134 00:04:46,959 --> 00:04:49,145 why at UpToDate is we're looking at some 135 00:04:49,145 --> 00:04:51,564 of these models is we're also really gauging 136 00:04:52,025 --> 00:04:53,725 how do we make sure that 137 00:04:54,105 --> 00:04:56,985 no hallucinations at 0 risk, so it's really 138 00:04:56,985 --> 00:04:58,845 grounded in the up to date content, 139 00:04:59,464 --> 00:05:02,389 as well as giving those nudges or giving 140 00:05:02,389 --> 00:05:03,769 that guidance to say, 141 00:05:04,069 --> 00:05:05,750 have you really thought about all the patient 142 00:05:05,750 --> 00:05:06,889 contextual factors? 143 00:05:07,750 --> 00:05:09,430 Well, it's fascinating to hear. You know, there's 144 00:05:09,430 --> 00:05:11,910 some really impressive results and things you're talking 145 00:05:11,910 --> 00:05:13,589 about that make a big difference in the 146 00:05:13,589 --> 00:05:14,714 clinical space. So 147 00:05:15,115 --> 00:05:16,714 excited for what's ahead, but I know as 148 00:05:16,714 --> 00:05:19,055 you mentioned, still things to work through. Yes. 149 00:05:19,274 --> 00:05:21,274 More hospitals and health systems right now are 150 00:05:21,274 --> 00:05:24,074 also getting serious about value based care. As 151 00:05:24,074 --> 00:05:26,634 a former care provider for pediatric patients, how 152 00:05:26,634 --> 00:05:27,834 do you view this shift, 153 00:05:28,235 --> 00:05:31,250 toward more proactive preventative care? Yeah. It's it's 154 00:05:31,250 --> 00:05:32,949 really interesting to me. So as a pediatrician, 155 00:05:33,250 --> 00:05:34,949 I mean, you're sort of grounded 156 00:05:35,330 --> 00:05:38,449 in proactive preventative care from the beginning. It's 157 00:05:38,449 --> 00:05:40,550 really the core of pediatric medicine. 158 00:05:41,090 --> 00:05:41,910 I mean, we 159 00:05:42,544 --> 00:05:44,625 do growth charts where you measure a patient's 160 00:05:44,625 --> 00:05:46,384 growth against the growth chart. And if they 161 00:05:46,384 --> 00:05:47,824 fall off the growth chart, that may be 162 00:05:47,824 --> 00:05:50,485 a sign of a problem. We do developmental 163 00:05:50,544 --> 00:05:52,865 screening to make sure kids are hitting their 164 00:05:52,865 --> 00:05:54,404 develop developmental milestones. 165 00:05:55,504 --> 00:05:59,149 We do anticipatory guidance to say, hey. Have 166 00:05:59,149 --> 00:06:00,350 you thought of you know, are you doing 167 00:06:00,350 --> 00:06:02,750 the right diet nutrition for your patients? And, 168 00:06:02,750 --> 00:06:03,970 of course, we do immunizations 169 00:06:04,350 --> 00:06:07,250 to make sure that you're preventing disease. So, 170 00:06:07,709 --> 00:06:09,730 so having that sort of proactive 171 00:06:10,574 --> 00:06:13,295 preventative mindset is very core to what I 172 00:06:13,295 --> 00:06:15,935 always did and what all pediatricians do. And 173 00:06:15,935 --> 00:06:17,694 so when you think about the broader scale 174 00:06:17,694 --> 00:06:19,694 of health care, you know, most physicians aren't 175 00:06:19,694 --> 00:06:21,454 trained in that model. It's really more, I'm 176 00:06:21,454 --> 00:06:23,535 gonna treat the patient that's sitting across from 177 00:06:23,535 --> 00:06:24,975 me in my office or that's in the 178 00:06:24,975 --> 00:06:25,475 hospital 179 00:06:26,069 --> 00:06:28,250 bed. So you're not thinking about the patient 180 00:06:28,470 --> 00:06:29,209 the other 181 00:06:29,509 --> 00:06:31,350 364 days of a year when you're not 182 00:06:31,350 --> 00:06:33,669 seeing them because there aren't economic drivers to 183 00:06:33,669 --> 00:06:34,569 do that, frankly. 184 00:06:34,949 --> 00:06:36,949 So I really that's why I've been an 185 00:06:36,949 --> 00:06:39,269 early proponent of moving toward the more value 186 00:06:39,269 --> 00:06:42,055 based care to really understand how do what 187 00:06:42,055 --> 00:06:43,595 do we need to do for our patients 188 00:06:43,894 --> 00:06:45,115 in those other 364 189 00:06:45,495 --> 00:06:47,814 days so that they aren't consuming health care 190 00:06:47,814 --> 00:06:50,055 services, so they are staying healthy, so they 191 00:06:50,055 --> 00:06:51,654 don't have to come see me. And that's 192 00:06:51,654 --> 00:06:53,354 what value based care is all about. 193 00:06:54,120 --> 00:06:55,480 That makes a lot of sense. You know, 194 00:06:55,480 --> 00:06:57,319 it is amazing to hear a great vision. 195 00:06:57,319 --> 00:06:59,180 And I I love the idea of pediatrics 196 00:06:59,319 --> 00:07:00,300 being so proactive, 197 00:07:00,920 --> 00:07:02,600 from the get go and then instilling that 198 00:07:02,600 --> 00:07:04,620 within other areas of the health care system. 199 00:07:05,240 --> 00:07:06,379 How can health care 200 00:07:06,845 --> 00:07:09,985 leverage systems thinking and design systems that facilitate 201 00:07:10,044 --> 00:07:12,125 the delivery of the highest quality care to 202 00:07:12,125 --> 00:07:13,264 generate better outcomes, 203 00:07:13,564 --> 00:07:15,345 increase safety, and improve efficiency, 204 00:07:15,725 --> 00:07:17,904 especially in this move towards value based care? 205 00:07:18,044 --> 00:07:20,720 Yeah. Those things these things are very intertwined. 206 00:07:20,939 --> 00:07:23,420 So what I would argue is that in 207 00:07:23,420 --> 00:07:25,199 how we call our health care system, 208 00:07:25,740 --> 00:07:27,360 it isn't a system at all. 209 00:07:27,740 --> 00:07:30,160 It's a very broken, fragmented, 210 00:07:30,860 --> 00:07:31,360 siloed 211 00:07:32,204 --> 00:07:34,944 with very, sort of perverse economic incentives. 212 00:07:35,564 --> 00:07:37,805 And and the data shows that the system 213 00:07:37,805 --> 00:07:39,324 isn't working. You know, if you look at 214 00:07:39,324 --> 00:07:40,704 the Commonwealth Fund, 215 00:07:41,485 --> 00:07:44,064 every year they publish that we spend 216 00:07:44,769 --> 00:07:46,849 3 to 4 times more in our health 217 00:07:46,849 --> 00:07:47,509 care dollars 218 00:07:47,810 --> 00:07:50,469 than other developed countries. And yet our outcomes 219 00:07:50,529 --> 00:07:53,009 are significantly worse. I mean, it's very, very 220 00:07:53,009 --> 00:07:53,509 sobering 221 00:07:53,889 --> 00:07:55,329 that we are just we are just not 222 00:07:55,329 --> 00:07:57,169 getting value for the health care dollars that 223 00:07:57,169 --> 00:07:57,829 we spend. 224 00:07:58,165 --> 00:07:59,845 And so that's why I've been talking a 225 00:07:59,845 --> 00:08:02,964 lot about how systems thinking needs to be 226 00:08:02,964 --> 00:08:05,045 addressed. Because if you if you throw in 227 00:08:05,045 --> 00:08:08,024 other issues like, you know, workforce burnout 228 00:08:08,645 --> 00:08:09,145 and, 229 00:08:09,605 --> 00:08:12,180 and bias, like, if you don't think about 230 00:08:12,180 --> 00:08:14,900 those issues with the systems thinking approach, it's 231 00:08:14,900 --> 00:08:17,139 almost impossible, because they're systemic issues. So how 232 00:08:17,139 --> 00:08:18,900 do you fix it without taking a systematic 233 00:08:18,900 --> 00:08:21,400 approach into a broken system? So 234 00:08:21,779 --> 00:08:25,185 what systems thinking is, it's all about designing 235 00:08:25,185 --> 00:08:27,745 a system to meet the outcomes you're looking 236 00:08:27,745 --> 00:08:28,245 for. 237 00:08:28,625 --> 00:08:31,024 So that's like, in health care, obviously, it's 238 00:08:31,024 --> 00:08:32,884 better care at lower costs. 239 00:08:33,584 --> 00:08:36,004 And, you know, so many health care systems 240 00:08:36,144 --> 00:08:38,245 do process improvement or quality improvement 241 00:08:38,679 --> 00:08:41,000 where the focus is really on fixing people. 242 00:08:41,000 --> 00:08:43,000 Like, how do you change their mindsets or 243 00:08:43,000 --> 00:08:45,399 how do you change their workflow? But, again, 244 00:08:45,399 --> 00:08:47,240 I'd argue that if the system that they're 245 00:08:47,240 --> 00:08:49,080 working in isn't designed to meet the outcome 246 00:08:49,080 --> 00:08:51,240 you want, just trying to fix the people 247 00:08:51,240 --> 00:08:53,995 is not gonna lead to sustainable or systemic 248 00:08:54,054 --> 00:08:55,975 change. So that's why I really advocate for 249 00:08:55,975 --> 00:08:58,535 this broader systems thinking to make sure that 250 00:08:58,535 --> 00:09:00,934 you're designing the entire system and all of 251 00:09:00,934 --> 00:09:02,634 the people in it and all the processes 252 00:09:02,855 --> 00:09:03,835 and all the technology 253 00:09:04,295 --> 00:09:06,570 to meet your intended outcomes. And there's an 254 00:09:06,570 --> 00:09:07,070 interesting, 255 00:09:07,529 --> 00:09:10,410 so the leaders at the IHI published an 256 00:09:10,410 --> 00:09:13,070 article in the Harvard Business Review in July, 257 00:09:13,769 --> 00:09:15,450 that talks about this. So if anybody wants 258 00:09:15,450 --> 00:09:17,050 to read more about it, I highly recommend 259 00:09:17,050 --> 00:09:19,294 the article. It's really well done. And they 260 00:09:19,294 --> 00:09:22,735 call out, James Clear's book, Atomic Habits, and 261 00:09:22,735 --> 00:09:24,735 he says, you do not rise to the 262 00:09:24,735 --> 00:09:26,495 level of your goals. You fall to the 263 00:09:26,495 --> 00:09:27,634 level of your systems. 264 00:09:28,975 --> 00:09:30,674 That's fascinating. What a poignant, 265 00:09:31,375 --> 00:09:32,894 thing to keep in mind in the health 266 00:09:32,894 --> 00:09:34,434 care space and truly 267 00:09:34,850 --> 00:09:36,610 overall just a really important, 268 00:09:36,930 --> 00:09:39,009 lesson, to think about on a regular and 269 00:09:39,009 --> 00:09:41,830 daily basis. Systems are definitely make a difference. 270 00:09:42,050 --> 00:09:43,730 Now I know you're currently working on a 271 00:09:43,730 --> 00:09:46,769 publication that's going to highlight the generative AI's 272 00:09:46,769 --> 00:09:49,649 potential in health care and practical applications for 273 00:09:49,649 --> 00:09:50,149 clinicians. 274 00:09:50,504 --> 00:09:52,504 Can you share your insights on how AI 275 00:09:52,504 --> 00:09:54,764 can disrupt the traditional health care models 276 00:09:55,144 --> 00:09:57,465 and what challenges health care organizations should really 277 00:09:57,465 --> 00:10:00,105 be aware of? Yeah. Absolutely. So the, the 278 00:10:00,105 --> 00:10:03,644 textbook is Digital Health AI and Generative AI, 279 00:10:03,899 --> 00:10:06,699 a concise practical guide for clinicians. It's gonna 280 00:10:06,699 --> 00:10:07,259 come out, 281 00:10:07,819 --> 00:10:10,940 Springer Nature's publishing it edited by Terri Adiram 282 00:10:10,940 --> 00:10:11,759 in in 2025. 283 00:10:12,699 --> 00:10:14,639 And my chapter is on implementing 284 00:10:14,940 --> 00:10:15,839 IA tools. 285 00:10:16,514 --> 00:10:18,915 Again, you have to take a systems thinking 286 00:10:18,915 --> 00:10:21,315 approach when you're implementing these tools. It's not 287 00:10:21,315 --> 00:10:23,075 a very simple plug and play, like, I'm 288 00:10:23,075 --> 00:10:24,855 gonna plug in an EKG machine, 289 00:10:25,315 --> 00:10:26,754 get a patient the leads on, and I'll 290 00:10:26,754 --> 00:10:28,434 get an EKG out of it. It's far 291 00:10:28,434 --> 00:10:30,370 more complex than that. You really need to 292 00:10:30,370 --> 00:10:32,389 think about all of the people and processes 293 00:10:32,450 --> 00:10:34,450 that are gonna be involved with the tool 294 00:10:34,450 --> 00:10:34,950 itself. 295 00:10:35,490 --> 00:10:37,809 So just to take an example, let's talk 296 00:10:37,809 --> 00:10:38,309 about, 297 00:10:39,169 --> 00:10:41,490 a a patient inbox where you're where you're 298 00:10:41,490 --> 00:10:44,230 using AI to do triage of patient messages. 299 00:10:44,975 --> 00:10:47,375 So I would say taking a systems thinking 300 00:10:47,375 --> 00:10:49,774 approach means you start with what metrics are 301 00:10:49,774 --> 00:10:52,574 you gonna follow to know that you're going 302 00:10:52,574 --> 00:10:53,074 to, 303 00:10:53,695 --> 00:10:56,495 meet the outcomes you're looking for, and who's 304 00:10:56,495 --> 00:10:58,334 tracking them, and who's owning them, and who's 305 00:10:58,334 --> 00:10:59,714 reporting out on those metrics, 306 00:11:00,269 --> 00:11:01,090 what resource 307 00:11:01,550 --> 00:11:03,230 is managing those, 308 00:11:03,550 --> 00:11:04,529 triage messages, 309 00:11:05,070 --> 00:11:06,050 how are messages 310 00:11:06,430 --> 00:11:09,070 rerouted to other clinical users when needed, and 311 00:11:09,070 --> 00:11:10,850 what are the criteria to reroute, 312 00:11:11,629 --> 00:11:14,985 and and auditing? So are there any messages 313 00:11:14,985 --> 00:11:17,464 falling through the crack? That's that's to cover 314 00:11:17,464 --> 00:11:17,964 risks. 315 00:11:18,345 --> 00:11:20,184 Is there AI drift? I mean, we know 316 00:11:20,184 --> 00:11:22,745 a lot about these AI models that they're 317 00:11:22,745 --> 00:11:24,985 trained in, you know, with the vendor side, 318 00:11:24,985 --> 00:11:27,404 but once they get into patient population models, 319 00:11:27,990 --> 00:11:29,669 that they can drift and they can change. 320 00:11:29,669 --> 00:11:31,450 So who the only way to really 321 00:11:31,830 --> 00:11:34,470 see that happening is through audits. And then 322 00:11:34,470 --> 00:11:37,290 finally, addressing issues of bias. I mean, doing 323 00:11:37,750 --> 00:11:40,745 subgroup population analysis to make sure that you're 324 00:11:40,745 --> 00:11:42,784 not seeing any bias in the in the 325 00:11:42,784 --> 00:11:45,585 AI response. And then finally, the most important 326 00:11:45,585 --> 00:11:47,345 is governance. Right? Who has governance over all 327 00:11:47,345 --> 00:11:49,184 of that? So you really need to be 328 00:11:49,184 --> 00:11:51,985 thinking people process and technology when you're implementing 329 00:11:51,985 --> 00:11:52,725 these tools. 330 00:11:53,470 --> 00:11:55,490 That's fascinating to hear and certainly 331 00:11:55,870 --> 00:11:58,350 a very poignant reminder in terms of how 332 00:11:58,350 --> 00:11:59,730 you can really 333 00:12:00,429 --> 00:12:02,990 most effectively implement these tools, get them out 334 00:12:02,990 --> 00:12:05,149 to the broader system, and make sure that 335 00:12:05,149 --> 00:12:07,375 they're used and optimized correctly. I know that 336 00:12:07,375 --> 00:12:09,294 can be a big challenge, and a lot 337 00:12:09,294 --> 00:12:10,514 goes into it. So 338 00:12:10,894 --> 00:12:13,294 fascinating to hear. Is there anything else you'd 339 00:12:13,294 --> 00:12:14,654 like to share with our audience that we 340 00:12:14,654 --> 00:12:16,735 haven't already touched on? No. I think that's 341 00:12:16,735 --> 00:12:18,735 great. Thank you again for the opportunity to 342 00:12:18,735 --> 00:12:20,834 chat today, and, I enjoyed the conversation. 343 00:12:21,490 --> 00:12:23,170 Doctor Urban, thank you so much for joining 344 00:12:23,170 --> 00:12:24,769 me on the podcast today. It has been 345 00:12:24,769 --> 00:12:26,769 such a pleasure speaking with you. I'd also 346 00:12:26,769 --> 00:12:29,090 like to thank our podcast sponsor, Wolters Kluwer 347 00:12:29,090 --> 00:12:30,930 Health. You can tune in to more podcasts 348 00:12:30,930 --> 00:12:33,570 from Becker's Healthcare by visiting the podcast page 349 00:12:33,570 --> 00:12:34,309 at beckershospitalreview.com.