AI Clinical Documentation Tools That Actually Save Time with Ed Stern
If you’ve ever wondered what AI clinical documentation tools actually look like once someone puts real hours into building and testing them, this episode of Run Your Private Practice with AI is for you. My guest, Ed Stern, a psychiatric nurse practitioner who owns Nothing Better Health and Queer Ear Mental Health, has spent years pulling apart what works, what doesn’t, and what he’s had to rebuild from scratch when a tool didn’t hold up.
Ed walks through the safeguard he built directly into his charting software, why he refuses to sign long-term contracts with any AI vendor no matter how much money it could save him, and the questions he now asks before trusting a tool with sensitive patient data. If you’re trying to figure out where AI actually earns a place in your documentation workflow and where it doesn’t, this conversation gives you a real look at how one clinician made those calls.
AI Tips You Don’t Want to Miss:
- The billing safeguard Ed built directly into his charting tool, and why it changes how he approaches CPT codes.
- Why a free, HIPAA-compliant tool with a BAA became his go-to for evidence-based clinical answers.
- The small language shift that keeps his notes cleaner without changing how his AI listens to sessions.
- How Ed traded hours of spreadsheet work for a five-minute business decision using data he already had.
- The one rule he will not break with any AI vendor, no matter what it might save him.
Connect with Ed Stern:
Website: https://www.queerearmentalhealth.com/
Website: https://www.nothingbetterhealth.com/
Email: ed@estern.com
Mentioned in this episode:
Run Your Private Practice with AI (free Facebook group)
https://www.facebook.com/groups/aiprivatepractice/
Clinical AI Club: https://www.clinicalaiclub.com
Thera AI Hub: https://kymtolson.kartra.com/page/AI-Tools-for-Therapists
Connect with me:
- Instagram: @thetravelingtherapist_kym
- Join the Clinical AI Club: www.clinicalaiclub.com
Sponsored by Alma: www.helloalma.com/kym
Sponsored by Berries:www.heyberries.com/therapists
Sponsored by Auralink Pro: Start your free trial and use code KYMTOLSON for 20% off www.auralink.com/pro
Transcript:
[00:01:22] Hey, everybody. Welcome back to another episode of Run Your Private Practice with AI. This is my second guest on this podcast. I mentioned this a couple weeks ago. I’m starting to bring in members of the TheraAI Hub and the Clinical AI Club that I have met up with, I’ve talked to, that are doing really cool things with AI, and I just got this idea.
[00:01:43] Kym: I was like, “Why don’t I bring them on the podcast so they can share what they’re doing, also share their experience in the memberships?” But really I hope it’s energizing for people listening to hear how other clinicians in the field are using AI, learning about AI, and building things. You know, getting ideas and building things.
[00:02:01] And Ed Stern is here with us. Ed has done a lot. There’s, like, a million things he’s already done, so I wanna talk about some of that, geek out a little bit about AI.
[00:02:09] But Ed, Could you introduce yourself and let people know? ‘Cause you’re not, like, a therapist like we are. You’re a nurse practitioner. Let’s talk about that a little bit, who you are and what you do.
[00:02:17] Edward: I would love to. So my name’s Ed Stern. I’m a psychiatric nurse practitioner. I own a practice that provides services in multiple states. I have multiple other nurse practitioners that are part of the practice, and then I also do IOP and PHP work for a third party as well.
[00:02:35] So do, do a variety. I am part of a subset of psychiatric nurse practitioners who probably take pride in the fact that we also have taken extra steps to make sure that we’re good therapy providers as well. While it’s not the primary space that I, you know, that I work in I prefer people much more to have their outside s- you know, therapist.
[00:02:57] But in crisis, if, you know, i- instead of me saying, “Well, go find a therapist,” if you’re in crisis and you need help, you know, we- we’ll, we’ll definitely do, you know, some good psychotherapy too. And I think that’s what attracted me originally to, you know, to the group is I was like, “Ooh, like, what are the therapy tools that are out there?”
[00:03:16] And then that started this, like, insane cascade in my own geeky brain about, “Oh my God, wait, what about, what about, what about?” And, you know, and all the other things. Yeah. You know, so thank you for making me stay up late at night playing with Claude and other things like that. I’ll make sure my husband knows to blame you.
[00:03:34] Kym: Send him to me. Say, “Well, take it up with Kim. It’s her
[00:03:37] Edward: I will do that in a heartbeat. Absolutely
[00:03:39] Kym: Because that’s true. I mean, I do have a tendency to dump… Every month I dump modules, and I do have a tendency to dump a lot of stuff, you know? It like, it kinda goes against the grain of most memberships. They say, “Don’t overwhelm your members,” but AI moves so fast, you know?
[00:03:52] It’s like from month to month, it’s like, ” Oh my gosh, well, I’ve gotta add this, and I’ve gotta make sure I tell them about this.” So I’m always adding stuff in and, you know… And then we have the mastermind. We have, like, three, well, two masterminds a month, and they’re 90 minutes, but I always go two hours and, you know, there, there’s a ton of content in there, too.
[00:04:07] So yeah, I, I hear you, and if you’re anything like me or anybody listening is like that, I don’t know. I like, I’m obsessed with this stuff. I go down this rabbit hole, and then it’s like I gotta keep building and figuring out more about it, and I get so creative with it. And I think you’re like that too, Ed, it seems like to me.
[00:04:22] Edward: I can be. I could blame my ADHD if I want But yeah, I c- I absolutely can be, and I think, you know, I’m, from a background perspective, I actually left healthcare for a bunch of years and ended up in the business world, and then ended up in the technology sector.
[00:04:38] So, you know, I look at it you know, I’ve been, I made it to a vice president level of a large company in the technology sector. And so, you know, I get enough of the tech that I’m probably a little bit dangerous. But as, as you said, Kim, right, you know, it changes so much, it leapfrogs. Like, you know, if you blink, it has gone far beyond where, where anybody would’ve thought.
[00:05:01] So the you know, the cool things are that if, you know, today, if you don’t know enough, you can really pretty much just say to AI, “Hey, I wanna do blank.” You know, so my practice y- you know, I built the practice in a very unique way. I- we have– The practice is, is n- called Nothing Better Health. We predominantly operate under a different brand called Queer Ear Mental Health.
[00:05:27] We work mostly with the LGBTQ community in the states that, that we work in, and it’s umbrella-ed in such a way to sort of protect Inadvertently outing people or, or stigma associated with it. So I was working with somebody recently, helping them. Uh, A colleague of mine reached out. They want to do gender-affirming letters and I sort of do it very organically because it’s so much part of what I do.
[00:05:53] And I just before I jumped in here, I almost was late to jump in here because I was in Claude basically saying, “Let’s put together a questionnaire using the WPATH guidelines, using this,” you know, and other things. I think, you know, there’s so many different tools that we have. Clinically, I spend a lot of time– There’s others out there, but I spend a lot of time in Open Evidence which is, you know, a great clinical AI.
[00:06:22] You know, and it works very nicely. It provides a lot. It’s, you know, it only works in the clinical space with evidence-based research as opposed to, “Hey, I found this on, you know, Reddit.” It’s like, no, no, no. You know? Like, I want the articles. I want to be able to open the article. I want to be able to validate and things, you know.
[00:06:41] And it does very powerful things for us clinically. Use Claude really more for structure and, and, you know, an organization. And then sometimes I’ve even done– Like, one of the tools I put together recently, because I teach also is I was trying to put together a tool for, like, side, side effect comparison for meds.
[00:07:02] Not necessarily something that therapists might do, but, you know… So what I did was I gathered all of the data and I had Open Evidence create the data and a basic heat map so that people could. And I took all of that hard clinical evidence stuff, all of the research, and then I dumped that into Claude and I said, “Now make me a tool.”
[00:07:26] So now my students can say, “I want a, I want a medication,” and this happened to be sec- SGAs, second-generation antipsychotics. “I’m looking for an antipsychotic that does this, this, this and has these few, like, like, low side effects for this and this. What are my choices?” And, you know, it’s gonna filter, it’s gonna through, and then it’s gonna provide the clinical evidence.
[00:07:51] So, you know, it’s, it’s sort of like a trainer’s version of a, evidentiary-based, you know, clinical decision support tool. you know, it’s nothing that like the tools that cost hundreds of thousands of dollars that big hospitals
[00:08:04] Kym: wow, but I actually saw it. I actually like, played around with it a little bit. You
[00:08:08] Edward: Oh, that’s right. I shared it. Yeah. Yeah. I shared it.
[00:08:11] Kym: it’s awesome. Oh
[00:08:12] Edward: Yeah. You know, and so like, why not use it? And you gave me some great advice of, of like some places to host it, and I’m gonna tap your brain about that because I kinda– like I have some of these others and I wanna find a place to sort of make these available.
[00:08:27] It’s not my data, so I’m not look- you know, I’m not, I’m like, but these are tools and they’re, you know, they’re kinda cool. And if we wanna help our colleagues learn that AI can do things for us as opposed to be, part of the evil empire, which maybe or maybe it isn’t, I don’t, I don’t know yet, But I think there’s so many things, you know, we can do. the AI that I use for charting, I’ve had great conversations. I, you know, we’re based here in the Washington DC area, and so I have policy wonks who are my patients, and I have spoken to them about like, “Where’s your comfort level with AI for charting?”
[00:09:03] Like, “What are you comfortable with?” And, you know, and those kinds of things. And so, the way we’ve structured our AI to work clinically is, a very specific way based on a very privacy-sensitive patient population I deal with, layered on top of a very privacy-centric, group of people who are government employees who are very concerned about, any number of different things. And, I’ve made some
[00:09:31] Kym: is the sentiment? I mean, what, w- what is their comfort level like? I’m
[00:09:35] Edward: I think it varies. You know, I mean, I, in, in all honesty, I think it varies. I have some, I have some people who are very linear thinkers, possibly because of some of the clinical diagnoses that they have and other things like that, and so they’re linear. Sometimes they’re more rigid. so I think it depends.
[00:09:53] We’ve taken certain steps with– So, like, for example, we use a third-party AI product. And among the things we’re, we’re doing is, so we made sure it was a product that doesn’t record audio. Audio is too you know, too able to be cross-checked. It creates a transcript. It automatically deletes the transcript after 14 days unless I delete it sooner.
[00:10:15] So, you know, what we’re ultimately left with is just a note, and we keep it. We could delete it sooner, but we keep it for 14 days so that I have the ability to, you know, manipulate and add and, and other things if, you know, for certain more sensitive s-scenarios. so for example, sometimes I’ll have, I’ll, I’ll interview– you know, we see minors, and so sometimes I’ll have an interview on day one with parents and then a couple of days later with the minor.
[00:10:40] And so I keep the transcripts so I can get a, you know, a cohesive. The transcripts don’t have patient names on them at all. They’re uh, it’s a code identifier, so it’s less like e-even though it’s a HIPAA-compliant system, that if for some reason something bad happened to their system, it’s not like it would say, “Oh, this is Kim.”
[00:11:01] You know? Um, So we take some moo- we take some movements like that.
[00:11:05] Kym: And you did that on purpose, right? Like, the, the system didn’t come like that, right? You
[00:11:10] Edward: No, we did it with– Well, I
[00:11:12] Kym: de-identified,
[00:11:13] Edward: we had to, yeah, I mean, we had to first find a place that wasn’t… Like, there were some of the products we experimented with that were recording audio and deleting the audio was, like, far too many steps.
[00:11:24] There wasn’t an auto process or anything like that. So, like, that was immediately were out of the question for us. So then we looked at ones that do transcripts only. What was important for me clinically was that it was able to discern the players in the transcript, right? You know, ’cause, like, and so this one advertised that they can handle group therapy sessions.
[00:11:46] It’s kind of designed for our com- so I was like, “Ooh, okay.” So that means it can discern. And, you know, then there were a couple of other factors as well. Being that we do medications, some of the products that are out there, the dictionaries that they use, the databases that they use for meds and, you know, is some of it’s public domain, others use commercial product.
[00:12:10] We found one that uses a commercial database, so it’s less likely to screw up the spelling or the misinterpretation of, h- like if I say, you know, s- this is not product placement, but if I say I’m gonna prescribe Ad-Addzenys to you or Qelbree to you, those are two ADHD meds, I-I want it to be as close as possible to the correct spelling if it, can be.
[00:12:35] And, if it’s pulling stuff out of a data file of medications as opposed to randomly just getting something that’s freeware, which some of the companies that we pay a lot of money to actually are doing behind the scenes, I’m like, “No.” Like, you know, so there was a lot of criteria ultimately I wanted a product that offered a great deal of flexibility in templates, and I can build a lot of big templates in it.
[00:13:03] You know, one thing I did with my product recently, and it was actually, it came from a conversation, I don’t remember the exact context, but it came from a conversation in the group. I am transitioning about two-thirds of my patient load over to some of my colleagues that are part of my practice ’cause I’m moving on to do some other things.
[00:13:21] And so I was like, “How can I do this intelligently and simply and streamline some things?” And there was a conversation in the group, and the idea popped into my head, “Oh my God, I have more than a year’s worth of notes in the AI for my patients who have been with me for, some of them have been with me for five plus years, whatever.” So can I ask the AI to summarize the care,
[00:13:45] Kym: Yes.
[00:13:47] Edward: their product allows me to reference up to 10 prior notes in one note generation,
[00:13:55] Kym: Like in one context? Uh-huh. Yeah. Nice
[00:13:58] Edward: for some people that’s, ages, right? Because if I’m seeing them on a 90-day period for med management, then 10 plenty. If I’m seeing some people every two weeks ’cause they’re acutely ill, then it’s a different situation.
[00:14:12] But 10 notes is a pretty good, you know, snapshot. So I was like, I, I went in, I said, “I wonder if it’ll do this.” It’s like, create a summary. and sort of create a, like a clinician handoff summary document. And I
[00:14:27] Kym: you know what it just, you know what that just made me think? Like, what if you said, “Here’s all my clinicians and here’s their specialties. Which clinician would be best for this client?” Like, yeah, that, I don’t know why that just popped in my head,
[00:14:37] Edward: Yeah, I mean, that’s a good point. I, I mean, I, you know, I have five clinicians, so I,
[00:14:42] Kym: Yeah, so you
[00:14:42] Edward: my AI, my, my very personal AI that, you know, has 62 years of programming knows the
[00:14:49] Kym: like, “I already know that.”
[00:14:50] Edward: right? You know, but, you know, yes, no, I think you’re right, right? I mean, that’s– and that’s the danger that we can get into with AI.
[00:14:57] And I, you know, I air quote “danger,” of course, because I think it’s really good. It’s like, I wonder if it can do this, and it did it. It was the same with that SGA side effects. I was like: I wonder if I can do this. And I was– I played it first, I played around with Claude and I was like, “Yeah, but like, am I really trusting the ref?”
[00:15:15] And I was like: Oh my God, wait, I forgot I have, you know, open evidence. And, you know, for anybody who doesn’t know open evidence, it’s any of us that have an NPI can have it and it’s free.
[00:15:27] Kym: That’s
[00:15:28] Edward: And it’s all, all clinically reference-based. There’s got, they’ve got, you know, massive libraries, lots of psychology li- you know, references and other things.
[00:15:38] And so it’s only searching that realm and not going out into, you know, the, the true ether. And it’s, you know, it’s pretty good. Like I’ll, I’ll– I have asked it before for, you know, letters and whatnot, and much like one of your presentations talked about, it’s HIPAA compliant with a BAA, right? Whereas others, whereas others are saying, “Oh, it’s for the clinical space,” and, you know, but like, boom, HIPAA compliant, BAA free.
[00:16:09] Kym: did I see that Claude has a connector to OpenEvidence? Like, you know how when you’re in Claude You can connect
[00:16:15] Edward: haven’t looked, but now that’s one of the things I’m gonna look for
[00:16:18] Kym: because I swear I thought I saw that, but maybe it was a different… There are some, like, clinical things that you can connect to in Claude. I just can’t remember which ones, but I think that might have been one of
[00:16:27] Edward: It might be. It might be. There’s a couple of other products out there. You know, I just I, I, you know, it, the Doximity has an AI. It’s not super, super powerful. There’s a newer company that’s out that’s also ad-driven like Open Evidence is that I haven’t played with as much. I mean, I think this is gonna be a moving target.
[00:16:45] So today we’re saying Open Evidence, tomorrow it could be Joe’s, you know, clinical psychology research tool. You know, who knows? And I think we just have to be willing to, you know, to, to, to move and shake and, you know, and shift w- with it. You know, I use like anything that’s– I, I still use de- de-identified info in Open Evidence, even though it talks about, you know, that it’s AI, you know, that it’s HIPAA, et cetera, et cetera.
[00:17:15] But I was doing something recently where I was, like a parent had sent me a couple hundred-page PDF and like, “Here.” You know? And I’m like,
[00:17:23] Kym: Oh my God.
[00:17:24] Edward: I am a pathetically empathic, practitioner,
[00:17:27] Kym: I will spend all night reading
[00:17:28] Edward: I am not gonna read… And this was, this was from a clinical space, and so it was, you know, a lot like, basically it was visit notes, and every visit note basically had like the same template, and it was like I was really looking for punchlines.
[00:17:41] And in my case, what I was looking for is indications of medications and why they were why they were stopped, right? And so I was like, I wonder if O- you know, if Open Evidence can help me with this.
[00:17:56] Kym: wow.
[00:17:57] Edward: I made sure that I, I used a I used a redaction tool to take that, like, ’cause the name was in the header.
[00:18:06] So I used a redaction tool for that. And so I fed into Open Evidence basically in you know, sort of, kind of unidentified. I mean, like, you know, if you f- if you’re a, a HIPAA purist, yeah, sure, there was probably identifying stuff in there, but again, it’s a HIPAA-compliant tool. And I sent it into that and I said, “Extract from,” and it, you know, it did better than I would’ve done on it with 100 and some odd pages, right?
[00:18:33] You know, it gave me, it gave me
[00:18:34] Kym: half half a second where it
[00:18:36] Edward: gave me a
[00:18:36] Kym: eight
[00:18:37] Edward: sense of what was going on and, you know, is it 100% accurate? Well, I wouldn’t know unless I put them side by side. But I was trying to get an understanding, an abstract, and so the best thing, you know, I mean, it gave me an abstract. You know, so there’s, like, I think there, that’s where the future is taking us.
[00:18:55] When electronic, like bouncing back, what, two-plus decades ago at this point, I guess. Longer? Have to have been longer. Bouncing back years ago, I sat on a… The, the benefit of being here in DC is sometimes you get suckered into something on, on, on Capitol Hill, and I was part of a panel that was talking about electronic medical records and privacy concerns.
[00:19:21] It was ages ago. And one of the one of the really impressive attorneys on the panel posed an interesting challenge and said, “What if you ha- like imagine if you will, if you had somebody in the same electronic record from birth into adulthood.”
[00:19:43] Kym: Mm-hmm.
[00:19:44] Edward: And now that’s very true. I mean, it’s very common in large hospital systems like Kaiser Permanente, for example.
[00:19:50] I mean, like, you know, you’re, you’re in there, right? If you’re… Right? Yeah. You know, so if you’re in Epic, you’re, you’re likely to be in it for the whole time. And his, what he posed was What if a physician, when you’re four years old, talks about the fact that you had a weird response to a medication? Not a classic allergic reaction, but an odd response, and he made note of it in his narrative.
[00:20:20] Kym: Mm-hmm
[00:20:21] Edward: you’re 21 years old and somebody prescribes that medicine or a cousin of that medicine, and you have a full-blown anaphylactic reaction and end up in the emergency room or worse, right? If you end up with a really, really negative outcome.
[00:20:39] Kym: Yeah
[00:20:40] Edward: What’s the liability? Should that practitioner who prescribed that
[00:20:45] Kym: Mm.
[00:20:45] Edward: known that, you know, 18 years ago you had a tickle in your throat from, from something?
[00:20:54] Kym: Wow
[00:20:55] Edward: And we all kind of went, ” you know, because it was early, right? And we didn’t, you know, like, we didn’t understand– You know, we, we were all like, “That’s a great hypothetical, but that’s not practical. We would never know that. And in a courtroom, how could we ever be expected to know that?” And weirdly, the answer is that we could, by today’s technology standards because of AI and other things, we could hypothetically know that.
[00:21:27] So what’s that m- what does that
[00:21:28] Kym: flagging that stuff for us, you know, so we’re not making mistakes.
[00:21:32] Edward: Yeah. What does that mean? So I think, like, where the AI is gonna go is, is awesome, and if as small practices, if we’re adopting tools, you know, right, you know, whether it’s just using it as a fancy Google search engine or, you know, building tools that work or pulling together libraries of, patient or clinic, you know, clin- patient or client references and information and, you know, and everything else.
[00:22:03] I think it’s, you know, the power is there as long as we continue to use our brain, you know, in the process, validating that it’s legit and, you know, and everything else. I think the biggest danger we have and sometimes I catch myself about to do it, is, you know, “Oh, the AI got this,” and then just not, you know, right, not check it.
[00:22:25] It’s
[00:22:25] Kym: Yeah.
[00:22:26] Edward: there yet, right? You know, one of the things I did, and I know it’s not the, you know, the focus of our conversation here, but the, the clinical charting tool AI that I use, I built an audit tool into my note for, you know, for CPT coding Oh, it’s, yeah, it’s called, it’s, it’s yeah, it’s Clinical Notes AI, which I think they recently rebranded as Clinically AI or
[00:22:51] Kym: Oh, okay. Okay. Yeah, I’ve heard
[00:22:53] Edward: and you know, but I was like, “I wonder if it’ll do this,” you know? And is it right all the time? No. You know? But you know, it’s, it’s making things easier and you know, allowing me– I’m weird, I like to bill ethically. I don’t, I don’t automatically… Like, I have some colleagues who automatically will do, you know, if it’s, they’ll automatically add a psychotherapy add-on code, 30-minute appointment psychotherapy add-on code with a nine, you know, with, with the appointment.
[00:23:24] And I’m like, “Nah, I wanna do it ethically.” So, you know, m- the, the audit tool that I built in says, “Here,” you know, “You spent in your…” Right? ‘Cause in order for us to do a psychotherapy add-on code, you ha- like, if you’re doing a 99214, which is a m- mid-level complexity me-medical case to do a 90833, which is an add-on code for 16, I think it’s, if I remember right, 16 to 30 minutes of psychotherapy add-on.
[00:23:55] So, like, you know, it’s gonna at least say, “Based on my…” Right? The AI says, “Well, I was listening,” you know? “And based on what you two talked about in your appointment, this, you know, s- the psychotherapy elements were X number of minutes, so it doesn’t qualify for an 833.” You know, every once in a while it’s wrong.
[00:24:17] Like, I’ve had a few times say, you know, “You spent 29 minutes in psychotherapy and one minute on medication management.” And I’m like, “No, we spent, we talked about meds for, like, you know, longer time.” So it’s AI. It’s still, it’s still, you know, flawable, but at the same token, you know, I feel now, ’cause, you know, my note structure is very specific to what we need, you know, there, if we’re doing psychotherapy, it automatically adds the psychotherapy component that’s needed for billing, et cetera, et cetera.
[00:24:53] You know, I make sure that the again, ’cause of the sensitivity of the things that we tend to talk about with my patient population, I make sure… Like, one of the things I saw AI doing originally, and this is all products until you train it, you know, it would say that, you know, like, if my patient said, “My boss is such a blankety-blank-blank,” you know?
[00:25:13] Then it would sometimes go, “Ooh, patient states, quote, ‘My boss is such a…'” And I’m like, “Well, that’s not really good in a note,” you know? So can we change that into conflict at work?
[00:25:25] Kym: Yeah.
[00:25:27] Edward: You know? Right? You know?
[00:25:27] Kym: Or like with Berries, there’s literally a button that you could, like, take all the quotes out, like re- unquote it,
[00:25:32] Edward: Right. Right. You know, and those kinds of things. So, you know, so working as clinicians, right, we have to work and still, like, review and sign and go, “Wait, no, what the heck was this thinking?” And some, you know, some AIs hallucinate more than others. And, you know, there are people out there who are looking at AI and they’re like, “What’s the least expensive solution that I can find?”
[00:25:57] Kym: Yeah
[00:25:58] Edward: And if that’s what you need for your business because of w- of how you operate your business, okay. But odds are you’re gonna be doing more review work,
[00:26:07] Kym: Right, right,
[00:26:08] Edward: And if you’re not, and if you’re not doing more review work, then you picked the wrong fee product,
[00:26:13] Kym: Exactly. Yeah, exactly. Oh, that’s such good advice.
[00:26:17] Edward: Yeah. And I think it’s also, like, it’s simple things.
[00:26:20] Like, and this is stuff that we’ve talked about in the group as well. It’s, you know, what I, what I joke about, and I now teach… This is, this is weird. I’m, I’m con- I, I’m a little atypical. When I teach my clinicians in school, we talk about AI, right? Whereas, like, universities are all like, “AI is bad.”
[00:26:39] Kym: Mm-hmm.
[00:26:39] Edward: Well, but you’re gonna go into a clinical space where, you know, AI is gonna be there, so understand how to use it right.
[00:26:47] And you know, when, when therapists, nurses, all of us have gone through school, we learned to verbalize when we were doing like, you know, mock, mock therapy sessions or whatever. We learned to verbalize certain things so that the instructor who was supposedly paying attention as opposed to being on their phone would know, “Oh, they’re now switching into blank,” right?
[00:27:13] So I’ve learned that, you know, I will– I now in my AI, I say certain things that I might not have said instead of, you know, while I was busy clicking doing. You know, so I’ll say things like, “Let’s verify your medications,” so that the AI knows that this is what we’re doing. Or, you know, I would say things like, you know, “From a psychological therapeutic perspective,” you know?
[00:27:36] Right? You know, so I’ll throw in key phrases so that I, you know, my AI’s like, “Oh, he’s telling me to pay attention. This is, you know, this is what this is.” You know, and things like that. Or, you know, “Let’s talk about safety.”
[00:27:51] Kym: Yeah
[00:27:52] Edward: that’s like saying, you know… It’s the same as when my pre– you know, when I’m precepting students to see what their skills are, they’ll say, “We’re now gonna have, we’re now gonna make sure we assess your safety,” right?
[00:28:02] ‘Cause that’s their cue to me to go, “Okay, yeah, they met that objective.” Right? So now I say, now I say it out loud so the AI goes, “Oh, wait, this I should be, like, assessing this part in,” you know, or placing this part of the conversation in the safety section of the note, you know, or, or
[00:28:19] Kym: And that’s where the, like you said before, the training really does come in. Th- you’re making your templates a specific way and using, you know, the carrots or whatever on your template that makes sure it emphasizes like we really wanna touch on safety or sleep or medication or whatever to make it super specific to you.
[00:28:35] So you’ve gotta train it. You can’t go into any of these systems cold and just think it’s gonna be perfect to begin with.
[00:28:41] Yeah
[00:28:42] Edward: It’s a matter of how you build the templates and how you train it and how you communicate to it moving forward, right? You know, there are some tools that are better at, at you training it, and then there are other tools that are better at allowing you to custom build the template. And, you know, that’s a moving target.
[00:29:01] You know, I made the mistake– It’s kind of funny. I made the mistake a coup- bunch, I guess when I– It’s now five years ago, I think. I bought a multi-year license to Dragon NaturallySpeak, right? Dragon Medical,
[00:29:16] Kym: Oh,
[00:29:16] Edward: right? Right, a dictation, because I was like, “This is it, man. This is gonna make my charting so much easier.”
[00:29:23] And I think, I think my contract with them expires, like, in a year, and I’m still, I’m still paying every month. And I’m like, you know, I shoot myself. I, like, I was so proud of the fact that I was saving, like, 10, $15 a month. It’s like, oh, over the course of, you know, and, and now I don’t sign any long-term contracts with any AI product.
[00:29:45] It’s not, not, you know, it’s like I respect them and what they’re doing and everything, but,
[00:29:49] Kym: Mm-hmm.
[00:29:50] Edward: you know, the, if there’s gonna be– I mean, you know, since AI is basically building AI now, right? You know, there’s gonna be something that’s gonna come and that’s gonna leapfrog Berries or Clinical Notes AI un-un-unless they are moving equally as fast.
[00:30:07] And if they snooze, I would hate to own an AI company right now, right? Because if they snooze for a moment, right, you know, my YouTube feed is full of AI stuff about Claude and OpenAI and, you know, it’s like you’ll hear, “This week, OpenAI released blah, blah, blah, blah, blah.” And then a week from now, “Claude AI released blah, blah, blah, which leapfrogged.”
[00:30:29] And you’re like, “Okay, really?” You know? I’m old enough to remember when there was, you know, Word and WordPerfect were your, quote, “word processing softwares,” right? And, you know, it would be like one week one would leapfrog the next and this is better, or Cerner versus Epic in large, you know, large EHRs and, you know, they were always playing leapfrog.
[00:30:52] So no long-term
[00:30:54] Kym: And that’s,
[00:30:55] Edward: I don’t care how much money I’m gonna save, I’m not gonna get locked into a long-term contract with any of
[00:31:00] Kym: Yeah, that’s what the Thera AI Hub… It’s like every month there’s so much cool stuff. Like, I literally have to, like, figure out what one thing to talk about so I’m not just, like, reinventing the AI wheel every single month. ‘Cause it’s, it’s just… I mean, it’s very exciting, but it’s
[00:31:14] Edward: You wouldn’t sleep.
[00:31:15] Kym: too.
[00:31:16] Yeah.
[00:31:16] Edward: sleep. You’d have to clone yourself with AI
[00:31:19] Kym: Totally, yeah, and just put a, a chip in my brain. Yeah, no, I, I love all that. So what what would you say, like, could you quantify… I mean, this is kind of a big question, but, like, time-wise or money-wise, like, all the AI systems that you’ve kind of built into your life, like, are you saving money?
[00:31:38] Are you saving time? If you really sat and thought about that, ’cause I’m just curious. You know, is it giving you your life back, or if you’re like me, you just fill it with more stuff?
[00:31:48] Edward: That’s the answer. Yeah, you know, so I think it depends. So, I– So my clinical charting tool, Clinical Notes AI, has given me time back. You know, it’s interesting. I mean, so, you know, so I, I– I mean, as I said earlier, right? So self-disclosure, I have ADHD, I have dyslexia, and I have a, a, you know, a pretty significant trauma background.
[00:32:11] So those three things really get in my way a lot. So what I found with using AI, like I’m actually a really good touch typist. My husband actually stopped, stopped in my office last night while I was talking and I was typing and looking at him, and he’s like, “How do you do that?” And I was like, “Well, first let’s see if I actually typed it right.”
[00:32:31] You know, and I, and I did. So that’s good, but relatively anyway, did type it right. But so my Achilles, when I finally… You know, so I was actually a pretty good touch typist and able to talk to a patient and, and create a note. The note may not have been perfect, but I was at least not, you know, trying to completely piecemeal my life together.
[00:32:55] When I started using AI, I went overboard and my note got like really, really big because it could. And then I remembered training that I had about, you know, sometimes less is more. The more you put in the note, the higher potential exposure you have and liability, ’cause it’s like, “Hey, five notes ago you wrote this.
[00:33:17] Why don’t you remember it now?” Like, just because it’s written doesn’t mean my brain had it on board. You know, and a few other things. So I’ve been tweaking my templates, but my challenge being dyslexic is also that my reading speed is, less. I’m one of those kids who was tested a million years ago, and I was actually diagnosed before they had the definition of dyslexia.
[00:33:39] Interesting thing in the mental health world before they had the definition of dyslexia, it was called minimum brain dysfunction.
[00:33:47] Kym: Oh, I’ve never even
[00:33:48] heard
[00:33:49] Edward: great stigma to be, you know, basically in s- I was in second grade and I was told I had minimum brain dysfunction,
[00:33:55] Kym: Nice.
[00:33:56] Edward: but it’s sometimes, some days it’s true. But you know, so like reading through the note was a bit much.
[00:34:04] So I now use a screen reading program at two– because all of my testing as a kid and everything else, my reading comprehension, funny enough, my reading comprehension is twice that if I read something upside down.
[00:34:19] Kym: Oh,
[00:34:20] Edward: a kid, I used to be able to, I used to test better… It was one of my accommodations as a kid that I was allowed to turn the paper upside down.
[00:34:28] Kym: Oh my gosh. That is
[00:34:30] Edward: and you know, and so I now, like I create the note. I have a s- Speechify is the software that I use. It’s a… And I pay for the service to get good quality voice from it, and I read it back to me at two times speed. And it, and as it reads, it highlights, you know, which is another quality thing for somebody with ADHD and everything else.
[00:34:51] And so it goes through, it’s like follow the bouncing ball if you got, if, you know, if you’re old enough to remember those sing-along things, right? And it reads the note to me at two times speed, and then I’m like, “Okay, I’m good.” Or I edit and I drop it, you know, and I drop it and sign it. You know, so it’s a matter of those tools, and now that I’m doing that, yes, I have saved time.
[00:35:13] But then I can be up until 2:00 in the morning, you know, going, “I wonder if the tool can do…” Right? You
[00:35:20] Kym: Yeah, totally. Down the rabbit hole.
[00:35:22] Edward: yeah. You know, and so,
[00:35:24] Kym: Mm-hmm
[00:35:25] Edward: you know, my assessment would be different than my husband’s As you know, so, but yeah. You know, it’s those, those different pieces. I think long term it’s better.
[00:35:37] You know, I now use, You know, there’s little things that we do. So we have to do prior authorizations for medications, and now I just ins- you know, I– It may take me a moment longer to do what I do now, but it saves me a lot of headache on the back end. So now I use AI to, since it has all of this data from prior visits and everything, I say to my AI, “Write a prior authorization note reviewing the last X number of charts for such and such a medication, cite clinical evidence, et cetera.”
[00:36:12] And so I now produce a cover letter, which most of us as clinicians don’t usually do. We usually just attach charts. But when we attach charts, the insurance company will just say, “Meh, didn’t see it,” because they didn’t read all of it. Now everything is consolidated into a few short paragraphs, cover letter, here’s all the info they need.
[00:36:31] So it takes me a touch longer to do that, but I don’t get denials as much as I used to,
[00:36:38] Kym: so it saves money in the long,
[00:36:39] in the
[00:36:40] Edward: so it saves energy in the long run maybe. You know, and then other things like when I’ll take a bunch of stuff and say, you know, recent- like eight– I have a university in my area that is very, very very, very, we’ll say politically politely conservative in how they require ADA mater- an ADA accommodation request.
[00:37:07] And the director of their disabilities office and I now have, like, are on, we’re on first name basis because I used to c- just, like, pitch a fit. Now I built something in my, my system that says, you know, “Here, prepare all of this.” I have all of their questions. I ask the questions or tell the AI, “Find the data in prior visit notes and produce the content.”
[00:37:38] And then originally I got kind of pissy and I said, “Produce the content in at least 30 pages or more.”
[00:37:46] Kym: Yeah.
[00:37:47] Edward: It’s like, “You’re gonna question me? Here’s 46 pages that meet your criteria.” Now I’ve, I’ve shortened it, you know. But,
[00:37:54] Kym: That’s
[00:37:55] awesome.
[00:37:55] Edward: know, again, that’s AI’s doing that and it’s pulling, because of how I use the system, it’s pulling the data that I need or I’m feeding it what it doesn’t have.
[00:38:06] So I think those are, clinically those are really big savers. From the business perspective, all of it’s savers. All of it’s savers. I used to do a lot of Excel pivot tables to look at my numbers and, and, you know, what’s, like, you know, my insurance reimbursements and everything else. A lot of us are scrambling right now because some insurance companies are down coding.
[00:38:28] Some insur- like, you know, trying to, trying to project my averages and all those other things. And I used to do these big fancy Excel pivot tables, and now I just feed stuff into Claude and I say, “Create a blank,” and here’s all of the data that you need from, you know, like three or four different spreadsheets from a multiple different systems.
[00:38:50] Tell me, you know, tell me what I need to know, you know, magic eight ball that it is. And that’s hugely business savings, you know, and sometimes that’s like, you know, that’s not app building, that’s just like I need this business decision and we in our world I, I recently brought on a fractional CFO for the company and they’re great, but I’m their first salvo into in a, a mental health company that does insurance
[00:39:25] Kym: Oh, wow. Yeah
[00:39:27] Edward: You know, they’re like, “Wait, you mean you can’t track your dollar to, to like your dollar bill to…”
[00:39:32] I was like, “No, it just doesn’t work that way in our, in our industry.” And, you know, he’s like, “What?” And I’m like, “Yeah.” You know, he– I mean, don’t get me wrong, he is exceptional at what he does, but he’s not, like, he’s like, he’s never worked in a business like that. And, you know, the fact that I, you know… So he’s the first person to judge me and, and the business that we’re in because AI is just like, “Here, you know, here’s your data, here’s your answer.
[00:39:57] Here’s the closest we can get.”
[00:39:59] Kym: Yeah
[00:39:59] Edward: been, that’s been super powerful for the business. And so I don’t spend you know, we take most commercial insurance plans, the big plans, and so all of their sub-plans. And one of the conversations I was looking at recently was what is my av- what is my mean and average per insurance brand per CPT code that we routinely use?
[00:40:28] Kym: Mm-hmm.
[00:40:29] Edward: And my o- my answer used to be, you know, pivot table this, do this, do that. I basically extracted a flat CSV out of my insurance billing thing, dropped it into Claude, and what would’ve taken me a couple of hours of pivot tables, I had in a matter of minutes, you know, out of, out of Claude.
[00:40:52] Kym: Crazy. Yeah. Oh my gosh, it’s just… Yeah, my episode before this one, it hasn’t come out yet, but I was complaining about Fable, Claude Fable, the new Fable 5. And, and then over the last, like, two days I’ve actually, like, really deep, like, brought in my financials and stuff and built out a dashboard like you’re talking about, and it’s, like, blowing my mind.
[00:41:14] Like, seriously, I’m like, “Okay, maybe I was wrong about Fable. Maybe it is the next level,”
[00:41:18] Edward: yeah. And there’s some really cool stuff, like again, if you wanna geek out into it, I haven’t done it myself, but there’s some live dashboard stuff that you can do. And, you know, I’m not gonna do that with my financials because I just kinda don’t wanna open up that, that, you know, risk exposure. But there’s definitely a couple of other things.
[00:41:39] You know, we’re moving we live out here in DC. We’re moving to Southern California in September and, like, so one of the live things that I did in Claude just was really more of a goof, and I blame you and the community for it, is I said, you know, so I have a live dashboard now o- on my computer showing me the, like, current temperatures in a couple of different places, and I built that in Claude.
[00:42:05] You know, I was just like, you know, “Here, this is what I want. This is the data I want in it.” And, you know, it’s just, it’s live updating from, you know, publicly available weather stuff. And I was like, “This is great. Now it’s here,” and, you know, it refreshes. And so when I’m, you know, it’s like it was in, you know, it was 102 here in DC the other day, and when it’s telling me that it’s like in Southern California where we’re looking at it’s gonna be 76, you know, it’s like 102 versus 76, overcast and like, you know, 94% humidity here, sunny and, you know, and like what is humidity over there?
[00:42:39] I was like, “Yay.”
[00:42:41] Kym: That’s awesome. Oh my gosh, I feel like I could just talk to you all day. So just to kind of wrap it up, if somebody’s listening to this and, and they’re like, “Maybe I should, should I join the Thera AI Hub or the Clinical AI Club?” Do you have anything you would say about that to somebody that’s listening about maybe how it’s helped you or
[00:42:58] Edward: Yeah. Why are you waiting so long to make the decision?
[00:43:01] Kym: Thank you
[00:43:02] Edward: no, I, you know, I mean, it’s you and I have talked about the fact that it’s a great space. I, you know, I want to encourage more psych NPs to become a part of, of the space, and I think we have a lot of great conversations in, in our future about that.
[00:43:15] But I spoke to a colleague recently who said I think I’m against AI for the wrong reasons,” right? And we ha- And it’s like if you are trying to figure– Even if you’re not like, “Yes. Oh my God, yes, I want AI,” if you’re trying to figure out where is a good place for AI or what should my thinking be in order to be evolving, it’s, it’s a great community for that, you know?
[00:43:39] I do, I have lots of conversations with my patients about it and everything else, and we’ve made the move, it’s probably controversial, we’ve made the move that, you know, it’s AI or you don’t need to be part of our practice, and that’s okay. And, and I’ve answered that. It’s kind of funny, I’ve answered that for a couple of patients is because it fits all of my other providers have some sort of neurodivergence or other things, you know?
[00:44:05] And so it’s how we created ADA accommodations internally. And it’s funny, some of my patients actually go, “Oh,” and then all of a sudden they’re okay with it. But yeah, you know, and but I think if you’re trying to figure out what it is, how to do it, you’re, you’re afraid of it, you know, everything else, you’re worried about the amount of water it’s using, whatever your concerns are, come and join the community and be amongst a bunch of people who have similar questions or have moved beyond those questions.
[00:44:39] Because you’re in a safe space of fellow mental health professionals who all have very, very similar needs and flavor of needs. And I think that’s the power of the community that you’ve built. I could sit in on a community you know, I am on a Facebook community of Claude users, and I, you know, I’m in it.
[00:44:58] I unfollowed it. It’s far too much to like, to digest. Be dangerous. I probably wouldn’t sleep. But to be with my colle- my mental health colleagues in a space where we all have sort of similar issues, we’re all dealing with similar patient bases, similar business needs, and similar desires to be successful with a technology that’s moving rapidly.
[00:45:24] It’s a great combination.
[00:45:26] Kym: It’s amazing.
[00:45:27] Edward: And that was a sh- that was a short, long answer.
[00:45:29] Kym: I love it. No, it’s so good, ’cause I feel the same way. I mean, I, I look f- so forward to those meetings. I know I’m, I’m running them, but I love it. I love to be able to just geek out and, like, all of us come up with, like, solutions and, like, to have somebody be like, “Oh,” and see that light bulb come on, that is, like, the best
[00:45:47] Edward: Yeah. Well, it’s like the thing, like the, like the session, I know we have to run, but it’s like the session that you did on HIPAA compliance, you know? And I think I was the one, you know, like
[00:45:56] Kym: Oh, yeah
[00:45:56] Edward: of the things I’m trying to kill for is better e- like, is an email solution, and I’m using a product that’s okay, that is HIPAA compliant for email management, but it’s really built to be geeky.
[00:46:10] And then in other words, you know, and so I don’t have the time to learn it, so that’s my challenge with it. But that one commercial company that came out and was like, “Oh, we’re HIPAA compliant,” and you know, blah, blah, blah. It’s like, oh my God, this is awesome and the pricing is de- you know, is reasonable.
[00:46:24] This is great. Oh, but we only do BAA for enterprise solutions at thousands of dollars. And I’m like, well, then you’re not HIPAA compliant. And then I noticed after I challenged somebody in their leadership, ’cause they’re a small company, they actually changed their advertising a little bit.
[00:46:44] Kym: Oh, interesting.
[00:46:46] Edward: still misleading, but it’s not as bad, right?
[00:46:49] You know, and so– But it’s like, where else could we have had that conversation? Because we all are s- we all are drowning in a, in, in email,
[00:47:01] Kym: Yeah.
[00:47:01] Edward: And we’re all… Like, so if you’re positive AI, you’re like, ooh, here– Like, ’cause I get marketed a bunch of AI email management tools. But to come into that session on, on, on HIPAA, BAA, et cetera, and to go, whoa, pump the brakes just because they say it’s HIPAA compliant, if they’re not giving you a BAA, it’s not gonna meet the legal threshold right?
[00:47:26] Where else would we have had that conversation but in with a room full of AI therapy people
[00:47:30] Kym: Yeah, I know. I know. Ed, thank you so much for being here today. I really appreciate it.
[00:47:33] Where do people find you if they wanna talk to you, if they’re nurse practitioners and they wanna know more about what you’re doing with AI and that sort of thing? Where would they
[00:47:39] Edward: Yeah, they they’re welcome to email me as long as it doesn’t get lost in my email that needs AI. So yeah, the, the company again is Nothing Better Health, so they can always go to nothingbetterhealth.com and find me through there. The easiest way to directly email me is ed@estern, E-S-T-E-R-N.com
[00:47:59] Kym: Perfect. Easy. Thanks
[00:48:02] Edward: Awesome. It was a pleasure. Thank you so much for including me. I’m glad I’m number two, right? Woo.
[00:48:07] Kym: Yay. Thank you
[00:48:09] Edward: Talk to you later, Kim. Take care.