Real-world data: The Good, the Bad and the Ugly with Jan Poleszczuk from CliniNote

Maya TrialHub (00:01)
Hello, hello, this is Maya Z and to be precise Zlatanova, but I know that my surname will be very hard to pronounce. And I have a guest that like whose surname is also not easy to meet to pronounce, but I'll give it a try. I already tried a couple of things. It's actually a fellow friend from an accelerator program that I did together with my company earlier this year. And I've been fascinated by the things that these guys are doing. So Jan

Poleszczuk welcome to Trials with Maya Z And I can't wait to dive into the world of standard of care and real world data. But before that, can you give me a little bit of your background?

Jan Poleszczuk (00:46)
Yeah, perfect. Thank you, Maya, for having me. So...

about myself a little bit of background. So I'm a mathematician by training. So, but please don't turn the podcast off, right? So I'm mathematician, but I quickly, I very quickly after obtaining my PhD, switched to modeling of cancer and trying to come up with the better models to, predict the outcomes of treatment, right? That was my main goal. So that's why I started working in the hospitals, right? So I did my postdoc in Moffitt Cancer Center in the States when I moved back to Poland.

Maya TrialHub (00:57)
Hahaha.

wow.

Jan Poleszczuk (01:21)
I started working in the National Cancer Center as well here and I was always about getting the data for my models, right? I mean, we cannot really have any model without the proper data, right? And by proper data, mean heterogeneous to come from reality, et cetera. But I quickly realized that there is a big issue with that data, right? And maybe I'll go by analogy, right? So...

Maya TrialHub (01:39)
Hmm.

Jan Poleszczuk (01:50)
When you think about real world data, not only in medicine, that's a crucial part to tweak your product. So the common knowledge is never buy a newly launched car. Wait for two years. When all the complaints come in, they're going to provide the facelift. That's the good moment to buy the car because you know all the troubles of the infancy. But imagine now that after a clinical trial,

Maya TrialHub (01:57)
Mm -hmm.

I love genealogy.

Jan Poleszczuk (02:14)
That really doesn't happen, right? I mean, there's a clinical trial, we launched a product on the market, and the real -world evidence about the usage of the drug is not that great, as in the case of other products, right? We know that...

Maya TrialHub (02:17)
Hmm.

Why is that not that great, Jan? Like tell me, because I spent quite some time in researching real world data. can't, I'm not an expert. Like I'm definitely not an expert, but let's say that I kind of understand the role of real world data, but like what makes it like not good enough today?

Jan Poleszczuk (02:45)
that's a simple question, but with a very complex answer, right? So I guess the source problem is very limited time that the doctor can spend with the patient, right? That's the origin of everything, right? And we know who's actually the real world data generator. It's not the hospital administrative staff, right? It's the doctor himself, right? So.

Maya TrialHub (02:50)
Yeah.

Hmm.

Yeah.

Jan Poleszczuk (03:10)
As long as the doctor that have very limited time to generate the data is not allowed to do so, I mean, we're not going to have a great data, right? Everyone knows that in a hospital database we've got plenty of structured data. For example, ICD -10 codes, for example, what kind of a procedure has been performed.

But that's the claims data. And actually, that data typically is not entered by the clinicians. And it's certain issues, right? We know the common term. Someone can Google afterwards, up coding.

Maya TrialHub (03:36)
I'm...

Jan Poleszczuk (03:43)
There are frequent mistakes, there plenty of anecdotes about how the WHO went to some region in a country because there was a higher incidence of some disease. But then it occurred that one doctor knew only one ICD -10 code and he was putting that code to every single patient that he had. Because from the doctor's perspective, that didn't really matter, right? For him, the medical note, the things he writes down is the crucial part right?

Maya TrialHub (03:59)
my god.

Jan Poleszczuk (04:10)
And actually, that's always typical when I work with the doctors. They always say that if you want to have the source of truth, if you want to have the truth, read my notes. Don't go to the claims data. And at least in Poland and many European countries, and I know I remember from the States, there is a lot of up -coding. There is a lot of additional cancers that actually are not cancer, for example.

Maya TrialHub (04:21)
Hmm

Jan Poleszczuk (04:35)
someone during the diagnostic process already entered the ICD -10 code of C50 of the breast cancer, example, and there are plenty of issues around that.

Maya TrialHub (04:35)
Hmm.

Is that the reason you started? So you are the CTO of CliniNote a very cool company in the space of real world data. Is that the reason why you started the company? Because you realized that the notes are actually the real source of real world data and not the claims data.

Jan Poleszczuk (05:03)
So I would rephrase that the doctors are the source of the real world data and their main mode of action is through notes. And we cannot really change that because if we're going to introduce the forms that like many attempts, like a lot of forms to be clicked through that actually increases time for them, it's way quicker to write in their own words, the medical note. And that's exactly why at CliniNote we focus on the medical note. That's how it started. So I realized that if I want to generate the data without

Maya TrialHub (05:05)
Okay, true.

Thank

Mm.

Jan Poleszczuk (05:32)
giving additional burden to the clinician, need to provide them the tool that will allow them to write faster and also in a way that I generate the data at the same time. Because like if you talk to the clinicians, like I sometimes hear that the worst thing that can happen to you as a patient, right, is that my doctor will go to the holidays and someone else will need to read through his notes.

There is a huge chance of medical, a lot of medical errors are associated with misinterpretation of others notes, right? We know the cool example of digital pathology in which there needs to be a consensus, for example, in which the interpretation can be different.

Maya TrialHub (06:09)
Mm.

Yeah.

I'm actually laughing now because I was in a founder's retreat the last week and then the coach was telling us that 90 % of conflicts start from misinterpretation of words. I can imagine in this case when you have like very important piece of information that can be translated differently and that can lead to a lot of confusion but also not proper decision -making. So, okay.

Let's go back. If we define real world data, and I agree with you, like claims data, know we are speaking about, like we use claims data for real world data as well. It's not complete, but sometimes we do that only because there is no other alternative. So actually let me rephrase that. How accessible real world data is. And when we speak real world data, like the real world data, the insights from the doctors, how accessible is that today worldwide?

Jan Poleszczuk (07:18)
That's a good question. So I think that when you think about real world data and you don't think about the quality, right? When you think simply, as you said about, it's quite accessible these days. I mean, there are a lot of sources. mean, the hospitals are getting idea infrastructure better and better of giving out the codes, right? We know a lot of companies that work on that. But when you actually do the project on real data, I did a lot of projects on the real data while working with the hospital.

Maya TrialHub (07:28)
Mmm.

Jan Poleszczuk (07:47)
getting that data is only like 3 % of the whole project of analyzing that data. In a sense, you start with the code, you start with the structure, but then you dig down and you see a lot of problems. And actually the quality impairs the quality of the project itself. You need to filter out a lot of patients. And sometimes you end up with a very few cases at the end. So the access is there, but when you like...

Maya TrialHub (08:10)
Hmm.

Jan Poleszczuk (08:15)
to some kind of data. would say access to data coming from the world is there, but that's a quality of that data that does not allow to successfully complete the study that you had in mind, right?

Maya TrialHub (08:27)
What are the top challenges contributing to lower quality of real world data?

Jan Poleszczuk (08:34)
So I would say.

Maya TrialHub (08:37)
If we're in the top for example, like the top three.

Jan Poleszczuk (08:40)
Top three, so the healthcare system setting, first of all, right? So the source is that's like, you have a three player game, right? You've got the payer, I mean, you've got the insurance company, you've got the hospital, you've got the patient. And that there's a game in which the hospital wants to maximize the income from the insurance claims and the insurance claims need to maximize the way to cut down on the costs.

Maya TrialHub (09:02)
you

Jan Poleszczuk (09:07)
And there is a game not about the quality of data, but about how to create their reimbursement claims data in a way that you're going to maximize that. The focus is different, right? So summing up the first big challenge is that currently in the healthcare system, there is absolutely no focus on the real world data, right? When you think about like a healthcare system on the top, right? The other...

Maya TrialHub (09:16)
Hmm.

Hmm. Yeah. Hmm.

Jan Poleszczuk (09:34)
The I would say is the limited time of doctors. Right? I mean, that's the...

Maya TrialHub (09:39)
Yeah. And this is becoming bigger and bigger problem. Unfortunately, I've been, that's actually a problem that's contributing to other, a lot of other issues, including patient recruitment and clinical trials. When you have 10 minutes per patient, how much do you actually know about this patient? How can you actually even discuss the different opportunities for this patient? And can you actually build trust with the patient? So that's the other thing.

Jan Poleszczuk (10:08)
Yeah, I mean, that's an important issue, especially that you don't have the access to the full documentations typically, right? Because it's not like back in the days, as we discussed, I think, on the day one meeting, I think back even 100 years ago, there was a doctor that was yours, right? He knew you from the child from the birth till death. There was a doctor in the village. He knew everything. These days, how many different doctors per year are you visiting?

Maya TrialHub (10:15)
Mmm.

Mm -hmm.

Yeah.

Yeah.

Yeah. Well, hopefully not many, but if I have to, if I have to, would visit multiple sometimes for different things, for same things, actually. That's other thing. The only one doctor that you visit all the time is actually Dr. Google, to be honest, which is, which is quite interesting, but that's the reality. Actually, let me stop you here, Jan, speaking about doctors not having time with patients and also speaking about real world data. In the past, let's say

Jan Poleszczuk (10:38)
Notion.

for the same thing, right?

Yes.

Maya TrialHub (11:05)
the last two years or so, maybe even like before that, I've heard that companies like Microsoft, I think even Google, but maybe I'm wrong, but one of these big corporations, they wanted to go into exactly notestaking of doctors, literally listening to the doctors and typing this automatically so that they can save time listening to the conversation and putting notes automatically for the doctors so that they can save them these efforts. mean, is that?

possible? Is it already up and running? How reliable is that today and do you think if not today this is the future?

Jan Poleszczuk (11:43)
Excellent question again. So those things are happening, right? We know companies, especially in US, there are a of companies that actually do that. And Microsoft followed with the idea and actually is providing a piece of software for that. But the usability of that thing really depends on the clinical setting, right? And when you're working on a high volume center, then...

Maya TrialHub (11:47)
Okay.

Mmm.

Jan Poleszczuk (12:09)
when you discuss with the glitch and it's typically faster to use the template to quickly fill the note rather than then correct automatically transcribed note, right? Even if it's a perfect solution, right? I mean,

After the thing has been transcribed, the doctor needs to read that, correct that, add some things, modify that. And it's not completely clear to me that image in every setting, that's a time saver. Because I observed a lot of clinicians that are treating, for example, cancer patients in which the summary of the visit is quasi -standardized. And it's way quicker for that person, for example, to use even a clinic note in which he just answered the specific note generator. And it's way quicker than to trans -

for the transcription and to correct the transcription. And also those notes, I mean, if they're written, it's not exactly structure, it's exactly data generation. There is no connection with the generated by that mechanism note and the data behind. It's not straightforward. So we have a different connection, but that depends.

Maya TrialHub (13:18)
Okay.

Mm -hmm. They're very interesting.

Jan Poleszczuk (13:22)
In Poland, for example, the environment is noisy and most of the commissioners will say there is no way we're going to use the microphone. There is a hardware limitation, et cetera.

Maya TrialHub (13:30)
Yeah, I guess confidentiality comes in as well. GDPR in Europe. it does sound complicated. I can see that happening in the future, like with, for example, sometimes we're discussing with some folks the opportunity of having a unified electronic health record system across the globe. It sounds amazing. it's the dream. But then how feasible it is in the next even 10 years. I mean, I want to be positive. I hope.

We'll figure it out, we'll find a business model or some sort of a model to make this work. But yeah, it is actually challenging. So you mentioned that there is no focus on real world data as the first challenge. And the second challenge is actually limited access to real world data, to these notes. And actually not limited access to the notes, but limited time of the doctor also causing limited notes as a side effect.

Jan Poleszczuk (14:29)
as a side effect.

Maya TrialHub (14:30)
Yes, as a side effect, anything else that's causing a challenge, like the challenge in not having quality real world data.

Jan Poleszczuk (14:39)
Well, mean, just to also refer to the first thing. So there is no focus of the healthcare system on real world data. don't want someone understand that the warmth has no focus because the people like us or the pharma companies, they really want that data, right? I'm only saying that the healthcare system is not focused on that mainly. I would say also there is an interesting transition in the hospitals, at least in Poland that I observed that...

Maya TrialHub (14:46)
Yeah.

Yeah.

Yeah.

Jan Poleszczuk (15:07)
the notes are becoming more and more vague. And that's on purpose. And sometimes the clinician don't want to explicitly various things for a very simple reason. They want to have the options for the patients to go with different things. For example, as an example.

Maya TrialHub (15:19)
Hmm.

Jan Poleszczuk (15:30)
If you think about the PD -1 blockade in the cancer treatment, right? So after there were clinical trials, so there was a hard criteria that if there is a resist progression, you should stop the patient from treatment, right? There was a very hard stop from the clinical trial and that was also embedded in the reimbursement claim system system in the prospective, in the normal setting.

But the clinicians sometimes believed and they were right because we know there is something like pseudo progression that we should keep the patient on that treatment for a little bit longer because it might actually work. And especially they didn't want to write in the documentation that the tumor was enlarged because they wanted to avoid the necessity of taking out the patient from the treatment. And also there is a, the law is...

Maya TrialHub (16:08)
Yeah.

are interesting.

Jan Poleszczuk (16:24)
the laws, for example, in pathology reports. I frequently observed recently in the raw data that they write maybe there's a new word maybe because that removes the liability from you if you misdiagnose someone. Right? mean, that's that makes them less prone to lawsuit, at least in several settings.

Maya TrialHub (16:27)
Yeah.

got it. Very interesting.

Hmm. Yeah, that's very interesting. Jan, tell me what's the vision with CliniNote for real world data and how can we actually make real world data more qualitative, not qualitative, I'm sorry, but like with better, higher quality and more accessible.

Jan Poleszczuk (17:09)
Right, so as we believe at CliniNote the most important part is to focus on the clinician.

And help them out to generate the data while they're writing, because I know a of clinicians that want to do the research. I received a lot of databases that they created on the side. Right. And that cleaning note, instead of them having a separate Excel or separate notes where they're taking notes about the patient, we give them the tool that they write the note and generate the data at the same time that they can use. Right. So our vision is to provide each and every doctor with a tool that they can share their

their terminology, their templates, the way of writing so they can start collaborating with each other globally generating the volume of high quality data, right? Because they are becoming the validators of the clinical evidence because the...

Maya TrialHub (17:57)
Hmm... I love that.

Jan Poleszczuk (18:05)
trick is about having them validating the suggestions that we are providing rather than after they've created the node, we can come up with the measures if that's correct or not, or we can validate using another doctor if our NLP method found something. No, here we are using each and every doctor as a validator of the NLP finding, right? Which makes the process play more efficient, right?

Maya TrialHub (18:25)
Yeah.

I love that Jan, because what we're saying is let's focus on the jobs to be done of the doctors and let's help them do their job better. And one of their jobs is actually creating these high quality data points out there because otherwise they can't do their job and they can't help the patients with their ecosystem to actually achieve a milestone.

disease, let's say improvement, quality of life improvement and so on and so forth. So how can we produce the tools for the doctors to actually do their job better? And I would just take these words and translate it to clinical trials because that's where my passion is. It's the same for clinical trials. We speak about side burden. Actually, maybe the thing that we need to focus on is how can we help patients, I'm sorry, doctors do their job faster, simpler.

quicker, in higher quality. And actually that will help them with both clinical trials and their normal practices. And at the end of the day, provide data that's both helpful to clinical trials, healthcare, medicine as a whole.

Jan Poleszczuk (19:44)
And that's actually what we are doing right now because we started, we already have clinical trial in which clinic the note is used because the big issue in clinical trial as well is that there is a protocol, there are ECRFs, there is plenty of data points that needs to be collected, but still the clinician works in his hospital information system with the patient with still limited time.

Maya TrialHub (19:57)
Hmm.

Yeah.

Jan Poleszczuk (20:05)
So what we do, provide, we take the ECRF and we help the doctor to fulfill the protocol and collect all the data points required by the trial. So exactly that's the same mode of action. So there is a ECRF and we simply take the structure and online help clinician, that's a cycle three of the therapy right now. So you should do the pregnancy test or you should do the blood pressure measurement.

Maya TrialHub (20:14)
nice

Hmm.

Jan Poleszczuk (20:28)
you've done that? Okay. So put the information to know that everything is fine, et cetera, et cetera. So we are guiding in the creating the note that will allow to fill the ECRF because that's a limiting factor for the clinical trials as well. Right. The doctors have three trials at the same time and they don't simply don't remember. would like to remember that they have time to look at the protocol. And then there is a hassle of data query. Right.

Maya TrialHub (20:34)
Christ.

Yeah.

Indeed.

Yeah.

Yeah, you're absolutely right. One of the things that any doctor and actually to be precise, investigators are always saying is that we hate the paperwork. And when I say paperwork, you can also like add this, the software work. Maybe it's not paper, but if you have to do the same type of repeating work and like do notes here and there and for this like same patient, it kind of doesn't make sense. These people, they spend so many years in education. Their focus is on

like one of the reasons going into medicine because that's like super hard is actually healing people only to find out that they will be administrative workers. So I love what you're doing with CliniNote Let's take away. like kind of a, like a conclusion to me is let's take away the burden from the shoulders of the doctors and help them focus on patient care and providing this insights and data that can make the patient care even better.

Jan Poleszczuk (21:35)
Mm -hmm.

Maya TrialHub (21:55)
Thank you so much, Jan, for joining me on this short but very insightful conversation. Good luck with CliniNote I'll be definitely following your company. yes, keep us posted on how real world data evolves because we all need that.

Jan Poleszczuk (22:12)
Thank you, Maya, for having me. It was an awesome conversation.

Real-world data: The Good, the Bad and the Ugly with Jan Poleszczuk from CliniNote
Broadcast by