Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Leveraging eSource in clinical data collection with Mats Sundgren
7 Plays4 months ago
Join Sylvain Berthelot as he welcomes Mats Sundgren, Senior Advisor at the European Institute for Innovation through Health Data. A pioneer in implementing eSource during his years at AstraZeneca, Mats shares how this approach is transforming clinical research.
Together, they explore what eSource really means and why reducing duplicate data entry at site level is such a game-changer. Mats highlights findings from a recent study with Memorial Sloan Kettering Cancer Center, which revealed the heavy time burden manual data entry places on coordinators, and how eSource can dramatically improve efficiency and data quality.
The discussion also covers the promise and challenges of scaling eSource adoption: from interoperability standards like HL7 FHIR, to dealing with unstructured data using natural language processing. Mats emphasizes the role of AI as an assistant rather than a replacement, and why true progress requires aligning value across sponsors, hospitals, vendors, and regulators.
Tune in to learn more about eSource, the cultural shifts needed for sustainable adoption, and the guiding principles Mats applies – rooted in Nash equilibrium.
Recommended
Transcript
00:00:13
Speaker
Welcome
Introduction to Clinical Data Talks
00:00:14
Speaker
to Clinical Data Talks, a podcast brought to you by CRS-Cube.
00:00:18
Speaker
I'm your host, Sylvain Bertelot.
00:00:20
Speaker
Join me and industry experts as we discuss the latest trends impacting the world of clinical data.
00:00:28
Speaker
My
Meet Mats Sönghan: eSource Innovator
00:00:29
Speaker
guest today has spent many years challenging where we get eSource data from.
00:00:37
Speaker
He worked for AstraZeneca for many years, which is where he first implemented eSource to clinical trials.
00:00:46
Speaker
And he's now a driving force behind industry initiatives that aim to make us use e-source data more, especially through his role at the European Institute for Innovation through Health Data.
00:01:04
Speaker
I'm very pleased to welcome Mats Sönghan.
00:01:08
Speaker
Hi Mats.
00:01:10
Speaker
Hello, I'm fine and thank you for having me.
00:01:15
Speaker
Well, thank you for joining us.
00:01:17
Speaker
I have high expectations because I've seen you talk and I know you're very passionate about this topic and we are going to learn a lot.
00:01:29
Speaker
And I feel like you have a privileged position in our industry because you've had the opportunity to implement eSource in clinical trials.
00:01:40
Speaker
And you've now decided to focus on this challenge specifically.
00:01:46
Speaker
But
What is eSource and How Does it Work?
00:01:47
Speaker
before we go into you sharing your experience, could you explain to our audience what you mean by eSource data?
00:01:57
Speaker
Certainly.
00:01:59
Speaker
eSource refers to the direct capture of clinical trial data from electronic health record systems into electronic data capture EDC system used in clinical research.
00:02:10
Speaker
So instead of manually re-entering data, eSource enables structured clinical data like
00:02:19
Speaker
lab results, diagnosis, medication, vital signs to be securely and accurately transferred within the patient consent.
00:02:28
Speaker
So this actually eliminates duplicate entry, reduces transcription errors and improves data quality.
00:02:38
Speaker
And importantly, eSource doesn't replace the role of the clinical research staff.
00:02:44
Speaker
It rather acts as a virtual research assistance
00:02:48
Speaker
that automates routine tasks and therefore frees up time for more critical work.
00:02:54
Speaker
So this process also aligns with GCP and regulatory expectation and has shown a major potential in streamlining prior workflows and reducing costs.
00:03:06
Speaker
And at its core, ESO's
00:03:09
Speaker
is about harnessing data already collected in care and to make research much more efficient and scalable across the institution.
00:03:21
Speaker
That's interesting.
00:03:23
Speaker
So for you, eSource is the use of existing records for patients, is that right?
00:03:31
Speaker
It is and eSource is a candy.
00:03:35
Speaker
So what I'm going to talk about today is the specific case of eSource and that is EHR to EDC.
00:03:43
Speaker
We are seeing more additional aspect of eSource where perhaps we can do the same for ePro, electronic patient reported outcomes and so on.
00:03:56
Speaker
But it is at its core reduced double entry.
00:04:01
Speaker
data is entered once is the absolutely most key principle.
00:04:08
Speaker
Yeah, yeah.
00:04:09
Speaker
And I'm all for that.
The Impact of eSource on Data Entry Efficiency
00:04:12
Speaker
So you wrote or you co-wrote a very interesting study in applied clinical trials last year.
00:04:21
Speaker
If I remember well, you co-authored that with the Memorial Sloan Kettering Cancer Center.
00:04:28
Speaker
And I found what you presented in that study extremely interesting.
00:04:34
Speaker
Could you remind me what the conclusion was regarding the time a clinical research coordinator spends entering data?
00:04:46
Speaker
Yes, yes.
00:04:48
Speaker
The study we published with the Memorial Sloan Kettering highlighted the substantial time burden that clinical research coordinators, CRCs, face during manual data entering clinical trials.
00:05:03
Speaker
So through the structured interviews and analysis, we found that the CRC spend at least three to five minutes per data point.
00:05:14
Speaker
And that's quite a lot.
00:05:15
Speaker
A number that scales rapidly in oncology, where each patient may generate 10,000 data points or more, which translates into thousands of hours per site, per study.
00:05:30
Speaker
The manual process is not only time consuming, it's also prone to errors.
00:05:36
Speaker
So CRCs must search across fragmented and often unstructured EHRs and then re-enter data into the complex EDC system.
00:05:46
Speaker
Query resolution alone could consume up to 50% of CRC's time, especially when data is in consistency arise.
00:05:57
Speaker
And
Benefits of eSource: Speed and Verification
00:05:58
Speaker
the study showed that e-source technology automating the transfer of structured EHR data into EDC can cut entry time by 50% and reduce transcriptions errors and free CRC's to focus on more
00:06:14
Speaker
critical and patient facing task.
00:06:17
Speaker
And equally for the sponsors, the benefits are absolutely clear.
00:06:22
Speaker
Fewer queries, reduced source data verification, which is a huge burden, and also accelerated study timelines, which ultimately could get medicines to patients much faster.
00:06:37
Speaker
So that's interesting actually.
00:06:40
Speaker
So reducing the data entry you think could have a positive impact on study timelines directly.
00:06:49
Speaker
Yeah, if I may expand on that, because what we are facing now is that we have hybrid studies with eSource, and that means that some of the sites are using eSource, but the majority of other sites are not using it yet.
00:07:09
Speaker
That means it is like the chain is as strong as the weakest point of it, right?
00:07:16
Speaker
But imagine, imagine when we have a eSource enabled trial, where all sites are using eSource, then, then we are talking about sort of cutting study timeline with at least three to six months.
00:07:31
Speaker
Wow, that is amazing.
00:07:33
Speaker
Yeah, yeah, yeah.
00:07:35
Speaker
You saw my reaction.
00:07:37
Speaker
So you mentioned that it reduces obviously the duplication of data entry because data is already there.
00:07:46
Speaker
So CRCs don't need to enter it again.
00:07:50
Speaker
But how easy or how hard is it to actually move data from EHR to an EDC system?
00:07:58
Speaker
It depends,
Challenges in EHR to EDC Data Transfer
00:07:59
Speaker
both technically and operationally.
00:08:01
Speaker
Transferring data from EHRs to EDC system, what we call ESOS, is increasingly feasible thanks to interoperability standards like HL7 FHIR, which stands for Fast Healthcare Interoperability Resources.
00:08:16
Speaker
In simple terms, FHIR acts like a universal translator.
00:08:20
Speaker
It is an interoperability vehicle that sort of helps
00:08:24
Speaker
healthcare system to speak the same language so data can move securely and accurately between them.
00:08:30
Speaker
And between them here is the middle layer, which the e-source.
00:08:36
Speaker
But challenges remain, of course, around 50 to 70 percent of clinical data, clinical trial relevant EHR data, that is.
00:08:47
Speaker
is fairly easy to map, but the remaining 30 to 50% of unstructured data often in text notes, PDF requiring manual validation.
00:08:57
Speaker
So it's more things to do.
00:09:02
Speaker
But beyond the technical layer, institutions must align to IT infrastructure, privacy safeguards and regulatory compliance.
00:09:10
Speaker
But eSource is nowadays sort of add on to all that kind of machinery.
00:09:18
Speaker
And the benefits are compelling.
00:09:20
Speaker
Hospital like MSK and Mayo Clinic and Cambridge University Hospital have shown.
00:09:25
Speaker
that with the right preparation and partnership, data transfer can be securely and compliant efficient.
00:09:31
Speaker
So ultimately, this is not just an IT upgrade.
00:09:35
Speaker
Essentially, it's a transformation how we do clinical trials.
00:09:42
Speaker
Yeah, and it's interesting that sites are ready because you've proven that with your studies and working with sites directly.
00:09:52
Speaker
But as you said earlier, not all sites currently use this process of using EHR data.
00:10:01
Speaker
So we don't necessarily see the full benefits.
00:10:05
Speaker
So if you think about the multicentered, potentially multi-country study, how likely are we to be able to harmonize data across multiple EHR systems?
Global Harmonization of EHR Data
00:10:23
Speaker
Harmonizing EHR data across countries is entirely possible, but it requires deliberate coordination.
00:10:31
Speaker
The core challenge lies probably in difference across hospitals in data models, terminologies, languages and perhaps also documentation practice.
00:10:41
Speaker
For instance, one site may use ICD-10 codes, another may rely on free text notes.
00:10:48
Speaker
So without harmonization, this is variability limits, consistent and high quality data analysis.
00:10:56
Speaker
That said, real progress is being made.
00:10:59
Speaker
Federated platforms now allow hospitals to share insights without moving raw data, and especially both privacy and governance requirements.
00:11:08
Speaker
So standards like HL7 via support structured data exchange and terminology mapping tools bridge gaps across systems.
00:11:18
Speaker
And we have seen this in practice through initiatives like EHR4CR and EHR2EDC in Europe, but also from vendors like Trinetix and Flatiron and Ignite Data, the latter enabling live automated EHR2EDC transfers at major sites.
00:11:38
Speaker
Many hospitals are already contributing harmonized query ready data.
00:11:43
Speaker
So while cross-border harmonization is complex,
00:11:47
Speaker
With the right to understand this and partnership, this is not only feasible, it's already happening.
00:11:53
Speaker
And I could also add that this morning I was talking to the German network of university hospitals, 46 hospitals in Germany is moving towards this initiative as well.
00:12:09
Speaker
So it's exciting.
00:12:12
Speaker
So I was under the impression that it was mostly compatible with the American or North American system of hospitals, because there are a lot of groups of hospitals that share similar systems.
00:12:28
Speaker
But in Europe, you see also progress in that direction.
00:12:32
Speaker
I think it's it's.
00:12:35
Speaker
we are going to face and we're going to live with heterogeneity when it comes to different EHR systems.
00:12:41
Speaker
That's fine.
00:12:42
Speaker
But as if we can work towards interoperability standards and and structure data, all that can be overcome.
00:12:53
Speaker
So, for example, the the German initiatives is working towards one federated
00:13:01
Speaker
repositories that data can be accessed.
00:13:06
Speaker
So it's quite interesting compared to how it was five years ago or 10 years ago.
00:13:16
Speaker
Yes, and I think you've alluded to the fact that you need so harmonization and it's not necessarily that you transfer data directly from an EHR to an EDC, but you need something in the middle to do this harmonization, is that correct?
00:13:37
Speaker
It is.
00:13:38
Speaker
It is.
00:13:39
Speaker
I think that lies what could be in a very broad term called the mapping, the mapping engines of those middleware layers is to secure that data.
00:13:51
Speaker
And I used to say this both in my time in AstraZeneca, but also in conferences.
00:13:58
Speaker
I remember in AstraZeneca when I said when I
00:14:05
Speaker
I informed about the progress with the first ESOURCE trial and my colleagues at Data Science and Artificial Intelligence were excited because they said, now, Matt, now we can begin to explore the data that resides in EHRs.
00:14:20
Speaker
And I said, sorry, couldn't be more wrong.
00:14:23
Speaker
The data that is subject for ESOURCE is data that is completely in synchronizing with the ECRF.
00:14:32
Speaker
No more, no less.
00:14:34
Speaker
It is only that data that is subject for the mapping and the transfer.
00:14:39
Speaker
And all the other data is impossible to access.
00:14:44
Speaker
Yeah, yeah, that makes sense.
00:14:47
Speaker
So based on what you've been doing for many years, looking into this topic and all the progress you're seeing in the industry, what do you think are the biggest challenges for the use of EHR data?
00:15:06
Speaker
I think
Scaling eSource: Challenges and Value
00:15:07
Speaker
scaling up use of EHR data in clinical trials faces several interconnected challenges.
00:15:14
Speaker
First and foremost, it is the data interoperability that
00:15:19
Speaker
That is the single most and I think the tipping point we have seen for the last couple of years is what I've talked about, the FIIA interoperability vehicle.
00:15:31
Speaker
The ability of systems across different hospitals, regions and vendors to communicate in a quick and a fast way and interpret data consistently.
00:15:43
Speaker
While standards like HL7-5 help, many sites still operate on legacy systems with varied levels of maturity.
00:15:51
Speaker
So it is... And having said that, that United States is ahead of the game because the Obama administration in 2008 actually
00:16:05
Speaker
put a standard, put fire that all hospitals in US should be fire ready.
00:16:12
Speaker
So that makes a big thing.
00:16:14
Speaker
Another compared to Europe, but Europe is keeping up now and
00:16:22
Speaker
Another major hurdle is the quality and structure of the data.
00:16:27
Speaker
Much of what's captured in routine care, especially physician notes or pathology reports, is unstructured.
00:16:35
Speaker
Extracting meaningful research-grade data from these sources requires very advanced tools like natural language processing, which is not widely deployed yet.
00:16:46
Speaker
And another thing with
00:16:49
Speaker
harnessing unstructured data is also the validation.
00:16:52
Speaker
So there's more, there's challenge there.
00:16:57
Speaker
But it is also a light in the tunnel, I would say.
00:17:00
Speaker
Finally, it is a change management issue.
00:17:03
Speaker
Adopting eSource workflows means rethinking site processes.
00:17:07
Speaker
It also means rethinking the sponsors' processes.
00:17:12
Speaker
Retraining staff and realigning IT infrastructure is not just about plug-and-play solution.
00:17:19
Speaker
It's the transformation that requires time and investment and trust between sponsors and healthcare organization.
00:17:27
Speaker
So, but I think the fantastic view with eSource is it has such tremendous value proposition, both for sponsors and for sites and ultimately for patients.
00:17:46
Speaker
Yes.
00:17:47
Speaker
And, you know, I'm always a bit frustrated with our industry, mainly because I want things to move faster.
00:17:59
Speaker
And I'm sure I share this frustration with many.
00:18:03
Speaker
But you've just made me realize that, especially for EHR to EDC, you're kind of moving a mountain, but you need to move that mountain a few rocks at a time in the right places.
00:18:20
Speaker
Do you agree with that?
00:18:23
Speaker
Yes, indeed.
00:18:24
Speaker
I mean, I've been trying to push for this technology since 2010.
00:18:28
Speaker
Yeah.
00:18:32
Speaker
So it's absolutely right and it is now sort of with the maturity of EHRs with the understanding and also the other thing of the fence is that
00:18:53
Speaker
clinical trials is becoming not cheaper.
00:18:57
Speaker
They becoming more data rich, they becoming more complex and therefore they're becoming much more expensive.
00:19:04
Speaker
This has come, this has reached, I think, an inflection point where sort of it's not sustainable.
00:19:12
Speaker
And this, this position eSource as something it's really important to
00:19:20
Speaker
to drive for even though it will take time and effort.
00:19:27
Speaker
When we set up our first pilot study in AstraZeneca back in 2022, that's before I left AstraZeneca, it was... I got the response from one of my
00:19:44
Speaker
major stakeholders and said that, Matt, if you could only transfer local lab, we will be so happy.
00:19:51
Speaker
Yeah.
00:19:53
Speaker
So this is also a very promising, it's not, it's incremental.
00:20:00
Speaker
And so there is a very good trajectory forward for this.
00:20:09
Speaker
Yes, and I think you just read the right point here that we shouldn't aim for everything to be done to start with.
00:20:18
Speaker
It can be an incremental process.
00:20:22
Speaker
So we have to talk about AI because everyone talks about AI.
00:20:27
Speaker
No,
AI's Role in Enhancing Data Quality
00:20:28
Speaker
more seriously, you mentioned that in some cases we may need natural language processing to analyze what's not necessarily structured data.
00:20:42
Speaker
Do you think AI can help in any other places with the EHR to EDC?
00:20:48
Speaker
Yes, yes, I think, but I do have a conservative view.
00:20:52
Speaker
I'm trying to be feet on the ground.
00:20:58
Speaker
And the reason is, I think absolutely AI is already beginning to play a role and its impact will grow.
00:21:07
Speaker
And either HR to EDC workflow, one of the biggest bottlenecks is high volume and unstructured data that I said.
00:21:14
Speaker
And
00:21:17
Speaker
I think here artificial intelligence, especially natural language processing, can add real value by converting narrative text into structured research-ready data.
00:21:27
Speaker
But having said that, it's also a very important validation step.
00:21:35
Speaker
It's not just it needs to be
00:21:38
Speaker
How should I put it?
00:21:39
Speaker
The process of harnessing unstructured data to meaningful structured data needs to be correct.
00:21:47
Speaker
But the AI can also assist what I think is less problematic thing and it can assist with data quality check.
00:21:56
Speaker
It can assist with pre-populating CRFs, flagging anonymous and anomalies and even predicting missing data points.
00:22:06
Speaker
This is much less problematic, I would say, because it is like an assisting steering in cars.
00:22:15
Speaker
It helps the driver.
00:22:16
Speaker
It helps CRCs, all which reduce manual burden.
00:22:21
Speaker
But it's important to emphasize that AI doesn't replace researchers or site staff.
00:22:26
Speaker
It augments.
00:22:27
Speaker
It augments their work.
00:22:29
Speaker
And I think in this second aspect, I think AI is definitely something that is going to play a big role.
00:22:41
Speaker
However, my view is that for AI to be effective, the underlying data must be clean, structured and interoperability.
00:22:52
Speaker
And as we often say, data first and AI later, without high quality input data, even the best AI will struggle.
00:22:59
Speaker
So AI is a powerful enabler, but only once we build trust for the data foundations.
00:23:06
Speaker
So you see, my answer is, it's mixed.
00:23:09
Speaker
I think the second part is already making difference because it's augment.
00:23:14
Speaker
It helps to
00:23:16
Speaker
to find things.
00:23:17
Speaker
The first thing, this is harnessing unstructured data is going to take a longer time and it's more complicated than one might think.
00:23:28
Speaker
Yes.
00:23:29
Speaker
Yeah.
00:23:29
Speaker
Yeah.
00:23:30
Speaker
But it's good.
00:23:31
Speaker
It's good to have a realistic view rather than thinking AI is going to change everything.
00:23:38
Speaker
It sounds like you have a much better understanding of the
00:23:44
Speaker
like the reach AI can have in this topic.
00:23:50
Speaker
And I like the fact that it could help not only, well, it could be an assistant rather than trying to use AI as something that replaces a full process.
00:24:02
Speaker
Absolutely.
00:24:02
Speaker
I used to say like this, and I said it in my last year in AstraZeneca as well, because the
00:24:11
Speaker
AI is in clinical research, it is a kind of a virtual research assistance, but that assistant should always be sitting in the back seat of the car and you should always be sitting in the front seat.
00:24:25
Speaker
That means that everything that comes from the back seat of the car, you need to understand in all cases, whether it's right or wrong.
00:24:34
Speaker
If you can't decide that, you should leave it
00:24:39
Speaker
and never ever let the research assistant sitting in the front seat.
00:24:44
Speaker
Yeah, I don't know if everyone agrees with that.
00:24:50
Speaker
I do.
00:24:53
Speaker
I'm sure there are people who see AI as being able to replace decision-making in some respect, but I agree with you.
00:25:02
Speaker
It should be an assistant.
00:25:03
Speaker
Well, I've got one last question for you today, and I can't wait to hear your answer.
Applying Nash Equilibrium in Clinical Research
00:25:09
Speaker
What's the best piece of advice you've received that you consistently apply in your work?
00:25:16
Speaker
And I've been thinking about this and I've also been talking about it in conferences because it has, it's something that really applied.
00:25:28
Speaker
It is a very impactful principle I apply in my work and it's inspired by Nash equilibrium after John Nash, the Nobel Laureate in economics, 1994.
00:25:42
Speaker
The idea that the best outcomes rise when every stakeholder acts in their own best interest while still contributing to the collective good.
00:25:54
Speaker
This is not just a theory of game economics.
00:25:57
Speaker
It's a powerful guide for collaboration in complex and multi-stakeholder environments like clinical research.
00:26:03
Speaker
So whether we are scaling e-source technology, aligning sponsors and hospitals, or introducing AI in real-world data, lasting progress only happens when parties sees clear value for themselves and recognize the benefits of shared success.
00:26:21
Speaker
And that is the balance
00:26:24
Speaker
I think I will always drive for.
00:26:27
Speaker
And by upholding this principle demands actually effort, courage, and also endurance.
00:26:34
Speaker
It means creating trust, being transparent, and sometimes taking the harder path to ensure mutual benefit.
00:26:43
Speaker
Yet when done well, it unlocks sustainable collaboration, innovation, and learning.
00:26:50
Speaker
And I've seen this firsthand.
00:26:53
Speaker
several times scaling technologies like eSource isn't just about efficiency gains it opens up for new avenues it's a partnership across pharma hospitals and academia and vendors in that sense Nash insights has become a practical compass for how I approach systems level transformation in healthcare research
00:27:15
Speaker
Nice, very nice.
00:27:17
Speaker
And especially true, I would say for what you're focusing on, because more and more we hear conflicts between various parties of clinical trials, like sites and CROs and sponsors don't necessarily all
00:27:37
Speaker
see the same benefits.
00:27:39
Speaker
So what you're saying essentially is that this approach needs to be beneficial for everyone.
00:27:45
Speaker
Otherwise, it's not going to last.
00:27:48
Speaker
Exactly.
00:27:50
Speaker
And it works actually and it's quite powerful.
00:27:59
Speaker
And as a matter of fact, it works almost universally, I think.
00:28:06
Speaker
Any kind of collaboration with different people and so on and so forth, taking, trying in every, every sense, every step, establishing Nash equilibrium.
00:28:18
Speaker
If you do that, it will go well.
00:28:21
Speaker
People will be happy.
00:28:22
Speaker
People will be motivated.
00:28:25
Speaker
Yeah.
00:28:26
Speaker
That's a very sound piece of advice.
00:28:28
Speaker
Thank you very much.
00:28:29
Speaker
Thank you.
00:28:32
Speaker
Now, it's been an absolute pleasure.
00:28:36
Speaker
I will definitely keep an eye on what you're doing because I see it as one of the futures of clinical trials contributing to better outcome for everyone.
00:28:52
Speaker
Thank you.
00:28:53
Speaker
Thank you everyone for watching us or listening to us today.
00:28:58
Speaker
You can find more clinical data talks on our website.