This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

  Today on Town Hall

At the end of the day, billions of dollars are spent in operations, but you know , a very small fraction are spent on actually getting access to that data, understanding what its impact can be and, they do that at their own peril, quite frankly.

My name is Bill Russell. I'm a former CIO for a 16 hospital system and creator of This Week Health.

Where we are dedicated to transforming healthcare, one connection at a time. Our town hall show is designed to bring insights from practitioners and leaders. on the front lines of healthcare. .

Alright, let's jump right into today's episode.

 Good morning. Welcome to this Week Health. I'm Matt Sullivan, and I'm here with Alan Forster. He's the Vice President of Innovation, transformation, and clinical Performance at the McGill University Health Center. Good morning, Alan.

Nice to meet you, Matt.

It's wonderful to talk and I think today we're gonna cover a topic that a lot of people probably understand enough about, but need to understand a lot more about, and that's really some of the work that you've been doing on data.

So tell us a little bit about how McGill is empowering clinicians to really go through sort of a continuous improvement process through a self-service data discovery platform.

we have a very proud academic tradition here at McGill University Health Center. And I was recruited to join this prestigious and fantastic organization.

One thing I noted when I got here was that, while there were amazing analytics skills, there wasn't great access to the hospital data. and that's not because they didn't invest. They had a data warehouse. They had a they had a team. They actually, they had several teams supporting access to data.

But really that there's still this insatiable demand for accessing data. So the CEO really wanted to create a data driven culture where our health system was making decisions based on evidence that we are generating ourselves, create that learning health system. And she recruited me to, to help do that.

And, with that recruitment, what I've been leading is the implementation of a self-service data strategy. And with the intention there that rather than having to go to a middleman to get access to data you can access the data directly yourself, whether you're a physician, a researcher, and a manager, a student, People can access and log on to a system themselves and really then learn from the data, learn from the, the system to then apply those lessons learned either in a clinical practice sense, in a sort of quality improvement process sense or in a research and analytics sense. So, so that's really where, where that's, that's the strategy.

Empower people to do their thing. I can tell you a little bit more about, about how we're doing that, but that's, that's the overall strategy.

No, No, it's great. I'd love to know how you, how you execute.

the first thing I did was to get investment for software and hardware solution called MDClone

It ingests your data to be available within their data lake organized in a way that makes it very intuitive for people to do analysis. So it, it has a fantastic, data model, very easy to build from your source systems. it then has a very nice user interface. So to build queries to create that your own data set is very intuitive. You don't need any special programming skills or language. And then I think fundamentally the most sort of exciting part about.

At MD Clones Adams platform is, is its ability to create synthetic data. so as a user you log into the system, you create a query you have an understanding of that query just from the sort of the process of generating it. But you don't actually see any personal health information at any point.

And then you, from that query, from the table and behind you create a synthetic data set that, that you can use to draw inferences from.

Now, I've talked to Ziv a little bit at MD Clone about that. Tell me how that actually works and sort of put that, put us in the mind of a new user.

Like, okay, so I've got an idea. I'm coming to you. You're now helping me understand this. And how do you make that decision of, do we use regular data, synthetic data? What's the value proposition there? That's curious to me.

First of all the real data are always there. It's not like you cannot get access to the real data.

So, Because oftentimes when you're doing like an analytical project, you don't really always know the, like the distinct question you're asking you're exploring the data and that's where the problem gets created.

Because if you're speaking with an analyst and you don't really know. What you need like you have a general sense as to what the question is, but you've never actually constructed an analytical data set. You don't know the data behind it. You, you can't really give instructions, clear instructions to that analyst for you, right?

So. However, if you're working with the data itself, you see it, you create it, then you can but you know, the problem is because of the privacy issues, you're not allowed to just go and play in real data, right? You're you're just not allowed. It's against the law so what this does is it allows you to create a set of data.

A synthetic data set, which is it's made up, it doesn't have any linkage back to the original data, but it has exactly the same statistical properties as the real data. And it looks like the real data, that the variable names, the variables themselves, that they look exactly the same. The missing data will be the same as it would be in your real dataset.

So like you asked the question, walk me through how it would work. Like, imagine you're doing a project to look at you, you know, mortality rates in your cardiac surgery patient population, for example, you or, or your length of stay or cost.

you could start looking and exploring factors that would predict that, you know, obviously the type of procedure. maybe baseline comorbidities. And you can add in laboratory tests, you could add in your treatments that you're giving. And then you could start to start exploring to see what's causing the variation.

Maybe there's some variation by the surgeon. Maybe there's variation by the comorbidity that the patients have maybe the day of the week. We've seen some instances where the day of the week predicts length of stay. In fact, it's stronger than most other things actually.

So, you can then start to drill in on, you know, potential associations that might warrant exploration, either from a causation factor or from a, you, you know, like really just an association. And so then, now you have this, these, this inferences that you've made with the synthetic data.

you can be very confident it's gonna be what the underlying data are. Now, if you're gonna do a research project, you probably want the real data

go through all the hoops, but you don't Yeah. And then run that time.

I see.

But then it's just like a click on a button. And you don't need to involve that middleman at all. Like, it is essentially a automated process to get the real data. But the really fundamental thing that's different is. Now that physician or that clinician or that manager,

they understand what you can do with the data, what you can't do with the data, and that actually creates, starts to create a virtuous cycle where they now understand why they need to document well, in the clinical record, they start to understand their business better because, the process of generating the data is from their clinical processes, from the hospital process, the health system processes.

And if you start to see variation in, things, fundamentally it could be very true variation in care or it could be just variations in how the data are collected. And then finally, I guess the other thing, which is more important is people start to care more about the processes themselves, because they can see now what's causing the problems that they're observing in the clinic.

Like whether the, the patients are showing up late or maybe there's a doctor booking all the patients at the same time. And therefore there's a big rush at one o'clock. But by three o'clock people have been waiting for two or three hours

the idea is that if by giving people access to data so they can understand their business better, they start to become more motivated to change their own behavior. They start to understand it.

That's great. Well, you've said in the past that you believe that data really is the lifeblood of what this is.

And I'm starting to tap into your deep passion here. And so tell me why you think that, and then maybe transition to, what we've been talking about in, in all of informatics over the last three to four years is all AI, but really AI sits on a really solid data platform.

So tell us about those two things. How they connect and what do you think for the future?

Well, I was explaining here just a moment ago that, data is generated through our business processes and through clinical practice. And if we're to draw insight from that data, whether that be through, statistical models of some kind or, other forms of machine learning.

If you don't actually have the data in a format that can be used by those models and then you can't make the models. You can't, and then you can't run the algorithms in, in practice. And, and, And so fundamentally it, it all comes down to this digitization of care.

that's for the AI component, right? But if you step back a bit and say, okay, We wanna look at this business of ours, healthcare business of ours. We wanna make sure that it's producing the most value, best outcomes lowest cost. And we wanna make sure our staff are well and engaged.

Well, how are you gonna do that? Are you gonna use people's opinions? Are you you're gonna talk to the last person who's discharged from your hospital and ask them? You gonna talk to the doctors? Well, you know, everyone's gonna have an opinion. At the end of the day, if you don't have quantifiable information to define what the outcomes are what you did to get those outcomes, there's no way you can actually do anything.

which means therefore you cannot do innovation because you don't know whether you're better than before. you can't actually uh, sort of actually state we need more resources to do a and less resources to do B. It just is impossible. And so for me, that's where that comment of the lifeblood, like, you know, I would just say health systems are not spending enough money on setting up their data for success.

At the end of the day, billions of dollars are spent in operations, but you know, a very small fraction are spent on actually getting access to that data, understanding what its impact can be and, they do that at their own peril, quite frankly.

Yeah.

Well, the way you've set it up, it sounds like if we didn't invest, then we've really not taken the very first step in understanding healthcare and the necessary changes. Alan the, insights you've given us are phenomenal. Thanks for your time and thanks for being on this with help.

Thanks so much.

No problem, Matt. Pleasure.

Yeah.

  📍 📍 Thanks for listening to this week's Town Hall. A big thanks to our hosts and content creators. We really couldn't do it without them. We hope that you're going to share this podcast with a peer or a friend. It's a great chance to discuss and even establish a mentoring relationship along the way.

One way you can support the show is to subscribe and leave us a rating. That would be really appreciated.

Thanks for listening. That's all for now..