HIStalk Interviews B. J. Moore, CIO, Providence

B. J. Moore is CIO and EVP of real estate strategy and operations at Providence of Renton, WA.

Tell me about yourself and the organization.

I’m the chief information officer at Providence and am also responsible for real estate strategy and operations, so I wear two very distinctive hats. I’ve been at Providence for four years. Before that, I was at Microsoft for 27 years in various roles, the last of which was vice president of engineering for the Azure group. I am well versed on the cloud and the transformation of the cloud, which has been helpful in my journey here at Providence.

Are your dual roles based on your personal interests, or is that an indicator that some fluidity exists in how Providence views its bricks and mortar footprint versus technology?

Absolutely there is some fluidity there in three areas. One, we have a big, bold goal to be carbon negative by 2030, and real estate and IT are the two biggest offenders on the carbon front. A big way that we can solve it for both spaces is through technology, so it makes sense to have that under one leader and manage both of those portfolios to drive it down, but also use technology to more efficiently reduce our carbon emissions.

Two, in this modern workplace post COVID, everything is now a hybrid work environment, with some remote and some in-person. How do we create these rich, collaborative work environments when people are at work to get the best of the workspace, and how do technology and real estate tie into that? 

The final bucket is that the future of healthcare is becoming more and more virtual. Before COVID, a bed was a very concrete thing. A hospital had X number of beds. Now a bed is more abstract. It could be virtual care, at home, or a physical bed in a hospital. Thinking of a bed in more of a virtual way, more of an abstract way, is helpful. Me being able to wear both hats helps us bridge that gap.

What do you think about Nuance DAX and its enhancement with ChatGPT integration?

The first implementation of DAX was good, but there are human beings in the middle doing quality assurance, so it does a good job of transcribing. There’s a two- to four-hour QA process for a human being on the Nuance side to review things before it gets posted to the medical record.

With generative AI, there’s a real opportunity to make that near real time, to cut that quality assurance person out of the middle and use generative AI, that large language modeling capability, to close that gap. We are the single biggest adopter of DAX and we’ve been a early partner with Nuance on that product, actually Nuance and Microsoft before Microsoft bought Nuance.

What is ChatGPT’s potential?

I see the potential as huge. If you would have asked me six or seven months ago what I thought of generative AI or ChatGPT, I probably would have given you a blank stare, or would have said that I think AI has been overused. What we’ve seen in the last six months is just incredible. From 3.0 to 4.0, it really opens your eyes to what’s possible on generative AI with images, video and the whole processing. It’s just absolutely incredible.

The downside is that what everybody sees is the consumer version of it. It was literally fed every piece of social media pop culture, from “I Love Lucy” to  “The Communist Manifesto.” It was fed everything, so it provides for a great model, but it also is easy for people to find examples where it has bias or answers in a misleading way or whatever.

I love ChatGPT and showing them the power, but I worry that people assume that it’s one size fits all versus it’s this large language model that we can apply to clinical settings. We are working with Microsoft and the Azure team to take that technology, not the generic ChatGPT, to train that against our own information here at Providence, our own medical data, so you don’t get the quirkiness of “I Love Lucy,” but get the solid domain of healthcare. I think we’ll see better outcomes than maybe some of the YouTube videos where there’s some funny scenarios with ChatGPT.

How can EHR vendors use ChatGPT to improve or extend their product?

We are an Epic shop and are actually a real example. We’ve taken that power and we are going to use it to train against the Epic inbox. Our doctors are overwhelmed with messages and maybe don’t get the messages until the end of the day. We are training the model to look at these messages, triage them, and bring the most important ones to the doctor’s attention.

It’s an example of something that can be done within the workflow of the EHR. It’s an example of a baby step, by using this technology that can help the productivity of a doctor and hopefully help a patient by getting those critical messages upfront.

Epic has been a great partner with Azure. They have some good computational capabilities that have partnered with Azure. When I hear of them wanting to partner on the ChatGPT side of things, it feels like a natural extension of that partnership.

We heard early on that providers who didn’t move to the cloud would miss out on tools and capabilities and we’ve seen the rise in low-code development tools, ChatGPT, and APIs. Will health systems that don’t have a big engineering group use these tools to do in-house development?

It helps to use some of these generic capabilities and see the art of the possible, but the advice I give to everybody when I’m speaking or on the conference circuit is that you have to get on these native cloud solutions. You can’t be a locked in on prem. You are really missing out on the innovation since and all the innovation is happening in the cloud. You can use ChatGPT without being in the cloud, but our example, where you need to train it on your own models and your own data, won’t be effective for these smaller systems that are locked on premise. The cloud journey is necessary.

Companies clamor for EHR data to use for AI training and to support life sciences research. How is that use of EHR data evolving?

It’s a no-brainer, and it’s much bigger than that. We talk about the big data EHR, but big data is not EHR. It’s  all of the information from the biomedical devices, from wearables, from social determinants of health, all these other things. When you have that data on premise, you’re really limited by the scale-up capability of hardware that you have on premise. Whereas in the cloud, you have basically unlimited storage and unlimited scale.

As part of our journey four years ago, we have already moved all of our data to the cloud. To me, that’s the only way you can connect all this data together, and then as stated earlier, that’s the only place these advanced analytic AI tools exist, is in the cloud. It’s a journey that everybody has to do. My advice to your readers is that it’s much bigger than EHR. EHR data is Step 1 of 20 as far as the interesting data sets that should be in the cloud.

Do we have the interoperability maturity as well as the motivation to connect all of these data sources of a patient’s longitudinal record?

It’s still a challenge. Even if you’re on the same two versions of Epic, you put that in the cloud, it’s still hard to integrate. People are seeing the value more, especially as you connect with other data sets. It’s easier in that you have more computational power, but there’s still some blocking and tackling issues. Bringing that data together, normalizing the data, cleaning the data, de-duping the data, making sure that you have that full 360-degree view of patient is still a challenge.

How will that change if the prediction comes true that consolidation will result in the country having just a few huge health systems?

I don’t see the consolidation trends. I don’t think we’re in a governmental environment where there’s energy to consolidate. Even when they do, look at somebody like CommonSpirit. They are a large health system that grew through acquisitions and they are still on 20 different electronic health records. I don’t know that it solves it.

I think what solves it is that you have the computational power, and where you went earlier in your question, you now have the imperative to do it. So I think you’re going to see more cloud-level integration, and that’s how you solve that 360 degree view of a patient versus necessarily hospitals consolidating to achieve that. Hospitals are consolidated to get efficiencies of scale, but I think the data problem is independent of that.

Both providers and vendors are being challenged to protect their bottom lines as we roll out of an economic environment of extensive investment and experiments with innovation. How will they weigh the adoption of technology that might be innovative with the need to protect margins?

I can only speak to what we’ve done at Providence. Luckily in my first four years, we really modernized that back office. We were already on a single instance of Epic. We are on a single instance of Oracle Cloud. We’ve done that heavy lift in our budget.

My budget is about 15% smaller this year. We have really had to tighten our belt –get rid of contractors, vendors, unfortunately lay off some employees, and reduce or cancel licenses and subscriptions to focus on shorter-term wins. We have that luxury because we have already consolidated, but the feedback I give partners is that if you don’t have a ROI in 12 months or less, we really are not in a position right now to make those bets.

This is where generative AI six months ago wasn’t even a tool in my toolbox, but it feels like a tool that I can add quickly that can have that easy 12-month or less return on investment. The key is productivity. We have nursing and caregiver shortages. There’s never going to be enough. How do we make them more productive? Right now, 50% of their time is spent doing administrative work. If generative AI can chip away at that and get rid of that burdensome administration and allow them to practice their craft, I think we can reduce costs, but also reduce burnout and attrition at the same time.

How will big tech companies that have made recent health IT acquisitions, such as Microsoft and Oracle, influence healthcare?

Although Microsoft bought Nuance, I don’t perceive them wanting to get into healthcare. I see that as adding to their AI capabilities around ambient artificial intelligence and voice recognition for improving their services. It just happened to be a healthcare company. I see that as different than Amazon, which clearly wants to get into healthcare, or Google that wants to get into healthcare. I would separate the two.

Frankly, when I look at partnerships, I look at that. When I was at Microsoft, some of our best customers were retailers that were leaving Amazon to come to Microsoft because they didn’t want to be hosted on a competitor’s infrastructure. I think it’s the same thing in healthcare. When I look at partnerships, is it a partner that is more altruistic, and I think Microsoft is more in that camp, or is it somebody that today may be a tech partner, but tomorrow may be a competitor? That certainly weighs into how I make technology decisions.

How do you as a CIO develop a strategic plan in an environment that changes dramatically month by month?

Like I said, I have the luxury of having closed the book on our big transformations last year. Our focus has shifted to how we optimize the investments we have. Great, we are on a single instance of Epic — how do we optimize that? We’re on a single ERP — how do we optimize that? How do we start chipping away at the holy grail, which is around patient experience, caregiver experience, caregiver productivity, and health outcomes? Our three- to five-year horizon is more looking at those.

Based on our budgets, we will be more opportunistic to chip away at that. Luckily I don’t I need to go to a single ERP or need to go to Epic. I don’t have that cloud hanging over my head any more. Our planning horizon probably looks markedly different than other large health systems.