Q&A with Providence CIO BJ Moore: “Don’t Try to Catch Up on 20 Years.”

When BJ Moore came to Providence in January 2019 after more than two decades with Microsoft, he made a common assumption: that healthcare was behind other industries because of resistance to technology adoption. As he quickly learned, “that wasn’t the case at all.” The problem was that “healthcare wasn’t getting the right technological leadership,” he said during an interview with Kate Gamble, Managing Editor at healthsystemCIO. Given the right leadership and the right strategies, “they were actually quick to adopt.”

Of course, it wasn’t easy – and it never will be, particularly as Providence “systematically marches toward the goal” of being carbon-negative by 2030 and moves closer to achieving a single platform.

In the interview, Moore talked about how they’re approaching these and other core objectives, and offers his thoughts on where the true value of ChatGPT lies, the questions he poses to vendors, and his constantly evolving role at Providence.

LISTEN HERE USING THE PLAYER BELOW OR SUBSCRIBE THROUGH YOUR FAVORITE PODCASTING SERVICE. 

Podcast: Play in new window | Download (Duration: 24:47 — 17.3MB)

Subscribe: Apple Podcasts | Spotify | Android | Pandora | iHeartRadio | Podchaser | Podcast Index | Email | TuneIn | RSS

Key Takeaways

• Although Providence’s goal to become carbon-negative by 2030 hit a few roadblocks because of Covid, the organization is “systematically marching” toward it by focusing first on non-capital-intensive initiatives such as putting printers and PCs in sleep move and using smart devices to manage thermostats.
• After a four-year journey, Providence completed its Epic migration in March of 2022. However, “getting on a single instance of Epic didn’t make anyone’s life better. You need to optimize,” Moore said.
• ChatGPT is “the hottest topic under the sun,” but the lack of real-world examples is concerning to Moore, particularly when it comes to patient care. The smart move, he believes, is leveraging large-language models to reduce the administrative burden and improve productivity.
• The most important advice Moore offers for CIOs and other leaders coming to healthcare from outside industries? Don’t make assumptions about why healthcare is tech-resistant and “don’t try to catch up on 20 years.”

Q&A with BJ Moore, CIO & EVP of Real Estate Strategy Operations, Providence

Gamble:  Hi BJ, thanks for taking to speak with us. The last time we spoke was in the fall of 2021 — it may not seem like a long time ago, but things change so fast. I wanted to talk about what you guys are doing at Providence, including the goal to become carbon negative. Where are you guys with that?

Moore:  We’re systematically marching toward the goal of being carbon negative by 2030. It’s something that we’ve mapped out. But because it’s a multi-year investment, unfortunately Covid put some pause to some of our efforts, especially the ones that are pretty capital intensive. And so, during Covid, we started doing non-capital-intensive things to reduce our costs. If you objectively look at the glide path toward negative, we’re actually doing well. The next batch will be tougher. We’re working with our executive leadership team and financing to do that.

The other thing we’re doing, which is maybe unique to us, is that bonuses for the top 15 or 20 execs are based on two goals. One is to be carbon negative by 2030. We think of it not just as a big, bold goal as a health system, but something that executive leadership should be compensated based on our achievement toward that goal.

If you look at the glide path, we’re on track. But we’re going to have to do more capital-intensive projects to stay on that path.

 

Starting small with green initiatives

Gamble:  What were some of the smaller projects you worked on?

Moore:  Some of that was just about more efficiently using our existing resources. For example, powering down PCs and printers and putting them to sleep. Using smart devices to manage lights and thermostats. As part of the normal maintenance break fix, as we replace boilers or AC equipment, we replace it with higher efficiency items. We’re looking at how to bake this into the natural rhythm and make sure we’re doing things to help achieve the harder, more capital-intensive things.

Take, for instance, LED lighting. We have 52 hospitals and around 1,000 clinics. That’s a lot of light bulbs. Even if the light bulbs are funded, if you think about the logistics of going and replacing every one of them, it is capital intensive. With some of the bigger capital projects, we’re looking at 10-, 20-, 30- and 40-million-dollar initiatives to replace some of these more expensive infrastructures. That’s going to be tough.

Some of it is selling off assets. Covid allowed us to get rid of a bunch of administrative space, and so we were able to achieve efficiency there. These are examples of ‘easy button’ things. But they weren’t easy. They all required thought and strategy. It’s easier to sell a building and take that carbon footprint away than it is to take a hospital that was built in 1890 and make it efficient.

Dual roles

Gamble:  Right. And I know we’ve talked about it in the past, but this is an area where it really makes sense having real estate under your title, especially from a logistics point of view.

Moore:  Yes. We talked prior to Covid about a hospital bed being a physical asset. A new hospital bed could take five to 10 years to add. And now, wearing my IT hat, it becomes about virtual care and care delivery at home, so that we can naturally do things that improve healthcare for our caregivers and patients, but tangentially help us

It’s easy to talk amongst my team about finding better ways to sleep PCs and how to replace them with models that are more conservative. Wearing both hats has been helpful. Those are small examples, but I think it helps make the point.

 

“Energy wasted” on inefficiencies

Gamble:  For sure. All those things add up, especially when you have such a large organization.

Moore:  On the inverse, think about the energy we’ve wasted because we didn’t do these disciplined things. There’s no advantage to having inefficient lighting, having PCs or printers on all day, or having having lights on in unoccupied rooms. It wasn’t a trade-off. There was never any benefit from the inefficiencies we had. And so, while this has helped us achieve our goals, it was also the right thing to do from a pure business perspective.

Change management – “We’ve had some learnings”

Gamble:  And most of these changes haven’t had a big impact on workflow, correct?

Moore:  Exactly. The change management has been relatively low. But in the example of the sleeping PC, if you’re in a clinical setting, there’s some learning there. In some cases, we were oversleeping PCs. Someone would just step away for 5 minutes and come back to complete the task, and they’d be locked out. We found that maybe it should be shift-based; for example, at 8 o’clock we can be aggressive, whereas in the shared services space, we were being overly aggressive and had to pull back a bit. But in general, the change management was relatively simple.

Gamble:  That’s probably going to happen anytime there’s new tools or processes being used.

Moore:  Right. And now that we know, it’s embarrassing to think about it. We had the best intentions; we thought, ‘nobody needs an idle PC for this period of time. Let’s put those on sleep.’ Well, there are clinical reasons for why it has to happen that way.

 

Migrating to the cloud – “We’re pretty much complete”

Gamble:  Let’s switch gears a bit and talk about cloud migration. Where are you on that journey?

Moore:  When I joined Providence 4 years ago, we were probably 2 percent through that journey. Now we’re 70 to 75 percent complete. There are a few reasons it’s not closer to a hundred. In terms of data, we’re probably 98 percent complete. But with Epic, we’re not there yet. We have the largest single largest Epic instance in the world, and so our demand for Epic on the cloud is greater than any pretty much anyone else. And so, we are further behind, but it isn’t because we’re not adopting the cloud; it’s that the capabilities aren’t there yet. But we have a good partnership with Epic.

We’re in the process of migrating our Epic environment to the cloud. Once we complete that, we’ll basically be at 95 percent. We’ll probably never be a hundred percent. There are things that need to remain locally and there are legacy applications we can’t move to the cloud. Or there’s fault tolerance or other reasons we need to keep it in a hospital or clinic. But, to summarize it, we’re pretty much complete with our journey to the cloud except for that kind of massive epic infrastructure that we’re in process of moving that to the cloud.

Gamble:  Being the largest instance of Epic comes with its challenges, I would think.

Moore:  Yes, but we have the largest single instance, just to clarify. There are other health systems on Epic that are larger than us, but they are segmented across multiple EHRs, versus us having a single EHR code base for all 52 hospitals. We’re unique in that way.

 

The “heavy lifting” of EHR optimization

Gamble:  What has been your approach to tackling something on that big of a scale?

Moore:  It was a 10-year journey, which we completed in March of last year. We were on multiple versions of Epic. We were on Meditech. We were on Allscripts. We were all over the board. The question was, how do we get everybody onto a single code base and single instance? Now, I don’t want to diminish what the team did, because it was a 10-year journey, but the unfortunate part is that was step 1 of 10.

What that allows us to do, on the process side, is to act as a single health system. But when that happened, most of my caregivers didn’t high five me. It was more like, ‘okay, I moved from Epic A to epic B, but my life hasn’t gotten better.’ Now that we’re on a single code base, however, we can optimize the experience for our caregivers and our patients. We can do it once and then apply it, and then everybody will benefit from it. That’s the heavy lifting we’re doing now. Epic, from a pure scale perspective, can meet our needs. And so, we’re not focusing a lot of energy on how to get the infrastructure to scale to meet our needs. Epic has been a good partner there.

“Blessing and curse” of multiple EHRs

It was both a blessing and a curse having multiple EHRs. The curse was that we had multiple EHRs. The blessing was that each hospital could do their own thing. Now, we’re looking at things from a governance perspective. We don’t want every hospital to do their own thing. We’re going to do things in a very systematic way. That’s become more of a business problem than a tech problem. How do you get everybody to act as a single health system? Because everybody thinks they have the best practice.

 

Getting standardized

Gamble:  It sounds like a case study in change management.

Moore:  It is. And this is where I feel we are uniquely set up. When I joined four years ago, we had multiple EHRs, we had multiple ERPs, and we had multiple networks and domains. We had Teams. We had Skype. We had WebEx. We had everything. Over the last four years, we’ve consolidated. Now, we’re on a single ERP system with Oracle Cloud. We’re on a single EHR with Epic. We have a single, modern network infrastructure with Cisco. We’re standardized.

It’s not, ‘we have 10 ERP systems. How do we do it 10 different ways?’ Now, we’re now on Oracle Cloud. How do we implement chat GPT to benefit our shared services employees and benefit everybody? How do we implement chat GPT in partnership with Epic and Microsoft and apply it to our single instance to get the benefits? It’s not, ‘how do we do it for Meditech or Allscripts.’

And so, I feel it’s been a tough four years to get to single platforms and processes. But we have this next wave of innovation with generative AI and these large language models, and I think we’re set up uniquely at Providence.

 

Gamble:  Right. You’ve done the legwork to get to that point. For the most part, was it accepted fairly well to move toward standardization?

Moore:  At an executive board level, absolutely. Unfortunately, we implemented a lot of this change during Covid. For our caregivers, it probably felt like an overwhelming amount of change. But everybody was willing to do it. I was at Microsoft for 27 years; my perception was that a tech company like Microsoft was adept at change, and that healthcare systems would be really resistant to change. My experience at Providence was that we were actually more open to change at Providence than we were at Microsoft.

Looking back at the heavy lifting we did over the last four years, there was a lot of pain and suffering. But compared to the experience Microsoft, it was actually much less than I expected. And now it’s over. It’s in the rearview mirror. We don’t have another ERP coming. We’re done. We’re going to optimize it. We’re not going to switch Epic on you again. We’re going to optimize it for you. That’s a much easier message. And luckily, most organizations don’t have a long memory. They forget what it was like a year ago when you put them through hell. They remember today when life is stable and pretty good.

 

ChatGPT use cases

Gamble:  I agree. Earlier, you mentioned large language models. Can you talk about what you guys are doing and what you hope to do in that space?

Moore:  We’re looking at a few meta points. Looking at healthcare in general, we don’t have enough nurses or doctors; the amount of time it’s going to take to get new nurses and doctors is just too long. Those same nurses and doctors spent 40 or 50 percent of their time doing administrative work — what can we do to improve their productivity and get rid of things they don’t enjoy? That’s what we’re looking at chat GPT to do. We’ll focus on the nonclinical settings first; maybe using it to help nurses and docs, but only with non-clinical experiences. When it comes to HIPAA and the regulatory environment, it’s uncharted territory.

Inbox management

One possible use case is inbox management. People send a lot of messages to their doctors, but managing the inbox is tough. How do we use these tools to triage messages for doctors? How do we put the most important messages up top and then help them craft responses to improve their productivity?

We’re looking at non-patient-care examples around shared services, HR, finance, and help desk. There are a lot of places where we can use the technology and look at what works and doesn’t work. And then as we build that muscle or strength, we can start putting in more complex environments or more environments where maybe we start to look at clinical or regulatory things. But we’re definitely not starting there. There’s already enough to learn with large language models; to then add regulation on top of it is just too complex.

 

Gamble:  Absolutely. But with something like the help desk, there’s a lot of potential.

Moore:  Right. And proactive measures. When we look at our logs, how many people call in with printer issues? Chat GPT-type intelligence could do that. As I remind my team, the best help desk experience is the one that never happens. We can use ChatGPT when someone has printing problems; we can evaluate the printer and say, ‘you need a new driver,’ and a human being doesn’t have to get involved.

We can use that technology to scan our environments; see who else has a driver that’s out of date and proactive update it. That way the next person never has a printer issue because ChatGPT already took care of it. Those are really good examples where we can improve and experience, but more importantly, prevent the issue from even happening.

“The hottest topic”

Gamble:  It’s interesting. We’ve seen things that are trending on such a high level, but this is beyond anything. All of a sudden, ChatGPT was everywhere. I’m sure you get a lot of questions like, ‘are we doing this? What’s up with ChatGPT?’
Moore:  It is by far the hottest topic. I thought it was hot, then I went to HIMSS, and it was 110 percent of the conversation. I thought maybe we were talking about it too much, but when I heard my peers talking about it, I thought maybe we’re not talking about it enough. It seems to be eclipsing the sun right now.

The counterbalance to that is something I brought up with my team, which is that if you would’ve asked me on December 31 of last year how I felt about where Providence was, I was say, ‘We’re world class. We’re ahead of everybody else. We’re on a, you know, modern everything, single instance of everything. We really have our stuff together. And then if you asked me a few months ago how we compare, I’d say we need to do more on the generative AI-GPT front.

When I went to HIMSS everyone was talking about it, and one person would say, ‘That’s awesome. But can you give me great examples of how other health systems are using chat GPT and how we can replicate it?’ And the truth is, there aren’t good examples. We’re all talking about it, but I really can’t give you a good example. So maybe we’re not that far behind.

 

Gamble:  You’re absolutely right. And I don’t think anybody wants to start on the patient side, especially with so many questions still.

Moore:  That’s what’s tough for me. We don’t want to introduce ChatGPT at the doctor’s office and have to make clinical recommendations or decisions. We’re not ready for that. It’s like when a patient goes home and starts to Google things with no context of their medical history and make decisions based on that. It’s actually a much worse case scenario than if I took ChatGPT in the context of my own health record and made recommendations. I’d love to find a way to use ChatGPT with my health record, but do it in a way that says, ‘Treat this the same way you would Google results.’ It could explain what it means to have low vitamin D, for example, and explain what you need to do. It would be much more helpful. But speaking for myself, I’m afraid to look at clinical settings just because of the regulatory hoops.

And honestly, we have so much opportunity to use ChatGPT for shared services and administrative work. We could spend the next two years innovating that.

 

Vetting vendors

Gamble:  Let’s talk about the vendor piece. When something like ChatGPT is hot, all of a sudden, a lot of vendors want to talk to you about it. What’s your approach there?

Moore:  I have a few quick questions that I ask. One is, are you trying to build your own language models or are you just leveraging what Amazon or Microsoft or Google have already done? It’s kind of an IQ test. If you’re a small company with 30 employees, and their answer is, ‘yeah, we’re developing our own lang language models,’ that’s the wrong answer. They don’t have the scale or the capabilities to do that.

Is it LLMs or machine learning?

So yes, I do have simple ways to weed people out, but it can be overwhelming, and you start to wonder what’s real. How many people are abusing the term? When we talked last, I told you my pet peeve, which was that everyone was misusing the terms ‘machine learning’ and ‘artificial intelligence.’ I think we’re seeing the same thing now. People are saying ‘large language models,’ when it might just be machine learning.

 

Advice for healthcare newcomers

Gamble:  Very true. Now, since you left Microsoft, we’ve seen more people come into healthcare from other industries. Based on your experience, do you have any advice to offer or is there anything you wish you would’ve known?

Moore:  I have a few pieces of advice. One is that 20 percent of what we do in healthcare is unique to healthcare — you need to learn quickly what that 20 percent is. The other 80 percent is basically the same with any Fortune 100 company.

Two is, don’t make the mistake I did, which was to assume that healthcare was behind on technology because they were resistant. That wasn’t the case at all. Healthcare wasn’t getting the right technological leadership. Given the right leadership, they’re actually quick to adapt. But because of my perceptions, IT probably moved slower than it should have because I feared they would be resistant. That was a bad assumption.

“Don’t try to catch up”

The final piece of advice that I would give is that if you believe healthcare is 20 years behind other industries, don’t spend time trying to catch up on those 20 years. Take, for example, the consumerization of IT. You have Uber, Expedia, and Amazon, and everyone knows healthcare is behind them. And yes, it would be great if we caught up there, but I would take your creative energy and instead focus it on this new wave of large language models and innovate there.

Landlines & cellular technology

The analogy I like to use is that if you look at countries like India, they didn’t have good landline infrastructure. And so, when cellular came along, they didn’t think, ‘okay, how do we do cellular and landlines?’ They basically said, ‘We missed the landline revolution. We’re never going to have landlines. We’re never going to have fax machines. Let’s go right to cell technology. And India ended up with better cell infrastructure than a lot of countries. The same is happening in healthcare. If you focus all your energy on catching up on the last 20 years, you may miss this new wave of generative AI where you could maybe leapfrog other industries. This is the cellular equivalent and the last 20 years is the landmine equivalent. Focus on cellular, not landline.

Gamble:  Right. You’re going to drive yourself crazy trying to work backwards.

Moore:  Exactly. All of our budgets are shrinking. We’re in tough economic times. We’re 20 years behind. If you encumber yourself to catch up on all that while also innovating on large language models, you’re basically tying your hands. Instead, focus on this new wave of technology. That would be my advice.