Q&A with Dr. David Kaelber, Part 2: “Don’t Come to Us with a Solution.”

When David Kaelber took on the role of MetroHealth System’s first CMIO, he believed that if he could identify one or two issues and solve them, he could help transform care across the organization. What he learned, however, is that healthcare is far more complex than that. “There is no one thing that’s going to change healthcare,” he said during an interview with Kate Gamble, Managing Editor at healthsystemCIO.com. However, “if you do a whole bunch of things well and intelligently, you’re going to move the needle in a way that was impossible to do without health IT.”

At MetroHealth, his team has led several initiatives to help improve care delivery, care quality, and clinician efficiency, whether it’s through medication adherence flags, electronic support for adverse event reporting, or sepsis prediction tools. It’s all part of the overarching goal of getting technology to work for the healthcare system, said Kaelber, who strongly believes “IT’s value is in being a strategic partner.”

During the discussion, he talked about his evolving strategy as CMIO; the critical role social determinants can play in shaping the health of the community — if leveraged properly; the enormous value he gleans from having regular meetings with stakeholders and listening to their concerns; and why imitation is a good thing, particularly now.

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

Podcast: Play in new window | Download (Duration: 20:50 — 11.2MB)

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

Key Takeaways

• When it comes to understanding population health, collecting data on social determinants is critical, said Dr. Kaelber. “You don’t want to make guesses; you want to make decisions based on the data.”
• Once social determinants have been established, the challenge for providers is connecting patients to social service agencies. One solution? Startup exchanges such as UniteUs and Aunt Bertha.
• For IT and clinical leaders, the goal isn’t to find the problems, but rather, to identify tools that can help solve problems that have already been defined by health system.
• “As the technology leader, have to take a very different psychological approach; even the way you interact with peers, VPs, or those above you, has to be different.”
• Although predictive analytics can be extremely useful, they’re often complicated to implement for one simple reason: “the rules are based on other peoples’ data.”

Q&A with CMIO David Kaelber, Part 2 [Click here to view Part 1]

Gamble:  You mentioned earlier social determinants; can you just talk a little bit about some of what MetroHealth is doing there?

Kaelber:  First, you have to have a good way to assess social determinants of health factors. It’s like the saying, ‘no data, no mission.’ Everybody thinks social determinants of health are really important, and yet, when you ask people what social determinants of health look like in their population, most have no idea.

MetroHealth’s SDOH strategy

The first step was to systematically come up a tool and a process to collect the information. Epic has what they call the social determinants of health Wheel with about a dozen different domains, including education, transportation, substance use, food insecurity, and social isolation. Each one only has one or two questions; the key is, if you already having a dozen different domains, that’s already a few dozen questions. It’s just supposed to be screening, so we built the tool as the foundation.

Now the question becomes, how do we then populate the tool? If you go to a doctor and say, ‘We’ve got this great new tool. All you have to do is ask a couple dozen more questions during your visit,’ — that’s going to go over like a lead balloon, because it’s not like doctors are just sitting around waiting to ask more questions during a visit. And most of the time, patients don’t come in asking for help with their social determinants of health issues. They ask for help with different issues. But now how do we collect the SDoH information?

“A complete picture”

We took an approach where we leverage that personal health record. As part of our strategy, we want everybody — both pediatric and adult patients — to be seen routinely, and to be screened for social determinants of health. About two-thirds of screening is done through the personal health record. Anyone can fill it out at a visit; in that case, it’s probably less for screening and more for cause. Whenever someone is referred to a social worker or for clinical case management, they fill it out so that the social worker or case manager has a complete picture of the social determinants, and can use that information to determine next steps. This way, we have a tool and a robust infrastructure that doesn’t rely solely on the physician or the provider to collect information.

Data-based decisions

The next question is, once you collect all the information, what are you going to do with it? There are different flavors to that. One, we want to analyze the data to figure out where the correlations and opportunities exist. Within the MetroHealth population, the number one social determinant of health risk factor reported by patients last year was social isolation. I’ll be honest; I thought the answer was financial insecurity. But when you think about what we’ve been through with the pandemic, it’s not surprising that social isolation was the top concern.

That’s the perfect reason why we need to collect data. If you would’ve asked the leaders of MetroHealth to figure out what we should do for social determinants, we literally would have chosen the wrong thing, because we didn’t have any data. We were just using our intuition. But you don’t want to make guesses; you want to make decisions based on the data.

Once we have that, it opens up more possibilities. Obviously, we want to develop programs for all social determinants, but now that we know social isolation is number one in our population, we can focus more on that.

Connecting patients with social services

The other big technology infrastructure piece is when we look at social determinants of health, of all the different topics, which one is the healthcare system really situated well to deal with? Is it food insecurity? Is it domestic violence? Is it housing insecurity? The healthcare system isn’t really well designed to deal with any of these — so who is?

We do have social service agencies and community-based organizations, including the Cleveland Food Bank, Habitat for Humanity, and women’s shelters, to name a few. Their entire purpose is to address the social determinants of health issues. And so, the question becomes, how do we put the peanut butter and the chocolate together?

On the healthcare system side, how do we connect our patients with documented needs to social service agencies?

The answer on the technology side is a health information exchange. There are a few companies popping up in this space: UniteUs, NowPow, and Aunt Bertha. We actually implemented UniteUs about a year ago. Now, in the same way that I might electronically refer a patient with bad asthma to a pulmonologist, we can do that for social services agencies or community-based organizations if those needs have been identified.

 

Bringing the right tools to the table

Gamble:  It’s amazing what you can learn, but the process of identifying the problem and addressing it really shows how important it is to have integration, or at least collaboration among different entities that weren’t always connected with the health system. It’s interesting to see how that has evolved.

Kaelber:  The other thing I want to frame is that I didn’t come up with the idea that, ‘social determinants are really important. Let’s figure out the wheel and let’s figure out the health information exchange for community-based organizations.’ It was our non-IT leadership, the CEO, the board, and others who said, ‘Our healthcare system is a public safety net essential healthcare system; you need to be doing much more in this space,’ and then it became an institutional priority. As soon as I see that, my technology brain starts to ask, what tools I can bring to the table that are really going to help move the ball forward for our healthcare system?

“It doesn’t start with technology”

And of course, other leaders throughout the system bring forth their non-technology tools and we’re able to make great things happen. My point is that it doesn’t start with technology people finding the problem; it starts with the system defining the problem. And then we, as technology leaders, bring our toolbelts to the meeting to figure out what solutions will help from a technology perspective.

 

The technology dyad

Gamble:  That speaks to what you mentioned earlier about the goal of getting IT to work for the health system.

Kaelber:  My CIO and I call ourselves the technology dyad. Obviously, we have our own priorities like a new data center, server upgrades, or a technology refresh. But that’s not really being strategic. Yes, we need those things, and we’re going to talk to our leadership to secure funding for them, but that’s not helping to move the healthcare system forward. That’s keeping the lights on.

One of the values our bosses would say we bring is that when they frame an opportunity or a priority for the healthcare system, we come in as a strategic partner to identify technology that can be a force multiplier or an enabler of that healthcare system priority.

 

CMIO evolution

Gamble:  Right. Now, having been in the CMIO role since 2008, has your approach to the role evolved since that time?

Kaelber:  Yes. I was actually our first chief medical informatics officer. And for the first few years, I wasn’t exactly sure what my role should be. So, if someone told me what to do, I would make it happen — which hopefully every employee does. If your boss tells you to do something that’s within your scope, you should be able to deliver on that.

But in terms of my evolution, there are always going to be things my boss tells me to do, and of course I have to do those things. The problem is that my bosses aren’t going to be technology experts. If I wait for them to tell me what to do with technology, that’s really not going to help the system, because I know more about technology than they do. And so, it’s much better for me to be proactive and reach out to find the problems we’re trying to fix, and then craft solutions knowing what technology can do.

Meeting with key stakeholders

One of the things I routinely do — which I didn’t do in the beginning — is to routinely meet with key stakeholders. I meet with our population health innovation institute. I meet with our Institute for Quality. I meet with our patient experience people. And I’m basically saying, ‘what are the problems you’re trying to solve? What are your goals for the next year? And of course, some of the goals will be outside of my realm, but sometimes I’ll say, ‘Thank you so much for telling me that.  Did you know that I and my team could offer you these things to help with your goals?’ It’s changing from being more reactive from a technology perspective to being more proactive.

Taking “a psychological approach”

But that means you, as the technology leader, have to take a very different psychological approach; even the way you interact with peers, VPs, or those above you, has to be different. You have to be out there understanding what their problems are instead of waiting for them to come to you.

“Come to us with a problem”

And the other thing — and my team knows this — is when a person comes to me to say, ‘I need your help building a letter in Epic to do X, Y, Z,’ I always say to them, ‘Please don’t come to us with a solution. If you come to us with a solution, you’re not allowing us to use our expertise. Come to us with a problem. We’re the technology experts; let us think about all the possible technological ways in which we might be able to solve your problem. And we might end up with the same solution that you recommended, but a lot of times we’re going to have a much broader understanding of what’s possible from the technology standpoint, and so we will probably come up with a better solution.

 

Epic sepsis study

Gamble:  That’s an interesting way to look at it. I know we’re short on time, but let’s talk about the sepsis study and the work your team has done there.

Kaelber:  Sure. And it’s not just about sepsis specifically, but rather, how we can use things like cognitive computing, AI, predictive analytics — whatever you want to call it. The basic concept is, when you have a whole bunch of data, how can you forecast what might happen to a patient in such a way that enables you to intervene earlier than you otherwise might have, and therefore change that predicted outcome?

“A system opportunity”

Epic has a sepsis model which is a good example of that. At MetroHealth, we recognized that it seemed like we had more people having morbidity and mortality from sepsis than we wanted. That has nothing to do with technology; that’s a system opportunity. And so, my team came forward in a meeting and said, ‘If we want to improve sepsis, Epic has a prediction tool. Let’s implement it.’

The challenge is that predictive tools are very complicated to implement because the prediction rule is based on other people’s data. And so, we implemented it in our system and did a test. We do this a lot when we have different types of alerts; we turn it on the background so we can determine where the tool is going to work well and where it’s not.

Case study: ICU

We looked at three different situations. First, we looked at the ICU. It worked great; the only problem is that all of the things we’d do for someone who we thought was going to have sepsis were already being done. Patients in the ICU are sick, and so, they’re already on fluids and IV antibiotics. And so, even though the tool worked great, there was no opportunity to intervene. Therefore, it wasn’t a good use case.

Case study: Med-surg

Then we looked at the regular medical surgical floor. What happened, though, was that it gave a bunch of false positives. And so, we decided it doesn’t make sense to use it there because we’d be firing all these alarms and doing extra stuff just to help a very small number of people.

Case study: ED

The third place we tried was the ED. The ED seemed to have pretty good sensitivity and specificity, so there weren’t too many false positives or false negatives. And because patients in the ED are early in their stay, there were things we could do, like trying to start IV antibiotics earlier. Based on the ability to turn it on silently, we thought it was a good place to do it.

We did a randomized quality improvement initiative where we turned it on for half the patients coming to the ED. The intervention was to alert the pharmacist, who would then evaluate the patient. If they agreed with the prediction, they’d start on antibiotics. What we found was that antibiotics were started several hours earlier than in the control group.

If you look at our intervention, the main outcome we looked at were days alive and out of the hospital during a 30-day period. It was statistically significant — we did decrease days alive and out of the hospital. But if you look at the details, the average days alive and out of the hospital was about 20 days, which means this intervention only decreased that by less than one day.

It goes back to my earlier point that if you’re one of those patients who just spent one less day in the hospital, or if you’re a payer and you now have to pay for one fewer day, that’s great. But it’s not like we went from 20 days to two days. If you look at the percentage difference, we only improved the number by about five percent.

It’s a nice bookend to the study that came out of University of Michigan, which found that the sepsis model didn’t work that well. Their findings were true and so were our findings. The difference is that Both their findings and our findings are true; the difference is that they didn’t try to customize their findings to the healthcare system, whereas we did.

 

Gamble:  That’s an important distinction.

Kaelber:  It is important. Part of the challenge is you have to understand your local problems, your local patients, and your local processes. You have to understand what the technology can do and you have to map the technology to, again, the local priorities, local operations, and local patients. If you don’t, that’s where you’re likely to run into problems. The sepsis studies are just one example of that.

Fast follower

One strategy I always try to emulate is the fast follower strategy. The idea is that if someone else has already found something that works really well, there’s no point in reinventing the wheel. Study what they did — it doesn’t mean you should take it lock, stock and barrel, but study it. And then, if needed, adapt it to your healthcare system. That’s easier to do than starting from scratch. I was once on a call with Judy Faulkner, CEO of Epic, and she said something like, ‘All of these healthcare systems keep talking to me about chief innovation officers. I think people should have chief imitation officers more than they should have chief innovation officers.’

 

Gamble:  I like that. Like you said, you don’t have to reinvent the wheel, but at the same time, you don’t have to do it exactly the same. There’s a lot of value in that.

Kaelber:  To me, sepsis is just one example of a whole strategy we’ve taken on during the past 12 years. It just so happens that the sepsis example is one of the more recent ones. It’s one of many examples.