Health care data isn’t just for analytics officers. It affects patients and hospital staff, too.
Medical providers across the country are investigating how to get doctors the information they need to best treat a patient. This type of data-driven care plan can keep the patient healthy so they don’t require crisis care. It also has the power to predict future health care problems.
In Idaho, St. Luke’s Health System is hard at work developing new analytics capacities.
The Idaho Business Out Loud podcast recently sat down with Dr. Neeraj Soni, chief medical informatics officer, and Onur Torusoglu, chief digital and analytics officer, of St. Luke’s Health System to discuss how these capabilities can be used to improve the quality of care for patients and the efficiency of how that care is provided.
This interview has been edited for length and clarity.
What was your previous system like, and why did you decide to pursue a different way of gathering analytics?
Soni: We’ve been on different kinds of electronic health records for a while, and there are certain parts of the organization that just moved to an electronic record three years ago when we implemented Epic as a system-wide solution. But in pockets, we had electronic health records dating back 20 plus years. The problem was that the care that was provided in those areas, maybe one emergency department or one labor and delivery unit really, couldn’t be seen by anybody else because the people caring for the patient in that area could only see that electronic record. It’s really hard to even have all of the information, much less know how to use the information if that’s the model you’re in.
Three years ago, we completed a multiyear project to bring all of our hospitals, all of our clinics, all of the care settings together into one common electronic health record. And so now all of us could at least see what’s happening to the patient in different settings. And that’s really the first step towards having all of the data to analyze. And that’s where Onur’s team comes in.
So these changes that you’re making in your analytics capabilities, what are they going to bring to the table? What specific changes are you making?
Torusoglu: One of the challenges in most organizations is knowing which data set to use where. We have common denominator data fields, but they will serve totally different purposes. For example, I’m a clinician and our EMR system may be serving one purpose or one physician field serving one purpose, but in the financial terms that may be completely different. So if those interpretation points don’t exist in a data environment, you may be utilizing the wrong data sets to try to make the right decisions for your organization. So this foundational layer of data management skills allows for aligning what data and which fields to be used in what setting. Who should be using it and what decisions should be made with that is step number one.
The second step to that comes from what we call the business intelligence field, which is the ability to take that filtered, cleaned and managed data and make it available in the right operational setting. For example, when Dr. Soni’s in a care setting and is taking care of patients actively, I don’t anticipate him going outside of Epic and logging into two, three other different systems to try to get to information in managing the care for a given patient. All that information should be at the fingertips of that physician.
And then the next layer to this is the advanced analytic capabilities. So going from hindsight of what happened in the past to what can happen – how do we predict, how do we allow people to prepare and plan for it? A great example here is the sepsis model that we’re actively working on deploying in our organization. The premise of our sepsis predictive model is we will give the ability to a physician to see that a patient they may have seen in recent history, as part of their rounding or whatever the case may be, is deteriorating.
Whether they’re going to do their rounding and see this patient in the next few hours or not can determine this patient becoming more adversely sick versus not. Building a sepsis model or building a predictive model allows for determining through predetermined variables. This patient’s about to get worse, so highlight the attending physician or the floor nurse or a specific team to be able to say, “Hey, this patient in the next couple of hours can turn septic. Let’s pay extra attention.”
If we can save the physician that effort and save a patient’s life by alerting the physician ahead of time, that’s a great example of being able to do things ahead of the curve before it actually take a turn for the negative. So that’s a good example of the three layers that I talked about: foundational to middle and now what we brought to the table with the advanced analytic skills.
It seems to affect every single level of business from the way the hospitals run to the care for the patients. Something that I would love to share with our readers and listeners is not just what a doctor at St. Luke’s can experience from these changes, but how this is going to affect the average patient?
Soni: The goal is to work with patients and work with the people caring for them to figure out how best to bring this data to that care setting in the right way. In the case of a sepsis, for example, it’s not enough to say that the patient might become septic. It’s also not good enough to just say when we’re sure the patient is going to become septic because, by the time we’re sure, doctors and nurses have already figured that out too. So what we’re working on now is presenting information to patients and to a physician and care team, asking them whether this looks like it’s useful information and then getting their input into how to build tools around it.
It’s hard enough and controversial enough to know what we should do for a patient who we all agree is septic. It’s definitely unknown what to do about a patient who might become something. And you definitely don’t want to do the same things for all of those patients. That wouldn’t reduce cost and it wouldn’t improve care. So using the tools that we have, the predictive tools to alert a team of care providers that something different is going on with the patient – that’s where we’re starting.
The point of this exercise is to help the care team do the things that they already want to do better for patients. There is a variety of ways in which predictive analytics and descriptive analytics can help. So descriptive analytics is just being sure that you understand what it is that you did and what it is that you are doing. The predictive analytics start to get into the realm of what is it that you should be doing. And what we should be doing is a lot harder to answer than what we are doing.
But patients stand to benefit a lot from this. So instead of having a set treatment tied to any patient with a certain disease, we can now start thinking about the kinds of treatments that might be most effective, not for any patient with that disease, but this particular patient, taking into account their personal history, their current medication list, the social issues that affect their health, rather than trying a one-size-fits-all treatment for pneumonia or sepsis or diabetes or hip fractures.
That’s the part, I think, that is most interesting for patients.
It sounds like it’s not just saving time and money but also saving lives.
Soni: If it’s done right, that’s what should happen. What we don’t want to do at St. Luke’s is start treating these analytic tools as just another shiny thing to put in front of doctors and patients. What we want to do is be really thoughtful and always be asking the question about how this information or this prediction actually help our care team do their job better, not just check off a box that we’ve turned on a new tool.
What are some of the challenges associated with gathering and implementing this data?
Torusoglu: The first step of our transformation was bringing the right technology and skills to the organization, which helps avoid or mitigate many of the risks that are associated with having to be able to get the right information or get it in a timely manner. One of the most common challenges of getting this information in a timely manner is it still is a minimum of a 24-hour cycle to be able to run that information through the system to make it available from a reporting or dashboarding perspective. One of our future state capabilities and thought processes is that real time ability to apply some of these advanced technologies to it so that the information is even more quickly and readily available to clinicians as well as non-clinicians.
So the focus of these is not just efficiency but also care. Can you tell us a little bit about how this affects preventative care, not just crisis care?
Soni: It used to be that there were two kinds of people who cared about this information and getting these reports: people whose responsibility it was to run the hospital or a department efficiently and people who are just trying to take care of the patients. And so those who are responsible for efficiency might care about how many patients are coming in, how quickly are they paying their bills, those kinds of things, and how well is our staffing matched to them. And the people who are trying to take care of patients might care about getting results quickly, getting orders in correctly, getting medications up from pharmacy quickly.
What we were really missing is the people who care about what’s happening to this patient but who might not be in a position to affect it right now. They need to start planning for the future. So as a group, this includes case managers and discharge planners. When the patient’s in the hospital, it includes care coordinators and what are called nurse navigators when the patient is at home in between episodes of care. Now we’re able to start thinking about the kind of data and information those groups of people could use and the way that they could get the data so that they can intervene.
If a patient is in the hospital and we have a model that says that they’re at risk of being readmitted, that’s important information. Right now they’re in the hospital, they’re getting care. What we’d really want is for them to get better, go home and stay at home and not get sick again. And it’s these models that can help tell us who’s at risk for needing to come back. And that by itself isn’t enough either. We need to know why the patient might be at risk.
So there’s a big difference in what we would do as a hospital if we knew that this patient might come back because they don’t believe that the plan that they’re getting is good. And that’s different than the patient who doesn’t understand the plan. And that’s different than the patient who can’t hear the plan. And that’s different from the patient who can’t afford the plan. But knowing which of those issues this patient has is going to be important for how we intervene.
Torusoglu: This is going into the realm of population health, as well as value-based care management. Where analytics comes into the picture from that standpoint is, with the advances in being able to share more data and having access to more data, we can go above and beyond just our actual patient population to available information that can be millions of patients’ records that are in our global system. It’s a de-identified way that allows for us to access those sorts of data sets. So if we had X number of patients, which is limited to our physical location, it creates a small sample size versus we can open up that population to the entire country and get information on the same or common type of patients and the treatments and what’s been most successful. That’s part of that picture of making it available to the case managers or the hospitalists or the overall care team. It allows for them to make better, faster decisions and being able to see what next step is available for that patient to keep them healthy versus treating them when they’re sick. So that’s where the analytics will come into play.