Healthcare Analytics, Population Health Management, Healthcare Big Data

Analytics in Action News

EHR Workflow, Big Data Combine for Population Health Management

By Jennifer Bresnick

- Enacting an effective and meaningful population health management program, especially when that population has a high proportion of patients with complex needs, has been one of the most challenging tasks for physicians and health systems seeking to transform themselves into epicenters of coordinated, data-driven accountable care. 

Big data analytics and population health management

At Advocate Health Care, a twelve-hospital health system based in Illinois, that task is complicated even further by a target patient pool approaching 700,000 individuals and a large number of care sites operating on a mix of EHR technologies that need to be integrated in order to create a seamless care continuum. 

With unplanned hospital readmissions and chronic disease management squarely in the sights of Advocate’s growing accountable care programs, Tina Esposito, Vice President of the Center for Health Information Services, spoke to about using predictive analytics, big data, and clinical information to support a continuum of value-based, high quality care for patients across this large and challenging community.

“As we headed into our first ACO contract with Blue Cross Blue Shield of Illinois in 2011, we needed to start thinking about more broad-based population health management across the care continuum,” Esposito said in an interview. “When we were taking a look at how we were going to help support the organization as we were moving into the ACO world, readmissions became a top priority for us for a couple of different reasons.  It was a very large opportunity for Advocate at the time.  It was also part of ensuring that we were acting like an ACO and an integrated health care system.”

“To some extent, any unplanned readmission is really a failure to ensure that we are proving the best transitions of care and the best patient management.  So it became a top priority to ensure that we had the data analytics capabilities that we needed to tackle that problem,” she said.  “We really wanted to make sure that we could identify patients at highest risk for readmissions while they were still in the hospital, so that we could intervene appropriately.  Other key metrics include ED visits per thousand, patients that stay in-network, and length of stay.  These are all measures we absolutely focus on to support high quality at a lower cost.”

READ MORE: PGHD, Genomic Data Predicted to Power Clinical Decision Support

But collecting data on all of these quality measures has not been an easy task.  Like many large health systems, Advocate Health Care has providers using a number of different EHR technologies, and interoperability is a major challenge.  While Cerner Corporation’s HealtheIntent platform supports the core of the health system’s population health management efforts, big data is streaming in from half a dozen different technology vendors, Esposito said.

“On the inpatient side, the EHR that we use is Cerner, for the most part,” she explained.  “We do have a couple of facilities on MEDITECH.  They were later acquisitions for the system.  On the outpatient side, our employed physicians are on the Allscripts system, and many of our aligned physicians are using eClinicalWorks.  We have a large medical group on Epic, and then we have another installation of Allscripts for our home care division.  So it’s a pretty complex environment.”

Any healthcare organization that has tried to engage in even the most basic level of health information exchange or health data interoperability will be familiar with the struggle that Advocate has faced when attempting to gather these disparate data sources into a harmonious foundation for big data analytics.  While EHR vendors are continuing to put a great deal of effort into solving the problems of conflicting data standards and incomprehensible outputs as new generations of technology evolve, healthcare organizations that want results right now are faced with a dauntingly fragmented big data landscape.

“Probably the greatest challenge in the industry right now is how to go beyond the four walls of the organization to get to a truly longitudinal and personal level of patient care,” acknowledged Bharat Sutariya, CMO and Vice President of Population Health Leadership at Cerner Corporation.  “A person gets his or her care in a community.  Even in the most concentrated health system, a patient doesn’t necessarily get all of his own care in a one single health system, let alone into the one single venue with one EHR.”

“So when you are trying to do population health management with 700,000 patients like Advocate is doing, you end up with a scenario where patients are being treated in multiple different venues and in multiple different electronic record systems or transaction systems.”

READ MORE: Unfiltered EHR Data Overloads Physicians, Perpetuates Burnout

Healthcare organizations have relied heavily on claims data to provide a relatively standardized basis for their clinical analytics work when EHR systems refuse to cooperate, but claims data is no longer enough to support the advanced level of population health management required to meet increasingly stringent quality goals and value-based reimbursement frameworks.  Real-time data is key to these efforts, and cannot be obtained by scanning claims that may be six months to a year old by the time they reach the informatics office.

Clinical data from the EHR must be a central part of any big data analytics program, a lesson that Esposito and her department have learned well, despite the difficulties of collecting and streamlining myriad streams of information.

“We have done a lot of work on the interoperability front to ease the process of exchanging data between EHRs, but that challenge actually gets multiplied several fold when you want to talk about population health management,” said Sutariya.  “You’re not just trying to exchange data.  You’re trying to aggregate it from every source that the patient touches.   And once you aggregate that data, you have to take three additional steps.”

“The first is to match it to the right person, which requires an industrial strength EMPI that goes above and beyond what people have done in the EHR, because your data on one patient may come from fifty different sources,” he said.  “You have to be sure that you are matching those sources appropriately.”

“Secondly, you have to get the terminology and ontology right. We still don’t have a consensus on standards for almost anything.  There’s LOINC, and CPT, and ICD-10…and many different terminology systems.  How do you take all of that and map it to a consistent standard? That’s always going to be a challenge.”

READ MORE: Lack of Talent, Direction Afflict Healthcare Data Analytics Plans

And the third, and probably the most difficult task, is formulating a truly longitudinal record,” he continued. “HIEs have been successful with exchanging data, but it isn’t always easy to read and digest.  There could be five different entries from five providers that include the same medications or same labs or same diagnosis in multiple different systems. When the CCDs come over, it might show five different brand names for aspirin with slightly different dosages.  When the physician looks at that, he’s not seeing a clean list of what the patient is truly taking, and that is frustrating for him.”

Those concerns have prompted a number of stakeholder groups to collaborate on large-scale solutions, from streamlining the patient history and intake process to tackling the critical task of ensuring that patient data is matched appropriately across new and existing records.  The challenges of big data analytics are being addressed at multiple levels across the industry, and much of the work is being done by individual health systems that hope to create tailored yet replicable infrastructures that bring together clinical, claims, and demographic data into a meaningful portrait of a patient’s life.

“When it comes to performing population health management analytics, I think we’ve done a very good job integrating all that disparate clinical data with our claims data,” said Esposito.  “The clinical data is really vital to support a level of specificity that just isn’t there with using claims data alone.  It’s really the marrying of both of these that we have been after, and that’s what we have been working to support.”

Their work has started to pay off in the realm of preventable readmissions, a popular quality metric that can be extremely costly for healthcare organizations that see high numbers of patients returning to the inpatient setting within 30 days of discharge.

“We have a predictive analytics model for readmissions that allows the care manger or the clinician to review a risk score as part of their EHR workflow,” she continued. “It’s updated every two hours during the patient’s stay.  It’s not a separate application they need to go to, and it’s not information that was calculated based on claims data that is six months old.  It’s real-time, and it leverages the clinical data to really identify what that readmission risk is for the patient so their clinicians can act appropriately.”

Clinicians have historically been wary of new tools that are stuffed into EHR workflows, which may be cumbersome, convoluted, and confusing while providers are attempting to complete even the most basic tasks.  Esposito and her team are well aware of the undercurrent of dissatisfaction that follows many EHR implementations, and they have taken significant steps to ensure that clinicians understand the need for these new tools and are happy with the way they are integrated and deployed.

“We have absolutely ensured that we are collecting feedback on making the tool more useable, friendlier, and definitely much more efficient,” Esposito stated.   “Prior to our collaboration with Cerner and the creation of readmission risk tool, Advocate had its own readmission risk algorithm, but it wasn’t automated.  It actually required care mangers to manually select check boxes and tick off variables that were true of the patient so that they could compute the score. By automating that, we were able to remove those added steps from the work flow, which has certainly saved time and effort.”

“We’ve gone from a 40 percent compliance rate to somewhere near 75 percent,” she added.  “We’re continuing to collect feedback from our end users and incorporate changes to ensure that the solution and the model are being utilized appropriately.  We have two clinical process designers working on our team.  They have a clinical background, and their sole purpose is to ensure that we understand the workflow from the clinicians’ point of view.  They do a lot of work to train the staff, and they’re very hands on.”

Engaging staff members to tweak usability concerns and enhance big data tools is an important part of enacting organizational change management.  Big data analytics is often viewed not so much as a project for the HIM department but a cultural shift in the way healthcare organizations provide care, and it isn’t always easy to get boots-on-the-ground clinicians to understand that if they are not involved in the process.

“When we are talking about big data, I think there needs to be a clear purpose,” said Esposito.  “There has to be a core need or a well-defined problem that you are trying to solve.   Big data is a means to an end for solving problems.  So you got to be very clear that you are not pulling this data together just to do put it together. There has got to be a focused effort from the right people to leverage that information so that ultimately you are supporting the business and your population health goals.”

“You need to be sure that what you are creating is usable in the most efficient and easiest way, and that it makes a positive impact on clinicians. Is the clinician leveraging that intelligence that you are providing as part of their workflow in the EHR?  Are they seeing a benefit from it?  That’s going to be the most important piece of any big data project.”

In the primary care setting or in the hospital, clinicians who do not see immediate return on their investment of time and effort may have difficulty embracing the principles of big data-driven population health management, and may not view EHRs and other health IT infrastructure as particularly helpful tools in the fight to navigate rapidly changing reimbursement structures and a seismic shift in the nature of patient care.  EHR systems must act as partners and aids for clinicians, not as barriers that come between a provider and her patient.

“Ensuring that providers are going to be successful is definitely the ultimate goal,” Sutariya agreed.  “We want to improve performance on a broader scale for health systems, but also at the level of the individual clinical provider.  To do that, you have to provide context about the person or population that you are responsible for.”

“It’s no longer sufficient for you just to know what’s in your EHR.  You have to know so much more than that to engage in population health management, from so many different perspectives.  Are there psychosocial barriers?  How motivated are they?  What is going to engage this person in their healthcare decisions?  Sometimes it’s about providing information that the patient doesn’t have, and sometimes it’s about removing an obstacle for them.”

“Secondly, we have to increasingly integrate intelligence into the clinical workflow.  A transactional system is no longer sufficient. As a clinician, I can read ten different articles every single day and read one book every single week and I’ll still be hundreds of years behind on the knowledge that is getting published in the literature every single day.  That’s where the computers and the systems come into play.   They have to integrate this knowledge and make it available to the clinician, because it just isn’t possible for humans to keep up.”

“And then lastly, taking a data-driven approach is key,” he concluded.  “Less than 25 percent of all decisions made in medicine are based on published evidence, because many, many times there just isn’t any evidence out there.  Big data analytics is changing that.  Now, we can look at what happens when people make tens of thousands of decisions, here are the outcomes from those decisions, and here is how those decisions may affect this similar type of patient.  That is definitely the power of big data.”

Healthcare organizations simply cannot engage in population health management without big data on their side and under their control, Esposito added.  “It’s so critical to have the ability to piece together any and all data about the patient,” she said. “In some ways, you even want to move beyond clinical and claims data to environmental data or other sources of information that really help you better describe the patient and her challenges.  Integrating multiple sources of data will help a clinician understand how best to intervene with that patient – and we are now entering the kind of care environment that makes providers responsible for that kind of population health management.”

“So this notion that somehow big data analytics strategies are simply an extension to your EHR strategy is a false one.  It needs to be so much broader than that, and understanding that is very necessary for realizing success with population health management on a large scale.”


Join 25,000 of your peers

Register for free to get access to all our articles, webcasts, white papers and exclusive interviews.

Our privacy policy

no, thanks

Continue to site...