Finding and Compiling Salient Population Health Data
Wouldn’t it be nice, when talking about population health, if there was only one payment contractor? In the real world, there are many: state-mandated programs, Medicaid, Medicare, commercial. ACOs are likely contracting with all of these. Whether it’s value-based, shared-risk, government, or commercial contracting, data is needed from many different sources in order to truly assess population risk and effectively manage their health.
The Progression of Population Health Data Needed
To effectively manage population health, ACOs need external data from multiple sources in addition to what they get from their EMRs. Most organizations agree that data is required from claims, as well as commercial and government sources. In the April 2016 issue of Inside Edge from Scottsdale Institute, John Glaser, the senior VP for population health at Cerner, said that “there’s a bunch of data from many, many different sources. Layered on the top should be a single view from which to construct patient registries, compare cost data and logic and analytics for reporting purposes and direction-setting of the organization.”
The Ups and Downs of the Primary External Data Source
The EMR is also a primary source of external data for population health management. In order to do process improvement, we need more than patient allergy and medication data. We need environmental data. What processes are being measured? Which nurse documented those processes? How long did it take? How many patients are being seen in a day? What are the flow rates through a given practice, exam room, or surgical room? Clinical process improvement, and real population health management, demands access to this type of data.
However, talk to the IT department of any large integrated delivery network (IDN), and you’ll likely hear that they have a single EMR system across their entire organization. If this were true, population health would be much easier. Talk to the leaders of the ACO run by that same IDN, and you’ll likely learn that their value-based, shared-risk contracts always involve health providers who are outside of the single EMR provided by the IDN IT department.
Often, the ACO works with a Clinically Integrated Network (CIN), Independent Physician Association (IPA), or some other external organization to provide the full continuum of care services required for value-based contracting. The physicians in the CIN or IPA have only a portion of their patients within the ACO and may even compete with the IDN in fee-for-service opportunities. Thus they are generally not willing to turn over all their data to the ACO. And if the ACO isn’t involved in the care or administration of some patients, and the CIN or IPA shared the data on those patients, then it would be in violation of HIPAA.
With multiple external organizations, the number of EMR systems suddenly jumps from one to 20 or 30. This represents a big challenge from the perspective of analytics in the management of population health. Assuming the data from all these EMRs is necessary, it’s an awful lot of places to go and get it.
Additionally, a single provider may only have five or 10 percent of his patients within the ACO, and he may be contracting with two or three value-based organizations. Another doctor may have fee-for-service patients in her EMR and, in some cases, may be competing with the IDN that owns the ACO. This part of the problem won’t change anytime soon even as networks begin to narrow.
Claims Data Better than No Data
Let’s consider claims data as data at the speed of crawling. For many years prior to value-based contracting, clinical people turned up their noses at using claims data because the diagnosis or procedure coding wasn’t right. Supposedly, claims data was optimized for billing purposes. Regardless, when it comes to a patient who’s visited a provider outside of the ACO or network, claims data is better than no data at all. At least it provides a record of the encounter and some information about the diagnosis.
Limited Health Information Exchange Data?
When taking data out of the EMR is problematic, a lot of people think a health information exchange (HIE) is the place to go. Many HIEs (or Health Information Organizations (HIOs), the operating entity of an HIE) have the capability to route ACO patient data to the ACO (or health plan) based on a membership list maintained by the ACO and periodically updated to the HIO.
HIE data comes out of the EMR, so it’s clinically more relevant than claims data. Though maybe not relevant enough to do process improvement across the ACO, HIE data is a good for helping with care coordination and care management activities. It’s also relevant for process improvement around health and wellness measures. For example, are patients with diabetes getting their annual hemoglobin A1C tests? Is the cohort of cardiology patients being screened for cholesterol? While this may not be enough data to determine if heart failure readmissions need to be reduced or if COPD patients require better post-acute care, it’s a step in the right direction while the ACO works out how to get a deeper set of data from the EMRs.
The limitation presented by HIE data is that it’s defined by national standards developed from the Continuity of Care Document (CCD), the template that allows healthcare providers to transfer patient data. The problem with the CCD is that it was designed for one care provider to communicate with another care provider about patient information, like medications, allergies, and lab values, for the purposes of transferring that patient from one facility to another. This is about 80 to 90 percent of what is needed, but it’s not adequate for meeting population health requirements.
Fast Healthcare Interoperability Resources: The Data Picker
Getting deeper and more specific data out of the EMR is the promise of Fast Healthcare Interoperability Resources (FHIR) being created by the HL7 organization. This standard is easily implemented and it holds a lot of promise, though it doesn’t yet define the data needed for clinical process improvement, nor is it sufficiently implemented by actual EMR systems. It’s still a research project at this point.
Other External Sources of Healthcare Data
Healthcare expenditures in the U.S. follow the Pareto Principle in that the top one percent of the population accounts for 20 percent of healthcare costs; the top five percent accounts for 50 percent of costs; and the top 20 percent accounts for almost 82 percent of healthcare expenditures. So the question on the mind of anyone responsible for managing population health has to be, “what’s going on with the other 80 to 99 percent of my population?” And 50 percent of the population rarely visits a doctor. They aren’t even classified as patients, so how does an ACO contract intelligently when it’s serving a population that it knows very little about?
This is why demographic data, available from national healthcare research organizations such as Nielsen, AHRQ, the Centers for Disease Control, and IMS Institute for Healthcare Informatics, is a necessary external source.
Demographic data may be blinded and not a 1-to-1 match to the contracted membership, but it will provide an idea of what’s going on within a defined geography. For example, Nielsen data can show what pharmacy a population favors. If a healthcare system sees that 80 percent of the contracted population in a given area is going to CVS, it might foster a relationship with that pharmacy to improve patients’ ability to get their medications.
Zip codes reveal economic class, which, in turn, reveals healthcare utilization. How many individuals have a smartphone and is that a way to interact with them? Where do people go for healthcare? Minute clinics are popping up everywhere. How much healthcare is being provided at these facilities, which, by the way, are not responsible for the overall health of the patient? Are patients going there because the facilities within the ACO aren’t wide enough or because it takes two days to get an appointment? This is all valuable external data.
I was talking with a healthcare system that was having trouble getting patients to show up for their post hospitalization follow-up visits. This was leading to high readmission rates. The system plotted out its population addresses on a map and then overlaid it with the city bus routes. It found that bus routes avoided a certain part of town, so the organization talked the city into changing some of the routes so patients could catch the bus to go see their physicians. Subsequently, follow-up visits increased by 30 percent, and this was accompanied by a drop in their readmissions. This is a great example of using external data to manage population health.
The Data Challenges Ahead
There are challenges ahead for integrating external EMR data, along with other external data sources, to get the full picture of the contracted population. What is the solution? In the short term, HIOs with large geographic bases will leverage the movement toward population health contracting mechanisms. In the long-term, FHIR will become the standard. In the meantime, it’s important to just recognize that the problem exists. An alternative to fix this challenge might be employing more of the independent physicians and allied professionals to gain access to the EMRs of their patients, which leads to better control, not only from a data perspective, but more so from a clinical process perspective. But the bottom line is that data from multiple external sources is vital to clinical process improvement and effectively managing population health.