Dimensions of Big Data for Healthcare

November 13, 2016  |  EHR mHealth

Now that the healthcare system is starting to look at “patients as consumers” and consider how to deliver and document quality care in which patients enjoy the experience, what changes in terms of data requirements and strategy? The short answer is lots of things, most notably more data elements on patients collected on a continuous basis.

EHRs were built to document delivered care and they do a good job of that. There have been a lot of critiques of the UX and design of EHRs but documenting care for billing in today’s healthcare system is extremely cumbersome. EHRs are the main tool, and do it well. Today, documented care is based on episodic care—ambulatory appointments, notes from acute stays, labs, etc. EHRs contain terabytes of episodic data plus the corresponding metadata about the episodes like charges, payments, schedules, and more. Thousands of tables worth of data in fact, most of which is not clinically relevant.

EHR documentation becomes relevant in volume-based systems where payment stems from patient visits, but becomes problematic with the shift to quality and outcome-based payment models. It certainly doesn’t have anything to do with what is best for the patient since, by design, EHRs were built and are optimized to document and bill for the visit and not to directly take into account the outcome, or experience, or patient financial impact.

I find this image from Epic’s website ironic. The notion of the patient at the heart, while accurate in the sense of the patient being at the center of the technology and different systems that Epic sells, is fundamentally misaligned with the patient, heart or otherwise. It’s not the fault of Epic. They solved a major industry problem by addressing the government-incentivized Meaningful Use era. But nothing about the EHR is about the patient or hearts of patients.

Now that the healthcare system is starting to look at “patients as consumers” and consider how to deliver and document quality care in which patients enjoy the experience, what changes in terms of data requirements and strategy? The short answer is lots of things, most notably more data elements on patients collected on a continuous basis.

  1. Patients need access to EHR data. Health systems have to provide patients access to their own data. This isn’t a major hurdle to overcome and standards like Blue Button Plus and CCD exist and meet this requirement. Access to data is really just an enabler of what should be tools that help patients use their own data to inform their decisions. Access to data is a precursor to the real ROI-driving tools.
  2. Health systems need to collect data from patients. Health systems, not the EHRs, need to be able to collect data from patients. This can be data on symptoms or adherence or information about patient satisfaction with health services rendered. Patient-sourced data is emerging as a key part of several new payment programs and is referred to broadly as patient reported outcomes (PRO).
  3. Health systems need data from digital health partners. For the first time at scale, payment for healthcare delivery is not about delivery and is about the patient. To optimize care around the patient, delivery organizations need to partner with digital health vendors that engage and interact with patients, and leverage data from those partners to make informed and proactive decisions about care.
  4. Health systems need data from health system “partners”. As health systems assume risk for episodes (30-90 day bundles are good immediate term examples) that span more than in person encounters, they need to be able to track patients outside of their own facilities. To accomplish this, health systems will need to have near real time access to data, or at the very least alerting, from other entities like pharmacies (for primary and secondary adherence), skilled nursing facilities, and other delivery organizations and community anchors where patients may present during certain periods of time in which systems have assumed the accountability for care for patients.
  5. Health systems need activity data. This is a specific form of PRO and “partner” data. Since writing my post 3 years ago we’ve seen the growth of companies like Validic to make it easier to collect activity and sensor data from a multitude of 3rd party apps and sensors. We’re really just scratching the surface here. The biggest challenge is not the technology but the integration of the data into clinical workflow.
  6. Health systems need socioeconomic data. There are lots of studies that have shown the correlation between socioeconomic data and health costs and outcomes. This is another source of data that needs to matched to patients to inform decisions by healthcare organizations.
  7. Health systems need genetics data. Genetics is the path to personalized medicine, though this is further off than the more immediate initiatives around quality and value-based care.
  8. Health systems need to know patient preferences. Healthcare is increasingly about the experience of patients. In order to delivery a positive experience, healthcare organizations need to understand the preferences of patients and need to be able to market to them. The aim of marketing in this context, mainly around quality care, is not to sell products but to influence decisions related to health and wellness.

If you look at the change that is required in terms of data strategy for healthcare delivery, you can see it is paradigm shift from today. This entirely new data strategy is predicated on gathering and leveraging big data on patients. Big Data in this context means the maximum number of dimensions, or data elements, on a patient, not 4,000-6,000 tables of EHR data in which only about 10% are relevant for clinical care and outcomes. This paradigm shift is driven with patients at the heart and quality outcomes as the goal.

Those who believe this is healthcare’s 5-10-year trajectory then ask the question “How do you start to work towards this strategy?” The industry currently thinks EHRs will be the hub of data. I understand why this is the case as health systems, and the federal government, have spent 100s of billions of dollars to implement EHRs and convince clinicians to adopt them in a meaningful way. The problem with this thinking is that it requires health systems to essentially bolt on a new, and extremely complex, data strategy, along with predictive analytics, to EHR - products built for an entirely different purpose and written with underlying technologies that were developed more than 40 years ago. This, to me, seems short sighted and reactive to what are now sunk costs for EHRs.

Episodes of care aren’t going away with bundled payments and accountable care, and neither should EHRs. They should continue to do the myriad of things they do today. The massive investment that has been made in EHRs, and the set of data EHRs were built to collect and store, should be leveraged as a part of this new quality care data strategy. We have a new, widely-hyped standard called FHIR that can feed clinically relevant EHR data from encounters so why not use it to do just that. It is being pushed by the major EHR companies as a way to feed EHR data to different 3rd party applications but could work equally well to feed EHR data into a new digital health hub.

Why are we shoehorning everything into an already clunky system purpose built for a different paradigm of care? Sunk costs aren’t a good reason. EHRs do EHR, why would you buy it to do pop health or data warehousing or predictive analytics? They aren’t the same things.