The Conundrum of Structured vs. Unstructured Data

As in all complex endeavors, the push for a healthcare system underpinned by structured and interoperable electronic health record (EHR) data has turned out to be more complicated than we might have anticipated when acceleration of EHR adoption was begun about a decade ago. This does not mean that anyone was right or wrong; it just shows the inherent complexities of trying to solve the real problems that motivate data-related problems in healthcare. These healthcare problems have been well-documented over the past couple decades by the Institute of Medicine (IOM) and others, and include incomplete and unavailable records [1], medical errors [2], and suboptimal quality of care [3]. These problems are still every bit as real as they were when the IOM and others first brought them to light, but the solutions have been more challenging to find.

It is almost a holy grail of informatics that the value of EHR data stems from structured and interoperable data, which in turn allows not only better primary use for patient safety, clinical decision support, and other benefits, but also secondary use, such as quality measurement, public health surveillance, and clinical research. Yet it has been known for some time that there is a "tension" between the entry and use of structured vs. unstructured data [4].

A few months ago, I wrote a post on what are realistic goals for EHR interoperability, based on what I saw was positive prioritization by the Office of National Coordinator for Health IT (ONC) on data interoperability within the EHR. There is no question that data flowing seamlessly, and maintaining its meaning, is critical to advance the value of health IT.

That discussion, however, uncovers a challenge of major magnitude within informatics, which is how much data should be structured, and how to best deploy that data. A number of commentators I greatly respect have weighed in on this issue.

My spurring to write on this topic was motivated by Wes Rishel, formerly of Gartner. Mr. Rishel used the challenges of patient summaries to avoid against "de-motivating" interoperability [5]. In particular, he noted the challenge between two views of the interoperable patient summary, one driven by a human-generated narrative that communicates the patient's situation succinctly and other generated by a computer with the goal of transfer of data. He (and others before him, such as Dr. Peter Basch [6]) have noted that clinicians have dissatisfaction and distrust with records generated from structured data.

Other groups have weighed in on this problem as well. Last year, the American Medical Association (AMA) had put forth a succinct piece on improving EHR usability, noting that while data "liquidity" is important, it takes a back seat to the primacy of clinician usage of the EHR for improving care to be its primary motivation [7]. And just recently, the American College of Physicians advocated in a similar manner, releasing a policy paper on clinical documentation also calling for the primary needs to be focused on meeting the needs of clinicians [8].

A major challenge for informatics is how to balance the desire for structured data to add value versus providing readable and succinct documentation to enable the best patient care. Unfortunately, the two can be at odds with each other. If physicians do not like, let alone trust, the kind of structured data that enables other value for EHR data, what is the solution?

When in situations like this, I always remember the words of an elder sage of informatics, Dr. Clement McDonald of the National Library of Medicine, who has noted, Informatics is a journey, not a destination.  We may never achieve the perfect solution, but must continually strive to find the right balance of structure and interoperability. Or, to quote from the decades-old Ten Commandments of Informatics [9], penned by another elder statesman of the field, Dr. Octo Barnett, who stated, Be optimistic about the future, supportive of good work that is being done, passionate in your commitment, but always be guided by a fundamental skepticism.

Or, to be guided by a quote often attributed to Voltaire, which is that we should not let perfect be the enemy of good. It is obvious that the EHR will never, like all of medicine, be perfect. Therefore, we should strive to find the best solution that balances the value of optimally devliered care balanced with the value that structured data can bring.

References

1. Dick, RS, Steen, EB, et al., Eds. (1997). The Computer-Based Patient Record: An Essential Technology for Health Care, Revised Edition. Washington, DC, National Academies Press.
2. Kohn, LT, Corrigan, JM, et al., Eds. (2000). To Err Is Human: Building a Safer Health System. Washington, DC, National Academies Press.
3. Anonymous (2001). Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC, National Academies Press.
4. Rosenbloom, ST, Denny, JC, et al. (2011). Data from clinical notes: a perspective on the tension between structure and flexible documentation. Journal of the American Medical Informatics Association. 18: 181-186.
5. Rishel, W (2015). How to Avoid DE-Motivating Interoperability? Retired Healthcare IT Nerd, January 5, 2015. http://rishel.com/blog/2015/01/avoid-de-motivating-interoperability/.

6. Basch, P (2014). ONC’s 10-Year Roadmap Towards Interoperability Requires Changes To The Meaningful Use Program. Health Affairs Blog, November 3, 2014. http://healthaffairs.org/blog/2014/11/03/oncs-10-year-roadmap-towards-interoperability-requires-changes-to-the-meaningful-use-program/.

7. Anonymous (2014). Improving Care: Priorities to Improve Electronic Health Record Usability. Chicago, IL, American Medical Association. https://download.ama-assn.org/resources/doc/ps2/x-pub/ehr-priorities.pdf.

8. Kuhn, T, Basch, P, et al. (2015). Clinical documentation in the 21st century: executive summary of a policy position paper from the American College of Physicians. Annals of Internal Medicine. Epub ahead of print.

9. Barnett, GO (1979). The use of computers in clinical data management: the ten commandments. Society for Computer Medicine Newsletter. 4: 6-8.