Recently, Dr. Jim Thompson wrote about the critical need for electronic health record (EHR) technology that works behind the scenes to match a provider’s clinical intent to the correct terminology. (Read Part 1 and Part 2 of Dr. Thompson’s blog). The heart of the issue is that the practice of medicine requires many languages: from the natural-speech style of the clinician’s differential diagnosis to the rigid computer-speak nature of coding systems such as ICD-10-CM, CPT®, SNOMED® and others. Without an effective translator at this medical Tower of Babel, language and nuance are lost and providers struggle with understanding the patient’s story based on the information in the chart.
This translator between clinical intent and administrative coding is a specific type of technology that works as an interface bridging providers and codes. It is this interface, or translator, that powers the Clinical Interface Terminology (CIT) that drives IMO’s software. And it is this type of translation technology that healthcare systems need in order to address issues like the ones raised in Dr. Thompson’s posts.
Without an effective translator at this medical Tower of Babel, language and nuance are lost and providers struggle with understanding the patient’s story.
The discussion around the importance of CIT at the initial point of documentation is a critical one. However, there is a second but perhaps lesser known difficulty involving the inefficient transmission of health data on a macro scale. Poor transmission of clinical intent can result in the loss of fidelity in the downstream use of this data, which may have further negative impacts on both provider and patient health.
The systems at work
While the promises offered by machine learning, population health, analytics and care management platforms are numerous, and can be quite compelling, all of these tools require a substrate of good data on which to operate. Developers, purchasers, and implementers of health IT tend to overlook or even misunderstand the effort required to clean and semantically normalize data prior to using these tools. This creates subsequent disappointment and frustration when “garbage in” leads to “garbage out.” Which leads me to an often overlooked problem: even if data is captured with full clinical intent — meaning physicians are able to document exactly what they want to say — it is possible for that intent to be lost when transmitting data to other information systems (like another EHR).
And, if this wasn’t complicated enough, there’s another language translation that must occur during transmission. Health Level-7 (HL7) is an international organization that puts forth a standard for how to transmit data between health IT systems. Governments frequently create standards aligned with HL7 that dictate how providers should share data between systems in their country. Yet these messaging standards, C-CDA and FHIR for example, can also contribute to degradation of the data’s semantic content if they are implemented improperly or without enough attention to detail. This often occurs when we are forced to reduce complex health data into a single, primary code, as required by these messaging standards. The recipients of the data get the clinical information in its coded and transferred form. Therefore, we end up working with inaccurate or incomplete information even though the clinical message was clearly articulated during the initial documentation.
The ‘heart’ of the issue
In a recent analysis, IMO’s terminology was used to retrospectively code data from the problems and diagnoses sourced from the EHRs of a regional health information exchange (HIE). The purpose of the study was to analyze if any specificity was lost as data was sent from EHRs to the HIE. Though I had a hypothesis that some information would be lost in this transaction, I was surprised by the magnitude of the problem. Using the Hierarchical Condition Category (HCC) Risk Adjustment Factor (RAF) model to gain visibility into population risk, we saw a nearly 20% improvement in the overall picture of the population when using the IMO software to facilitate the translation instead of typical “cross-mapping” solutions that are often used to translate standard industry codes.
Certainly, this lack of accurate visibility to risk has major implications to any organization wanting to understand risk or compete in a risk-based system, but consider the following example and its implications to patient care. In this case, just one of the thousands analyzed, a Medicare patient had a diagnosis of “breast cancer metastasized to pelvis.” Because only a single SNOMED®CT code was transmitted from the clinical application to the HIE (via C-CDA specifically), only the primary breast cancer diagnosis was communicated between systems. Any analytics performed using the information in the downstream system would therefore lose the staging (metastatic) component of the original diagnosis. Moreover, any application trying to assess this patient’s clinical situation at a future point in her care would likewise be compromised.
Preventing such loss of data fidelity to ensure that healthcare professionals can appropriately understand the patient’s story is at the heart of what IMO does. We often (rightfully so) think about the impact that our solutions have on providers and systems as a whole, while knowing that our products help patients, too. The story of this patient, though, allowed that reality to resonate with me on a more personal level and connect to the heart of why IMO’s team works so hard to address the problems produced when creating terabytes of data within electronic chart storage.
I am cognizant of the impact that IMO’s technology can have on patients, providers and health systems as a whole, but all too often, I tend to think about these problems abstractly. However, cases like the one described above are what drive me to figure out how to build even more powerful solutions. Understanding the story behind that patient’s care, not patient care in the abstract, brings renewed invigoration to my work and that of my colleagues, as we strive to create a more complete and accurate clinical picture for physicians and patients alike.
Dr. Matt Cardwell joined IMO in 2010 and now leads the company’s product management team. Over the years, Matt has built and led IMO’s technical and clinical services teams through the company’s rapid growth during Meaningful Use and the ICD-10-CM transition in the US. Prior to IMO, Matt completed his doctoral work in mathematics at Northern Illinois University.