Home  |   Subscribe  |   Resources  |   Reprints  |   Writers' Guidelines

July 2016

Language Barriers Hamper Analytics Efforts
By Jon Melling, FHIMSS
For The Record
Vol. 28 No. 7 P. 6

If you could peel back the layers of any hospital EHR, you'd find hundreds—even thousands—of codes that help convey patients' unique stories. But these aren't only HTML codes—they're also clinical codes.

These shorthand expressions of diagnoses and procedures, together with patient information culled from EHRs, pave the way for the standardization of health care data that leads to better informed clinical decisions and accurate billing for services rendered.

However, data based on clinical codes (coded data) is now being used more widely, including to report quality of care, predict outcomes, compare performance, inform risk-based contracts, and enable personalized medicine. As the use of coded data continues to expand, HIM and IT professionals are being forced to deal with information ingestion and analytics hurdles.

Multiple Codes Create Data Conundrum
Health and Human Services recognizes the need to unravel health data's dilemma. The agency recently announced two funding opportunities to develop better standards for data transfer. According to a statement by Karen DeSalvo, MD, MPH, MSc, national coordinator for HIT, the funding will "advance the use of common, interoperable standards, particularly for key pieces of data like medications and lab results." However, HIM professionals know that multiple code sets present many challenges on the road to achieving interoperability and meeting data analytics goals, including the following:

• Because coded data depend on how various rules and guidelines are interpreted, they can be inconsistent.

• Coded data vary by setting. The ICD-10-PCS classification system denotes inpatient procedures, CPT tags outpatient procedures, and the HCPCS codifies supplies, products, and services not covered by CPT.

• There are state-specific code sets to enable data analysis and research at a more granular level.

• Disparate health care registries use external cause codes with varying degrees of specificity.

In addition, the United States made clinical modifications to the World Health Organization's (WHO) ICD-10 diagnosis classification system and created its own procedural classification system (ICD-10-PCS). Using a modified system makes it difficult to compare clinical data with other countries that haven't modified ICD-10. It also makes it more difficult to ensure consistency on a national level among providers in the care continuum.

LOINC and SNOMED are other commonly used clinical health terminologies. In fact, the International Health Terminology Standards Development Organisation (IHTSDO) claims SNOMED CT is the "most comprehensive and precise clinical health terminology product in the world."

Beyond multiple clinical coding schemes, there are various data transmission standards, such as Health Level 7, Fast Healthcare Interoperability Resources, clinical document architecture, continuity of care document, and Digital Imaging and Communications in Medicine.

The sheer volume of standards and terminologies is overwhelming and will undoubtedly continue to be difficult to maintain over time.

Taming Clinical Data
All of these overlapping clinical codes create a large pool of data that makes it difficult to navigate effectively or efficiently. Multiplicity complicates attempts to obtain a consistent view of health care delivery. In other words, data are lost in translation because each stakeholder essentially speaks a different—and inconsistent—language.

Problem lists are one example of how multiple codes create confusion. In some organizations, even the definition of "problem list" lacks consistency. Is a problem list a tally of active problems for each patient or a synopsis of potential problems from which a physician can select the correct one?

The handoff between health care stakeholders is another data utilization hurdle to clear. Even if data are coded—and therefore the patient episode is codified—differences in EHR systems can lead to misdocumentation and miscoding. Imprecise data lead to mistrust and denigrate their value in the eyes of the clinician. Also, liability issues become a concern.

As health care moves closer to alternative payment methods such as bundled payments and accountable care organizations, health systems must aggregate data across care settings to help manage costs and improve outcomes. This requires consistent definitions and constant monitoring both pre- and postdischarge, as well as the ability to perform data analytics, including predictive analytics, to identify patients at risk for readmission or development of complications.

Historical patient data that don't conform to modern, updated coding standards can be challenging. As a result, analyzing data across patient histories will become increasingly difficult in the years ahead.

How can the industry accomplish this degree of monitoring without speaking a common language that transcends specific care settings?

The Case for Unification
In an ideal world, the industry will meet the challenge of standardizing language across all settings to truly enable data analytics and population health management. Unfortunately, attempts thus far to unify the code sets have failed.

For example, in 1986, the National Institutes of Health created the Unified Medical Language System, a translation tool to serve as an overarching scheme to bridge the gap between various terminology systems. However, differences in hierarchical structure among the various coding schemes caused the system to be fraught with errors.

SNOMED, which many consider the most comprehensive and multilingual clinical health care terminology, also has limitations. First, SNOMED codes are often applied in the background of an EMR using an automated methodology, making it difficult to ensure accuracy. Also, SNOMED is a reference scheme rather than a classification system, making it impossible to use the codes for data analytics or even cross-reference with ICD-10.

Interestingly, WHO is currently collaborating with IHTSDO to ensure alignment between SNOMED and ICD-11. As a result, HIM professionals must begin to familiarize themselves with SNOMED even if they don't currently reference or rely on these codes. In addition, EHRs must include all the codes and progress of codes over time to provide more effective clinical information.

Recent developments in natural language processing (NLP) and technologies such as IBM's Watson may enable a more practical progression to bridge the gaps in the decade ahead to perform the following:

• Measure achievement of outcomes.
• Enable new quality measures.
• Demonstrate commitments to quality.
• Improve case mix index.

Can HIM Professionals Help?
Looking ahead, there are ways in which HIM professionals can assist IT departments and health systems conquer the data revolution. Consider the following strategies:

Advocate for strong clinical documentation improvement (CDI) programs. CDI drives the accuracy of documentation from which all medical codes are derived. Without quality documentation, an organization cannot generate quality data.

Embrace NLP-related technology. Computer-assisted coding and computer-assisted physician documentation improve the effectiveness of CDI and increase coding accuracy. Don't be afraid of these technologies and the new workflows they require. When used with appropriate oversight, these tools can enhance the quality of data generated by any organization.

Pave the way for multidisciplinary, care coordination. Integrated care pathways based on consistent terms and phrases should be assigned to specific patients (by diagnosis), and used consistently. By analyzing clinical data, care teams can identify the correct pathway for each patient to deliver the right care, to the right patient, at the right time.

Ask questions. Join discussions about data analytics strategies. Define what analytics are most important. What data are necessary to achieve those analytics? And what is the quality of those data?

Enhance data analytics skills. Review AHIMA's health data analysis toolkit, and consider obtaining AHIMA's certified health data analyst (CHDA) credential. According to AHIMA, the CHDA credential demonstrates one's ability to "acquire, manage, analyze, interpret, and transform data into accurate, consistent, and timely information, while balancing the 'big picture' strategic vision with day-to-day details."

Take the time to learn about each of the different coding schemes and standards. This is new territory for many HIM professionals. However, considering the industry's continued focus on data analytics, it would be wise to devote time and effort to expanding knowledge and expertise in this area.

Moving Forward
When it comes to data analytics, the health care industry must ensure that codes, context, and meaning are harmonized. While this goal doesn't require a single common coding language, it does necessitate more effective mapping and coordinated use of the different coding schemes.

HIM professionals possess an intimate knowledge of data's significance. Thus, they are well positioned to help guide the conversation and overcome language barriers once and for all.

— Jon Melling, FHIMSS, is a partner at Pivot Point Consulting.