Representing Disease Concepts in Health Information Systems

A Comparison of the ICD-10-CM and SNOMED CT Terminologies

By Michael Stearns MD, CPC, CRC, CFPC

Introduction

The ability to share accurate and complete clinical information across disparate stakeholders has improved through more than three decades of sustained effort. However, progress continues to be constrained by competing clinical, regulatory, financial, and operational priorities, including the mandated use of the diagnosis-oriented classification system known as International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM). This blog examines the key differences between Systematized Nomenclature of Medicine-Clinical Terms (SNOMED CT®) and ICD-10-CM, with a focus on the knowledge representation limitations inherent in ICD-10-CM. In doing so, it aims to help health information technology professionals better understand how these limitations affect the accurate representation of disease concepts, clinical specificity, and the effective sharing of health information.

In the United States ICD-10-CM is mandated under the Health Insurance Portability and Accountability Act (HIPAA) for reporting diagnoses in administrative and reimbursement transactions. As a result, ICD-10-CM has become the primary coding system used for capturing disease concepts across healthcare systems in the US. However, ICD-10-CM was not designed to support many contemporary requirements of electronic health information systems, including semantic interoperability, clinical decision support, queries, clinical research, and advanced analytics.

By contrast, SNOMED CT® was developed specifically to represent medical conditions in electronic healthcare systems with a high degree of specificity and accuracy, including features that allow it to be optimized for semantic interoperability. However, the full adoption of SNOMED CT®’s potential has, in general, not been realized in most settings of care, including EHRs, hospitals, physician practices, health information exchanges, and state and local public health systems, SNOMED CT® is not used as the primary coding system in the clinical settings where it could be most beneficial, i.e., patient encounters including hospitalizations, emergency room visits, and physician offices, etc.

What is a Clinical Concept in a Terminology?

A clinical concept in a terminology such as SNOMED CT® is an abstract representation of a real world clinical idea. This includes medical conditions, symptoms, procedures, lab tests, diagnostic imaging tests, medications, etc. Each concept is represented by a unique code. Concepts may have synonyms and relationships to other concepts in the same terminology, including hierarchical relationships (e.g., Viral hepatitis B is a “type of” Viral hepatitis). Electronic healthcare applications are dependent on the codes that represent clinical concepts. In general, they do not have the ability to interpret text created by clinicians during patient care. SNOMED CT® and ICD-10-CM are both coding systems, however, SNOMED CT® has advantages we will detail further later in this article.

ICD-10-CM: Background and Current Role

ICD-10 was released by the World Health Organization (WHO) in the 1990s. The United States developed a clinically modified version (ICD-10-CM) that was implemented in 2015. It became the mandated code set for electronic transactions under the Health Insurance Portability and Accountability Act (HIPAA). ICD-10-CM was designed primarily as a statistical and administrative classification system. Since its adoption, ICD-10-CM has become deeply embedded in healthcare operations and clinician workflows. It is used extensively for reimbursement, quality measurement, registry reporting, and advanced payment models. It is also used for risk adjustment, case mix determination and in some settings clinical decision support.

To be reimbursed, healthcare systems need to report ICD-10-CM codes based on provider documentation and specific coding guidelines. Significantly, ICD-10-CM guidelines and other limitations of ICD-10-CM may result in codes being reported that may not consistently align with how clinicians represent disease concepts. For example, to avoid potential reimbursement denials and rejections, coding professionals are required to use official ICD-10-CM coding resources, starting with an “Alphabetic Index” to identify terms that are mapped to a specific ICD-10-CM code. However, the Alphabetic Index requires that certain distinct clinical conditions be reported using the same ICD-10-CM code. Representing different clinical concepts with the same ICD-10-CM code later limits the coding system’s value in healthcare applications.

As shown in Table 1, the ICD-10-CM code D73.89 “Other diseases of spleen” is used to represent multiple clinically distinct conditions. By comparison, SNOMED CT® has specific codes (including synonyms) for eight of the ten clinical concepts that ICD-10-CM represents using a single code. 

Table 1. Comparison of coding certain splenic disorder terms in ICD-10-CM and SNOMED CT®.

ICD-10-CM Code	ICD-10-CM Index Term	SNOMED CT® Code	SNOMED CT® Description
D73.89	Splenic calcification	442360003	Calcification of spleen (disorder)
D73.89	Postinfectional splenic disease	None	No matching concepts
D73.89	Episplenitis	25167006	Perisplenitis (disorder)
D73.89	Splenic fibrosis	56338005	Splenic fibrosis (disorder)
D73.89	Floating spleen	191384005	Wandering spleen (disorder)
D73.89	Inflammation of the splenic capsule (Perisplenitis):	25167006	Perisplenitis (disorder)
D73.89	Non-traumatic splenic lesion	None	Related term: Nontraumatic splenic rupture (disorder)
D73.89	Lien migrans	191384005	Wandering spleen (disorder)
D73.89	Splenocele	79040006	Cyst of spleen (disorder)
D73.89	Splenosis	38831007	Splenosis (disorder)

Once the correct ICD-10-CM code has been identified in the Alphabetic Index, coders turn to the ICD-10-CM “Tabular List” which provides additional coding rules in the form of instructional notes, which include laterality (i.e., left, right or bilateral) and specific coding “conventions” (e.g., when two codes may or may not be reported together), and other requirements.

The ICD-10-CM Tabular List file also includes what are referred to as ICD-10-CM “inclusion terms.”  These terms are listed below a code or a category section in ICD-10-CM.  They may or may not represent true synonyms but rather provide guidance as to which clinical concepts are intended to be within the scope of a code.  An example of this is the ICD-10-CM code A08.39 “Other viral enteritis.”  The Tabular List informs the user that the following unique conditions should be reported using the same ICD-10-CM code: 

• Coxsackie virus enteritis
• Echovirus enteritis
• Enterovirus enteritis NEC
• Torovirus enteritis

This is another example of ICD-10-CM using the same code to represent different medical conditions.

When this occurs, the meaning of the code becomes ambiguous. This in turn limits the reliability and usability of codified data for a wide range of applications, including clinical decision support, analytics, population health, clinical research and interoperability.

When additional coding guidance is needed, CMS officially recognizes the “AHA Coding Clinic for ICD-10-CM/PCS” as an official interpretive authority. As these rules reflect a potentially different set of requirements for reporting ICD-10-CM codes, they have the potential to introduce a degree of “semantic drift” away from what was documented by the provider during patient care.  

ICD-10-CM’s structure reflects pre-digital origins, and it was not intended to support the full range of clinical care objectives required by modern health information technology systems. As discussed in the section below, SNOMED CT® does meet these objectives. This includes but is not limited to more accurate clinical querying and support for computable meaning, clinical decision support, post-coordination (i.e., linking two or more codes together to fully describe a clinical concept), and the comprehensive coverage of clinical diagnoses.

Post-coordination: the ability of an application or terminology to represent additional clinical detail by combining a base concept with one or more additional concepts to create a more precise clinical expression. For example, the base term ‘abdominal pain’ is combined with the qualified value ‘severe’ to create the expression ‘Severe abdominal pain.’

Pre-coordination: the representation of a detailed clinical concept using a single, predefined code that includes multiple unique concepts. ICD-10-CM uses heavily precoordinated expressions to represent certain concept types, such as code S72.001A – Fracture of unspecified part of neck of right femur, initial encounter for closed fracture.

SNOMED CT® Background and Intended Use

SNOMED CT® is a clinical terminology that originated from the merger of SNOMED Reference Terminology, developed by the College of American Pathologists, and the U.K.’s Clinical Terms (Read Codes). It is an internationally governed standard that is maintained by SNOMED International and used to enable semantic interoperability, clinical decision support, and advanced health data analytics.

SNOMED CT® includes over 350,000 unique clinical concepts spanning the following healthcare domains:

Approximately 32% of SNOMED CT® codes represent clinical findings and disorders. As noted previously, SNOMED CT® was designed explicitly for integration into electronic healthcare systems, including an emphasis on support for the ability to exchange clinical information that is complete and accurate. It provides a standardized, concept based approach for capturing, sharing and receiving clinical information entered by clinicians. Unlike ICD-10-CM, SNOMED CT® supports rich relationships between concepts and enables more precise representation of clinical meaning. 

The ability to use SNOMED CT® as the required or the preferred terminology is required for certain national Healthcare IT use cases (e.g., EHR certification, Version 3 of the US Core Data Interoperability (USCDI) and HL7 Fast Healthcare Interoperability Resources (FHIR) guidelines). It is not mandated for HIPAA transactions. Historically, this has limited organizational incentives to adopt SNOMED CT® as the primary terminology for diagnosis capture. In practice, SNOMED CT® is mostly used as a terminology for EHR problem lists and clinical decision support applications. Certified EHRs and provider organizations are required to demonstrate the ability to exchange problem list codes using SNOMED CT® via the USCDI. However, further implementation and daily use of SNOMED CT® is not required.

When fully integrated, SNOMED CT can significantly improve data quality by enabling consistent indexing, storage, retrieval, and aggregation of clinical information.

Comparative Analysis: ICD-10-CM and SNOMED CT®

The following comparison examines ICD-10-CM and SNOMED CT® using selected principles derived in part from Dr. James Cimino’s seminal article titled “Desiderata for Controlled Medical Vocabularies,” that which outlines foundational characteristics for high-quality clinical terminologies.

Concept-Centric Design

A fundamental requirement of clinical terminologies is that each code must represent a single, distinct clinical meaning. SNOMED CT® adheres strictly to this principle. But as demonstrated above, ICD-10-CM allows codes to have more than one meaning. This can compromise the integrity of healthcare data.  

The Value of a Polyhierarchy

A polyhierarchy is a hierarchical structure where clinical concepts may have more than one parent term. Both ICD-10-CM and SNOMED CT® employ hierarchical structures. But only SNOMED CT® utilizes a polyhierarchy, in which a single concept may have multiple logical parents. ICD-10-CM has a monohierarchical structure – each concept is assigned a to single parent.

For instance, herpes encephalitis is both a disorder of the nervous system and an infectious disease. In ICD-10-CM, B00.4 “Herpesviral encephalitis” appears only within the infectious disease chapter. In SNOMED CT®, the corresponding concept (428638009 “Encephalitis caused by Herpesvirus (disorder)”) is classified as an infectious disorder and as disorder of the central nervous system. Polyhierarchies allow for precise representation of clinical concepts and enable more accurate querying, analytics, and clinical decision support.  They are also a key component of formal definitions, described below.

Formal Definitions

SNOMED CT® supports formal, computable definitions for many concepts through logical relationships with other concepts, including multiple parent-child relationships. A key component of this model are attributes, which are formally defined relationships between SNOMED CT® concepts. They are used to “define” concepts in SNOMED CT® by assigning specific characteristics to clinical concepts, such as body structure, causative agent, severity, or temporal context.

For example, the SNOMED CT® concept “Pneumonia caused by Streptococcus (disorder)” (code 34020007) has the following attributes:

• Causative agent (attribute) -> Genus Streptococcus (organism)
• Pathological process (attribute) -> Infectious process (qualifier value)
• Associated morphology (attribute) -> Inflammation and consolidation (morphologic abnormality)
• Finding site (attribute) -> Structure of parenchyma of lung (body structure)

Each attribute (e.g., “Causative Agent”) and its associated value (e.g., “Genus Streptococcus”) are represented by a unique SNOMED CT® code.

Not all concepts in SNOMED CT® are defined to this level. However, the structure of SNOMED CT® allows for automated inference of hierarchical relationships and this supports efficient and consistent terminology maintenance. ICD-10-CM does not provide comparable formal definitions, limiting its capacity for computational reasoning and automated classification.

Comprehensive Concept Coverage

Comprehensive concept coverage is essential to prevent the proliferation of local or proprietary codes that undermine interoperability. SNOMED CT® contains over one million synonyms for approximately 350,000 clinical concepts. ICD-10-CM includes approximately 70,000 diagnosis codes.

Despite the intuitive advantages of comprehensive coverage, the breadth of SNOMED CT® may introduce implementation challenges, particularly at the point of care, where reduced lists of SNOMED CT® codes may be needed improve usability at the point of care. The need to develop and maintain subsets has been cited as a barrier to wider clinical adoption, despite the inherent advantages of using SNOMED CT® in electronic applications.

Post-Coordination

SNOMED CT® supports post-coordination, allowing multiple concepts to be combined to express detailed clinical meaning, such as severity or temporality, when a single pre-coordinated concept is unavailable. While the practical use of post-coordination in healthcare settings remains limited, it is a key component of SNOMED CT®’s design, as it allows for detailed representation of clinical concepts.

SNOMED CT and ICD-11

ICD-10 was developed in the paper era, is a rigid, pre-coordinated classification designed mainly for statistical reporting, but mandated in the US for reimbursement and some types of reporting. ICD-11 represents a redesign of ICD-10 that includes polyhierarchies, improvements in clinical concept coverage, and sanctioned post-coordination. This allows clinical detail to be expressed more flexibly and aligns ICD-11 more closely with SNOMED CT.

SNOMED CT®’s adoption has gradually been increasing, notably with its expanded use in FHIR communications. However, full adoption of SNOMED CT® would require it to be mandated for reporting in all aspects of healthcare, including reimbursement. This is not likely to occur. Alternatively, ICD-11 was created by the World Health Organization (WHO) as the first version of ICD designed for use in electronic applications. Similarities to SNOMED CT® include the use of polyhierarchies, richer clinical content, and support for post-coordination. ICD-11 has been adopted by several countries, but implementation in the US is not currently planned.

Conclusion

ICD-10-CM and SNOMED CT®serve fundamentally different purposes within health information systems. ICD-10-CM remains mandated for reimbursement and administrative reporting in the United States, but its design limits its effectiveness as a clinical terminology. SNOMED CT® was designed specifically to support clinical care, analytics, and semantic interoperability in health information technology applications, but its adoption has been limited by the mandated reporting status of ICD-10-CM. ICD-11 shares several valuable features with SNOMED CT® that make it superior to ICD-10-CM, however, its implementation in the U.S. is not likely to occur in the near future.   

The main intent of this article is to highlight the differences between ICD-10-CM and SNOMED CT® in the context of how their features affect the capture, use and sharing of medical diagnoses in healthcare systems. ICD-10-CM is ubiquitous in healthcare. Health information technology professionals and clinicians may benefit from an understanding of the differences between ICD-10-CM and SNOMED CT®. Awareness of ICD-10-CMs inherent limitations may influence decisions that impact healthcare IT system design, analytics, clinical decision support, quality measure performance, interoperability, compliance, and clinical research.

---

About Dr. Stearns