Principles of Health Informatics: Terminologies and classification systems
0 likes121 views
This lecture focuses on defining and classifying clinical terminologies, specifically ICD-10, Read Codes, and SNOMED CT. ICD-10 primarily addresses disease classification, while Read Codes and SNOMED CT are designed for broader clinical applications with hierarchical structures. The lecture highlights the importance of mapping these terminologies to one another and discusses their limitations, especially related to attempting to cover too wide a range of clinical information.
Principles of Health Informatics: Terminologies and classification systems
- 1. Lecture 12: Terminologies and Classification Systems Dr. Martin Chapman Principles of Health Informatics (7MPE1000). https://martinchapman.co.uk/teaching
- 2. Lecture structure This will mostly be a practical lecture, giving you the opportunity to independently gain an understanding of different clinical terminologies. We will use online software to look at ICD-10 codes, Read codes and SNOMED CT codes. Look out for URLs throughout the lecture and use a laptop or other mobile device to access each platform.
- 3. Lecture limitations Coding systems are detailed and constantly evolving. Similarly, the pieces of browsing software used to present different coding systems are detailed and constantly evolving. As such, there may be features – of both the terminologies and browsing software – that aren’t covered here or, indeed, that I do not know about. Let’s learn together!
- 4. Learning outcomes 1. Be able to understand the key characteristics of three different (clinical) terminologies.
- 5. International Classification of Diseases (ICD) Tenth revision (ICD-10)
- 7. Questions 1. Which (clinical) domain is covered by this terminology? 2. How are the terms in the terminology structured? 1. What range of codes are available? 2. Is there a relationship between the codes? 3. Are there any provisions made for mapping to other terminologies? 4. Are there any obvious limitations to the terminology? https://icd.who.int/browse10/2019/en
- 8. Which (clinical) domain is covered by this terminology? Covers disease and other health problems. Allows morbidity (suffering from a disease) and mortality (death) information from around the world to be systematically recorded.
- 9. How are the terms in the terminology structured? In their simplest form, ICD-10 codes are three alphanumeric characters from A00 to Z99. ‘U’ codes are used for emerging diseases like COVID (at the time), as we have seen.
- 10. How are the terms in the terminology structured? When required, a fourth character can be added to create sub- categories: We may also see examples of a fifth character being added, to create additional categories. This creates a form of hierarchy.
- 11. How are the terms in the terminology structured? Chapters Separation into blocks depends upon what the focus of the current chapter is (multi-axial). Here, for example, the focus is disease, so each block represents different modes of transmission The relationship between codes, blocks and chapters forms an additional hierarchy.
- 12. Are there any provisions made for mapping to other terminologies? No immediate evidence of mapping (at least not on this browser). This may be because ICD-10 is considered the single standard for recording disease prevalence around the world, and thus other terminologies map to it.
- 13. Are there any obvious limitations to the terminology? Many adjustments have been made to try and cover a broad range of areas (beyond morbidity and mortality), resulting in a terminology that perhaps tries to cover too much, and its remit becomes confusing as a result. May not be as useful in a clinical setting. (Remember we were broadly referring to these terminologies as ‘clinical’, but, in fact, terminologies do have a wider remit). Note: ICD-10-CM, an extension, does have codes for diagnoses and procedures.
- 14. Aside: The wider ICD family
- 16. https://bioportal.bioontology.org/ontologies/RCD/?p=summary Note: Because Read Codes are retired, we do not have a direct browser, but viewing the terminology as an ontology should still be useful.
- 17. Questions 1. Which (clinical) domain is covered by this terminology? 2. How are the terms in the ontology structured? 1. What range of codes are available? 2. Is there a relationship between the codes? 3. Are there any provisions made for mapping to other terminologies? 4. Are there any obvious limitations to the terminology? For all these questions, how does the current terminology differ from those seen previously? https://bioportal.bioontology.org/ontologies/RCD/?p=summary
- 18. Which (clinical) domain is covered by this terminology? Predominantly designed for general clinical use, in order to code events in the EHR. This contrasts the (broader) focus on disease taken by ICD-10. Like ICD-10, we again encounter a hierarchical structure that allows us to explore logically. We can already see more terms that would be relevant to a clinical setting.
- 19. How are the terms in the terminology structured? Unlike ICD-10, hierarchy is not built into the codes themselves (e.g. A00 vs A00.1), nor is there a finite range of codes, but there are links between the terms, across which properties can be inherited. Type Unique Identifier. This is inherited between terms that connect to each other. Concept Unique Identifier
- 20. How are the terms in the terminology structured? Later versions of the Read codes terminology have a compositional aspect that involves combining terms with qualifiers (that can have a certain value) to create new terms. An entity called a template (rather than a ontology) controls combinations. Term Qualifier Qualifier value Hand fixation of fracture using intramedullary nail
- 21. Are there any provisions made for mapping to other terminologies? Lots! Particularly to ICD-10. Read codes in particular focus on the integrity of mappings, by having a set of quality insurance rules that check structure.
- 22. Are there any obvious limitations to the terminology? Read codes were revised multiple times before ultimately being retired because of many of the limitations of models we’ve seen, including being a snapshot that becomes less useful over time. There were other limitations to Read codes, including the inability of the templates we’ve seen to properly control the composition process.
- 23. SNOMED CT
- 25. Questions 1. Which (clinical) domain is covered by this terminology? 2. How are the terms in the ontology structured? 1. What range of codes are available? 2. Is there a relationship between the codes? 3. Are there any provisions made for mapping to other terminologies? 4. Are there any obvious limitations to the terminology? For all these questions, how does the current terminology differ from those seen previously? https://termbrowser.nhs.uk/?perspective=full&edition=uk- edition&release=v20230215&server=https://termbrowser.nh s.uk/sct-browser-api/snomed
- 26. Which (clinical) domain is covered by this terminology? Much like Read codes, SNOMED CT aims to provide a means to code a broad range of clinical events in the EHR. Indeed, as discussed, its use now supersedes the use of Read codes. Like ICD-10 and Read, we again encounter a hierarchical structure that allows us to explore logically. We can again see terms that would be relevant to a clinical setting. SNOMED CT is cited as having a >90% coverage of common patient problems.
- 27. How are the terms in the terminology structured? Similar to Read codes, we once again have connections between our concepts creating a hierarchy (parents and children). This is also the first time we see connected terms, concepts and codes (Lecture 11) in practice.
- 28. How are the terms in the terminology structured? We also see a more explicit example of multi-axial classification (the block separation in ICD-10 was less overt).
- 29. Are there any provisions made for mapping to other terminologies? We again see a focus on mapping to ICD-10.
- 30. SNOMED CT is large (~300,000 concepts), meaning that some redundancy is inevitable. Such a large number of terms can also result in errors, such as ambiguous terms, terms with illogical hierarchies, or terms that perhaps do not fall within the scope of the terminology. Are there any obvious limitations to the terminology?
- 31. Summary ICD-10, Read codes and SNOMED CT are three examples of (clinical) terminologies. ICD-10 is primarily designed to focus on disease, but does have clinical extensions. Read and SNOMED CT aim to directly cover a broad range of clinical problems. All three terminologies employ some hierarchical structure to organise their terms, with SNOMED CT utilising multi-axial connections most explicitly. Both Read and SNOMED CT map to ICD 10. Trying to be too broad is a common source of issues with clinical terminologies.
- 32. References and Images Enrico Coiera. Guide to Health Informatics (3rd ed.). CRC Press, 2015.