Toward a National Health Information Infrastructure: A Key Strategy for Improving Quality in Long-Term Care. B. Content Coverage for Terms Provided by Domain Experts


After generating the lists of terms from the domain experts, we proceeded to code those terms using procedures described earlier. Table 4 compares the content coverage of the MDS and the three coding systems (SNOMED CT, ICF, ICNP) for the terms identified by the domain experts. As expected, SNOMED CT provided the most complete content coverage, consistent with the far more extensive compilation of terms within that terminology coding system. For the domains of chronic pain and urinary incontinence, SNOMED CT provided coverage for all the terms suggested by domain experts and a complete match was possible for the overwhelming majority of terms. For the domain of pressure ulcers, SNOMED CT provided a complete match with 80% coverage observed. The types of terms for which no SNOMED CT code could be identified included abbreviations (e.g., MR/DD, ADL, and a variety of qualifiers such as none of the above). The ICF and ICNP each provided either complete or partial content coverage for approximately half of the domain expert terms related to pressure ulcers and urinary incontinence, but only a small percentage of content coverage for the terms related to chronic pain.

Coverage provided by the MDS varied significantly across domains. The MDS provided a complete match on 70 percent of the pressure ulcer terms, but provided a complete match on only 8 percent of the terms recommended by the domain experts to assess of quality of care relate to pain and urinary incontinence. This is not an unexpected result, and is probably explained in part by the nature of three conditions selected for this review. As noted earlier, pressure ulcers are directly observable and were widely recognized as an issue of concern when the MDS was originally developed. Chronic pain is not directly observable and the presence of chronic pain is either directly reported by the person experiencing the pain or inferred by persons providing care to the person experiencing the pain. In the latter case, a correct inference requires that the provider observe for a wide variety of non-specific behaviors and then correctly interpret those behaviors as pain behaviors. This is the sort of information that should be recorded in clinical documentation systems (whether paper or electronic). Similarly for urinary incontinence, the results of diagnostic and evaluation tests such as laboratory work and procedure results are typically recorded in clinical documentation systems. The MDS was not developed as a clinical documentation system.

Appendix G includes detailed lists of terms provided by the domain experts that do not have a complete "match" in either MDS or SNOMED CT (i.e., the "partial match" or "no match" columns of Table 4). It will be difficult to achieve the goal of "interoperable" or reusable data across applications without these terms.

For example, if a person has an existing pressure ulcer and is at known risk for the development of additional pressure ulcers, practice guidelines recommend that a skin inspection be completed at least once a day with particular attention to bony prominences and that either static or dynamic support surfaces be used depending on the person's ability to assume a variety of positions without weight bearing. Terms related to the concept of "bony prominences" were not found in either the MDS or SNOMED CT. The MDS includes two items related to pressure relieving devices; one is "pressure relieving devices for chair" and the other, "pressure relieving devices for bed". However our domain expert mentioned five more specific terms related to pressure relieving devices: air-fluidized beds, characteristics of support surfaces, dynamic support surface, low air-loss bed, and static support surface. SNOMED CT does provide terms related to these more specific concepts. The MDS use of the concept appears to be related to whether a pressure relieving device was used in a chair or a bed. More detailed information is needed in order to use the concept in an automated clinical decision support system (i.e., the type of pressure relieving device would need to be made explicit). Persons actually completing point of care clinical documentation would likely provide information at the more detailed level. A reporting interest in whether that device was used for a chair or bed could then be algorithmically derived from the more specific information within an electronic clinical documentation system.

Another example illustrating the need for terminology completeness relates to the names of specific clinical assessment scales. SNOMED CT includes names of many scales used in health care. For example, the McGill Pain Questionnaire and the FACES pain scale are terms within SNOMED CT, however neither the Braden Scale nor the Norton Scale were included in SNOMED CT at the time of this review, nor is pressure ulcer risk quantified in the MDS. Automated clinical alerts when critical values of the Braden Scale or Norton Scale are observed would likely result in the earlier deployment of preventive interventions (such as pressure relieving devices) and therefore a reduction in the incidence of new pressure ulcers. This means that the concept of a Braden Scale score or a Norton Scale score must be available within a given system. The developers of SNOMED CT have a defined procedure for updating the content of this terminology. Researchers at Mayo will submit the terms that were found to not have a complete match in SNOMED CT to the developers of SNOMED for their consideration as terms that should be added to SNOMED.

In the present analysis we included only the specific name of the clinical assessment scale recommended by the domain expert. We did not retrieve specific content within any given scale and further determine the content coverage of that scale. Readers should not automatically assume that because the name of the scale is missing that the content of scale is missing. Researchers at Mayo Clinic are working to apply natural language processing techniques to discover "kernel concepts" in such clinical scales, and will focus future content coverage studies on identifying the terms that may be lacking in clinical terminology systems.

View full report


"toward.pdf" (pdf, 237.05Kb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®