ORIGINAL ARTICLE
Evaluation of the Content Coverage of SNOMED CT: Ability of SNOMED Clinical Terms to Represent Clinical Problem Lists

https://doi.org/10.4065/81.6.741Get rights and content

OBJECTIVE

To evaluate the ability of SNOMED CT (Systematized Nomenclature of Medicine Clinical Terms) version 1.0 to represent the most common problems seen at the Mayo Clinic in Rochester, Minn.

MATERIAL AND METHODS

We selected the 4996 most common nonduplicated text strings from the Mayo Master Sheet Index that describe patient problems associated with inpatient and outpatient episodes of care. From July 2003 through January 2004, 2 physician reviewers compared the Master Sheet Index text with the SNOMED CT terms that were automatically mapped by a vocabulary server or that they identified using a vocabulary browser and rated the “correctness” of the match. If the 2 reviewers disagreed, a third reviewer adjudicated. We evaluated the specificity, sensitivity, and positive predictive value of SNOMED CT.

RESULTS

Of the 4996 problems in the test set, SNOMED CT correctly identified 4568 terms (true-positive results); 36 terms were true negatives, 9 terms were false positives, and 383 terms were false negatives. SNOMED CT had a sensitivity of 92.3%, a specificity of 80.0%, and a positive predictive value of 99.8%.

CONCLUSION

SNOMED CT, when used as a compositional terminology, can exactly represent most (92.3%) of the terms used commonly in medical problem lists. Improvements to synonymy and adding missing modifiers would lead to greater coverage of common problem statements. Health care organizations should be encouraged and provided incentives to begin adopting SNOMED CT to drive their decision-support applications.

Section snippets

Problem Set

Mayo Clinic has a long history of manual record-keeping. The terms used in the problem set for this study were from the Mayo Master Sheet Index (MSI). The MSI documents the diagnoses associated with clinical episodes of care (both outpatient and inpatient),31 which, at Mayo Clinic, are defined as ending when the ongoing care of a patient is changing minimally. Often this is after the diagnoses have been established or after the major interventions have been completed. At this time, the

RESULTS

SNOMED CT was judged as to how well 2 expert reviewers could use it to represent the content contained in 4996 problem statements. SNOMED CT correctly represented 4568 terms (true-positive [TP] results) (Figure 2, A). Thirty-six terms were not believed to be sensible expressions or were misspelled and were not matched by SNOMED CT (true-negative [TN] results), 9 were believed not to be sensible but were matched by SNOMED CT (FP results), and 383 terms were believed to be sensible but were not

DISCUSSION

Clinically meaningful interoperable data require codification with a robust reference terminology and methods for supporting composition. Compositional expressions are organizations of concepts connected using relationships (eg, “pneumonia” “has etiology” “Streptococcus pneumoniae”). Problem list data are an important type of data that require a robust general medical reference terminology.

In the current study, we showed that SNOMED CT performed well, representing 92.3% of common clinical

CONCLUSION

Our study shows that SNOMED CT, when used as a compositional terminology, can exactly represent a large portion (92.3%) of the terms used commonly in medical problem lists. Improvements to synonymy and the addition of missing modifiers would lead to greater coverage of common problem statements. The use of composition significantly improved the ability of SNOMED CT to exactly represent terms from the MSI problem list (51.4% vs 92.3%; P<.001). Health care organizations should be encouraged and

REFERENCES (39)

  • LL Weed

    Medical records, patient care, and medical education

    Ir J Med Sci

    (1964)
  • LL Weed

    Medical records that guide and teach

    N Engl J Med

    (1968)
  • JW Hurst

    Ten reasons why Lawrence Weed is right

    N Engl J Med

    (1971)
  • JW Hurst

    How to implement the Weed system (in order to improve patient care, education, and research by improving medical records)

    Arch Intern Med

    (1971)
  • JA Mehigan

    Patient care and medical case notes: a plea for the wider use of the Weed system of problem orientated medical records

    Ir Med J

    (1975)
  • BW Lloyd et al.

    Use of problem lists in letters between hospital doctors and general practitioners

    BMJ

    (1993)
  • AR Feinstein

    The problems of the “problem-oriented medical record”

    Ann Intern Med

    (1973)
  • LL Weed

    The problem-oriented record

  • AR Feinstein

    An analysis of diagnostic reasoning, I: the domains and disorders of clinical macrobiology

    Yale J Biol Med

    (1973)
  • Cited by (0)

    This work was supported in part by grant LM06918-A2 from the National Library of Medicine and by a Mayo Clinic Department of Internal Medicine MIDAS Award.

    View full text