Structured product labeling improves detection of drug-intolerance issues

Abstract

Objectives: This study sought to assess the value of the Health Level 7/U.S. Food and Drug Administration Structured Product Labeling (SPL) drug knowledge representation standard and its associated terminology sources for drug-intolerance (allergy) decision support in computerized provider order entry (CPOE) systems.

Design: The Regenstrief Institute CPOE drug-intolerance issue detection system and its knowledge base was compared with a method based on existing SPL label content enriched with knowledge sources used with SPL (NDF-RT/MeSH). Both methods were applied to a large set of drug-intolerance (allergy) records, drug orders, and medication dispensing records covering >50,000 patients over 30 years.

Measurements: The number of drug-intolerance issues detected by both methods was counted, as well as the number of patients with issues, number of distinct drugs, and number of distinct intolerances. The difference between drug-intolerance issues detected or missed by either method was qualitatively analyzed.

Results: Although <70% of terms were mapped to SPL, the new approach detected four times as many drug-intolerance issues on twice as many patients.

Conclusion: The SPL-based approach is more sensitive and suggests that mapping local dictionaries to SPL, and enhancing the depth and breadth of coverage of SPL content are worth accelerating. The study also highlights specificity problems known to trouble drug-intolerance decision support and suggests how terminology and methods of recording drug intolerances could be improved.

Figure 1

Structured Product Labeling (SPL) knowledge structures serve as a hub connecting different key terminologies: UNII for ingredient Substances and active moieties, NDF-RT Ingredient/Chemical or MeSH Chemicals for chemical structure classes, NDC for Products with specific dose form and strength (Ingredient.quantity), and SNOMED-CT for observations (of type medical problem or intolerance identified by LOINC). SNOMED-CT has the “causative agent” relationship, which exists as a Participation type in the Health Level 7 (HL7) Reference Information Model (RIM) and represents the same linkage of the SNOMED-CT allergy concept to the NDF-RT/MeSH chemical structure class. A drug-intolerance Issue finally connects an IntoleranceObservation and a conflicting SubstanceAdministration.

Figure 2

Example of how the Health Level 7 (HL7) Reference Information Model (RIM) model works for detecting a penicillin allergy issue for a particular patient. The Patient is subject of a Penicillin allergy observation, coded using a SNOMED allergy concept. SNOMED describes allergies with the causative agent role, which is represented directly in the HL7 RIM participation of type causative agent connecting to the Penicillin Entity kind coded using the NDF-RT/MeSH concept. The patient is also the subject to a SubstanceAdministration, with the consumable being a drug coded with an NDC. Through SPL the NDC code is described as a product with INGRedient ampicillin (coded with a UNII) and of chemical class Ampicillin (NDF-RT/MeSH). The NDF-RT/MeSH terminology defines the IS-A relationships which transitively classify Ampicillin as a Penicillin. For efficient reasoning using a small fixed number of joins the transitive closure IS-A* is materialized. This approach may be extended by materializing even higher order semantic links such as the INGR*IS-A link (HAS-A*).

Figure 3

Warshall's algorithm in SQL to compute the materialized transitive closure for IS-A hierarchies—the first statement CREATEs the transitive closure table and primes it with the direct relationships. The second statement INSERTs transitive relationships by joining the relationship table with itself to turn a.source – a.target = b.source – b.target into a.source – b.target. This algorithm spans the depth of the IS-A hierarchy exponentially converging usually after about 2 to 5 iterations. The final statement INSERTs the reflexive relationships (a IS-A a).

Figure 4

Drug-intolerance issue detection is as easy as joining a patient's drug with the causative agent of a patient's intolerance through the transitive and reflexive closure table.

Figure 5

Top-level MeSH classes useful for coding allergies. The bold-faced classes are included; the classes shown in italics are specifically not included because they did not seem to contain items that would be used for coding allergies.

Figure 6

MeSH headings under the β-lactam family are an example how MeSH concepts are a taxonomy only if we understand the MeSH chemical structures as derivatives. It is not true that the molecule amoxicillin is-a penicillin-G molecule, but we can say that amoxicillin is a penicillin-G derivative.

Figure 7

MeSH hierarchy for sulfonamides. Both allergenic sulfonamide antibiotics and nonallergenic compounds are under the same heading Sulfanilamides. The term Sulfonyl-aryl-amine is used in the literature to describe the allergen class more appropriately. Thus when the amino group is substituted to an amide as in furosemide, the compound apparently loses its allergenity. This suggests that allergenity does not inherit along all derivative relationships in the MeSH hierarchy, but that in some cases the taxonomy must be restructured to separate those compounds with allergen properties from others without such property.