Using Knowledge Rules for Pharmacy Mapping

Abstract

The 3M Health Information Systems (HIS) Healthcare Data Dictionary (HDD) is used to encode and structure patient medication data for the Electronic Health Record (EHR) of the Department of Defense’s (DoD’s) Armed Forces Health Longitudinal Technology Application (AHLTA). HDD Subject Matter Experts (SMEs) are responsible for initial and maintenance mapping of disparate, standalone medication master files from all 100 DoD host sites worldwide to a single concept-based vocabulary, to accomplish semantic interoperability. To achieve higher levels of automation, SMEs began defining a growing set of knowledge rules. These knowledge rules were implemented in a pharmacy mapping tool, which enhanced consistency through automation and increased mapping rate by 29%.

Background

Pharmacy data is different from many other clinical terminology domains in that clinical drug concepts can be readily defined by a limited set of attributes [1, 2]. This allows for a more rigorous terminology model whereby a clinical drug can be defined as having ingredient, strength, form, and route as core attributes. In addition, drugs have many brand names, abbreviations, synonyms, and packaging information that sometimes need to be taken into account for unique identification. Instead of matching representations for the entire drug concept at once, attributes are matched. Parsing out the correct attributes from a drug representation is challenging, but once all the attributes have been identified, they can be used to find exact concept matches. This enforces mapping consistency by removing human variability due to case-by-case judgment calls.

Methods

The HDD pharmacy mapping tool implements an automated search algorithm that is run to pre-select matches. This is followed by a structured manual review that adheres to the same searching methodology as the automated process. Each drug concept is sent through a parser, which references a knowledge base to identify ingredient, strength, form and route. The knowledge base of the pharmacy mapping tool supplies rules for parsing and matching each drug attribute. It is organized to support multiple synonyms, brand name to generic ingredient conversions, and form and route hierarchies.

Ingredient

When a brand name is identified in a DoD drug representation, it is translated into the appropriate generic name by referencing the knowledge base. The generic ingredient is then matched against a comprehensive list of ingredients. If no match is found, the tool will attempt to switch the generic and brand names and search again.

Form and Route

The form and route hierarchies and synonyms are used to broaden the scope of candidate matches. For example, the term CAPS in a DoD drug refers to capsule. Some synonyms used for matching include: CAP, CAPSULE, CAPSULES, etc. Additional potential matches are identified by referencing the hierarchy for all of the more specific forms of capsule: CAP SEQ, CAP SPRINK, CAP W/DEV, CAP DS PK, CAP MPHASE, CAPSULE DR, CAPSULE CR, CAPSULE SA, etc.

Drawing a clear distinction between form and route is challenging. Often form implies route and vice versa. Preferred attribute scores are used to assign form and route for matching purposes, when there is confusion in differentiating them. For example, injection is often used as both a form and a route. If injection and intravenous both appear in a drug representation, the knowledge base is referenced to determine that intravenous is a preferred route and injection is a preferred form.

Results

In the first review of the DoD formularies, 151,854 unique drug representations were evaluated. Using the pharmacy mapping tool, 50.8% were identified as exact matches to existing HDD concepts, 35.5% were approximate matches, and 13.7% were unmatched. Mapping speed was improved by 29% over the previous fully manual mapping process and consistency among SMEs was enhanced because of rules enforced by the tool.