Menu 返回 Poster-Presentations-Details

W-13: Comparing Two Drug Treatment Coding Approaches: Coding Challenges and Lessons Learned

Poster Presenter

      Sherry Chang

      • Scientific Data Reporting Analyst
      • FDA
        United States


To compare the drug dictionary used by PatientsLikeMe (PLM), a platform for tracking and sharing patient-generated health data (PGHD), to the FDA Adverse Event Reporting System (FAERS) Product Dictionary (FPD), and outline challenges in linking reported drug names to a standard drug terminology.


A qualitative review was conducted between the structure and treatment coding processes for the two databases, followed by a quantitative review to understand the concordance between treatment-name-and-ingredient pairs across the two databases for prescription, over-the-counter, and supplements.


The degree of matching between the two datasets was defined as follows: validated matches were records with agreement in both the treatment name and active ingredient, probable matches were those in which either the PLM treatment name or the ingredient matched the FPD, but not both; and non-matches were those where there was no agreement across the two databases for the record, whether it consisted of only one or both data fields. The initial review of the PLM treatment database showed that of the 11,144 individual entries (records), a total of 3,473 records (31%) were identified as validated full matches, 2,319 records (21%) as probable matches, and 5,352 records (48%) as non-matching in comparison to the FPD. Of the 5,352 records identified as non-matching, the majority (64%) were from the supplements category, followed by prescription drugs (22%) and over-the-counter (14%) products. Within the prescription drug category, majority (54%) were validated matches and 25% were probable matches; within the over-the-counter category, 22% were validated matches and 25% probable matches.


PGHD contains valuable information on a multitude of treatments. Most useful mapping of reported drug treatments includes both the drug name and the active ingredient. The analysis supports the recommendation to map patient reported prescription drug names or over-the-counter treatment names to validated name-substance pairs. Subsequently PLM selected the RxTerms database and conducted the mapping of its treatment records. Additional fields were added to PLM's database to identify and categorize records that do not have an appropriate mapping to RxTerms, such as foreign drug names, ambiguous entries (e.g. - chemotherapy) and prescribed therapies that do not contain a drug product (e.g. - supplemental oxygen). Currently, there is no single public drug dictionary to capture prescription, over-the-counter, and supplemental treatments in the US and global markets. Accurately linking reported drug names to a standard terminology in a uniform and consistent manner requires cross checking with various references to validate the reported treatments, presenting a tremendous challenge for mapping completeness and consistency.