(146) Validation of New Medication User Definitions in Single-Specialty EHR Data

Publication Number: 477

Background: New user designs are standard in many pharmacoepidemiology studies, but a first prescription appearance in electronic health record (EHR) data may not always represent a new medication start. The optimal implementation of new user designs in single-specialty EHR data is unknown.

Objectives: To identify the optimal implementation of new user designs in single-specialty EHR data.

Methods: We identified patients with new prescriptions for biologics or targeted therapies (e.g. janus kinase inhibitors) for rheumatoid arthritis (RA) from rheumatologists in the Illumination Health Real-World Evidence Platform (EHR data) linked to either fee-for-service Medicare or commercial health plan data. The first prescription in EHR data for each medication anchored the index date. We required ≥ 12 months of medical and pharmacy coverage in claims data prior to index.

We tested 6 new user definitions that required a rheumatologist office visit, any prescription in EHR data, or both, either ≥ 12 or ≥ 24 months prior to index. We evaluated positive predictive values (PPV) of new user definitions, using claims data as the gold standard with either a “fixed” 12-month baseline, and separately, a “variable” baseline (minimum 12-months, all available prior data).

Results: There were 1131 medication initiations in patients linked to Medicare (mean [SD] patient age 69.2 [10.3] years, 82.8% women) and 455 in patients linked to commercial claims (mean [SD] patient age 65.3 [12.7] years, 81.1% women). In Medicare and commercial health, the mean (SD) amount of prior claims data was 5.3 (3.4) years (Medicare) and 4.8 (3.3) years (commercial insurance), and patients had previously filled a mean (SD) of 1.2 (1.1) and 1.3 (1.1) biologics or tsDMARDs previously, respectively.

The PPVs of the 6 new user definitions ranged from 87-94% (fixed baseline), and 77-87% (variable baseline) in Medicare-linked patients. PPVs ranged from 92-98% (fixed baseline) and 87-95% (variable baseline) in commercial claims-linked patients. Requiring both visits and medications increased PPVs by 6% (Medicare) and 3% (commercial) and had greater yield than requiring a 24-month baseline.

Conclusions: Requiring visits and/or medications at least 12 or 24 months in an EHR data source prior to new medication initiation is a valid approach to identify new medication users.