((A-220)) (A-220) Concordance of Electronic Health Record Data and Medicare Part D Claims Data among Beneficiaries Prescribed Antihypertensive Medications

Background: Electronic health records (EHRs) are increasingly being utilized for regulatory decision-making and evaluation of medication safety. However, there are concerns regarding misclassification of medication exposure using EHR data alone, and few studies have characterized the concordance between EHR and Medicare Part D claims data to capture antihypertensive use in older adults.

Objectives: To evaluate the agreement between EHR prescribing data and Medicare Part D claims data to classify antihypertensive medication exposure among a population of older adults.

Methods: We conducted a retrospective cross-sectional study among patients with UF EHR prescribing data linked to Medicare Part D Event claims from 2014-2017. Beneficiaries were included if they had an antihypertensive medication prescribed in the EHR and were continuously enrolled in Medicare Parts A, B and D during a 12-month follow-up period. We defined the index date as the last prescribing date of a given antihypertensive in the outpatient EHR data for a given year. The follow-up period was defined as a 12-month assessment window including and immediately following the index date. We then calculated mean medication possession ratio (MPR) and positive predictive value (PPV) for each antihypertensive medication individually and by class, and we further stratified by year.

Results: We evaluated validity metrics for 40 antihypertensive medications with sufficient counts to calculate validity metrics between 2014-2017. The mean MPRs for each year across all antihypertensive medications were 53.3 (2014), 73.3 (2015), 73.3 (2016), and 76.0 (2017). The PPVs for each year across all antihypertensive medications were 70.7 (2014), 87.7 (2015), 84.0 (2016), and 87.8 (2017). Among the five main classes of antihypertensive medications, ACE inhibitors had the highest mean MPR (82.56), while thiazide diuretics had the lowest mean MPR (66.9). PPVs across the five main antihypertensive classes were more similar, with ACE inhibitors having the highest PPV (90.0) and angiotensin receptor blockers having the lowest PPV (84.4).

Conclusions: Overall, there was good agreement between electronic health record data and Medicare Part D event prescription data across years, despite some variability amongst different antihypertensive medications. Further research is warranted via stratified analyses to better understand factors associated with decreased concordance amongst these datasets when assessing antihypertensive exposure.