As researchers in Pharmacoepidemiology and Pharmacovigilance focused on studying the use, adherence, effectiveness and safety of drugs in different conditions and populations in Primary Care using data proceeding from the Electronic Health Records (EHR), understanding and managing such massive data poses a significant challenge for our research group. We have access to a large amount of data on drug exposure in the EHR obtained from the electronic prescription and invoice data generated from the dispensing of drugs reimbursed by the National Health system.
Our focus on the use and adherence to drugs involves determining the extent to which patients collect their medications from the pharmacy as prescribed. However, it remains challenging to ascertain the actual intake of medications at home. Identifying when a patient switches, discontinues or intensifies a specific drug treatment is often intricate due to complex treatment patterns, making analysis and interpretation difficult. Failure to assign the correct drug treatments to patient data can introduce errors when studying medication adherence.
To address this issue, our team here at the BEAMER project has designed a novel algorithm known as Smooth, aiming to obtain the most probable treatment patterns of a single drug in monotherapy or combined drugs over time to study treatment switch, addition, discontinuation and adherence using EHR. The Smooth algorithm is an automated approach that standardizes, simplifies and improves data processing in pharmacoepidemiological studies analyzing drug exposure data.
The algorithm has undergone validation using clinical data and was recently published in the JMIR Medical Informatics journal. Its implementation will facilitate the analysis of drug exposure and adherence data for BEAMER’s task 3.1 and task 2 within our specific setting.
We also developed an R package called smoothy which can be used to model drug exposure in pharmacoepidemiological studies, thereby allowing the study of adherence with real-world data available in our setting
A recent example of our research using the Smooth algorithm is an observational study in a large cohort of atrial fibrillation patients treated with oral anticoagulants (OAC) for stroke prevention in Catalonia, Spain, spanning from 2011 to 2020.
Our first article published within this study described sex and gender differences in patients starting OAC treatment, calculating adherence and persistence for those receiving direct OAC. We found a frequency of underdosing (ranging from 15-39%, depending on the drug), yet over 80% of patients exhibited adherence to these drugs.
In the second article published, we found that being adherent to direct OAC in comparison to those people who were non-adherent to medication resulted in a lower risk of suffering a stroke and having cerebral bleeding. We concluded that considering patients’ baseline characteristics and comorbidities when prescribing OAC is crucial for maximising the clinical benefit, particularly in stroke prevention.
The Smooth algorithm’s validation with different types of medications allows for its generalization to model drug exposure data and study treatment switches, additions, discontinuations, or adherence across various treatments.
About the authors:
Maria Giner-Soriano, PharmD, PhD. Pharmacoepidemiology Researcher IDIAPJGol, Barcelona, Spain.
Rosa Morros Pedrós, Silvia Fernández García, IDIAPJGol, Barcelona, Spain.