Close this search box.

How can we apply pharmacoepidemiology methods to deprescribing research questions?

Moriarty F, Thompson W, Boland F. Methods for evaluating the benefit and harms of deprescribing in observational research using routinely collected data. Res Social Adm Pharm. 2022;18(2):2269-2275. doi:10.1016/j.sapharm.2021.05.007

Recent advances in pharmacoepidemiology have allowed researchers to apply randomized controlled trial (RCT) principles to observational data. These advances have resulted in better estimates of medication harms and benefits using existing datasets. While these methods are more frequently applied to prescribing studies, they are rarely applied to deprescribing research. A recent study published in Research in Social and Administrative Pharmacy explored the applications of pharmacoepidemiology and causal inference methods to deprescribing research. Authors Frank Moriarty, Wade Thompson, and Fiona Boland framed their research using case studies of benzodiazepines and low-dose aspirin, both of which are often targeted for deprescribing. They provide an overview of research considerations when applying pharmacoepidemiological methods to deprescribing research using observational datasets. Their findings are presented in three tables, summarized below:

  1. Rationale for using observational data to address deprescribing questions
  • RCTs are not suitable or ethical for the study of rare outcomes, such as fractures caused by prolonged benzodiazepine use
  • Observational data provides a longer time frame to investigate questions about prolonged aspirin use and cardiovascular risk
  1. Considerations for emulating target trials using observational data
  • Incorporate a measure of adherence, such as medication possession ratio of >80%, to establish regular use among eligible participants
  • Apply a grace period to avoid misclassification of gaps in coverage as discontinuation of medication
  • Consider a range of ICD condition and procedure codes for outcome classification
  1. Mitigation of biases in observational studies
  • Propensity score methods can reduce the impact of confounding by balancing the distribution of covariates between those continuing medication and those discontinuing medication
  • Introduce billing codes to accurately reflect deprescribing decisions

While many methods in pharmacoepidemiology research can be tailored to deprescribing research, significant methodological challenges remain. The greatest challenge is capturing intentional deprescribing as a treatment approach. In their discussion of this barrier to deprescribing research, the authors suggest the establishment of deprescribing billing codes to capture purposeful deprescribing in large databases. 

Lead author Frank Moriarty summarizes this point:

“We hope that our paper highlights some of the opportunities in using observational research to study the effects of deprescribing. A major challenge, identification of deprescribing in routine data, could be addressed by the introduction of an activity or billing code for deprescribing. This has the potential to increase the validity and robustness of future observational research.”

Read the full article

Contact Us

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.