Adverse Drug Events in Ageing Populations (ADDRESS-AP): Who, why and which drugs to stop?

Research question Can we use data from routine electronic health records to better understand who is likely to experience adverse drug events, which medications are most problematic and whether deprescribing leads to benefit or harm?

Background As individuals age, they often develop multiple health conditions that necessitate the prescription of numerous medications. While this may be appropriate, some individuals are prescribed an excessive number of medications.

This is known as 'inappropriate polypharmacy' and is associated with an increased risk of harm, including delirium and falls. Medication-related harm accounts for 1 in 10 hospital admissions.

Presently, doctors lack knowledge regarding which patients are most susceptible to experiencing such harms and the most effective preventive measures. One solution is to 'deprescribe' medications that may cause harm.

However, deprescribing has not been evaluated widely in clinical trials, and while some practice it as a routine clinical strategy, outcomes remain unclear.

Aims and objectives The primary objective of this study is to enhance our understanding of the balance of harms and benefits that can arise from prescription of multiple medications to certain patients.

Specifically, the research aims to: Identify which patients are at the highest risk of important adverse drug events Discover which medications are most strongly associated with adverse drug events Explore whether stopping medications is associated with greater benefit or harm in patients at high risk of adverse drug events: focusing on delirium Methods This proposal will utilise anonymised data from the Clinical Practice Research Datalink (CPRD) and involve the following work packages: Work-package 1 will develop and externally validate two clinical prediction models to identify individuals at the highest risk of developing delirium or experiencing falls.

Work-package 2 will use the same dataset to examine the association between the top 25 most commonly prescribed medication classes in primary care, and delirium or falls.

Analyses will utilise multivariable Cox regression and supervised deep learning models designed to uncover previously unidentified associations between medication changes over time (including different combinations of medications) and delirium/falls. Work-package 3 will examine the effects of medication discontinuation.

Analyses will focus on patients undergoing medication reviews in primacy care and outcomes in those who had medications discontinued and those who did not.

A large number of medications, outcomes and analytical methods will be explored using a rigorous, pre-specified and transparent approach. Timelines The first year will involve obtaining approvals, curating variables, and managing the data. In years 2-3, I will develop and externally validate two prediction models.

From years 2-4, I will examine associations between specific medications and delirium.

In years 4-5, I will undertake the large-scale causal inference study examining the effects of medication discontinuation.

Anticipated impact and Dissemination The findings of this study will be communicated through scientific journals and conferences, patient summaries, a bespoke website, social media, and community engagement events.

All outputs will be made freely available, and tools will be developed for integration into routine electronic health record systems.