Background: The term “polypharmacy” is used to describe the prescription/usage of multiple medications. Many definitions have been established for polypharmacy; however, polypharmacy is usually defined based on a single day, despite it naturally changing over time. Studying the variation of polypharmacy over time instead of a single value could be useful in understanding its long-term effects on health.
Objectives: To characterise changes in polypharmacy over time, to describe patterns/trajectories, and to understand the association between these patterns and overall mortality.
Design: We will use previously recorded GP records from people aged 50 and above in the UK Clinical Practice Research Datalink (CPRD) AURUM and GOLD. We will measure variation of polypharmacy over a five-year period using novel longitudinal statistical methods.
Applicability: After describing and characterising polypharmacy, trends over time and their association with mortality in older and middle-aged people will be studied. The results of the study will enable NHS staff to identify people with polypharmacy, and to design and use treatments for those who need them the most.
Polypharmacy has been defined using cross-sectional thresholds, like for example the use of 10 or more medicines in a given year. Longitudinal data allow for a more accurate alternative definition, incorporating the element of time and trajectories over time.
We will perform a population-based network cohort study using CPRD (AURUM and GOLD) mapped to the Observational Medical Outcomes Partnership (OMOP) common data model. In this study, we aim to evaluate the long-term use and effects of multiple concomitant treatments, also called polypharmacy; and to assess the association between polypharmacy and mortality. We also aim to assess whether polypharmacy in middle-aged people develops similar to those in the older population, has a similar association with mortality and is predictive of polypharmacy trends in later life.
We will characterise polypharmacy over time in the older population (65 or older). Participants will be followed from 01 January 2015 until 31 December 2019. Polypharmacy will be calculated annually, based on the total number of drugs/ingredients prescribed every year during the follow-up period. Changes over time will be modelled using an unsupervised machine learning method; k-means for longitudinal data in patients with complete observations to capture the existing polypharmacy trajectories over time. Then, joint models and latent class joint models will be used for all patients (complete cases and censored). These models incorporate both longitudinal trajectories and survival analysis into one model while allowing for censoring. All-cause mortality will be inserted as the outcome for both models and association between polypharmacy trajectories and death will be assessed. The analysis will be run in CPRD GOLD first and replicated/validated in CPRD AURUM. The same analysis will then be run for middle-aged population (50-64).
The main outcome of interest for this study will be all-cause mortality
Daniel Prieto-Alhambra - Chief Investigator - University of Oxford
Leena Elhussein - Corresponding Applicant - Nuffield Dept of Orthopaedics
Antonella Delmestri - Collaborator - University of Oxford
Edward Burn - Collaborator - University of Oxford
Patient Level Index of Multiple Deprivation;Practice Level Index of Multiple Deprivation