Development and validation of novel data science tools for quantifying drug exposure using routinely collected data

Marshall, Alex Douglas (2021) Development and validation of novel data science tools for quantifying drug exposure using routinely collected data. PhD thesis, University of Glasgow.

Full text available as:
Download (4MB) | Preview


Routinely collected healthcare data are increasingly being used as a source for research into the effectiveness and safety of drugs. Although these data have great potential, they require preparation before they can be used in research, a process which, amongst a number of other tasks, typically involves quantifying patients’ exposure to the drug of interest. The aim of this thesis was to develop and validate a set of flexible, reusable functions for generating common drug exposure variables based on routinely collected prescribing data.
Six main classes of method for quantifying drug exposure were identified through a review of pharmacoepidemiological research; ever use vs. never use, use at a specified time point, daily dose or duration, persistence and discontinuation, adherence, and population level measures. The information obtained on these methods, their applications and the potential variations within each class formed the basis for developing an R package, prescribeR, which contains a range of functions designed to simplify and standardise the generation of drug exposure variables, and to provide a structure for reporting how these variables were produced.
The utility of the package was then demonstrated by applying it to two exemplar clinical studies, using a cohort of 5,571 patients with epilepsy constructed using linked data within the NHS Greater Glasgow and Clyde Safe Haven environment. In the first, prescribeR was used to quantify persistence to anti-epileptic drugs over the first 365 days of follow-up for cohort patients in order to assess differences in persistence across different drugs, as well as to compare persistence in new and existing users and patients prescribed monotherapy and combination therapy. All of the required persistence measurements for this study were generated using the prescribeR package, highlighting the relative ease of generating exposure data for a large cohort of patients and a number of different drugs.
In the second, the package was used to examine the effects of adjusting drug exposure definition on the estimated number of patients exposed to various drugs, the estimated exposure durations. The association between levetiracetam exposure and all-cause mortality was estimated using a range of time-fixed and time-varying exposure definitions, and a wide range of hazard ratios and significance levels were observed across the resulting models, highlighting that the selected definition of drug exposure can potentially have a large impact on the results observed in clinical research.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Keywords: pharmacoepidemiology, data science, routinely collected data, r, drug exposure.
Subjects: R Medicine > RA Public aspects of medicine
R Medicine > RA Public aspects of medicine > RA0421 Public health. Hygiene. Preventive Medicine
Colleges/Schools: College of Medical Veterinary and Life Sciences > Institute of Health and Wellbeing > Public Health
Funder's Name: Medical Research Council (MRC)
Supervisor's Name: McCowan, Professor Colin and Pell, Professor Jill
Date of Award: 2021
Depositing User: Dr Alex D Marshall
Unique ID: glathesis:2021-81982
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 02 Mar 2021 08:33
Last Modified: 02 Mar 2021 08:41
Thesis DOI: 10.5525/gla.thesis.81982

Actions (login required)

View Item View Item


Downloads per month over past year