CRIS Blog: Answering real-world questions about medication and mental health through pharmacoepidemiology

This blog is written by Katrina Davis, a Research Associate at the BRC, she tweets from @KDavis_kasd2

Pharmacoepidemiology is the study of the use and effects of medication in large groups of people, where researchers use epidemiological methods, such as population cohorts to investigate pharmacological outcomes.

Earlier this year the @MQmentalhealth Data Science Group enabled me to meet with a group of researchers to discuss what was important for non-specialists to know about pharmacoepidemiology in the mental health field. The result was a paper that was recently published in Lancet Psychiatry.  Three of the authors are supported by the NIHR Maudsley Biomedical Research Centre.

Our paper has three important takeaway messages:

  • Firstly, while randomised-controlled trials may be gold-standard for researching medicine efficacy, pharmacoepidemiology research can offer a quick and relatively cheap alternative way to study medicines or supplement trial evidence with "real-world evidence". This is particularly the case for groups of people under-represented in trials (such as people from ethnic minority backgrounds and those with comorbid conditions) and for outcomes that are rare, complex or delayed.
  • Secondly, although there are potential problems with bias and confounding with these study methodologies, design and analysis can be planned using "causal inference" techniques to improve confidence in the findings.
  • Thirdly, while a data-rich environment provides feasibility for pharmacoepidemiology projects, for these projects to flourish, researchers also need to build trust – both with patients whose care produces data, and with policy makers who will interpret the evidence.

The Clinical Records Interactive Search (CRIS) platform was highlighted at multiple points in the article. For instance, the guiding principles of the governance model of CRIS (full-text, pseudonymised records with patient-led access committee), were held up as an example of good practise.

The use of text and natural language processing was demonstrated with a project that identified suicidality from free-text in order to identify this as a potential side-effect of antidepressant medication, and the use of data linkage to extend CRIS was demonstrated with a project that looked at whether stopping medication while pregnant affected the risk of relapse.

Furthermore, our paper predicted that the NIHR Maudsley BRC, as one of the Medical Research Council Pathfinder sites, might be involved in the next step of creating a UK-wide mental health platform to facilitate data science projects. The BRC will likely remain at the cutting edge of pharmacoepidemiology in mental health due to its wealth of data resources and skills ranging from patient involvement, to clinical knowledge, to informatics. However, continued engagement with wider colleagues from networks such as the MQ data science group is imperative, to enable us to continue to answer the questions that patients and clinicians have about medication and mental health in the real-world.

Tags: CRIS blog -

By NIHR Maudsley BRC at 11 Feb 2020, 12:06 PM

Back to Blog List