Mobile Menu

AI and automation across the pharmacovigilance lifecycle

Despite technological advancements within data capture, storage, linking and analysis, little attention has been paid to the impact of these technologies on routine pharmacovigilance. We summarise a recent review, published in Drug Safety, that discussed emerging research in the use of artificial intelligence (AI), machine learning (ML) and automation across the pharmacovigilance lifecycle.


Pharmacovigilance is the practice of monitoring the effects of medical drugs after they have been licensed for use. This is important to identify and evaluate previously unreported adverse drug events. We currently have more data than ever, from which we can gain further scientific knowledge. Coupled with advances in computational power, this data has increased our capacity to innovate. As a result, there has been intense discussion of using AI/ML and automation within pharmacovigilance. This has garnered a lot of interest as routine pharmacovigilance has not changed for many decades.

The current role of AI/ML in healthcare

There is a wide range of applications for AI/ML within healthcare to gain new insights into disease and improve healthcare delivery. AI/ML is mainly having an impact on repetitive data-rich tasks. For example, ML has been used to analyse images from mammograms with levels comparable to radiologists. Other examples include predicting asthma exacerbations, monitoring insulin through mobile phone apps and monitoring adherence with anticoagulation therapy. Other areas of relevance include the use of ML to more effectively identify and potentially recruit patients into randomised clinical trials.

AI/ML and automation in pharmacovigilance

The use of AI/ML and automation in pharmacovigilance is not new. For example, in the early 1990s, neural networks were trained on side-effect profiles to distinguish between tricyclic antidepressants and selective serotonin reuptake inhibitors. ML is currently being explored in toxicology and also the understanding of safety in early pre-clinical drug development. Here are just some potential uses of AI/ML and automation in pharmacovigilance:

  • Automation of work in case intake, processing and reporting of spontaneous reports
  • Detecting useful patterns/trends in pharmacovigilance data
  • Extracting underlying medical concepts to support poor safety surveillance
  • Extracting meaning or concepts from free text narrative
  • Converting handwritten text to machine readable text for pattern discovery
  • Classifying disease/outcomes from pictures or scans
  • Data linkage
  • Better characterisation, prediction and prevention of adverse events


Emerging literature on pharmacovigilance and AI/ML algorithms are showing promise. However, there remains challenges in how this would be translated into routine effectively. Below are some of the challenges for widespread adoption:

  • Ensuring credibility and interpretability of outputs
  • Ensuring there is enough training and test data for developing effective ML algorithms
  • Limited standards/reference sets for testing and training algorithms
  • AI/ML-specific issues such as algorithmic transparency, replicability of findings, heterogeneity across approaches, workflow, skill set needs, policies and procedures


AI/ML and automation are tools that can enable more effective pharmacovigilance. However, widespread adoption of AI/ML and other technological advances will require the use of the right data methods, tools and processes in order to fulfil their true promise. Additionally, realistic expectations and transparent collaborative efforts will be important to overcome current challenges and ensure widespread use of these technologies within routine pharmacovigilance.

Image credit: By –

Share this article