Drug Safety invites the submission of original research articles (up to 6000 words) on the Role of Artificial Intelligence/Machine Learning in Pharmacovigilance for a themed issue of the journal to be published in 2021. The issue is guest edited by Dr. Andrew Bate (GSK) and Dr. Yuan Luo (Northwestern University). With a mix of invited review articles and original research articles, the theme issue will provide a comprehensive coverage of how emerging technologies can affect work tasks and further improve the field of pharmacovigilance.
We welcome proposals for original research articles on the following topics:
- Methods for identifying safety signals in any or combinations of data source
Author Brief: The article will showcase novel research on how pharmacovigilance can benefit from the continuously increasing number and form of data sources, especially electronic health record data and patient self-tracked/self-collected data. In addition, the article will discuss how machine learning has helped improve identifying safety signals from these data sources for pharmacovigilance.
- Prediction, in future numbers of adverse events or types of safety outcomes, or change in severity in RCT, PSPs, RWD or other data sources e.g. prospective surveys
Author Brief: The article will showcase novel machine learning methods for predicting outcomes such as future numbers of adverse events or types of safety events, or change in severity in randomized controlled trials (RCTs), patient support programs (PSPs), real world data (RWD) or other data sources e.g. prospective surveys. Of particular interests is how the methods will account for longitudinality and missing data from the above data sources for pharmacovigilance prediction.
- Visual pattern recognition in safety
Author Brief: This article will describe the novel application of visual pattern recognition to support signal detection in pharmacovigilance.
- Developing test sets for assessing machine learning algorithmic performance
Author Brief: The article will showcase efforts for developing test datasets for assessing machine learning performance in pharmacovigilance tasks such as predicting future numbers of adverse events or types of safety events. The article will need to demonstrate rigorous annotation process e.g., inter-annotator agreement evaluation and quality control. Of particular interests is the actual release of a dataset for pharmacovigilance prediction.
- Developing frameworks for increasing transparency or explainability of machine learning outputs (including visualization) as well as assessing how to best implement into routine use
Author Brief: The article will showcase efforts for developing novel methods to assess and ideally increase the transparency of machine learning models in predicting pharmacovigilance outcomes either at systemic cohort level or at individual patient level. Adoption and/or implementation of such efforts in real world pharmacovigilance is a plus. Integration of multiple data sources to collate on each other and improve transparency is also a plus. Of particular interests is the open-source release of such a system.
- Methods work in building capabilities to facilitate case processing and intake
Author Brief: Both regulators and pharmaceutical companies have very complex processes with many steps to get reports into their databases. The article will focus on robotic process automation and machine learning-based approaches to make report entry more efficient and effective.
- Approaches for assessing quality, reliability, credibility and trust in machine learning outputs
Author Brief: The article will showcase efforts for evaluating quality and reproducibility of machine learning models in predicting pharmacovigilance outcomes either at systemic cohort level or at individual patient level. Evaluation on a large multi-institute dataset is ideal. Detailed error analysis on generalizability and reproducibility is a plus. Of particular interests is the open-source release of such a system to establish some “industrial standard”.
Please submit an abstract describing your proposed paper by 30 April 2021 to email@example.com. Full papers will be invited by 30 May 2021 and manuscripts due by 31 August 2020.