Objective

Post-market adverse drug reaction surveillance is a mandatory regulatory responsibility of pharmaceutical companies. The adverse reaction data is hidden in the scientific content published in research journals.

Thus, a leading pharmacovigilance platform required labeled datasets for training AI models based on this information in order to help its clients improve their drug safety workflow processes.

Solution

• Innodata setup a team of bio-medical subject matter experts to label unstructured data; the adverse drug reactions are published across various journals and written as scientific research abstracts.
• Innodata’s medical experts annotated the abstracts from scientific journals broken into individual words with a customized labeling application.
• The SMEs studied the abstracts and labeled individual words to a controlled ontology of medical terms to create an annotated dataset for machine learning.

Results

• Innodata’s labeled datasets have helped the company build machine learning models that predict adverse drug events with high accuracy and enables pharma companies to navigate the increasingly complex and ever-changing regulatory environment.
• Today, our client can confidently offer its customers a fully compliant, configurable, easy-to-adopt and audit-ready solution for scientific literature monitoring and pharmacovigilance.