top of page

ChatGPT could optimize the generation of ICSRs

Updated: Jan 27

The processing of ICSRs (individual case safety reports) is a critical component of PV.

Yet delivering natural human-like text summarizing the vast volumes of reports submitted by healthcare professionals, patients, or other individuals who have observed or experienced an adverse reaction to a medication, can be a challenge.

The newly launched ChatGPT may help the process.

A Promising Addition

According to PV Consultant and Business Analyst at Insife Ronan Wheeler OpenAI’s ChatGPT may further optimize the PV process:

“NLP is a technology with a constant curve of impressive advancement. At Insife we have already introduced the SPARK NLP ML/AI which is the most widely used NLP library in the world, as part of HALOPV. Supplementing this with ChatGPT’s ability to summarise large amounts of text into natural human-like text could strengthen the final generation of ICSRs and potentially assist with the generation of aggregate report analysis”, says Ronan Wheeler, elaborating:

“SPARK is still the most comprehensive NLP technology in the market and the most accurate in terms of solving classification problems. However, using ChatGPT developed by the open-source AI research and deployment company OpenAI, it may be possible to optimize the process of generating the ICSRs by feeding the system relevant data, such as patient medical histories, prescribed medications, and adverse reactions. The system could then use this data to generate structured ICSR narratives in a recipient-friendly language, including relevant details such as patient demographics, medication details, and descriptions of the adverse reaction”.


SPARK NLP ML/AI has features such as word segmentation, sentence detector (DL models), coreference resolution, unsupervised keywords extraction, language detection & identification (up to 375 languages), multi-class text classification, multi-label text classification, named entity recognition and much more.

Read more about HALOPV using SPARK NLP ML/AI here

bottom of page