Blog
Automating Safety Case Processing with GenAI
Uwe Trinks, Ph.D, Practice Leader, IQVIA Technology Solutions
Oct 24, 2024

The world of pharmacovigilance, the crucial process of monitoring adverse drug reactions, is primed for transformation with new developing technologies. Currently, case processing relies heavily on manual data collection. This leads to lengthy timelines and potential inconsistencies. However, a form of advanced artificial intelligence (AI) known as generative AI (GenAI) holds immense promise for streamlining and enhancing this process. Uwe Trinks, Global Practice Lead, PV Technologies at IQVIA, recently spoke on this topic at the Drug Information Association meeting in San Diego.


Implementing GenAI into Pharmacovigilance Case Processing

Currently, when life science organizations collect pharmacovigilance data, it's mostly initiated passively, resulting in incomplete collection. Follow-up queries focusing only on crucial database fields that lack information, can be automated, however, those data searches also often come up short. With the capabilities of AI, IQVIA hypothesizes that queries could be much more holistic and data could be completed with derived information based on multiple other sources.

For example, if a patient reports a stroke, organizations require additional information about the patient to determine the cause and relationship of the stroke. Case processors would need access to the patient’s family history of strokes. They would inquire if the patient had experienced any cardiovascular events, if they have high cholesterol or if they are at risk of a heart attack. With the help of AI or GenAI, this information could be sourced from other patient documentation. This would then provide enough information to better determine the event and its relation to the drug's safety.


Current Case Processing Procedures

Current case processing procedures are supported by a significant amount of traditional code, which relies heavily on human intervention, increasing timelines and adding additional verification steps.

The first step of case processing can be automated using the Detect module of IQVIA Vigilance Platform, which utilizes AI to detect adverse events from large data and voice files. Though incredibly effective, this is only one of the required steps in the case processing protocol. Various additional steps, such as source document translation, case extraction, event coding and case assessment, currently use traditional code with the assistance of human verification to process these potential safety events.

The next step, source document translation and personal identification information redaction, is not currently fully automated with GenAI. Due to the language handicaps of the technology at this stage, it cannot always fully process languages. At the same time, the quality checks at current rates sometimes take longer than building the case itself. As technology develops and source documents increase, there is greater potential for automation of document translation and redaction.

For the validity of data, the current code only examines for missing data in empty fields, rather than the entire case. With the implementation of GenAI, there is potential to look at the case, compare it with other cases that have similar product event combinations and holistically examine the event. Then, the intelligence could determine which questionnaire to send to the patient to source targeted questions for certain diseases. This helps to rule out any anomalies that could be interfering with the detection of an adverse event.

One of the core steps of case processing, case assessment and causality, is the deciding factor to finding whether the event is serious, previously known, or related to the product. This step, currently carried out by nurses, pharmacists and doctors, is particularly challenging to replace with software, but in time, it may be accomplished. Then, the final step would be human verification in medical review, added by a graphic interface presenting a holistic view of the entire case.

The entire case processing workflow currently takes about five days because of many handovers between specialized personnel. If organizations were able to automate this workflow with the help of GenAI and human verification, this process could be reduced to less than a day. Streamlining this process to produce globally reported outcomes in a single day would drastically reduce the cost associated with this process and improve the quality of assessments.

As detailed above, most of case processing is dependent on manual work or traditional code. However, moving forward, IQVIA has assessed that GenAI could be rapidly implemented in about half of all case processing steps. This would include source document translation, case extraction, validation of missing data, active query, case assessment and causality as well as narrative and comments creation.


New GenAI Governance Regulations

One challenge when using GenAI is our inability to traditionally validate outcomes or data points. Because GenAI is constantly evolving and adapting, there is no feasible way to traditionally validate the code.  This implies that organizations will need to have a flexible and adaptable governance plan to account for the evolving nature of GenAI.

The Food and Drug Administration (FDA) is scheduled to release a guidance document later this year, based on Executive Order 14110 which contains over 150 detailed requirements on the use of GenAI. The FDA’s “Consideration for the Use of Artificial Intelligence to support Regulatory Decision Making for Drugs and Biological Products” will also include guidance specific to the implementation of GenAI in drug safety processes. As more federal regulations are slated for release worldwide, organizations must detail and specify the use GenAI in processing drug safety events.


The Future of Safety Automation

While we may still be a few years away from this technology being implemented to its fullest capabilities, we remain optimistic about its potential to streamline and improve case processing. Adhering to regulations in the face of this development will be critical to success in pharmacovigilance safety detection. If you are interested in learning more about the potential of GenAI in automating case processing, please contact SafetyPV@iqvia.com.

Related solutions

Contact Us