
Automating Adverse Event detection with AI/ML and aiding to timely FDA reporting of adverse events

The current system of identifying adverse events from customer complaints is labor intensive and time consuming. FDA guidelines require health care providers and manufacturers to submit voluntary reports of adverse events associated with products within 15 days of complaint. However, the main challenge companies face today is to swift through the massive volumes of the complaints which require highly trained professionals to review, validate and identify the adverse events to complete FDA reporting of adverse events in timely fashion (<15 days of complaint).
Goal: A fortune 100 Pharma/MD&D client approached Intuceo® team to automate and improve the process efficiency of detecting adverse events leveraging AI/ML.
Solution: Applying AI effectively to determine whether a complaint is potentially an adverse event or not requires not only base level prediction capability but also ability of the AI model to rationalize its prediction. A Yes/No decision of AI model may help in reducing the volume of complaints, but it doesn’t eliminate the painstakingly time-consuming efforts from experts in manually annotating the explanation to why that Event is an Adverse Event or not. While most AI cognitive models provide base level Prediction capabilities, they seldom come with “Explainable AI intelligence”. Intuceo® team of experts along with patented Intuceo® Auto Ml tools for pattern recognition and predictive models, partnered with the client and implemented a AI/ML solution that accomplished both the goals, “Accurately identify potential Adverse Event” along with rationale on WHY with the explainable AI feature.
- Any adverse event is reportable to FDA if the event: Is fatal, is life-threatening, is permanently or significantly disabling, Requires or prolongs hospitalization, Causes a congenital anomaly
- These adverse events must be reported to FDA as soon as possible but no later than within 15 calendar days following the initial receipt of the information

Results: Having an AI/ML driven solution with both Adverse event detection and effective explanation resulted in high throughput to process more complaints in given time window, significantly reducing the expert professional’s time and number, resulting in timely FDA reporting of adverse events.