Smart event encoding

Summary

Event auto-encoding using AI/ML in Argus Safety (Enhancement 36147492)

Description

In this release, Argus Safety introduces the Smart Encoder, an AI/ML-powered feature that enhances the auto-coding process for medical events. The Smart Encoder uses advanced language processing and pattern recognition to accurately interpret medical event descriptions, even with variations in terminology, spelling, or phrasing.

A new profile switch, Smart - Event Encoding, is introduced under Argus Console > System Management (Common Profile Switches) > Case Form Configuration > Modify Case Form Configuration > Auto-Encoding, Dictionary & Central Encoding. This switch enables you to set a customizable threshold for the auto-coding process. The Smart Encoder output is linked to a confidence level; when this confidence meets or exceeds the threshold, the system automatically codes the event with the correct MedDRA term.

For matches that do not meet the threshold, the Smart Encoder offers a selection of suggested terms for the case processors to choose from. If multiple matches are found with top confidence, a blue i icon appears for you to review the entry, allowing any necessary adjustments to the coded term.

To enable Smart Event - Encoding, you must upload the Smart Event Encoding Model. The smart encoder relies on MedDRA terminology to accurately code verbatim. After each MedDRA release, a new Smart Event Encoding Model will be provided and made available for download in the zip format. For more information, see:

• Oracle Argus Safety Administration Guide
• Oracle Argus Cloud Service Administration Guide

To upload the Smart Event Encoding Model, a new dictionary type option, Smart - Event Encoder Dataset, is added in the self-service dictionary loader. Oracle recommends to apply the latest Smart Event Encoding Model to ensure optimal coding accuracy.

Note:

The smart MedDRA term search using the Smart Event Encoding feature is for evaluation purpose only. Hence, it is disabled by default.

When to enable this feature

To make sure the search results generated via this feature meet your quality standards, Oracle recommends you perform thorough testing and evaluation with your real-world data to assess its performance in the test environment. This will help you understand how the feature works and identify any areas for improvement. You can compare our AI results with your own quality standards.

How it works

Raise a ticket with Oracle Support to get the following information to proceed further with testing and evaluation:

• Guidelines on what to expect.
• Key metrics to measure performance.
• Tips for debugging.
• Model Card details, including the model's intended use, performance, limitations, and AI governance.

What next?

When the search results generated via Smart Event Encoder are validated and meet your expectations, the product team will work with you on the next steps to deploy this feature for product usage.

We will also provide a formal validation pack when the feature is fully released to make sure it meets all regulations and is based on the real-world performance data. We are committed to working with you to make sure our AI feature generate optimal suggestions.