FDA RFI: Next-Generation Imports Screening Decision Support Risk Targeting System

Notice ID:  FDA-RFI-75F40124Q00178

The purpose of this RFI is to broaden FDA’s understanding of the market’s capability, either as solutions requiring potential development of a flexible system using current technologies, or existing commercially available applications, augmented by artificial intelligence and machine learning (AL/ML), to provide a next generation imports screening system utilizing big data.

Electronic data on imported products is received by FDA via the Automated Customs Environment (ACE) which is maintained by Customs and Border Protection (CBP). Shipments are transmitted to FDA as “entries”, each of which may contain multiple “entry lines” of FDA‐regulated products. FDA has processed over 50 million imported entry lines in Fiscal Year 2023. Over the last decade the agency witnessed a steady 5% annual increase in the number of entry lines of imported FDA‐regulated products. FDA’s Office of Regulatory Affairs (ORA) is in‐charge of import operations which is one of its key functional areas. ORA is responsible for screening imported FDA regulated products that are entering the country, so any highrisk products that can potentially cause health risk to the public are identified before they are released to the domestic supply‐chain. FDA receives large quantities of product and product‐associated information, and, upon receipt in the form of import entries, a best course of action needs to be decided to protect public health. The legacy FDA imports systems handling these data are generally siloed and limited, with the creation of interfaces or expansions being time‐consuming.

The PREDICT component of FDA Imports provides screening support for Customs entry reviewers for the FDA.  It serves to advise entry reviewers of action to take on imported commodities based on the standard business processes of the FDA.  It is integrated with the Entry Review application which allows entry reviewers to view detailed information about an entry.

Entries are sent to PREDICT by the OASIS (Operational and Administrative System for Import Support) PREDICT Interface (OPI).  PREDICT processes the entries against thousands of rules in its system, which determine the risk level of the entry.  The results are then sent back to OPI and subsequently to the FDA Imports Entry Review application where entry reviewers are presented with the results.

A web‐based user interface enables users to manage the rules, which control the business decisions that PREDICT makes when determining how to handle entries.

There are various rules in PREDICT that are used to decide the systems’ final recommendations which are maintained in a variety of ways.  Part of the PREDICT user interface is the Rule Studio, which provides the ability to manage the rules based on evolving business requirements.  Monthly, the Track Record and Data Anomaly rule sets are manually updated by a system administrator based on historical data. Other rule sets such as those classified as “expert” rules are manually updated on an ad‐hoc basis. PREDICT is comprised of the following four major applications:

1. Administration: The Administration application is used to manage users and roles in the system. In addition, it has the capability to manage organizations, contacts, and address information.
2. PREDICT: The PREDICT application allows users to view the detailed results of all requests sent from OPI. Rule maintenance is performed in this module and rule engine management is performed within this application.
3. Monitor: The Monitor application is a Business Activity Monitoring and Business Process Automation solution. It includes a complete business intelligence dashboard.
4. Auditor: The Auditor application enables critical auditing functionality of business data.  Given advancing technologies, it is sought to transition to more flexible decision support systems for screening these entries that are capable of handling large amounts of data while supporting the general operational processes already in place.

The next generation Decision Support System (DSS) shall have the following capabilities:

1. inputs and warehouses, structured /semi‐structured / un‐structured big data streams, from both private and public data sources.
2. includes decision support features, where needed, that can provide descriptive, prescriptive, and predictive analytic functionalities: a. application development and data sharing via application programming interfaces (API) b. data analysis c. forecasting d. decision tree analysis e. performance metrics f. rules‐based workflow g. sensitivity analysis h. version control i. thematic mapping j. Monte Carlo simulation k. Artificial Intelligence / Machine Learning modeling
3. provides a proof‐of‐concept (POC) environment for testing.
4. provides a secure private or hybrid platform that will support both a rules engine and artificial intelligence capabilities.
5. provides for archiving, tracking, reporting, and reversibility/portability of stored data …

Read more here.

🔊 Listen to Story