1CDP (SEDRIC) AIP for Advanced Foodborne Outbreak Investigation (AI Summarization and Receipt Reading)

Overview

Development stage

DeployedSource: c) Deployed – The use case is being actively authorized or utilized to support the functions or mission of an agency.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

2023-10 (month precision)Source: Oct-23

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Emergency Management

Operational functions

Summarization, Clinical and medical Derived by keyword rules; see methodology

Problem the AI is intended to solve

This AI solution impacts the process of investigating foodborne disease outbreaks. These foodborne outbreaks require cooperative efforts from CDC staff, FDA, USDA, and local agencies and the AI system is used through a centralized data platform System for Enteric Disease Response, Investigation, and Coordination (also known as SEDRIC). For more information on SEDRIC, please go to our website: https://www.cdc.gov/foodborne-outbreaks/php/foodsafety/tools/index.html

Expected benefits

SEDRIC's AIP use case provides CDC epidemiologists the ability to accelerate their investigations of multi-state foodborne disease outbreaks by more effectively leveraging data available in a rich data source, such as receipts from grocery stores, that otherwise requires extensive time and human effort to parse through. In addition, this workflow would free up epidemiologists' time and, potentially, increase the frequency with which both CDC and STLT partners could utilize shopper card, receipts, and free text responses to support investigations. There are two main expected benefits from this use case. The first addresses manually entering receipt information, shopper card information, or other such free text field is traditionally error prone and time intensive for a variety of information. This AI system provides a human-in-the-loop opportunity to review and update data entry points while reducing time spent by staff to gain these insights. Having a set structured output as well increased standardization of this information and eases reporting in situations requiring cooperation from multiple organizations. The second benefit revolves around using the summarization capability, the extensive process of mapping common names of different food items is done automatically, greatly reducing the human labor time to generate dashboards of information regarding current foodborne investigations to serve as a decision point to aid in outbreak response.

System outputs

The Artificial Intelligence Platform (AIP) available within SEDRIC provides CDC epidemiologists the power to accelerate their investigations of multi-state foodborne disease outbreaks. It can extract structured data from grocery receipts, shopper card records, and free-text responses in order to catalog the food items purchased by affected patients. It can also map those items to SEDRIC-defined vehicles which categorize the items and highlight commonalities across patients, helping to pinpoint potential outbreak vehicles. AIP can summarize these results to provide insights from information pulled from shopper receipts. Given ingredients can be found in multiple food products, and some ingredients such as herbs like coriander/cilantro may go by multiple names or be reported in multiple languages, this summarization tool provides a faster way to gather summary information from receipts on different food items which may be part of a foodborne investigation.

Technology

AI classification

Natural language processingSource: Natural Language Processing: AI that processes, interprets, and shares information in human language.

System name(s)

1 CDC Data Platform (1CDP)

Custom-developed code

Yes

Public source code

Not reported

Sourcing

How it was built

Contract and in-houseSource: c) Developed with both contracting and in-house resources

Vendor (as reported)

Palantir Technologies

Vendors (standardized)

Palantir

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

No

Privacy Impact Assessment

Not reported

Authorization to Operate

Yes

Demographic features

k) None of the above

Training and evaluation data

Data are used in outbreak/response scenarios, such as foodborne illness outbreak response. Data used is dependent on the situation and outbreak, and may be owned by CDC, FDA, USDA, State Health Departments, Tribal Health Departments, Local Health Departments, Territorial Health Departments, or other entities.

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported