AI-Assisted Extraction of Circumstance Information from National Violent Death Reporting System Narratives

Overview

Development stage

In development or acquisitionSource: a) Pre-deployment – The use case is in a development or acquisition status.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

Not reported

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Health and Medical

Operational functions

No function tag matched Derived by keyword rules; see methodology

Problem the AI is intended to solve

The National Violent Death Reporting System (NVDRS) compiles both quantitative and qualitative data on the circumstances surrounding violent deaths, including homicides and suicides, from three key data sources related to each death: the death certificate, the coroner/medical examiner report (including toxicology results), and the law enforcement report from the law enforcement agency that investigated the death. Data abstractors in state health departments across the U.S. abstract relevant information about each death from CME and LE reports, creating narratives that describe the most notable circumstances that contributed to the deaths captured in the system. Thus, much of the valuable contextual information about these deaths, such as details about chronic pain, interpersonal arguments, or other contributing factors, is embedded in free-text narratives and is not routinely abstracted into structured quantitative data fields. Manually extracting this information is labor-intensive, time-consuming, and subject to variability. The problem we are addressing with AI is the automated extraction of specific circumstance information from these unstructured narratives, enabling more comprehensive and systematic data analysis.

Expected benefits

Automating the extraction of circumstance information from NVDRS narratives using AI brings several important benefits. First, it significantly reduces the time required to process narrative data. While manual abstraction is resource-intensive and can take hours or days to review thousands of records, AI can accomplish this task in a matter of minutes. This efficiency is especially critical given the scale of the challenge: the NVDRS captures data on over 70,000 violent deaths annually, making manual analysis of detailed free-text information in the CME and LE narratives for each of these incidents impractical. In addition to saving time, AI improves data quality and consistency by applying uniform criteria across all records, which helps to minimize human error and variability. Furthermore, by extracting additional details, such as information about chronic pain or the presence of arguments, AI enhances the surveillance capabilities of public health officials. This richer data enables a better understanding of risk factors and circumstances surrounding violent deaths, which in turn informs more effective prevention strategies. Collectively, these outcomes directly support CDC’s mission to strengthen public health surveillance, guide prevention efforts, and ultimately reduce the incidence of violent deaths.

System outputs

The AI system produces structured data outputs derived from the free-text narratives in the NVDRS. For each narrative, the system identifies and extracts predefined circumstance categories (e.g., presence of chronic pain, evidence of an argument, substance use) and outputs them as structured variables (e.g., binary indicators, extracted text snippets, or coded categories). These outputs can be integrated into existing NVDRS datasets, enabling further quantitative analysis and reporting.

Technology

AI classification

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

System name(s)

Not reported

Custom-developed code

Not reported

Public source code

Not reported

Sourcing

How it was built

Not reported

Vendor (as reported)

Not reported

Vendors (standardized)

None on the standard list

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

Data and privacy

Involves PII

Not reported

Privacy Impact Assessment

Not reported

Authorization to Operate

Not reported

Demographic features

Not reported

Training and evaluation data

Not reported

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