Table 3 Recommended approaches to address challenges in utility of text narratives for research and practice

Approach 1:

Employ data science methods for large textual data to improve case identification and characterization of contexts around deaths

Use topic modeling to characterize salient risk factors and circumstances around violent deaths

Use natural language processing algorithms in addition to traditional textual analysis methods to optimize case identification using narratives, particularly for rare and emergent risk factors/outcomes

Broadly employ algorithm-based data science tools as an alternative to less efficient methods such as manual review to increase sample size and improve reliability of inferences

Approach 2:

Leverage new and existing linkages to data sources beyond NVDRS

Link to data sources beyond NVDRS to obtain additional circumstances and risk factors (e.g., population characteristics from US Census and health history from medical records)

Supplement case finding with other publicly available data sources, including media reports (e.g., The Washington Post database for fatal police shootings), Vital Statistics, and other violent death databases (e.g., Federal Bureau of Investigation's Uniform Crime Reporting Program)

Construct comparison groups matched to NVDRS/VDRS decedents from large population-based surveys, such as the National Survey on Drug Use and Health, National Health Interview Survey, and Youth Risk Behavior Surveillance System, to strengthen the ability to make inferences about risk and protective factors

Approach 3:

Address inconsistencies and improve data integration within the NVDRS

Provide concrete guidance to address discrepancies between coroners/medical examiners and law enforcement narratives, as well as between abstractor-coded variables and text narratives

Identify and investigate how discrepancies across different sources (death certificates, coroners/medical examiners reports, law enforcement reports) arise

Strengthen state and federal death investigation systems to improve the quality of the NVDRS

Improve transparency in data quality assurance procedures and results

Approach 4:

Engage in dialogue with NVDRS creators and establish best practices for maximizing the research potential of text narratives

Conduct focus groups with users and creators of the NVDRS/VDRS narratives to identify opportunities for data retrieval and sharing that maximizes the research usefulness of the NVDRS narratives

Encourage meaningful collaboration among NVDRS/VDRS data users as well as between data users and creators

Facilitate partnerships/collaborations between stakeholders across states and jurisdictions on death investigations, data collection, and abstraction

Transparent documentation of data completeness of the text narratives (e.g., report missingness of narratives in published studies, examine systematic differences in length and content of narratives by characteristics of decedents, etc.)