Deep Neural Networks for Fine-Grained Surveillance of Overdose Mortality

Patrick J. Ward; April M. Young; Svetla Slavova; Madison Liford; Lara Daniels; Ripley Lucas; Ramakanth Kavuluru

doi:10.1093/aje/kwac180

Deep Neural Networks for Fine-Grained Surveillance of Overdose Mortality

Patrick J. Ward, April M. Young, Svetla Slavova, Madison Liford, Lara Daniels, Ripley Lucas, Ramakanth Kavuluru

Research output: Contribution to journal › Article › peer-review

3 Scopus citations

Abstract

Surveillance of drug overdose deaths relies on death certificates for identification of the substances that caused death. Drugs and drug classes can be identified through the International Classification of Diseases, Tenth Revision (ICD-10), codes present on death certificates. However, ICD-10 codes do not always provide high levels of specificity in drug identification. To achieve more fine-grained identification of substances on death certificate, the free-text cause-of-death section, completed by the medical certifier, must be analyzed. Current methods for analyzing free-text death certificates rely solely on lookup tables for identifying specific substances, which must be frequently updated and maintained. To improve identification of drugs on death certificates, a deep-learning named-entity recognition model was developed, utilizing data from the Kentucky Drug Overdose Fatality Surveillance System (2014–2019), which achieved an F1-score of 99.13%. This model can identify new drug misspellings and novel substances that are not present on current surveillance lookup tables, enhancing the surveillance of drug overdose deaths.

Original language	English
Pages (from-to)	257-266
Number of pages	10
Journal	American Journal of Epidemiology
Volume	192
Issue number	2
DOIs	https://doi.org/10.1093/aje/kwac180
State	Published - Feb 1 2023

Bibliographical note

Publisher Copyright:
© The Author(s) 2022. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved.

Funding

The Kentucky Injury Prevention and Research Center (KIPRC) is not the owner of the data used in our study; the data is owned by the Kentucky State Office of Vital Statistics (OVS). In accordance with KIPRC’s Memorandum of Understanding with the Kentucky OVS, we are legally prohibited from releasing line-level death certificate data. Death certificates contain identifying information (including name, location of residence, and social security numbers), and in some cases the free text describing how an individual died can be identifying. Each state houses their own OVS, and data may be requested for research purposes from this body; the address for the Kentucky Office of Vital Statistics is 275 E. Main St, 1E-A, Frankfort, KY 40621, and data may be requested from here for research purposes. Additionally, the National Center for Health Statistics has recently made available a death certificate literal-text file, which can be requested and contains the free-text cause-of-death information for all US resident deaths: https://www.cdc.gov/rdc/b1datatype/rdcltf. html. Documented code for the methods described are available at https://github.com/pjward5656/DC_flair. We thank the Kentucky Office of Vital Statistics for providing the data for this study.

Funders	Funder number
Kentucky Injury Prevention and Research Center
Kentucky OVS
Kentucky Office of Vital Statistics	1E-A
Kentucky State Office of Vital Statistics

Keywords

deep learning
drug overdose
machine learning
surveillance

ASJC Scopus subject areas

General Medicine

Access to Document

10.1093/aje/kwac180

Cite this

@article{fb035fa310cb4b15b6277594ea90e5d3,

title = "Deep Neural Networks for Fine-Grained Surveillance of Overdose Mortality",

abstract = "Surveillance of drug overdose deaths relies on death certificates for identification of the substances that caused death. Drugs and drug classes can be identified through the International Classification of Diseases, Tenth Revision (ICD-10), codes present on death certificates. However, ICD-10 codes do not always provide high levels of specificity in drug identification. To achieve more fine-grained identification of substances on death certificate, the free-text cause-of-death section, completed by the medical certifier, must be analyzed. Current methods for analyzing free-text death certificates rely solely on lookup tables for identifying specific substances, which must be frequently updated and maintained. To improve identification of drugs on death certificates, a deep-learning named-entity recognition model was developed, utilizing data from the Kentucky Drug Overdose Fatality Surveillance System (2014–2019), which achieved an F1-score of 99.13\%. This model can identify new drug misspellings and novel substances that are not present on current surveillance lookup tables, enhancing the surveillance of drug overdose deaths.",

keywords = "deep learning, drug overdose, machine learning, surveillance",

author = "Ward, \{Patrick J.\} and Young, \{April M.\} and Svetla Slavova and Madison Liford and Lara Daniels and Ripley Lucas and Ramakanth Kavuluru",

year = "2023",

month = feb,

day = "1",

doi = "10.1093/aje/kwac180",

language = "English",

volume = "192",

pages = "257--266",

number = "2",

}

TY - JOUR

T1 - Deep Neural Networks for Fine-Grained Surveillance of Overdose Mortality

AU - Ward, Patrick J.

AU - Young, April M.

AU - Slavova, Svetla

AU - Liford, Madison

AU - Daniels, Lara

AU - Lucas, Ripley

AU - Kavuluru, Ramakanth

PY - 2023/2/1

Y1 - 2023/2/1

N2 - Surveillance of drug overdose deaths relies on death certificates for identification of the substances that caused death. Drugs and drug classes can be identified through the International Classification of Diseases, Tenth Revision (ICD-10), codes present on death certificates. However, ICD-10 codes do not always provide high levels of specificity in drug identification. To achieve more fine-grained identification of substances on death certificate, the free-text cause-of-death section, completed by the medical certifier, must be analyzed. Current methods for analyzing free-text death certificates rely solely on lookup tables for identifying specific substances, which must be frequently updated and maintained. To improve identification of drugs on death certificates, a deep-learning named-entity recognition model was developed, utilizing data from the Kentucky Drug Overdose Fatality Surveillance System (2014–2019), which achieved an F1-score of 99.13%. This model can identify new drug misspellings and novel substances that are not present on current surveillance lookup tables, enhancing the surveillance of drug overdose deaths.

AB - Surveillance of drug overdose deaths relies on death certificates for identification of the substances that caused death. Drugs and drug classes can be identified through the International Classification of Diseases, Tenth Revision (ICD-10), codes present on death certificates. However, ICD-10 codes do not always provide high levels of specificity in drug identification. To achieve more fine-grained identification of substances on death certificate, the free-text cause-of-death section, completed by the medical certifier, must be analyzed. Current methods for analyzing free-text death certificates rely solely on lookup tables for identifying specific substances, which must be frequently updated and maintained. To improve identification of drugs on death certificates, a deep-learning named-entity recognition model was developed, utilizing data from the Kentucky Drug Overdose Fatality Surveillance System (2014–2019), which achieved an F1-score of 99.13%. This model can identify new drug misspellings and novel substances that are not present on current surveillance lookup tables, enhancing the surveillance of drug overdose deaths.

KW - deep learning

KW - drug overdose

KW - machine learning

KW - surveillance

UR - https://www.scopus.com/pages/publications/85147389930

UR - https://www.scopus.com/inward/citedby.url?scp=85147389930&partnerID=8YFLogxK

U2 - 10.1093/aje/kwac180

DO - 10.1093/aje/kwac180

M3 - Article

C2 - 36222700

AN - SCOPUS:85147389930

SN - 0002-9262

VL - 192

SP - 257

EP - 266

JO - American Journal of Epidemiology

JF - American Journal of Epidemiology

IS - 2

ER -

Deep Neural Networks for Fine-Grained Surveillance of Overdose Mortality

Abstract

Bibliographical note

Funding

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this