Comparison of three information sources for smoking information in electronic health records

Liwei Wang, Xiaoyang Ruan, Ping Yang, Hongfang D Liu

Research output: Contribution to journalReview article

6 Citations (Scopus)

Abstract

Objective: The primary aim was to compare independent and joint performance of retrieving smoking status through different sources, including narrative text processed by natural language processing (NLP), patient-provided information (PPI), and diagnosis codes (ie, International Classification of Diseases, Ninth Revision [ICD-9]). We also compared the performance of retrieving smoking strength information (ie, heavy/light smoker) from narrative text and PPI. Materia ls and methods: Our study leveraged an existing lung cancer cohort for smoking status, amount, and strength information, which was manually chart-reviewed. On the NLP side, smoking-related electronic medical record (EMR) data were retrieved first. A pattern-based smoking information extraction module was then implemented to extract smoking-related information. After that, heuristic rules were used to obtain smoking status-related information. Smoking information was also obtained from structured data sources based on diagnosis codes and PPI. Sensitivity, specificity, and accuracy were measured using patients with coverage (ie, the proportion of patients whose smoking status/strength can be effectively determined). Results: NLP alone has the best overall performance for smoking status extraction (patient coverage: 0.88; sensitivity: 0.97; specificity: 0.70; accuracy: 0.88); combining PPI with NLP further improved patient coverage to 0.96. ICD-9 does not provide additional improvement to NLP and its combination with PPI. For smoking strength, combining NLP with PPI has slight improvement over NLP alone. Co nclusio n: These findings suggest that narrative text could serve as a more reliable and comprehensive source for obtaining smoking-related information than structured data sources. PPI, the readily available structured data, could be used as a complementary source for more comprehensive patient coverage.

Original languageEnglish (US)
Pages (from-to)237-242
Number of pages6
JournalCancer Informatics
Volume15
DOIs
StatePublished - 2016

Fingerprint

Electronic Health Records
Natural Language Processing
Smoking
Information Storage and Retrieval
International Classification of Diseases
Sensitivity and Specificity
Lung Neoplasms

Keywords

  • ICD-9
  • Natural language processing
  • Patient-provided information
  • Smoking status
  • Smoking strength

ASJC Scopus subject areas

  • Oncology
  • Cancer Research

Cite this

Comparison of three information sources for smoking information in electronic health records. / Wang, Liwei; Ruan, Xiaoyang; Yang, Ping; Liu, Hongfang D.

In: Cancer Informatics, Vol. 15, 2016, p. 237-242.

Research output: Contribution to journalReview article

@article{7198d78893b2417a931030f4fb3b226b,
title = "Comparison of three information sources for smoking information in electronic health records",
abstract = "Objective: The primary aim was to compare independent and joint performance of retrieving smoking status through different sources, including narrative text processed by natural language processing (NLP), patient-provided information (PPI), and diagnosis codes (ie, International Classification of Diseases, Ninth Revision [ICD-9]). We also compared the performance of retrieving smoking strength information (ie, heavy/light smoker) from narrative text and PPI. Materia ls and methods: Our study leveraged an existing lung cancer cohort for smoking status, amount, and strength information, which was manually chart-reviewed. On the NLP side, smoking-related electronic medical record (EMR) data were retrieved first. A pattern-based smoking information extraction module was then implemented to extract smoking-related information. After that, heuristic rules were used to obtain smoking status-related information. Smoking information was also obtained from structured data sources based on diagnosis codes and PPI. Sensitivity, specificity, and accuracy were measured using patients with coverage (ie, the proportion of patients whose smoking status/strength can be effectively determined). Results: NLP alone has the best overall performance for smoking status extraction (patient coverage: 0.88; sensitivity: 0.97; specificity: 0.70; accuracy: 0.88); combining PPI with NLP further improved patient coverage to 0.96. ICD-9 does not provide additional improvement to NLP and its combination with PPI. For smoking strength, combining NLP with PPI has slight improvement over NLP alone. Co nclusio n: These findings suggest that narrative text could serve as a more reliable and comprehensive source for obtaining smoking-related information than structured data sources. PPI, the readily available structured data, could be used as a complementary source for more comprehensive patient coverage.",
keywords = "ICD-9, Natural language processing, Patient-provided information, Smoking status, Smoking strength",
author = "Liwei Wang and Xiaoyang Ruan and Ping Yang and Liu, {Hongfang D}",
year = "2016",
doi = "10.4137/CIN.S40604",
language = "English (US)",
volume = "15",
pages = "237--242",
journal = "Cancer Informatics",
issn = "1176-9351",
publisher = "Libertas Academica Ltd.",

}

TY - JOUR

T1 - Comparison of three information sources for smoking information in electronic health records

AU - Wang, Liwei

AU - Ruan, Xiaoyang

AU - Yang, Ping

AU - Liu, Hongfang D

PY - 2016

Y1 - 2016

N2 - Objective: The primary aim was to compare independent and joint performance of retrieving smoking status through different sources, including narrative text processed by natural language processing (NLP), patient-provided information (PPI), and diagnosis codes (ie, International Classification of Diseases, Ninth Revision [ICD-9]). We also compared the performance of retrieving smoking strength information (ie, heavy/light smoker) from narrative text and PPI. Materia ls and methods: Our study leveraged an existing lung cancer cohort for smoking status, amount, and strength information, which was manually chart-reviewed. On the NLP side, smoking-related electronic medical record (EMR) data were retrieved first. A pattern-based smoking information extraction module was then implemented to extract smoking-related information. After that, heuristic rules were used to obtain smoking status-related information. Smoking information was also obtained from structured data sources based on diagnosis codes and PPI. Sensitivity, specificity, and accuracy were measured using patients with coverage (ie, the proportion of patients whose smoking status/strength can be effectively determined). Results: NLP alone has the best overall performance for smoking status extraction (patient coverage: 0.88; sensitivity: 0.97; specificity: 0.70; accuracy: 0.88); combining PPI with NLP further improved patient coverage to 0.96. ICD-9 does not provide additional improvement to NLP and its combination with PPI. For smoking strength, combining NLP with PPI has slight improvement over NLP alone. Co nclusio n: These findings suggest that narrative text could serve as a more reliable and comprehensive source for obtaining smoking-related information than structured data sources. PPI, the readily available structured data, could be used as a complementary source for more comprehensive patient coverage.

AB - Objective: The primary aim was to compare independent and joint performance of retrieving smoking status through different sources, including narrative text processed by natural language processing (NLP), patient-provided information (PPI), and diagnosis codes (ie, International Classification of Diseases, Ninth Revision [ICD-9]). We also compared the performance of retrieving smoking strength information (ie, heavy/light smoker) from narrative text and PPI. Materia ls and methods: Our study leveraged an existing lung cancer cohort for smoking status, amount, and strength information, which was manually chart-reviewed. On the NLP side, smoking-related electronic medical record (EMR) data were retrieved first. A pattern-based smoking information extraction module was then implemented to extract smoking-related information. After that, heuristic rules were used to obtain smoking status-related information. Smoking information was also obtained from structured data sources based on diagnosis codes and PPI. Sensitivity, specificity, and accuracy were measured using patients with coverage (ie, the proportion of patients whose smoking status/strength can be effectively determined). Results: NLP alone has the best overall performance for smoking status extraction (patient coverage: 0.88; sensitivity: 0.97; specificity: 0.70; accuracy: 0.88); combining PPI with NLP further improved patient coverage to 0.96. ICD-9 does not provide additional improvement to NLP and its combination with PPI. For smoking strength, combining NLP with PPI has slight improvement over NLP alone. Co nclusio n: These findings suggest that narrative text could serve as a more reliable and comprehensive source for obtaining smoking-related information than structured data sources. PPI, the readily available structured data, could be used as a complementary source for more comprehensive patient coverage.

KW - ICD-9

KW - Natural language processing

KW - Patient-provided information

KW - Smoking status

KW - Smoking strength

UR - http://www.scopus.com/inward/record.url?scp=85012241833&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85012241833&partnerID=8YFLogxK

U2 - 10.4137/CIN.S40604

DO - 10.4137/CIN.S40604

M3 - Review article

VL - 15

SP - 237

EP - 242

JO - Cancer Informatics

JF - Cancer Informatics

SN - 1176-9351

ER -