Automated feature selection of predictors in electronic medical records data

Jessica Gronsbell, Jessica Minnier, Sheng Yu, Katherine Liao, Tianxi Cai

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

The use of Electronic Health Records (EHR) for translational research can be challenging due to difficulty in extracting accurate disease phenotype data. Historically, EHR algorithms for annotating phenotypes have been either rule-based or trained with billing codes and gold standard labels curated via labor intensive medical chart review. These simplistic algorithms tend to have unpredictable portability across institutions and low accuracy for many disease phenotypes due to imprecise billing codes. Recently, more sophisticated machine learning algorithms have been developed to improve the robustness and accuracy of EHR phenotyping algorithms. These algorithms are typically trained via supervised learning, relating gold standard labels to a wide range of candidate features including billing codes, procedure codes, medication prescriptions and relevant clinical concepts extracted from narrative notes via Natural Language Processing (NLP). However, due to the time intensiveness of gold standard labeling, the size of the training set is often insufficient to build a generalizable algorithm with the large number of candidate features extracted from EHR. To reduce the number of candidate predictors and in turn improve model performance, we present an automated feature selection method based entirely on unlabeled observations. The proposed method generates a comprehensive surrogate for the underlying phenotype with an unsupervised clustering of disease status based on several highly predictive features such as diagnosis codes and mentions of the disease in text fields available in the entire set of EHR data. A sparse regression model is then built with the estimated outcomes and remaining covariates to identify those features most informative of the phenotype of interest. Relying on the results of Li and Duan (1989), we demonstrate that variable selection for the underlying phenotype model can be achieved by fitting the surrogate-based model. We explore the performance of our methods in numerical simulations and present the results of a prediction model for Rheumatoid Arthritis (RA) built on a large EHR data mart from the Partners Health System consisting of billing codes and NLP terms. Empirical results suggest that our procedure reduces the number of gold-standard labels necessary for phenotyping thereby harnessing the automated power of EHR data and improving efficiency.

Original languageEnglish (US)
JournalBiometrics
DOIs
StatePublished - Jan 1 2018

Fingerprint

Electronic medical equipment
Electronic Health Records
Feature Selection
electronics
Feature extraction
Predictors
Health
Phenotype
Electronics
phenotype
Gold
gold
Natural Language Processing
Labels
Natural Language
Translational Medical Research
Unsupervised Clustering
Rheumatoid Arthritis
rheumatoid arthritis
artificial intelligence

Keywords

  • electronic medical records
  • feature selection
  • prediction accuracy
  • regularized regression
  • risk prediction

ASJC Scopus subject areas

  • Statistics and Probability
  • Biochemistry, Genetics and Molecular Biology(all)
  • Immunology and Microbiology(all)
  • Agricultural and Biological Sciences(all)
  • Applied Mathematics

Cite this

Automated feature selection of predictors in electronic medical records data. / Gronsbell, Jessica; Minnier, Jessica; Yu, Sheng; Liao, Katherine; Cai, Tianxi.

In: Biometrics, 01.01.2018.

Research output: Contribution to journalArticle

Gronsbell, Jessica ; Minnier, Jessica ; Yu, Sheng ; Liao, Katherine ; Cai, Tianxi. / Automated feature selection of predictors in electronic medical records data. In: Biometrics. 2018.
@article{65e5c7918fd64420b3db38d02f75218f,
title = "Automated feature selection of predictors in electronic medical records data",
abstract = "The use of Electronic Health Records (EHR) for translational research can be challenging due to difficulty in extracting accurate disease phenotype data. Historically, EHR algorithms for annotating phenotypes have been either rule-based or trained with billing codes and gold standard labels curated via labor intensive medical chart review. These simplistic algorithms tend to have unpredictable portability across institutions and low accuracy for many disease phenotypes due to imprecise billing codes. Recently, more sophisticated machine learning algorithms have been developed to improve the robustness and accuracy of EHR phenotyping algorithms. These algorithms are typically trained via supervised learning, relating gold standard labels to a wide range of candidate features including billing codes, procedure codes, medication prescriptions and relevant clinical concepts extracted from narrative notes via Natural Language Processing (NLP). However, due to the time intensiveness of gold standard labeling, the size of the training set is often insufficient to build a generalizable algorithm with the large number of candidate features extracted from EHR. To reduce the number of candidate predictors and in turn improve model performance, we present an automated feature selection method based entirely on unlabeled observations. The proposed method generates a comprehensive surrogate for the underlying phenotype with an unsupervised clustering of disease status based on several highly predictive features such as diagnosis codes and mentions of the disease in text fields available in the entire set of EHR data. A sparse regression model is then built with the estimated outcomes and remaining covariates to identify those features most informative of the phenotype of interest. Relying on the results of Li and Duan (1989), we demonstrate that variable selection for the underlying phenotype model can be achieved by fitting the surrogate-based model. We explore the performance of our methods in numerical simulations and present the results of a prediction model for Rheumatoid Arthritis (RA) built on a large EHR data mart from the Partners Health System consisting of billing codes and NLP terms. Empirical results suggest that our procedure reduces the number of gold-standard labels necessary for phenotyping thereby harnessing the automated power of EHR data and improving efficiency.",
keywords = "electronic medical records, feature selection, prediction accuracy, regularized regression, risk prediction",
author = "Jessica Gronsbell and Jessica Minnier and Sheng Yu and Katherine Liao and Tianxi Cai",
year = "2018",
month = "1",
day = "1",
doi = "10.1111/biom.12987",
language = "English (US)",
journal = "Biometrics",
issn = "0006-341X",
publisher = "Wiley-Blackwell",

}

TY - JOUR

T1 - Automated feature selection of predictors in electronic medical records data

AU - Gronsbell, Jessica

AU - Minnier, Jessica

AU - Yu, Sheng

AU - Liao, Katherine

AU - Cai, Tianxi

PY - 2018/1/1

Y1 - 2018/1/1

N2 - The use of Electronic Health Records (EHR) for translational research can be challenging due to difficulty in extracting accurate disease phenotype data. Historically, EHR algorithms for annotating phenotypes have been either rule-based or trained with billing codes and gold standard labels curated via labor intensive medical chart review. These simplistic algorithms tend to have unpredictable portability across institutions and low accuracy for many disease phenotypes due to imprecise billing codes. Recently, more sophisticated machine learning algorithms have been developed to improve the robustness and accuracy of EHR phenotyping algorithms. These algorithms are typically trained via supervised learning, relating gold standard labels to a wide range of candidate features including billing codes, procedure codes, medication prescriptions and relevant clinical concepts extracted from narrative notes via Natural Language Processing (NLP). However, due to the time intensiveness of gold standard labeling, the size of the training set is often insufficient to build a generalizable algorithm with the large number of candidate features extracted from EHR. To reduce the number of candidate predictors and in turn improve model performance, we present an automated feature selection method based entirely on unlabeled observations. The proposed method generates a comprehensive surrogate for the underlying phenotype with an unsupervised clustering of disease status based on several highly predictive features such as diagnosis codes and mentions of the disease in text fields available in the entire set of EHR data. A sparse regression model is then built with the estimated outcomes and remaining covariates to identify those features most informative of the phenotype of interest. Relying on the results of Li and Duan (1989), we demonstrate that variable selection for the underlying phenotype model can be achieved by fitting the surrogate-based model. We explore the performance of our methods in numerical simulations and present the results of a prediction model for Rheumatoid Arthritis (RA) built on a large EHR data mart from the Partners Health System consisting of billing codes and NLP terms. Empirical results suggest that our procedure reduces the number of gold-standard labels necessary for phenotyping thereby harnessing the automated power of EHR data and improving efficiency.

AB - The use of Electronic Health Records (EHR) for translational research can be challenging due to difficulty in extracting accurate disease phenotype data. Historically, EHR algorithms for annotating phenotypes have been either rule-based or trained with billing codes and gold standard labels curated via labor intensive medical chart review. These simplistic algorithms tend to have unpredictable portability across institutions and low accuracy for many disease phenotypes due to imprecise billing codes. Recently, more sophisticated machine learning algorithms have been developed to improve the robustness and accuracy of EHR phenotyping algorithms. These algorithms are typically trained via supervised learning, relating gold standard labels to a wide range of candidate features including billing codes, procedure codes, medication prescriptions and relevant clinical concepts extracted from narrative notes via Natural Language Processing (NLP). However, due to the time intensiveness of gold standard labeling, the size of the training set is often insufficient to build a generalizable algorithm with the large number of candidate features extracted from EHR. To reduce the number of candidate predictors and in turn improve model performance, we present an automated feature selection method based entirely on unlabeled observations. The proposed method generates a comprehensive surrogate for the underlying phenotype with an unsupervised clustering of disease status based on several highly predictive features such as diagnosis codes and mentions of the disease in text fields available in the entire set of EHR data. A sparse regression model is then built with the estimated outcomes and remaining covariates to identify those features most informative of the phenotype of interest. Relying on the results of Li and Duan (1989), we demonstrate that variable selection for the underlying phenotype model can be achieved by fitting the surrogate-based model. We explore the performance of our methods in numerical simulations and present the results of a prediction model for Rheumatoid Arthritis (RA) built on a large EHR data mart from the Partners Health System consisting of billing codes and NLP terms. Empirical results suggest that our procedure reduces the number of gold-standard labels necessary for phenotyping thereby harnessing the automated power of EHR data and improving efficiency.

KW - electronic medical records

KW - feature selection

KW - prediction accuracy

KW - regularized regression

KW - risk prediction

UR - http://www.scopus.com/inward/record.url?scp=85061025010&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85061025010&partnerID=8YFLogxK

U2 - 10.1111/biom.12987

DO - 10.1111/biom.12987

M3 - Article

C2 - 30353541

AN - SCOPUS:85061025010

JO - Biometrics

JF - Biometrics

SN - 0006-341X

ER -