A combined strategy of feature selection and machine learning to identify predictors of prediabetes

Kushan De Silva; Daniel Jönsson; Ryan T. Demmer

doi:10.1093/jamia/ocz204

A combined strategy of feature selection and machine learning to identify predictors of prediabetes

Kushan De Silva, Daniel Jönsson, Ryan T. Demmer

Epidemiology & Community Health

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

23 Scopus citations

Abstract

OBJECTIVE: To identify predictors of prediabetes using feature selection and machine learning on a nationally representative sample of the US population.

MATERIALS AND METHODS: We analyzed n = 6346 men and women enrolled in the National Health and Nutrition Examination Survey 2013-2014. Prediabetes was defined using American Diabetes Association guidelines. The sample was randomly partitioned to training (n = 3174) and internal validation (n = 3172) sets. Feature selection algorithms were run on training data containing 156 preselected exposure variables. Four machine learning algorithms were applied on 46 exposure variables in original and resampled training datasets built using 4 resampling methods. Predictive models were tested on internal validation data (n = 3172) and external validation data (n = 3000) prepared from National Health and Nutrition Examination Survey 2011-2012. Model performance was evaluated using area under the receiver operating characteristic curve (AUROC). Predictors were assessed by odds ratios in logistic models and variable importance in others. The Centers for Disease Control (CDC) prediabetes screening tool was the benchmark to compare model performance.

RESULTS: Prediabetes prevalence was 23.43%. The CDC prediabetes screening tool produced 64.40% AUROC. Seven optimal (≥ 70% AUROC) models identified 25 predictors including 4 potentially novel associations; 20 by both logistic and other nonlinear/ensemble models and 5 solely by the latter. All optimal models outperformed the CDC prediabetes screening tool (P < 0.05).

DISCUSSION: Combined use of feature selection and machine learning increased predictive performance outperforming the recommended screening tool. A range of predictors of prediabetes was identified.

CONCLUSION: This work demonstrated the value of combining feature selection with machine learning to identify a wide range of predictors that could enhance prediabetes prediction and clinical decision-making.

Original language	English (US)
Pages (from-to)	396-406
Number of pages	11
Journal	Journal of the American Medical Informatics Association
Volume	27
Issue number	3
DOIs	https://doi.org/10.1093/jamia/ocz204
State	Published - Mar 1 2020

Bibliographical note

Funding Information:
This work was supported by a Swedish Institute scholarship granted to Kushan De Silva for studies at Lund University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Swedish Institute.

Publisher Copyright:
© 2019 The Author(s) 2019. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved.

Keywords

NHANES
feature selection
machine learning
prediabetes
predictors

PubMed: MeSH publication types

Research Support, Non-U.S. Gov't
Journal Article

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Publisher link

10.1093/jamia/ocz204

Find It

Find It at U of M Libraries

Cite this

@article{9a7450db1fd14da39ae65c9717ae9d38,

title = "A combined strategy of feature selection and machine learning to identify predictors of prediabetes",

abstract = "OBJECTIVE: To identify predictors of prediabetes using feature selection and machine learning on a nationally representative sample of the US population.MATERIALS AND METHODS: We analyzed n = 6346 men and women enrolled in the National Health and Nutrition Examination Survey 2013-2014. Prediabetes was defined using American Diabetes Association guidelines. The sample was randomly partitioned to training (n = 3174) and internal validation (n = 3172) sets. Feature selection algorithms were run on training data containing 156 preselected exposure variables. Four machine learning algorithms were applied on 46 exposure variables in original and resampled training datasets built using 4 resampling methods. Predictive models were tested on internal validation data (n = 3172) and external validation data (n = 3000) prepared from National Health and Nutrition Examination Survey 2011-2012. Model performance was evaluated using area under the receiver operating characteristic curve (AUROC). Predictors were assessed by odds ratios in logistic models and variable importance in others. The Centers for Disease Control (CDC) prediabetes screening tool was the benchmark to compare model performance.RESULTS: Prediabetes prevalence was 23.43%. The CDC prediabetes screening tool produced 64.40% AUROC. Seven optimal (≥ 70% AUROC) models identified 25 predictors including 4 potentially novel associations; 20 by both logistic and other nonlinear/ensemble models and 5 solely by the latter. All optimal models outperformed the CDC prediabetes screening tool (P < 0.05).DISCUSSION: Combined use of feature selection and machine learning increased predictive performance outperforming the recommended screening tool. A range of predictors of prediabetes was identified.CONCLUSION: This work demonstrated the value of combining feature selection with machine learning to identify a wide range of predictors that could enhance prediabetes prediction and clinical decision-making.",

keywords = "NHANES, feature selection, machine learning, prediabetes, predictors",

author = "{De Silva}, Kushan and Daniel J{\"o}nsson and Demmer, {Ryan T.}",

note = "Funding Information: This work was supported by a Swedish Institute scholarship granted to Kushan De Silva for studies at Lund University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Swedish Institute. Publisher Copyright: {\textcopyright} 2019 The Author(s) 2019. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved.",

year = "2020",

month = mar,

day = "1",

doi = "10.1093/jamia/ocz204",

language = "English (US)",

volume = "27",

pages = "396--406",

journal = "Journal of the American Medical Informatics Association : JAMIA",

issn = "1067-5027",

publisher = "Oxford University Press",

number = "3",

}

TY - JOUR

T1 - A combined strategy of feature selection and machine learning to identify predictors of prediabetes

AU - De Silva, Kushan

AU - Jönsson, Daniel

AU - Demmer, Ryan T.

N1 - Funding Information: This work was supported by a Swedish Institute scholarship granted to Kushan De Silva for studies at Lund University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Swedish Institute. Publisher Copyright: © 2019 The Author(s) 2019. Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved.

PY - 2020/3/1

Y1 - 2020/3/1

N2 - OBJECTIVE: To identify predictors of prediabetes using feature selection and machine learning on a nationally representative sample of the US population.MATERIALS AND METHODS: We analyzed n = 6346 men and women enrolled in the National Health and Nutrition Examination Survey 2013-2014. Prediabetes was defined using American Diabetes Association guidelines. The sample was randomly partitioned to training (n = 3174) and internal validation (n = 3172) sets. Feature selection algorithms were run on training data containing 156 preselected exposure variables. Four machine learning algorithms were applied on 46 exposure variables in original and resampled training datasets built using 4 resampling methods. Predictive models were tested on internal validation data (n = 3172) and external validation data (n = 3000) prepared from National Health and Nutrition Examination Survey 2011-2012. Model performance was evaluated using area under the receiver operating characteristic curve (AUROC). Predictors were assessed by odds ratios in logistic models and variable importance in others. The Centers for Disease Control (CDC) prediabetes screening tool was the benchmark to compare model performance.RESULTS: Prediabetes prevalence was 23.43%. The CDC prediabetes screening tool produced 64.40% AUROC. Seven optimal (≥ 70% AUROC) models identified 25 predictors including 4 potentially novel associations; 20 by both logistic and other nonlinear/ensemble models and 5 solely by the latter. All optimal models outperformed the CDC prediabetes screening tool (P < 0.05).DISCUSSION: Combined use of feature selection and machine learning increased predictive performance outperforming the recommended screening tool. A range of predictors of prediabetes was identified.CONCLUSION: This work demonstrated the value of combining feature selection with machine learning to identify a wide range of predictors that could enhance prediabetes prediction and clinical decision-making.

AB - OBJECTIVE: To identify predictors of prediabetes using feature selection and machine learning on a nationally representative sample of the US population.MATERIALS AND METHODS: We analyzed n = 6346 men and women enrolled in the National Health and Nutrition Examination Survey 2013-2014. Prediabetes was defined using American Diabetes Association guidelines. The sample was randomly partitioned to training (n = 3174) and internal validation (n = 3172) sets. Feature selection algorithms were run on training data containing 156 preselected exposure variables. Four machine learning algorithms were applied on 46 exposure variables in original and resampled training datasets built using 4 resampling methods. Predictive models were tested on internal validation data (n = 3172) and external validation data (n = 3000) prepared from National Health and Nutrition Examination Survey 2011-2012. Model performance was evaluated using area under the receiver operating characteristic curve (AUROC). Predictors were assessed by odds ratios in logistic models and variable importance in others. The Centers for Disease Control (CDC) prediabetes screening tool was the benchmark to compare model performance.RESULTS: Prediabetes prevalence was 23.43%. The CDC prediabetes screening tool produced 64.40% AUROC. Seven optimal (≥ 70% AUROC) models identified 25 predictors including 4 potentially novel associations; 20 by both logistic and other nonlinear/ensemble models and 5 solely by the latter. All optimal models outperformed the CDC prediabetes screening tool (P < 0.05).DISCUSSION: Combined use of feature selection and machine learning increased predictive performance outperforming the recommended screening tool. A range of predictors of prediabetes was identified.CONCLUSION: This work demonstrated the value of combining feature selection with machine learning to identify a wide range of predictors that could enhance prediabetes prediction and clinical decision-making.

KW - NHANES

KW - feature selection

KW - machine learning

KW - prediabetes

KW - predictors

UR - http://www.scopus.com/inward/record.url?scp=85079353320&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85079353320&partnerID=8YFLogxK

U2 - 10.1093/jamia/ocz204

DO - 10.1093/jamia/ocz204

M3 - Article

C2 - 31889178

AN - SCOPUS:85079353320

SN - 1067-5027

VL - 27

SP - 396

EP - 406

JO - Journal of the American Medical Informatics Association : JAMIA

JF - Journal of the American Medical Informatics Association : JAMIA

IS - 3

ER -

A combined strategy of feature selection and machine learning to identify predictors of prediabetes

Abstract

Bibliographical note

Keywords

PubMed: MeSH publication types

UN SDGs

Publisher link

Find It

Other files and links

Fingerprint

Cite this