Predictive modeling of hypoglycemia for clinical decision support in evaluating outpatients with diabetes mellitus

Xiaochun Li, Shengsheng Yu, Zuoyi Zhang, Larry Radican, Jonathan Cummins, Samuel S. Engel, Kristy Iglay, Jon Duke, Jarod Baker, Kimberly G. Brodovicz, Ramachandra G. Naik, Jeremy Leventhal, Arnaub K. Chatterjee, Swapnil Rajpathak, Michael Weiner

Research output: Contribution to journalArticle

Abstract

Objective: Hypoglycemia occurs in 20–60% of patients with diabetes mellitus. Identifying at-risk patients can facilitate interventions to lower risk. We sought to develop a hypoglycemia prediction model. Methods: In this retrospective cohort study, urban adults prescribed a diabetes drug between 2004 and 2013 were identified. Demographic and clinical data were extracted from an electronic medical record (EMR). Laboratory tests, diagnostic codes and natural language processing (NLP) identified hypoglycemia. We compared multiple logistic regression, classification and regression trees (CART), and random forest. Models were evaluated on an independent test set or through cross-validation. Results: The 38,780 patients had mean age 57 years; 56% were female, 40% African-American and 39% uninsured. Hypoglycemia occurred in 8128 (539 identified only by NLP). In logistic regression, factors positively associated with hypoglycemia included infection, non-long-acting insulin, dementia and recent hypoglycemia. Negatively associated factors included long-acting insulin plus sulfonylurea, and age 75 or older. The models’ area under curve was similar (logistic regression, 89%; CART, 88%; random forest, 90%, with ten-fold cross-validation). Conclusions: NLP improved identification of hypoglycemia. Non-long-acting insulin was an important risk factor. Decreased risk with age may reflect treatment or diminished awareness of hypoglycemia. More complex models did not improve prediction.

Original languageEnglish (US)
Pages (from-to)1885-1891
Number of pages7
JournalCurrent Medical Research and Opinion
Volume35
Issue number11
DOIs
StatePublished - Nov 2 2019

Fingerprint

Clinical Decision Support Systems
Hypoglycemia
Diabetes Mellitus
Outpatients
Natural Language Processing
Logistic Models
Long-Acting Insulin
Insulin
Electronic Health Records
Routine Diagnostic Tests
African Americans
Area Under Curve
Dementia
Cohort Studies
Retrospective Studies
Demography

Keywords

  • diabetes mellitus
  • Hypoglycemia
  • predictive value of tests

ASJC Scopus subject areas

  • Medicine(all)

Cite this

Predictive modeling of hypoglycemia for clinical decision support in evaluating outpatients with diabetes mellitus. / Li, Xiaochun; Yu, Shengsheng; Zhang, Zuoyi; Radican, Larry; Cummins, Jonathan; Engel, Samuel S.; Iglay, Kristy; Duke, Jon; Baker, Jarod; Brodovicz, Kimberly G.; Naik, Ramachandra G.; Leventhal, Jeremy; Chatterjee, Arnaub K.; Rajpathak, Swapnil; Weiner, Michael.

In: Current Medical Research and Opinion, Vol. 35, No. 11, 02.11.2019, p. 1885-1891.

Research output: Contribution to journalArticle

Li, X, Yu, S, Zhang, Z, Radican, L, Cummins, J, Engel, SS, Iglay, K, Duke, J, Baker, J, Brodovicz, KG, Naik, RG, Leventhal, J, Chatterjee, AK, Rajpathak, S & Weiner, M 2019, 'Predictive modeling of hypoglycemia for clinical decision support in evaluating outpatients with diabetes mellitus', Current Medical Research and Opinion, vol. 35, no. 11, pp. 1885-1891. https://doi.org/10.1080/03007995.2019.1636016
Li, Xiaochun ; Yu, Shengsheng ; Zhang, Zuoyi ; Radican, Larry ; Cummins, Jonathan ; Engel, Samuel S. ; Iglay, Kristy ; Duke, Jon ; Baker, Jarod ; Brodovicz, Kimberly G. ; Naik, Ramachandra G. ; Leventhal, Jeremy ; Chatterjee, Arnaub K. ; Rajpathak, Swapnil ; Weiner, Michael. / Predictive modeling of hypoglycemia for clinical decision support in evaluating outpatients with diabetes mellitus. In: Current Medical Research and Opinion. 2019 ; Vol. 35, No. 11. pp. 1885-1891.
@article{1f9475f6f4ae42f3801f21cfa2479cc4,
title = "Predictive modeling of hypoglycemia for clinical decision support in evaluating outpatients with diabetes mellitus",
abstract = "Objective: Hypoglycemia occurs in 20–60{\%} of patients with diabetes mellitus. Identifying at-risk patients can facilitate interventions to lower risk. We sought to develop a hypoglycemia prediction model. Methods: In this retrospective cohort study, urban adults prescribed a diabetes drug between 2004 and 2013 were identified. Demographic and clinical data were extracted from an electronic medical record (EMR). Laboratory tests, diagnostic codes and natural language processing (NLP) identified hypoglycemia. We compared multiple logistic regression, classification and regression trees (CART), and random forest. Models were evaluated on an independent test set or through cross-validation. Results: The 38,780 patients had mean age 57 years; 56{\%} were female, 40{\%} African-American and 39{\%} uninsured. Hypoglycemia occurred in 8128 (539 identified only by NLP). In logistic regression, factors positively associated with hypoglycemia included infection, non-long-acting insulin, dementia and recent hypoglycemia. Negatively associated factors included long-acting insulin plus sulfonylurea, and age 75 or older. The models’ area under curve was similar (logistic regression, 89{\%}; CART, 88{\%}; random forest, 90{\%}, with ten-fold cross-validation). Conclusions: NLP improved identification of hypoglycemia. Non-long-acting insulin was an important risk factor. Decreased risk with age may reflect treatment or diminished awareness of hypoglycemia. More complex models did not improve prediction.",
keywords = "diabetes mellitus, Hypoglycemia, predictive value of tests",
author = "Xiaochun Li and Shengsheng Yu and Zuoyi Zhang and Larry Radican and Jonathan Cummins and Engel, {Samuel S.} and Kristy Iglay and Jon Duke and Jarod Baker and Brodovicz, {Kimberly G.} and Naik, {Ramachandra G.} and Jeremy Leventhal and Chatterjee, {Arnaub K.} and Swapnil Rajpathak and Michael Weiner",
year = "2019",
month = "11",
day = "2",
doi = "10.1080/03007995.2019.1636016",
language = "English (US)",
volume = "35",
pages = "1885--1891",
journal = "Current Medical Research and Opinion",
issn = "0300-7995",
publisher = "Informa Healthcare",
number = "11",

}

TY - JOUR

T1 - Predictive modeling of hypoglycemia for clinical decision support in evaluating outpatients with diabetes mellitus

AU - Li, Xiaochun

AU - Yu, Shengsheng

AU - Zhang, Zuoyi

AU - Radican, Larry

AU - Cummins, Jonathan

AU - Engel, Samuel S.

AU - Iglay, Kristy

AU - Duke, Jon

AU - Baker, Jarod

AU - Brodovicz, Kimberly G.

AU - Naik, Ramachandra G.

AU - Leventhal, Jeremy

AU - Chatterjee, Arnaub K.

AU - Rajpathak, Swapnil

AU - Weiner, Michael

PY - 2019/11/2

Y1 - 2019/11/2

N2 - Objective: Hypoglycemia occurs in 20–60% of patients with diabetes mellitus. Identifying at-risk patients can facilitate interventions to lower risk. We sought to develop a hypoglycemia prediction model. Methods: In this retrospective cohort study, urban adults prescribed a diabetes drug between 2004 and 2013 were identified. Demographic and clinical data were extracted from an electronic medical record (EMR). Laboratory tests, diagnostic codes and natural language processing (NLP) identified hypoglycemia. We compared multiple logistic regression, classification and regression trees (CART), and random forest. Models were evaluated on an independent test set or through cross-validation. Results: The 38,780 patients had mean age 57 years; 56% were female, 40% African-American and 39% uninsured. Hypoglycemia occurred in 8128 (539 identified only by NLP). In logistic regression, factors positively associated with hypoglycemia included infection, non-long-acting insulin, dementia and recent hypoglycemia. Negatively associated factors included long-acting insulin plus sulfonylurea, and age 75 or older. The models’ area under curve was similar (logistic regression, 89%; CART, 88%; random forest, 90%, with ten-fold cross-validation). Conclusions: NLP improved identification of hypoglycemia. Non-long-acting insulin was an important risk factor. Decreased risk with age may reflect treatment or diminished awareness of hypoglycemia. More complex models did not improve prediction.

AB - Objective: Hypoglycemia occurs in 20–60% of patients with diabetes mellitus. Identifying at-risk patients can facilitate interventions to lower risk. We sought to develop a hypoglycemia prediction model. Methods: In this retrospective cohort study, urban adults prescribed a diabetes drug between 2004 and 2013 were identified. Demographic and clinical data were extracted from an electronic medical record (EMR). Laboratory tests, diagnostic codes and natural language processing (NLP) identified hypoglycemia. We compared multiple logistic regression, classification and regression trees (CART), and random forest. Models were evaluated on an independent test set or through cross-validation. Results: The 38,780 patients had mean age 57 years; 56% were female, 40% African-American and 39% uninsured. Hypoglycemia occurred in 8128 (539 identified only by NLP). In logistic regression, factors positively associated with hypoglycemia included infection, non-long-acting insulin, dementia and recent hypoglycemia. Negatively associated factors included long-acting insulin plus sulfonylurea, and age 75 or older. The models’ area under curve was similar (logistic regression, 89%; CART, 88%; random forest, 90%, with ten-fold cross-validation). Conclusions: NLP improved identification of hypoglycemia. Non-long-acting insulin was an important risk factor. Decreased risk with age may reflect treatment or diminished awareness of hypoglycemia. More complex models did not improve prediction.

KW - diabetes mellitus

KW - Hypoglycemia

KW - predictive value of tests

UR - http://www.scopus.com/inward/record.url?scp=85073576810&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85073576810&partnerID=8YFLogxK

U2 - 10.1080/03007995.2019.1636016

DO - 10.1080/03007995.2019.1636016

M3 - Article

C2 - 31234649

AN - SCOPUS:85073576810

VL - 35

SP - 1885

EP - 1891

JO - Current Medical Research and Opinion

JF - Current Medical Research and Opinion

SN - 0300-7995

IS - 11

ER -