Towards phenotyping stroke: Leveraging data from a large-scale epidemiological study to detect stroke diagnosis

Yizhao Ni, Kathleen Alwell, Charles J. Moomaw, Daniel Woo, Opeolu Adeoye, Matthew L. Flaherty, Simona Ferioli, Jason Mackey, Felipe De Los Rios La Rosa, Sharyl Martini, Pooja Khatri, Dawn Kleindorfer, Brett M. Kissela

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Objective 1) To develop a machine learning approach for detecting stroke cases and subtypes from hospitalization data, 2) to assess algorithm performance and predictors on real-world data collected by a large-scale epidemiology study in the US; and 3) to identify directions for future development of high-precision stroke phenotypic signatures. Materials and methods We utilized 8,131 hospitalization events (ICD-9 codes 430±438) collected from the Greater Cincinnati/Northern Kentucky Stroke Study in 2005 and 2010. Detailed information from patients' medical records was abstracted for each event by trained research nurses. By analyzing the broad list of demographic and clinical variables, the machine learning algorithms predicted whether an event was a stroke case and, if so, the stroke subtype. The performance was validated on gold-standard labels adjudicated by stroke physicians, and results were compared with stroke classifications based on ICD-9 discharge codes, as well as labels determined by study nurses. Results The best performing machine learning algorithm achieved a performance of 88.57%/ 93.81%/92.80%/93.30%/89.84%/98.01% (accuracy/precision/recall/F-measure/area under ROC curve/area under precision-recall curve) on stroke case detection. For detecting stroke subtypes, the algorithm yielded an overall accuracy of 87.39% and greater than 85% precision on individual subtypes. The machine learning algorithms significantly outperformed the ICD-9 method on all measures (P value<0.001). Their performance was comparable to that of study nurses, with better tradeoff between precision and recall. The feature selection uncovered a subset of predictive variables that could facilitate future development of effective stroke phenotyping algorithms. Discussion and conclusions By analyzing a broad array of patient data, the machine learning technologies held promise for improving detection of stroke diagnosis, thus unlocking high statistical power for subsequent genetic and genomic studies.

Original languageEnglish (US)
Article numbere0192586
JournalPLoS One
Volume13
Issue number2
DOIs
StatePublished - Feb 1 2018

Fingerprint

stroke
epidemiological studies
Learning systems
Epidemiologic Studies
Stroke
phenotype
Learning algorithms
artificial intelligence
International Classification of Diseases
Labels
nurses
Epidemiology
Nurses
Gold
Feature extraction
Hospitalization
physicians
ROC Curve
gold
Area Under Curve

ASJC Scopus subject areas

  • Biochemistry, Genetics and Molecular Biology(all)
  • Agricultural and Biological Sciences(all)

Cite this

Ni, Y., Alwell, K., Moomaw, C. J., Woo, D., Adeoye, O., Flaherty, M. L., ... Kissela, B. M. (2018). Towards phenotyping stroke: Leveraging data from a large-scale epidemiological study to detect stroke diagnosis. PLoS One, 13(2), [e0192586]. https://doi.org/10.1371/journal.pone.0192586

Towards phenotyping stroke : Leveraging data from a large-scale epidemiological study to detect stroke diagnosis. / Ni, Yizhao; Alwell, Kathleen; Moomaw, Charles J.; Woo, Daniel; Adeoye, Opeolu; Flaherty, Matthew L.; Ferioli, Simona; Mackey, Jason; La Rosa, Felipe De Los Rios; Martini, Sharyl; Khatri, Pooja; Kleindorfer, Dawn; Kissela, Brett M.

In: PLoS One, Vol. 13, No. 2, e0192586, 01.02.2018.

Research output: Contribution to journalArticle

Ni, Y, Alwell, K, Moomaw, CJ, Woo, D, Adeoye, O, Flaherty, ML, Ferioli, S, Mackey, J, La Rosa, FDLR, Martini, S, Khatri, P, Kleindorfer, D & Kissela, BM 2018, 'Towards phenotyping stroke: Leveraging data from a large-scale epidemiological study to detect stroke diagnosis', PLoS One, vol. 13, no. 2, e0192586. https://doi.org/10.1371/journal.pone.0192586
Ni, Yizhao ; Alwell, Kathleen ; Moomaw, Charles J. ; Woo, Daniel ; Adeoye, Opeolu ; Flaherty, Matthew L. ; Ferioli, Simona ; Mackey, Jason ; La Rosa, Felipe De Los Rios ; Martini, Sharyl ; Khatri, Pooja ; Kleindorfer, Dawn ; Kissela, Brett M. / Towards phenotyping stroke : Leveraging data from a large-scale epidemiological study to detect stroke diagnosis. In: PLoS One. 2018 ; Vol. 13, No. 2.
@article{4ac41e89e3fb46f88ee48ab3450ba37d,
title = "Towards phenotyping stroke: Leveraging data from a large-scale epidemiological study to detect stroke diagnosis",
abstract = "Objective 1) To develop a machine learning approach for detecting stroke cases and subtypes from hospitalization data, 2) to assess algorithm performance and predictors on real-world data collected by a large-scale epidemiology study in the US; and 3) to identify directions for future development of high-precision stroke phenotypic signatures. Materials and methods We utilized 8,131 hospitalization events (ICD-9 codes 430±438) collected from the Greater Cincinnati/Northern Kentucky Stroke Study in 2005 and 2010. Detailed information from patients' medical records was abstracted for each event by trained research nurses. By analyzing the broad list of demographic and clinical variables, the machine learning algorithms predicted whether an event was a stroke case and, if so, the stroke subtype. The performance was validated on gold-standard labels adjudicated by stroke physicians, and results were compared with stroke classifications based on ICD-9 discharge codes, as well as labels determined by study nurses. Results The best performing machine learning algorithm achieved a performance of 88.57{\%}/ 93.81{\%}/92.80{\%}/93.30{\%}/89.84{\%}/98.01{\%} (accuracy/precision/recall/F-measure/area under ROC curve/area under precision-recall curve) on stroke case detection. For detecting stroke subtypes, the algorithm yielded an overall accuracy of 87.39{\%} and greater than 85{\%} precision on individual subtypes. The machine learning algorithms significantly outperformed the ICD-9 method on all measures (P value<0.001). Their performance was comparable to that of study nurses, with better tradeoff between precision and recall. The feature selection uncovered a subset of predictive variables that could facilitate future development of effective stroke phenotyping algorithms. Discussion and conclusions By analyzing a broad array of patient data, the machine learning technologies held promise for improving detection of stroke diagnosis, thus unlocking high statistical power for subsequent genetic and genomic studies.",
author = "Yizhao Ni and Kathleen Alwell and Moomaw, {Charles J.} and Daniel Woo and Opeolu Adeoye and Flaherty, {Matthew L.} and Simona Ferioli and Jason Mackey and {La Rosa}, {Felipe De Los Rios} and Sharyl Martini and Pooja Khatri and Dawn Kleindorfer and Kissela, {Brett M.}",
year = "2018",
month = "2",
day = "1",
doi = "10.1371/journal.pone.0192586",
language = "English (US)",
volume = "13",
journal = "PLoS One",
issn = "1932-6203",
publisher = "Public Library of Science",
number = "2",

}

TY - JOUR

T1 - Towards phenotyping stroke

T2 - Leveraging data from a large-scale epidemiological study to detect stroke diagnosis

AU - Ni, Yizhao

AU - Alwell, Kathleen

AU - Moomaw, Charles J.

AU - Woo, Daniel

AU - Adeoye, Opeolu

AU - Flaherty, Matthew L.

AU - Ferioli, Simona

AU - Mackey, Jason

AU - La Rosa, Felipe De Los Rios

AU - Martini, Sharyl

AU - Khatri, Pooja

AU - Kleindorfer, Dawn

AU - Kissela, Brett M.

PY - 2018/2/1

Y1 - 2018/2/1

N2 - Objective 1) To develop a machine learning approach for detecting stroke cases and subtypes from hospitalization data, 2) to assess algorithm performance and predictors on real-world data collected by a large-scale epidemiology study in the US; and 3) to identify directions for future development of high-precision stroke phenotypic signatures. Materials and methods We utilized 8,131 hospitalization events (ICD-9 codes 430±438) collected from the Greater Cincinnati/Northern Kentucky Stroke Study in 2005 and 2010. Detailed information from patients' medical records was abstracted for each event by trained research nurses. By analyzing the broad list of demographic and clinical variables, the machine learning algorithms predicted whether an event was a stroke case and, if so, the stroke subtype. The performance was validated on gold-standard labels adjudicated by stroke physicians, and results were compared with stroke classifications based on ICD-9 discharge codes, as well as labels determined by study nurses. Results The best performing machine learning algorithm achieved a performance of 88.57%/ 93.81%/92.80%/93.30%/89.84%/98.01% (accuracy/precision/recall/F-measure/area under ROC curve/area under precision-recall curve) on stroke case detection. For detecting stroke subtypes, the algorithm yielded an overall accuracy of 87.39% and greater than 85% precision on individual subtypes. The machine learning algorithms significantly outperformed the ICD-9 method on all measures (P value<0.001). Their performance was comparable to that of study nurses, with better tradeoff between precision and recall. The feature selection uncovered a subset of predictive variables that could facilitate future development of effective stroke phenotyping algorithms. Discussion and conclusions By analyzing a broad array of patient data, the machine learning technologies held promise for improving detection of stroke diagnosis, thus unlocking high statistical power for subsequent genetic and genomic studies.

AB - Objective 1) To develop a machine learning approach for detecting stroke cases and subtypes from hospitalization data, 2) to assess algorithm performance and predictors on real-world data collected by a large-scale epidemiology study in the US; and 3) to identify directions for future development of high-precision stroke phenotypic signatures. Materials and methods We utilized 8,131 hospitalization events (ICD-9 codes 430±438) collected from the Greater Cincinnati/Northern Kentucky Stroke Study in 2005 and 2010. Detailed information from patients' medical records was abstracted for each event by trained research nurses. By analyzing the broad list of demographic and clinical variables, the machine learning algorithms predicted whether an event was a stroke case and, if so, the stroke subtype. The performance was validated on gold-standard labels adjudicated by stroke physicians, and results were compared with stroke classifications based on ICD-9 discharge codes, as well as labels determined by study nurses. Results The best performing machine learning algorithm achieved a performance of 88.57%/ 93.81%/92.80%/93.30%/89.84%/98.01% (accuracy/precision/recall/F-measure/area under ROC curve/area under precision-recall curve) on stroke case detection. For detecting stroke subtypes, the algorithm yielded an overall accuracy of 87.39% and greater than 85% precision on individual subtypes. The machine learning algorithms significantly outperformed the ICD-9 method on all measures (P value<0.001). Their performance was comparable to that of study nurses, with better tradeoff between precision and recall. The feature selection uncovered a subset of predictive variables that could facilitate future development of effective stroke phenotyping algorithms. Discussion and conclusions By analyzing a broad array of patient data, the machine learning technologies held promise for improving detection of stroke diagnosis, thus unlocking high statistical power for subsequent genetic and genomic studies.

UR - http://www.scopus.com/inward/record.url?scp=85042157386&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85042157386&partnerID=8YFLogxK

U2 - 10.1371/journal.pone.0192586

DO - 10.1371/journal.pone.0192586

M3 - Article

C2 - 29444182

AN - SCOPUS:85042157386

VL - 13

JO - PLoS One

JF - PLoS One

SN - 1932-6203

IS - 2

M1 - e0192586

ER -