Enhanced neonatal surgical site infection prediction model utilizing statistically and clinically significant variables in combination with a machine learning algorithm

Marisa A. Bartz-Kurycki, Charles Green, Kathryn T. Anderson, Adam C. Alder, Brian T. Bucher, Robert A. Cina, Ramin Jamshidi, Robert T. Russell, Regan Williams, Kuo Jen Tsao

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Background: Machine-learning can elucidate complex relationships/provide insight to important variables for large datasets. This study aimed to develop an accurate model to predict neonatal surgical site infections (SSI) using different statistical methods. Methods: The 2012–2015 National Surgical Quality Improvement Program-Pediatric for neonates was utilized for development and validations models. The primary outcome was any SSI. Models included different algorithms: full multiple logistic regression (LR), a priori clinical LR, random forest classification (RFC), and a hybrid model (combination of clinical knowledge and significant variables from RF) to maximize predictive power. Results: 16,842 patients (median age 18 days, IQR 3–58) were included. 542 SSIs (4%) were identified. Agreement was observed for multiple covariates among significant variables between models. Area under the curve for each model was similar (full model 0.65, clinical model 0.67, RF 0.68, hybrid LR 0.67); however, the hybrid model utilized the fewest variables (18). Conclusions: The hybrid model had similar predictability as other models with fewer and more clinically relevant variables. Machine-learning algorithms can identify important novel characteristics, which enhance clinical prediction models.

Original languageEnglish (US)
Pages (from-to)764-777
Number of pages14
JournalAmerican Journal of Surgery
Volume216
Issue number4
DOIs
StatePublished - Oct 1 2018

Fingerprint

Surgical Wound Infection
Logistic Models
Quality Improvement
Area Under Curve
Newborn Infant
Pediatrics
Machine Learning

All Science Journal Classification (ASJC) codes

  • Surgery

Cite this

Enhanced neonatal surgical site infection prediction model utilizing statistically and clinically significant variables in combination with a machine learning algorithm. / Bartz-Kurycki, Marisa A.; Green, Charles; Anderson, Kathryn T.; Alder, Adam C.; Bucher, Brian T.; Cina, Robert A.; Jamshidi, Ramin; Russell, Robert T.; Williams, Regan; Tsao, Kuo Jen.

In: American Journal of Surgery, Vol. 216, No. 4, 01.10.2018, p. 764-777.

Research output: Contribution to journalArticle

Bartz-Kurycki, Marisa A. ; Green, Charles ; Anderson, Kathryn T. ; Alder, Adam C. ; Bucher, Brian T. ; Cina, Robert A. ; Jamshidi, Ramin ; Russell, Robert T. ; Williams, Regan ; Tsao, Kuo Jen. / Enhanced neonatal surgical site infection prediction model utilizing statistically and clinically significant variables in combination with a machine learning algorithm. In: American Journal of Surgery. 2018 ; Vol. 216, No. 4. pp. 764-777.
@article{5bf662d3cee245e1975eb35f4983d231,
title = "Enhanced neonatal surgical site infection prediction model utilizing statistically and clinically significant variables in combination with a machine learning algorithm",
abstract = "Background: Machine-learning can elucidate complex relationships/provide insight to important variables for large datasets. This study aimed to develop an accurate model to predict neonatal surgical site infections (SSI) using different statistical methods. Methods: The 2012–2015 National Surgical Quality Improvement Program-Pediatric for neonates was utilized for development and validations models. The primary outcome was any SSI. Models included different algorithms: full multiple logistic regression (LR), a priori clinical LR, random forest classification (RFC), and a hybrid model (combination of clinical knowledge and significant variables from RF) to maximize predictive power. Results: 16,842 patients (median age 18 days, IQR 3–58) were included. 542 SSIs (4{\%}) were identified. Agreement was observed for multiple covariates among significant variables between models. Area under the curve for each model was similar (full model 0.65, clinical model 0.67, RF 0.68, hybrid LR 0.67); however, the hybrid model utilized the fewest variables (18). Conclusions: The hybrid model had similar predictability as other models with fewer and more clinically relevant variables. Machine-learning algorithms can identify important novel characteristics, which enhance clinical prediction models.",
author = "Bartz-Kurycki, {Marisa A.} and Charles Green and Anderson, {Kathryn T.} and Alder, {Adam C.} and Bucher, {Brian T.} and Cina, {Robert A.} and Ramin Jamshidi and Russell, {Robert T.} and Regan Williams and Tsao, {Kuo Jen}",
year = "2018",
month = "10",
day = "1",
doi = "10.1016/j.amjsurg.2018.07.041",
language = "English (US)",
volume = "216",
pages = "764--777",
journal = "American Journal of Surgery",
issn = "0002-9610",
publisher = "Elsevier Inc.",
number = "4",

}

TY - JOUR

T1 - Enhanced neonatal surgical site infection prediction model utilizing statistically and clinically significant variables in combination with a machine learning algorithm

AU - Bartz-Kurycki, Marisa A.

AU - Green, Charles

AU - Anderson, Kathryn T.

AU - Alder, Adam C.

AU - Bucher, Brian T.

AU - Cina, Robert A.

AU - Jamshidi, Ramin

AU - Russell, Robert T.

AU - Williams, Regan

AU - Tsao, Kuo Jen

PY - 2018/10/1

Y1 - 2018/10/1

N2 - Background: Machine-learning can elucidate complex relationships/provide insight to important variables for large datasets. This study aimed to develop an accurate model to predict neonatal surgical site infections (SSI) using different statistical methods. Methods: The 2012–2015 National Surgical Quality Improvement Program-Pediatric for neonates was utilized for development and validations models. The primary outcome was any SSI. Models included different algorithms: full multiple logistic regression (LR), a priori clinical LR, random forest classification (RFC), and a hybrid model (combination of clinical knowledge and significant variables from RF) to maximize predictive power. Results: 16,842 patients (median age 18 days, IQR 3–58) were included. 542 SSIs (4%) were identified. Agreement was observed for multiple covariates among significant variables between models. Area under the curve for each model was similar (full model 0.65, clinical model 0.67, RF 0.68, hybrid LR 0.67); however, the hybrid model utilized the fewest variables (18). Conclusions: The hybrid model had similar predictability as other models with fewer and more clinically relevant variables. Machine-learning algorithms can identify important novel characteristics, which enhance clinical prediction models.

AB - Background: Machine-learning can elucidate complex relationships/provide insight to important variables for large datasets. This study aimed to develop an accurate model to predict neonatal surgical site infections (SSI) using different statistical methods. Methods: The 2012–2015 National Surgical Quality Improvement Program-Pediatric for neonates was utilized for development and validations models. The primary outcome was any SSI. Models included different algorithms: full multiple logistic regression (LR), a priori clinical LR, random forest classification (RFC), and a hybrid model (combination of clinical knowledge and significant variables from RF) to maximize predictive power. Results: 16,842 patients (median age 18 days, IQR 3–58) were included. 542 SSIs (4%) were identified. Agreement was observed for multiple covariates among significant variables between models. Area under the curve for each model was similar (full model 0.65, clinical model 0.67, RF 0.68, hybrid LR 0.67); however, the hybrid model utilized the fewest variables (18). Conclusions: The hybrid model had similar predictability as other models with fewer and more clinically relevant variables. Machine-learning algorithms can identify important novel characteristics, which enhance clinical prediction models.

UR - http://www.scopus.com/inward/record.url?scp=85050874585&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85050874585&partnerID=8YFLogxK

U2 - 10.1016/j.amjsurg.2018.07.041

DO - 10.1016/j.amjsurg.2018.07.041

M3 - Article

VL - 216

SP - 764

EP - 777

JO - American Journal of Surgery

JF - American Journal of Surgery

SN - 0002-9610

IS - 4

ER -