TY - JOUR
T1 - Using machine learning to predict neurologic injury in venovenous extracorporeal membrane oxygenation recipients
T2 - An ELSO Registry analysis
AU - HERALD group
AU - Kalra, Andrew
AU - Bachina, Preetham
AU - Shou, Benjamin L.
AU - Hwang, Jaeho
AU - Barshay, Meylakh
AU - Kulkarni, Shreyas
AU - Sears, Isaac
AU - Eickhoff, Carsten
AU - Bermudez, Christian A.
AU - Brodie, Daniel
AU - Ventetuolo, Corey E.
AU - Whitman, Glenn J.R.
AU - Abbasi, Adeel
AU - Cho, Sung Min
AU - Kim, Bo Soo
AU - Hager, David
AU - Keller, Steven P.
AU - Bush, Errol L.
AU - Stephens, R. Scott
AU - Khanduja, Shivalika
AU - Kang, Jin Kook
AU - Chinedozi, Ifeanyi David
AU - Darby, Zachary
AU - Rando, Hannah J.
AU - Brown, Trish
AU - Kim, Jiah
AU - Wilcox, Christopher
AU - Leng, Albert
AU - Geeza, Andrew
AU - Akbar, Armaan F.
AU - Feng, Chengyuan Alex
AU - Zhao, David
AU - Sussman, Marc
AU - Mendez-Tellez, Pedro Alejandro
AU - Sun, Philip
AU - Capili, Karlo
AU - Riojas, Ramon
AU - Alejo, Diane
AU - Stephen, Scott
AU - Flaster, Harry
N1 - Publisher Copyright:
© 2024 The Author(s)
PY - 2024/10
Y1 - 2024/10
N2 - Background: Venovenous extracorporeal membrane oxygenation (VV-ECMO) is associated with acute brain injury (ABI), including central nervous system (CNS) ischemia (defined as ischemic stroke or hypoxic-ischemic brain injury [HIBI]) and intracranial hemorrhage (ICH). Data on prediction models for neurologic outcomes in VV-ECMO are limited. Methods: We analyzed adult (age ≥18 years) VV-ECMO patients in the Extracorporeal Life Support Organization (ELSO) Registry (2009-2021) from 676 centers. ABI was defined as CNS ischemia, ICH, brain death, and seizures. Data on 67 variables were extracted, including clinical characteristics and pre-ECMO/on-ECMO variables. Random forest, CatBoost, LightGBM, and XGBoost machine learning (ML) algorithms (10-fold leave-one-out cross-validation) were used to predict ABI. Feature importance scores were used to pinpoint the most important variables for predicting ABI. Results: Of 37,473 VV-ECMO patients (median age, 48.1 years; 63% male), 2644 (7.1%) experienced ABI, including 610 (2%) with CNS ischemia and 1591 (4%) with ICH. The areas under the receiver operating characteristic curve for predicting ABI, CNS ischemia, and ICH were 0.70, 0.68, and 0.70, respectively. The accuracy, positive predictive value, and negative predictive value for ABI were 85%, 19%, and 95%, respectively. ML identified higher center volume, pre-ECMO cardiac arrest, higher ECMO pump flow, and elevated on-ECMO serum lactate level as the most important risk factors for ABI and its subtypes. Conclusions: This is the largest study of VV-ECMO patients to use ML to predict ABI reported to date. Performance was suboptimal, likely due to lack of standardization of neuromonitoring/imaging protocols and data granularity in the ELSO Registry. Standardized neurologic monitoring and imaging are needed across ELSO centers to detect the true prevalence of ABI.
AB - Background: Venovenous extracorporeal membrane oxygenation (VV-ECMO) is associated with acute brain injury (ABI), including central nervous system (CNS) ischemia (defined as ischemic stroke or hypoxic-ischemic brain injury [HIBI]) and intracranial hemorrhage (ICH). Data on prediction models for neurologic outcomes in VV-ECMO are limited. Methods: We analyzed adult (age ≥18 years) VV-ECMO patients in the Extracorporeal Life Support Organization (ELSO) Registry (2009-2021) from 676 centers. ABI was defined as CNS ischemia, ICH, brain death, and seizures. Data on 67 variables were extracted, including clinical characteristics and pre-ECMO/on-ECMO variables. Random forest, CatBoost, LightGBM, and XGBoost machine learning (ML) algorithms (10-fold leave-one-out cross-validation) were used to predict ABI. Feature importance scores were used to pinpoint the most important variables for predicting ABI. Results: Of 37,473 VV-ECMO patients (median age, 48.1 years; 63% male), 2644 (7.1%) experienced ABI, including 610 (2%) with CNS ischemia and 1591 (4%) with ICH. The areas under the receiver operating characteristic curve for predicting ABI, CNS ischemia, and ICH were 0.70, 0.68, and 0.70, respectively. The accuracy, positive predictive value, and negative predictive value for ABI were 85%, 19%, and 95%, respectively. ML identified higher center volume, pre-ECMO cardiac arrest, higher ECMO pump flow, and elevated on-ECMO serum lactate level as the most important risk factors for ABI and its subtypes. Conclusions: This is the largest study of VV-ECMO patients to use ML to predict ABI reported to date. Performance was suboptimal, likely due to lack of standardization of neuromonitoring/imaging protocols and data granularity in the ELSO Registry. Standardized neurologic monitoring and imaging are needed across ELSO centers to detect the true prevalence of ABI.
KW - acute brain injury
KW - machine learning
KW - neurologic complications
KW - venovenous extracorporeal membrane oxygenation
UR - http://www.scopus.com/inward/record.url?scp=85200222965&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85200222965&partnerID=8YFLogxK
U2 - 10.1016/j.xjon.2024.06.013
DO - 10.1016/j.xjon.2024.06.013
M3 - Article
C2 - 39534333
AN - SCOPUS:85200222965
SN - 2666-2736
VL - 21
SP - 140
EP - 167
JO - JTCVS Open
JF - JTCVS Open
ER -