Machine learning based algorithms to impute PaO 2 from SpO 2 values and development of an online calculator

We created an online calculator using machine learning (ML) algorithms to impute the partial pressure of oxygen (PaO )/fraction of delivered oxygen (FiO ) ratio using the non-invasive peripheral saturation of oxygen (SpO ) and compared the accuracy of the ML models we developed to published equation...

Full description

Saved in:
Bibliographic Details
Published in:Scientific reports 2022-05, Vol.12 (1), p.8235
Main Authors: Ren, Shuangxia, Zupetic, Jill A, Tabary, Mohammadreza, DeSensi, Rebecca, Nouraie, Mehdi, Lu, Xinghua, Boyce, Richard D, Lee, Janet S
Format: Article
Language:English
Subjects:
Algorithms
Humans
Machine Learning
Oximetry - methods
Oxygen
Severity of Illness Index
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We created an online calculator using machine learning (ML) algorithms to impute the partial pressure of oxygen (PaO )/fraction of delivered oxygen (FiO ) ratio using the non-invasive peripheral saturation of oxygen (SpO ) and compared the accuracy of the ML models we developed to published equations. We generated three ML algorithms (neural network, regression, and kernel-based methods) using seven clinical variable features (N = 9900 ICU events) and subsequently three features (N = 20,198 ICU events) as input into the models. Data from mechanically ventilated ICU patients were obtained from the publicly available Medical Information Mart for Intensive Care (MIMIC III) database and used for analysis. Compared to seven features, three features (SpO , FiO and PEEP) were sufficient to impute PaO from the SpO . Any of the ML models enabled imputation of PaO from the SpO with lower error and showed greater accuracy in predicting PaO /FiO  ≤ 150 compared to the previously published log-linear and non-linear equations. To address potential hidden hypoxemia that occurs more frequently in Black patients, we conducted sensitivity analysis and show ML models outperformed published equations in both Black and White patients. Imputation using data from an independent validation cohort of ICU patients (N = 133) showed greater accuracy with ML models.
ISSN:2045-2322