SCI 발표논문

SCI 발표논문

← SCI 발표논문 목록

JMIR Form Res. 2023;7:e44763.

Machine Learning Algorithms Predict Successful Weaning From Mechanical Ventilation Before Intubation: Retrospective Analysis From the Medical Information Mart for Intensive Care IV Database.

, , , , , , , ,

제1저자 김진철, 김윤관 · 교신저자 이한, 김형준

PMID 37962939 PMC10685278 DOI 10.2196/44763 IF 2.4 호흡기내과
소속 (9)
  1. Division of Hematology-Oncology, Department of Internal Medicine, Inha University College of Medicine and Hospital, Incheon, Republic of Korea.
  2. Department of the Technology Development, Seers Technology Co, Ltd, Seongnam, Republic of Korea.
  3. Crowdworks Co, Ltd, Seoul, Republic of Korea.
  4. Healthcare Artificial Intelligence Team, National Cancer Center, Goyang, Republic of Korea.
  5. Department of Preventive Medicine, Yonsei University Wonju College of Medicine, Wonju, Republic of Korea.
  6. Department of Biomedical Systems Informatics, Yonsei University College of Medicine, Yongin, Republic of Korea.
  7. Asan Medical Center, Asan Institute for Life Sciences, Seoul, Republic of Korea.
  8. Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, Seoul National University Bundang Hospital, Seongnam, Republic of Korea.
  9. Department of Internal Medicine, Seoul National University College of Medicine, Seoul, Republic of Korea.

Abstract

BACKGROUND The prediction of successful weaning from mechanical ventilation (MV) in advance of intubation can facilitate discussions regarding end-of-life care before unnecessary intubation.

OBJECTIVE We aimed to develop a machine learning-based model that predicts successful weaning from ventilator support based on routine clinical and laboratory data taken before or immediately after intubation.

METHODS We used the Medical Information Mart for Intensive Care IV database, which is an open-access database covering 524,740 admissions of 382,278 patients in Beth Israel Deaconess Medical Center, United States, from 2008 to 2019. We selected adult patients who underwent MV in the intensive care unit (ICU). Clinical and laboratory variables that are considered relevant to the prognosis of the patient in the ICU were selected. Data collected before or within 24 hours of intubation were used to develop machine learning models that predict the probability of successful weaning within 14 days of ventilator support. Developed models were integrated into an ensemble model. Performance metrics were calculated by 5-fold cross-validation for each model, and a permutation feature importance and Shapley additive explanations analysis was conducted to better understand the impacts of individual variables on outcome prediction.

RESULTS Of the 23,242 patients, 19,025 (81.9%) patients were successfully weaned from MV within 14 days. Using the preselected 46 clinical and laboratory variables, the area under the receiver operating characteristic curve of CatBoost classifier, random forest classifier, and regularized logistic regression classifier models were 0.860 (95% CI 0.852-0.868), 0.855 (95% CI 0.848-0.863), and 0.823 (95% CI 0.813-0.832), respectively. Using the ensemble voting classifier using the 3 models above, the final model revealed the area under the receiver operating characteristic curve of 0.861 (95% CI 0.853-0.869), which was significantly better than that of Simplified Acute Physiology Score II (0.749, 95% CI 0.742-0.756) and Sequential Organ Failure Assessment (0.588, 95% CI 0.566-0.609). The top features included lactate and anion gap. The model's performance achieved a plateau with approximately the top 21 variables.

CONCLUSIONS We developed machine learning algorithms that can predict successful weaning from MV in advance to intubation in the ICU. Our models can aid the appropriate management for patients who hesitate to decide on ventilator support or meaningless end-of-life care.

©Jinchul Kim, Yun Kwan Kim, Hyeyeon Kim, Hyojung Jung, Soonjeong Koh, Yujeong Kim, Dukyong Yoon, Hahn Yi, Hyung-Jun Kim. Originally published in JMIR Formative Research (https://formative.jmir.org), 14.11.2023.

Keywords

algorithmsclinical decision-makingintensive care unitsnoninvasive ventilationorgan dysfunction scores

Figures

Figure 1
Figure 1 Flowchart of the patient selection process. Patients with evidence of endotracheal intubation were identified in the Medical Information Mart for Intensive Care IV database. After careful selection, patients were divided into 2 groups according to whether they had been successfully weaned from MV within 14 days of intubation or not. MV: mechanical ventilation; SAPS II: Simplified Acute Physiology Score II; SOFA: Sequential Organ Failure Assessment.
Figure 2
Figure 2 (A) Receiver operating characteristic curves and (B) precision-recall curves of the developed machine learning models to predict successful weaning from mechanical ventilation within 14 days of intubation. The area under receiver operating characteristics curves (AUROCs) and area under precision-recall curves (AUPRCs) and their 95% CIs for all 6 models are shown in the legends. For comparability, the random classifier is indicated by a black dashed-dotted line. CBC: CatBoost classifier; RFC: random forest classifier; RLRC: regularized logistic regression classifier; SAPS II: Simplified Acute Physiology Score II; SOFA: Sequential Organ Failure Assessment; VC: voting classifier.
Figure 3
Figure 3 Important explanatory variables for the ensemble voting classifier model. The means and SD of the top 21 important explanatory variables for the voting classifier model using 5-fold cross-validation and test set permutation are indicated by bars and error bars, respectively. BUN: blood urea nitrogen; GCS: Glasgow Coma Scale; ICU: intensive care unit; NL: neutrophil-to-lymphocyte; WBC: white blood cell.
Figure 4
Figure 4 The Beeswarm summary plot in the SHAP package with a randomly selected test set among a 5-fold cross-validation. This figure indicates which explanatory variables have the most significant influence on the model’s predicted value and the tendency of the model’s outcome probability as a function of the original value of an explanatory variable simultaneously. Each colored dot represents one variable value for a patient, for which larger values are presented in red and smaller values in blue. The SHAP value, a simplified and computer-friendly Shapley value, is shown on the horizontal axis and is a surrogate of the contribution of a variable value to the model output. BUN: blood urea nitrogen; GCS: Glasgow Coma Scale; SHAP: Shapley additive explanations.
Figure 5
Figure 5 Performance metrics of the ensemble voting classifier model according to the addition of explanatory variables. (A) Cohen κ, (B) AUROC, (C) F1-score, and (D) balanced accuracy of the ensemble voting classifier are shown as the number of explanatory variables in the training sets in descending order of feature importance coefficients. Mean (dot) and SD (error bar) were computed from the data sets by 5-fold cross-validation. The model reached its plateau performance in all 4 metrics with approximately 21 variables. AUROC: area under the receiver operating characteristics curve; RFI: relative feature importance.

그림 출처: PubMed Central 공개 논문 (PMC10685278)

Conflict of interest statement

Conflicts of Interest: None declared.

그 밖의 정보
출판 형태
Journal Article
저널
JMIR formative research · eISSN 2561-326X
전자 출판
2023-11-14
참고문헌
48편