In:
PLOS ONE, Public Library of Science (PLoS), Vol. 17, No. 5 ( 2022-5-5), p. e0262895-
Abstract:
Improving the Intensive Care Unit (ICU) management network and building cost-effective and well-managed healthcare systems are high priorities for healthcare units. Creating accurate and explainable mortality prediction models helps identify the most critical risk factors in the patients’ survival/death status and early detect the most in-need patients. This study proposes a highly accurate and efficient machine learning model for predicting ICU mortality status upon discharge using the information available during the first 24 hours of admission. The most important features in mortality prediction are identified, and the effects of changing each feature on the prediction are studied. We used supervised machine learning models and illness severity scoring systems to benchmark the mortality prediction. We also implemented a combination of SHAP, LIME, partial dependence, and individual conditional expectation plots to explain the predictions made by the best-performing model (CatBoost). We proposed E-CatBoost, an optimized and efficient patient mortality prediction model, which can accurately predict the patients’ discharge status using only ten input features. We used eICU-CRD v2.0 to train and validate the models; the dataset contains information on over 200,000 ICU admissions. The patients were divided into twelve disease groups, and models were fitted and tuned for each group. The models’ predictive performance was evaluated using the area under a receiver operating curve (AUROC). The AUROC scores were 0.86 [std:0.02] to 0.92 [std:0.02] for CatBoost and 0.83 [std:0.02] to 0.91 [std:0.03] for E-CatBoost models across the defined disease groups; if measured over the entire patient population, their AUROC scores were 7 to 18 and 2 to 12 percent higher than the baseline models, respectively. Based on SHAP explanations, we found age, heart rate, respiratory rate, blood urine nitrogen, and creatinine level as the most critical cross-disease features in mortality predictions.
Type of Medium:
Online Resource
ISSN:
1932-6203
DOI:
10.1371/journal.pone.0262895
DOI:
10.1371/journal.pone.0262895.g001
DOI:
10.1371/journal.pone.0262895.g002
DOI:
10.1371/journal.pone.0262895.g003
DOI:
10.1371/journal.pone.0262895.g004
DOI:
10.1371/journal.pone.0262895.g005
DOI:
10.1371/journal.pone.0262895.g006
DOI:
10.1371/journal.pone.0262895.g007
DOI:
10.1371/journal.pone.0262895.t001
DOI:
10.1371/journal.pone.0262895.t002
DOI:
10.1371/journal.pone.0262895.t003
DOI:
10.1371/journal.pone.0262895.t004
DOI:
10.1371/journal.pone.0262895.t005
DOI:
10.1371/journal.pone.0262895.t006
DOI:
10.1371/journal.pone.0262895.s001
DOI:
10.1371/journal.pone.0262895.s002
DOI:
10.1371/journal.pone.0262895.s003
DOI:
10.1371/journal.pone.0262895.s004
DOI:
10.1371/journal.pone.0262895.s005
DOI:
10.1371/journal.pone.0262895.s006
DOI:
10.1371/journal.pone.0262895.s007
DOI:
10.1371/journal.pone.0262895.s008
DOI:
10.1371/journal.pone.0262895.s009
DOI:
10.1371/journal.pone.0262895.s010
DOI:
10.1371/journal.pone.0262895.s011
DOI:
10.1371/journal.pone.0262895.s012
DOI:
10.1371/journal.pone.0262895.s013
DOI:
10.1371/journal.pone.0262895.s014
DOI:
10.1371/journal.pone.0262895.s015
DOI:
10.1371/journal.pone.0262895.s016
DOI:
10.1371/journal.pone.0262895.s017
DOI:
10.1371/journal.pone.0262895.s018
DOI:
10.1371/journal.pone.0262895.s019
DOI:
10.1371/journal.pone.0262895.s020
DOI:
10.1371/journal.pone.0262895.s021
DOI:
10.1371/journal.pone.0262895.s022
DOI:
10.1371/journal.pone.0262895.s023
DOI:
10.1371/journal.pone.0262895.s024
DOI:
10.1371/journal.pone.0262895.s025
DOI:
10.1371/journal.pone.0262895.s026
DOI:
10.1371/journal.pone.0262895.s027
DOI:
10.1371/journal.pone.0262895.s028
DOI:
10.1371/journal.pone.0262895.s029
DOI:
10.1371/journal.pone.0262895.s030
DOI:
10.1371/journal.pone.0262895.s031
DOI:
10.1371/journal.pone.0262895.s032
DOI:
10.1371/journal.pone.0262895.s033
DOI:
10.1371/journal.pone.0262895.s034
DOI:
10.1371/journal.pone.0262895.s035
DOI:
10.1371/journal.pone.0262895.s036
DOI:
10.1371/journal.pone.0262895.s037
DOI:
10.1371/journal.pone.0262895.s038
DOI:
10.1371/journal.pone.0262895.s039
DOI:
10.1371/journal.pone.0262895.s040
DOI:
10.1371/journal.pone.0262895.s041
DOI:
10.1371/journal.pone.0262895.s042
Language:
English
Publisher:
Public Library of Science (PLoS)
Publication Date:
2022
detail.hit.zdb_id:
2267670-3
Permalink