Prediction of prolonged mechanical ventilation in the intensive care unit via machine learning: a COVID-19 perspective

Marianna Weaver; Dylan A Goodin; Hunter A Miller; Dipan Karmali; Apurv A Agarwal; Hermann B Frieboes; Sally A Suliman

doi:10.1038/s41598-024-81980-0

Prediction of prolonged mechanical ventilation in the intensive care unit via machine learning: a COVID-19 perspective

Sci Rep. 2024 Dec 4;14(1):30173. doi: 10.1038/s41598-024-81980-0.

Authors

Marianna Weaver^#¹, Dylan A Goodin^#², Hunter A Miller², Dipan Karmali¹, Apurv A Agarwal¹, Hermann B Frieboes^#^{3

4

5

6}, Sally A Suliman^#^{7

8}

Affiliations

¹ Division of Pulmonary Medicine, University of Louisville, Louisville, KY, 40292, USA.
² Department of Bioengineering, University of Louisville, Lutz Hall 419, Louisville, KY, 40292, USA.
³ Department of Bioengineering, University of Louisville, Lutz Hall 419, Louisville, KY, 40292, USA. hbfrie01@louisville.edu.
⁴ Department of Pharmacology/Toxicology, University of Louisville, Louisville, KY, 40292, USA. hbfrie01@louisville.edu.
⁵ Center for Predictive Medicine, University of Louisville, Louisville, KY, 40292, USA. hbfrie01@louisville.edu.
⁶ James Graham Brown Cancer Center, University of Louisville, Louisville, KY, 40292, USA. hbfrie01@louisville.edu.
⁷ University of Arizona Medical Center Phoenix, Phoenix, AZ, 85004, USA.
⁸ Formerly at: Division of Pulmonary Medicine, University of Louisville, Louisville, KY, 40292, USA.

^# Contributed equally.

Abstract

Early recognition of risk factors for prolonged mechanical ventilation (PMV) could allow for early clinical interventions, prevention of secondary complications such as nosocomial infections, and effective triage of hospital resources. This study tested the hypothesis that an ensemble machine learning (ML) analysis of clinical data at time of intubation could identify patients at risk of PMV, using a COVID-19 dataset to classify patients into PMV (> 14 days) and non-PMV (≤ 14 days) groups. While several factors are known to cause PMV, including acid-base, weakness, and delirium, lesser-utilized but routinely measured parameters such as platelet count, glucose levels and fevers may also be relevant. Patient data from a single University Hospital were analyzed via the ML workflow to predict patients at risk of PMV and identify key clinical markers. Model performance was evaluated on a chronologically distinct cohort. The ML workflow identified patients at risk of PMV with AUROC_TRAIN=0.960 (F1_TRAIN = 0.935) and AUROC_TEST=0.804 (F1_TEST = 0.800). Top key features for classification included glucose, platelet count, temperature, LVEF, bicarbonate (arterial blood gas), and BMI. Data analysis at intubation time via the proposed workflow offers the potential to accurately predict patients at risk of PMV, with the goal to improve patient management and triage of hospital resources.

Keywords: COVID-19; Machine learning; Mechanical intubation; Predictive modeling.

MeSH terms

Aged
COVID-19* / epidemiology
Female
Humans
Intensive Care Units*
Machine Learning*
Male
Middle Aged
Respiration, Artificial*
Risk Factors
SARS-CoV-2 / isolation & purification