Prediction of Duration of Mechanical Ventilation in Acute Hypoxemic Respiratoty Failure (PREMIER)

Jesus Villar

Status: Active, not recruiting

Conditions: Acute Hypoxemic Respiratory Failure

Treatments: Other: Machine learning and logistic regression for the training/testing cohort and validation cohort

Study type: Observational

Funder types: Other

Identifiers

NCT06815523

PIFIISC24

Details and patient eligibility

About

Acute hypoxemic respiratory failure (AHRF) is a common cause of admission in intensive care units (ICUs) worldwide. We will assess machine learning (ML) techniques for prediction of prolonged duration (> or = to 7 days) of mechanical ventilation (MV) in 1,241 patients enrolled in the PANDORA study in Spain. The study was registered with ClinalTrials.gov (NCT03145974). Our aim is to identify a model with the minimum number of variables that predict duration of prolonged ventilation in AHRF patients using data as early as from the first 48 hours with machine learning algorithms.

Full description

Acute hypoxemic respiratory failure (AHRF) is the most common cause of admission in intensive care units (ICUs) worldwide. The investigators will assess the value of machine learning (ML) techniques for prediction of prolonged duration (> or equeal to 7 days) of mechanical ventilation (MV) in 1,241 patients enrolled in the PANDORA study in Spain. Few studies have investigated the prediction of prolonged MV in patients with AHRF.

For model training and testing, the investigators will extract data from random pateints from the first 2 days after diagnosis of AHRF. The investigators had a database with 2,000,000 anonymized and dissociated demographics and clinically relevant data from 1,241 patients with AHRF from 22 hospitals in Spain. The investigators will follow the TRIPOD guidelines for prediction models. The investigators will screen relevant collected variables using a genetic algorithm variable selection to achieve parsimony. We will use 5-fold corss-validation in the data set of patients with data at T0, T24 and T48. We will use 25% of patients randomly selected for evaluation of the model.

Enrollment: 1,241 patients

Sex: All

Ages: 18+ years old

Volunteers: No Healthy Volunteers

Inclusion criteria

• enotracheal intubation puls mechanical ventilation
• PaO2/FiO2 ratio <or = 300 mmHg under MV with PEEP >or =5 and FiO2 >or = 0.3

Exclusion criteria

• Brain death patients

Trial design

1,241 participants in 2 patient groups

Derivation/testing cohort

Description:

The investigators will use a chort of 75% of patients, randomly selected, with data at T0, T24 and T48 after diagnosis of acute hypoxemic respiratory failure (AHRF). We will apply machine learning approaches.

Treatment:

Other: Machine learning and logistic regression for the training/testing cohort and validation cohort

Validation hohort

Description:

we will use 25% of unseen patients, randomly selected, with data at T0, T24 and T48 after diagnosis of AHRF.

Treatment:

Other: Machine learning and logistic regression for the training/testing cohort and validation cohort