Respiratory Signal Research Articles

Transfer Learning‐Assisted Porous Polymer Humidity Sensor for Powered Air‐Purifying Mask

Masks protect respiratory health in the coal, oil, and gas industries. However, prolonged exposure to high humidity inside masks can cause discomfort and increase the risk of respiratory diseases. To address the issue, in this work, a porous polymer humidity sensor suitable for monitoring respiratory changes in high humidity environment is prepared and integrated into the power air‐purifying mask. Combined with the transfer learning algorithm, the problem with respiratory resistance caused by delayed air supply due to signal processing in the traditional power air‐purifying mask is overcome, and the respiratory signal is effectively predicted in advance, so as to achieve real‐time on‐demand air supply. The brief process is as follows: A porous polymer humidity sensor with fast response/recovery times (2.94/4.86 s) at 95% relative humidity (RH) is developed for monitoring respiratory rate changes in high humidity environments. By integrating this sensor with a powered air‐purifying system and employing a transfer learning algorithm, the system predicts respiratory signals and adjusts the air supply in real‐time. This reduces mask humidity from 95% RH to 40–50% RH in 1.8 s, ensuring comfortable, low‐resistance breathing for workers.This work will be conductive to the development of comfortable poweredair‐purifying respirators with low resistance and humidity.

Use of respiratory signal analysis to assess severity of Brachycephalic Obstructive Airway Syndrome (BOAS) in dogs

Brachycephalic Obstructive Airway Syndrome (BOAS) is a potentially life-threatening condition that can be challenging to diagnose and grade objectively. The aim of this study was to investigate the use of respiratory signal analysis to assess severity of BOAS in dogs. Hundred and seventeen client-owned dogs of brachycephalic and non-brachycephalic breeds were enrolled. Respiratory sounds were recorded using an electronic stethoscope before and after a 3-minute exercise test (ET). Dogs were assigned a BOAS severity grade (BOAS 0–3) using a validated respiratory functional grading scheme. Signal analysis techniques were used to identify seven sound variables. Analysis of variance (ANOVA) was used to investigate associations between variables and BOAS severity and receiver operating characteristic (ROC) curves to assess the diagnostic efficacy of each sound variable. For each sound variable, there was a significant association with BOAS grade. An increase in BOAS grade resulted in greater sound magnitude in the frequency spectrum (0–1000 Hz), and in a greater contribution of lower frequencies (170–260 Hz). The variable “Peak 1” had the best performance in predicting BOAS negative (BOAS 0 +1) versus BOAS positive dogs (BOAS 2 + 3) before the ET; area under the curve (AUC) = 76.6 % (95 % confidence interval 67.4–85.8 %), whereas the variable “Valley 1” had the highest predictive value after the ET; AUC = 87.8 % (95 % confidence interval 81.4–94.3 %). Respiratory signal analysis has good potential for assessing BOAS severity and could be valuable for clinicians in clinical decision processes and for breeders when selecting suitable breeding dogs.

A high precision vital signs detection method based on millimeter wave radar

Millimeter wave (mmWave) radar technology has potential applications in vital signs detection and medicine. In order to minimize the influence of human micro-movements and respiratory harmonics on heart rate estimation, the vital signs detection method based on mmWave radar is studied in this paper. First, we use median filtering to eliminate baseline drift caused by human micromotion. Next, a differential recursive least squares multiple classification (DR-MUSIC) algorithm is proposed based on the combination of recursive least squares-based adaptive filter (RLS) and multiple signal classification (MUSIC) algorithm. This algorithm effectively suppresses respiratory harmonics and separates respiratory and heartbeat signals. Finally, heart rate value can be precisely estimated using spectral peak search. We invite a number of people to participate in the experiment, which demonstrate that the method successfully suppresse the impact of respiratory harmonics at low SNR. The error rate between the estimated heart rate and the reference heart rate is only 1.69% to 2.61%, which is significantly better than the existing algorithms.

Vital signs detection of moving targets using FMCW radar

Abstract Respiratory and heartbeat rates are crucial indicators for human health assessment. Compared to contact-based measurements, millimeter-wave radar detection of these vital signs avoids the discomfort caused by physical contact and better protects personal privacy, making it highly valuable for home health monitoring. The use of millimeter-wave radar for vital sign detection of the human body is mostly focused on targets in a stationary state at present. However, the human body may sway or even move during actual detection. This article proposes a non-contact vital sign detection method for moving targets. Compared with methods for detecting vital signs of stationary targets, detecting vital signs of moving targets requires determining the range bin where the targets are located continuously and extracting target phase information. The noise components such as movement and sway contained in the phase signal need to be removed. In this paper, moving target indication is used to remove static components, an adaptive range bin selection method is proposed to determine the range bin where the targets are located, and range bin selection fluctuation is smoothed using a moving average filter. The wavelet transform is used to decompose the phase signal, remove swaying noise based on autocorrelation function, and reconstruct the life signal for different scale factors. A bandpass filter is used to separate the respiratory and heartbeat signals, and a notch filter is designed to suppress respiratory harmonic signals. The experimental results show that the proposed method can separate vital signs signals from the phase signals of moving targets, achieving detection of respiration and heartbeat rate. The average accuracy of respiration and heartbeat rate detection is 94.7% and 95.5%, respectively.

Electrocardiogram and respiration recordings show a reduction in the physical burden on professional caregivers when performing care tasks with a transfer support robot

ABSTRACT In this study, we assessed the physical burden on professional caregivers when using a transfer support robot, “Hug,” to transfer and move a care recipient. We compared heart rate (HR), heart rate variability (HRV), and the time-series synchronization between HRV and respiration in caregivers using the robot or a normal wheelchair as control. Under both conditions, a commercial wearable device was used to simultaneously obtain electrocardiogram and respiration signals while performing care tasks, which comprised transfer from a bed to a wheelchair or to the robot, movement to a remote location, and transfer to a chair. We found that HR was significantly higher and HRV and wavelet coherence were significantly lower in the control than with Hug during at least the first 30 seconds of rest after the task (p < 0.05). This suggests that Hug could reduce the increase in HR and thereby maintain a more constant rhythm between HRV and respiration during care tasks. A post-task questionnaire also revealed that caregivers felt the physical burden of the tasks was reduced using Hug. Our results suggest that the introduction of transfer support robots can reduce the physical burden for caregivers of daily transfer assistance tasks.

Remote physiological signal recovery with efficient spatio-temporal modeling.

Contactless physiological signal measurement has great applications in various fields, such as affective computing and health monitoring. Physiological measurements based on remote photoplethysmography (rPPG) are realized by capturing the weak periodic color changes. The changes are caused by the variation in the light absorption of skin surface during systole and diastole stages of a functioning heart. This measurement mode has advantages of contactless measurement, simple operation, low cost, etc. In recent years, several deep learning-based rPPG measurement methods have been proposed. However, the features learned by deep learning models are vulnerable to motion and illumination artefacts, and are unable to fully exploit the intrinsic temporal characteristics of the rPPG. This paper presents an efficient spatiotemporal modeling-based rPPG recovery method for physiological signal measurements. First, two modules are utilized in the rPPG task: 1) 3D central difference convolution for temporal context modeling with enhanced representation and generalization capacity, and 2) Huber loss for robust intensity-level rPPG recovery. Second, a dual branch structure for both motion and appearance modeling and a soft attention mask are adapted to take full advantage of the central difference convolution. Third, a multi-task setting for joint cardiac and respiratory signals measurements is introduced to benefit from the internal relevance between two physiological signals. Last, extensive experiments performed on three public databases show that the proposed method outperforms prior state-of-the-art methods with the Pearson's correlation coefficient higher than 0.96 on all three datasets. The generalization ability of the proposed method is also evaluated by cross-database and video compression experiments. The effectiveness and necessity of each module are confirmed by ablation studies.

Respiratory Rate Estimation from Thermal Video Data Using Spatio-Temporal Deep Learning.

Thermal videos provide a privacy-preserving yet information-rich data source for remote health monitoring, especially for respiration rate (RR) estimation. This paper introduces an end-to-end deep learning approach to RR measurement using thermal video data. A detection transformer (DeTr) first finds the subject's facial region of interest in each thermal frame. A respiratory signal is estimated from a dynamically cropped thermal video using 3D convolutional neural networks and bi-directional long short-term memory stages. To account for the expected phase shift between the respiration measured using a respiratory effort belt vs. a facial video, a novel loss function based on negative maximum cross-correlation and absolute frequency peak difference was introduced. Thermal recordings from 22 subjects, with simultaneous gold standard respiratory effort measurements, were studied while sitting or standing, both with and without a face mask. The RR estimation results showed that our proposed method outperformed existing models, achieving an error of only 1.6 breaths per minute across the four conditions. The proposed method sets a new State-of-the-Art for RR estimation accuracy, while still permitting real-time RR estimation.

Implementation of Wearable Wireless Chest Patch Monitoring Terminal

A wearable wireless chest patch monitoring terminal is designed to realize the acquisition, processing, and wireless transmission of ECG, respiration, and body temperature signals. The analog front-end ADS1292R, which integrates respiratory impedance and ECG front-end, is utilized to collect human ECG and respiratory signals. The body temperature is collected using a low-power, high-precision digital temperature sensor MAX30208. A filter algorithm for signal processing and wireless transmission is designed through a low-power nRF52840 Bluetooth SoC with an Arm Cortex-M4F kernel. The experimental results show that the designed monitoring terminal can monitor the ECG, respiration, and body temperature parameters of the human body in real-time and send the monitoring results via Bluetooth, with a continuous working time of more than 13 hours. The wearable wireless chest patch monitoring terminal features good portability, long standby time, and high measurement accuracy, and it has promising application prospects in the fields of family health monitoring, mobile medical treatment, and smart healthcare.

Functionalized Face Masks as Smart Wearable Sensors for Multiple Sensing.

Wearable sensors provide continuous physiological information and measure deviations from healthy baselines, resulting in the potential to personalize health management and diagnosis of diseases. With the emergence of the COVID-19 pandemic, functionalized face masks as smart wearable sensors for multimodal and/or multiplexed measurement of physical parameters and biochemical markers have become the general population for physiological health management and environmental pollution monitoring. This Review examines recent advances in applications of smart face masks based on implantation of digital technologies and electronics and focuses on respiratory monitoring applications with the advantages of autonomous flow driving, enrichment enhancement, real-time monitoring, diversified sensing, and easily accessible. In particular, the detailed introduction of diverse respiratory signals including physical, inhalational, and exhalant signals and corresponding associations of health management and environmental pollution is presented. In the end, we also provide a personal perspective on future research directions and the remaining challenges in the commercialization of smart functionalized face masks for multiple sensing.

PPG2RespNet: a deep learning model for respirational signal synthesis and monitoring from photoplethysmography (PPG) signal.

Breathing conditions affect a wide range of people, including those with respiratory issues like asthma and sleep apnea. Smartwatches with photoplethysmogram (PPG) sensors can monitor breathing. However, current methods have limitations due to manual parameter tuning and pre-defined features. To address this challenge, we propose the PPG2RespNet deep-learning framework. It draws inspiration from the UNet and UNet + + models. It uses three publicly available PPG datasets (VORTAL, BIDMC, Capnobase) to autonomously and efficiently extract respiratory signals. The datasets contain PPG data from different groups, such as intensive care unit patients, pediatric patients, and healthy subjects. Unlike conventional U-Net architectures, PPG2RespNet introduces layered skip connections, establishing hierarchical and dense connections for robust signal extraction. The bottleneck layer of the model is also modified to enhance the extraction of latent features. To evaluate PPG2RespNet's performance, we assessed its ability to reconstruct respiratory signals and estimate respiration rates. The model outperformed other models in signal-to-signal synthesis, achieving exceptional Pearson correlation coefficients (PCCs) with ground truth respiratory signals: 0.94 for BIDMC, 0.95 for VORTAL, and 0.96 for Capnobase. With mean absolute errors (MAE) of 0.69, 0.58, and 0.11 for the respective datasets, the model exhibited remarkable precision in estimating respiration rates. We used regression and Bland-Altman plots to analyze the predictions of the model in comparison to the ground truth. PPG2RespNet can thus obtain high-quality respiratory signals non-invasively, making it a valuable tool for calculating respiration rates.

Morphological Prediction of CPAP Associated Acute Respiratory Instability.

Multiple mechanisms are involved in the pathogenesis of obstructive sleep apnea (OSA). Elevated loop gain is a key target for precision OSA care and may be associated with treatment intolerance when the upper airway is the sole therapeutic target. Morphological or computational estimation of LG is not yet widely available or fully validated - there is a need for improved phenotyping/endotyping of apnea to advance its therapy and prognosis. This study proposes a new algorithm to assess self-similarity as a signature of elevated loop gain using respiratory effort signals and presents its use to predict the probability of acute failure (high residual event counts) of continuous positive airway pressure (CPAP) therapy. Effort signals from 2145 split-night polysomnography studies from the Massachusetts General Hospital were analyzed for SS and used to predict acute CPAP therapy effectiveness. Logistic regression models were trained and evaluated using 5-fold cross-validation. Receiver operating characteristic (ROC) and precision-recall (PR) curves with AUC values of 0.82 and 0.84, respectively, were obtained. Self-similarity combined with the central apnea index (CAI) and hypoxic burden outperformed CAI alone. Even in those with a low CAI by conventional scoring criteria or only mild desaturation, SS was related to poor therapy outcomes. The proposed algorithm for assessing SS as a measure of expressed high loop gain is accurate, non-invasive, and has the potential to improve phenotyping/endotyping of apnea, leading to more precise sleep apnea treatment strategies.

Application study of apnea-hypopnea duration for assessing adult obstructive sleep apnea.

Obstructive sleep apnea (OSA) is a common sleep disordered breathing disorder, which can cause serious damage to multiple human systems. Although polysomnography (PSG) is the current gold standard for diagnosis, it is complex and expensive. Therefore, it is of great significance to find a simple, economical and rapid primary screening and diagnosis method to replace PSG for the diagnosis of OSA. The purpose of this study is to propose a new method for the diagnosis and classification of OSA, which is used to automatically detect the duration of sleep apnea hypopnea events (AHE), so as to estimate the ratio(S) of the total duration of all-night AHE to the total sleep time only based on the sound signal of sleep respiration, and to identify OSA. We performed PSG tests on participants and extracted relevant sleep breathing sound signal data. This study is carried out in two stages. In the first stage, the relevant PSG report data of eligible subjects were recorded, the total duration of AHE in each subject's data was extracted, and the S value was calculated to evaluate the severity of OSA. In the second stage, only the sleep breath sound signal data of the same batch of subjects were used for automatic detection, and the S value in the sleep breath sound signal was extracted, and the S value was compared with the PSG diagnosis results to calculate the accuracy of the experimental method. Among 225 subjects. Using PSG as the reference standard, the S value extracted from the PSG diagnostic data report can accurately diagnose OSA(accuracy rate 99.56%) and distinguish its severity (accuracy rate 95.11%). The accuracy of the S value detected in the sleep breathing sound signal in the diagnosis of severe OSA reached 100%. The results show that the experimental parameter S value is feasible in OSA diagnosis and classification. OSA can be identified and evaluated only by sleep breathing sounds. This method helps to simplify the diagnostic grading of traditional OSA and lays a foundation for the subsequent development of simple diagnostic grading equipment.

Free-Running Cardiac and Respiratory Motion-Resolved Imaging: A Paradigm Shift for Managing Motion in Cardiac MRI?

Cardiac magnetic resonance imaging (MRI) is widely used for non-invasive assessment of cardiac morphology, function, and tissue characteristics due to its exquisite soft-tissue contrast. However, it remains time-consuming and requires proficiency, making it costly and limiting its widespread use. Traditional cardiac MRI is inefficient as signal acquisition is often limited to specific cardiac phases and requires complex view planning, parameter adjustments, and management of both respiratory and cardiac motion. Recent efforts have aimed to make cardiac MRI more efficient and accessible. Among these innovations, the free-running framework enables 5D whole-heart imaging without the need for an electrocardiogram signal, respiratory breath-holding, or complex planning. It uses a fully self-gated approach to extract cardiac and respiratory signals directly from the acquired image data, allowing for more efficient coverage in time and space without the need for electrocardiogram gating, triggering, navigators, or breath-holds. This review provides a comprehensive overview of the free-running framework, detailing its history, concepts, recent improvements, and clinical applications.

Respiratory-swallow patterning and oropharyngeal swallowing impairment in patients undergoing evaluation for lung transplant.

Although swallowing impairment is prevalent following lung transplantation, baseline respiratory and swallowing characteristics are often overlooked. Respiratory disease processes may predispose lung transplant candidates to altered respiratory-swallow patterning and swallowing impairment. This cross-sectional study included patients referred for a Modified Barium Swallow Study during lung transplant evaluation. Swallowing impairment was measured using the Modified Barium Swallow Impairment Profile and Penetration-Aspiration Scale. Respiratory plethysmographic signals synchronized with videofluoroscopy were analyzed to determine phase patterning, pause duration, and rate. Mixed-effects logistic regression was used to identify linkages between respiratory and swallowing measures. Fifty patients were included and demonstrated delayed swallow initiation (49/50), oral residue (37/50), incomplete pharyngoesophageal segment opening (35/50), and esophageal retention (43/50). Airway invasion occurred infrequently (10/50). Atypical respiratory patterning was significantly associated with impairment in pharyngeal swallow initiation (OR [95% CI] = 1.76 [1.16, 2.68], p = 0.009), laryngeal elevation (OR [95% CI] = 1.45 [1.01, 2.07], p = 0.044), and laryngeal vestibular closure (OR [95% CI] = 2.57 [1.48, 4.46], p < 0.001). Increased pause duration was associated with impaired initiation (OR [95% CI] = 2.24 [1.20, 4.16], p = 0.011), laryngeal elevation (OR [95% CI] = 1.18 [1.03, 1.36], p = 0.018), laryngeal closure (OR [95% CI] = 1.28 [1.9, 1.50], p = 0.003), and tongue base retraction (OR [95% CI] = 1.33 [1.13, 1.56], p < 0.001). Patients undergoing evaluation for lung transplant demonstrated impaired swallowing and phase patterning. Preliminary findings implicate the need for further evaluation of respiratory-swallow coordination and its potential role in swallowing impairment before and after lung transplantation.

Overnight Sleep Staging Using Chest-Worn Accelerometry.

Overnight sleep staging is an important part of the diagnosis of various sleep disorders. Polysomnography is the gold standard for sleep staging, but less-obtrusive sensing modalities are of emerging interest. Here, we developed and validated an algorithm to perform "proxy" sleep staging using cardiac and respiratory signals derived from a chest-worn accelerometer. We collected data in two sleep centers, using a chest-worn accelerometer in combination with full PSG. A total of 323 participants were analyzed, aged 13-83 years, with BMI 18-47 kg/m2. We derived cardiac and respiratory features from the accelerometer and then applied a previously developed method for automatic cardio-respiratory sleep staging. We compared the estimated sleep stages against those derived from PSG and determined performance. Epoch-by-epoch agreement with four-class scoring (Wake, REM, N1+N2, N3) reached a Cohen's kappa coefficient of agreement of 0.68 and an accuracy of 80.8%. For Wake vs. Sleep classification, an accuracy of 93.3% was obtained, with a sensitivity of 78.7% and a specificity of 96.6%. We showed that cardiorespiratory signals obtained from a chest-worn accelerometer can be used to estimate sleep stages among a population that is diverse in age, BMI, and prevalence of sleep disorders. This opens up the path towards various clinical applications in sleep medicine.

Body mass index associated with respiration predicts motion in resting-state functional magnetic resonance imaging scans.

Decreasing body mass index (BMI) reduces head motion in resting-state fMRI (rs-fMRI) data. Yet, the mechanism by which BMI affects head motion remains poorly understood. Understanding how BMI interacts with respiration to affect head motion can improve head motion reduction strategies. A total of 254 patients with back pain were included in this study, each of whom had two visits (interval time = 13.85 ± 7.81 weeks) during which two consecutive re-fMRI scans were obtained. We investigated the relationships between head motion and demographic and pain-related characteristics-head motion was reliable across scans and correlated with age, pain intensity, and BMI. Multiple linear regression models determined that BMI was the main determinant in predicting head motion. BMI was also associated with two features derived from respiration signal. Anterior-posterior and superior-inferior motion dominated both overall motion magnitude and the coupling between motion and respiration. BMI interacted with respiration to influence motion only in the pitch dimension. These findings indicate that BMI should be a critical parameter in both study designs and analyses of fMRI data.

An IoT-based contactless neonatal respiratory monitoring system for neonatal care assistance in postpartum center

According to previous studies, one of the major causes of 20 % to 25 % of neonatal deaths is respiratory distress syndrome (RDS). Early identification, progressive monitoring, and treatment and/or management of neonatal RDS can substantially increase the rate of survival in neonates. However, global research indicates frequent shortages and burnout among nursing staff, especially in postpartum units, contributing to the difficulty in early identification of RDS in neonates. Clinicians currently use breathing sounds and frequency as key criteria in the Neonatal Resuscitation Program (NRP) for identifying and treating RDS. In practice, the monitoring of respiratory signal abnormalities relies on sensor patches, which frequently detach from the neonates’ slippery skin, leading to potential skin injuries and unstable signal reception. This paper presents an Internet of Things (IoT)-based contactless neonatal respiratory monitoring system that integrates computer vision (CV), beamforming microphone array (BFMA), and millimeter Wave (mmWave) radar, all connected to a cloud platform. Clinical trials revealed that CV-based neonatal feature identification achieved over 96 % accuracy within 40 cm to 120 cm. The neonatal breathing sound strengthening, utilized CV and BFMA, achieved an average sound-to-noise ratio (SNR) of 5.07 dB, and CV with mmWave radar reduced chest displacement signal error from 0.66 to 0.26 BPM. Additionally, survey results showed that doctors and clinical personnel were satisfied with the system's functionality and usability. This demonstrates the system's ability to assist in monitoring respiratory signals of swaddled neonates and in the early identification of neonatal RDS, with further applications in neonatal care at postpartum centers.

Leveraging Attention-reinforced UWB Signals to Monitor Respiration during Sleep

The respiration state during overnight sleep is an important indicator of human health. However, existing contactless solutions for sleep respiration monitoring either perform in controlled environments and have low usability in practical scenarios or only provide coarse-grained respiration rates, being unable to accurately detect abnormal events in patients. In this article, we propose Respnea, a non-intrusive sleep respiration monitoring system using an ultra-wideband device. Particularly, we propose a profiling algorithm, which can locate the sleep positions in non-controlled environments and identify different subject states. Further, we construct a deep learning model that adopts a multi-head self-attention mechanism and learns the patterns implicit in the respiration signals to distinguish sleep respiration events at a granularity of seconds. To improve the generalization of the model, we propose a contrastive learning strategy to learn a robust representation of the respiration signals. We deploy our system in hospital and home scenarios and conduct experiments on data from healthy subjects and patients with sleep disorders. The experimental results show that Respnea achieves high temporal coverage and low errors (a median error of 0.27 bpm) in respiration rate estimation and reaches an accuracy of 94.44% on diagnosing the severity of sleep apnea-hypopnea syndrome.

RF-GymCare: Introducing Respiratory Prior for RF Sensing in Gym Environments

Exercise-induced sudden death has become an increasingly serious health issue in recent years. Fortunately, respiratory monitoring can serve as an effective early prevention solution. In particular, radio frequency (RF) sensing has emerged as a promising solution due to its contactless nature and privacy preservation. Existing RF-based respiratory monitoring solutions for homes, offices, or vehicles would suffer from significant performance degradation when deployed in typical gym workouts, where respiratory and physical movements are mixed in the RF signals in an unpredictable manner, leading to a blind source separation (BSS) problem. What is worse, the prior assumptions of source independence in existing BSS algorithms no longer hold, leaving respiratory monitoring in gym scenarios as an open issue. In this paper, we propose RF-GymCare, which introduces respiratory priors into the BSS problem and utilizes knowledge distillation to transfer these priors, aiming to learn the demixing weight from mixed signals. With the demixing weight, the respiratory signal can be obtained through a linear combination of the mixed signals. Our experiments with 13 hours of data across nine types of exercises demonstrate its superiority over existing methods, achieving a median similarity of 0.75 and a median rate error of 0.5 respiration per minute (RPM), respectively.

Derivation and external validation of predictive models for invasive mechanical ventilation in intensive care unit patients with COVID-19

BackgroundThis study aimed to develop prognostic models for predicting the need for invasive mechanical ventilation (IMV) in intensive care unit (ICU) patients with COVID-19 and compare their performance with the Respiratory rate-OXygenation (ROX) index.MethodsA retrospective cohort study was conducted using data collected between March 2020 and August 2021 at three hospitals in Rio de Janeiro, Brazil. ICU patients aged 18 years and older with a diagnosis of COVID-19 were screened. The exclusion criteria were patients who received IMV within the first 24 h of ICU admission, pregnancy, clinical decision for minimal end-of-life care and missing primary outcome data. Clinical and laboratory variables were collected. Multiple logistic regression analysis was performed to select predictor variables. Models were based on the lowest Akaike Information Criteria (AIC) and lowest AIC with significant p values. Assessment of predictive performance was done for discrimination and calibration. Areas under the curves (AUC)s were compared using DeLong’s algorithm. Models were validated externally using an international database.ResultsOf 656 patients screened, 346 patients were included; 155 required IMV (44.8%), 191 did not (55.2%), and 207 patients were male (59.8%). According to the lowest AIC, arterial hypertension, diabetes mellitus, obesity, Sequential Organ Failure Assessment (SOFA) score, heart rate, respiratory rate, peripheral oxygen saturation (SpO2), temperature, respiratory effort signals, and leukocytes were identified as predictors of IMV at hospital admission. According to AIC with significant p values, SOFA score, SpO2, and respiratory effort signals were the best predictors of IMV; odds ratios (95% confidence interval): 1.46 (1.07–2.05), 0.81 (0.72–0.90), 9.13 (3.29–28.67), respectively. The ROX index at admission was lower in the IMV group than in the non-IMV group (7.3 [5.2–9.8] versus 9.6 [6.8–12.9], p < 0.001, respectively). In the external validation population, the area under the curve (AUC) of the ROX index was 0.683 (accuracy 63%), the AIC model showed an AUC of 0.703 (accuracy 69%), and the lowest AIC model with significant p values had an AUC of 0.725 (accuracy 79%).ConclusionsIn the development population of ICU patients with COVID-19, SOFA score, SpO2, and respiratory effort signals predicted the need for IMV better than the ROX index. In the external validation population, although the AUCs did not differ significantly, the accuracy was higher when using SOFA score, SpO2, and respiratory effort signals compared to the ROX index. This suggests that these variables may be more useful in predicting the need for IMV in ICU patients with COVID-19.ClinicalTrials.gov identifier:NCT05663528.