Dataset Information

Predicting responders to prone positioning in mechanically ventilated patients with COVID-19 using machine learning.

ABSTRACT:

Background

For mechanically ventilated critically ill COVID-19 patients, prone positioning has quickly become an important treatment strategy, however, prone positioning is labor intensive and comes with potential adverse effects. Therefore, identifying which critically ill intubated COVID-19 patients will benefit may help allocate labor resources.

Methods

From the multi-center Dutch Data Warehouse of COVID-19 ICU patients from 25 hospitals, we selected all 3619 episodes of prone positioning in 1142 invasively mechanically ventilated patients. We excluded episodes longer than 24 h. Berlin ARDS criteria were not formally documented. We used supervised machine learning algorithms Logistic Regression, Random Forest, Naive Bayes, K-Nearest Neighbors, Support Vector Machine and Extreme Gradient Boosting on readily available and clinically relevant features to predict success of prone positioning after 4 h (window of 1 to 7 h) based on various possible outcomes. These outcomes were defined as improvements of at least 10% in PaO₂/FiO₂ ratio, ventilatory ratio, respiratory system compliance, or mechanical power. Separate models were created for each of these outcomes. Re-supination within 4 h after pronation was labeled as failure. We also developed models using a 20 mmHg improvement cut-off for PaO₂/FiO₂ ratio and using a combined outcome parameter. For all models, we evaluated feature importance expressed as contribution to predictive performance based on their relative ranking.

Results

The median duration of prone episodes was 17 h (11-20, median and IQR, N = 2632). Despite extensive modeling using a plethora of machine learning techniques and a large number of potentially clinically relevant features, discrimination between responders and non-responders remained poor with an area under the receiver operator characteristic curve of 0.62 for PaO₂/FiO₂ ratio using Logistic Regression, Random Forest and XGBoost. Feature importance was inconsistent between models for different outcomes. Notably, not even being a previous responder to prone positioning, or PEEP-levels before prone positioning, provided any meaningful contribution to predicting a successful next proning episode.

Conclusions

In mechanically ventilated COVID-19 patients, predicting the success of prone positioning using clinically relevant and readily available parameters from electronic health records is currently not feasible. Given the current evidence base, a liberal approach to proning in all patients with severe COVID-19 ARDS is therefore justified and in particular regardless of previous results of proning.

SUBMITTER: Dam TA

PROVIDER: S-EPMC9583049 | biostudies-literature | 2022 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Predicting responders to prone positioning in mechanically ventilated patients with COVID-19 using machine learning.

Dam Tariq A TA Roggeveen Luca F LF van Diggelen Fuda F Fleuren Lucas M LM Jagesar Ameet R AR Otten Martijn M de Vries Heder J HJ Gommers Diederik D Cremer Olaf L OL Bosman Rob J RJ Rigter Sander S Wils Evert-Jan EJ Frenzel Tim T Dongelmans Dave A DA de Jong Remko R Peters Marco A A MAA Kamps Marlijn J A MJA Ramnarain Dharmanand D Nowitzky Ralph R Nooteboom Fleur G C A FGCA de Ruijter Wouter W Urlings-Strop Louise C LC Smit Ellen G M EGM Mehagnoul-Schipper D Jannet DJ Dormans Tom T de Jager Cornelis P C CPC Hendriks Stefaan H A SHA Achterberg Sefanja S Oostdijk Evelien E Reidinga Auke C AC Festen-Spanjer Barbara B Brunnekreef Gert B GB Cornet Alexander D AD van den Tempel Walter W Boelens Age D AD Koetsier Peter P Lens Judith J Faber Harald J HJ Karakus A A Entjes Robert R de Jong Paul P Rettig Thijs C D TCD Arbous Sesmu S Vonk Sebastiaan J J SJJ Machado Tomas T Herter Willem E WE de Grooth Harm-Jan HJ Thoral Patrick J PJ Girbes Armand R J ARJ Hoogendoorn Mark M Elbers Paul W G PWG

Annals of intensive care 20221020 1

<h4>Background</h4>For mechanically ventilated critically ill COVID-19 patients, prone positioning has quickly become an important treatment strategy, however, prone positioning is labor intensive and comes with potential adverse effects. Therefore, identifying which critically ill intubated COVID-19 patients will benefit may help allocate labor resources.<h4>Methods</h4>From the multi-center Dutch Data Warehouse of COVID-19 ICU patients from 25 hospitals, we selected all 3619 episodes of prone ...[more]

PMID: 36264358

Similar Datasets

Project description:BackgroundProne positioning improves survival in moderate-to-severe acute respiratory distress syndrome (ARDS) unrelated to the novel coronavirus disease (COVID-19). This benefit is probably mediated by a decrease in alveolar collapse and hyperinflation and a more homogeneous distribution of lung aeration, with fewer harms from mechanical ventilation. In this preliminary physiological study we aimed to verify whether prone positioning causes analogue changes in lung aeration in COVID-19. A positive result would support prone positioning even in this other population.MethodsFifteen mechanically-ventilated patients with COVID-19 underwent a lung computed tomography in the supine and prone position with a constant positive end-expiratory pressure (PEEP) within three days of endotracheal intubation. Using quantitative analysis, we measured the volume of the non-aerated, poorly-aerated, well-aerated, and over-aerated compartments and the gas-to-tissue ratio of the ten vertical levels of the lung. In addition, we expressed the heterogeneity of lung aeration with the standardized median absolute deviation of the ten vertical gas-to-tissue ratios, with lower values indicating less heterogeneity.ResultsBy the time of the study, PEEP was 12 (10-14) cmH2O and the PaO2:FiO2 107 (84-173) mmHg in the supine position. With prone positioning, the volume of the non-aerated compartment decreased by 82 (26-147) ml, of the poorly-aerated compartment increased by 82 (53-174) ml, of the normally-aerated compartment did not significantly change, and of the over-aerated compartment decreased by 28 (11-186) ml. In eight (53%) patients, the volume of the over-aerated compartment decreased more than the volume of the non-aerated compartment. The gas-to-tissue ratio of the ten vertical levels of the lung decreased by 0.34 (0.25-0.49) ml/g per level in the supine position and by 0.03 (- 0.11 to 0.14) ml/g in the prone position (p < 0.001). The standardized median absolute deviation of the gas-to-tissue ratios of those ten levels decreased in all patients, from 0.55 (0.50-0.71) to 0.20 (0.14-0.27) (p < 0.001).ConclusionsIn fifteen patients with COVID-19, prone positioning decreased alveolar collapse, hyperinflation, and homogenized lung aeration. A similar response has been observed in other ARDS, where prone positioning improves outcome. Therefore, our data provide a pathophysiological rationale to support prone positioning even in COVID-19.

Project description:The prone position (PP) is increasingly used in mechanically ventilated coronavirus disease 2019 (COVID-19) acute respiratory distress syndrome (ARDS) patients. However, studies investigating the influence of the PP are currently lacking in these patients. This is the first study to investigate the influence of the PP on the oxygenation and decarboxylation in COVID-19 patients. A prospective bicentric study design was used, and in mechanically ventilated COVID-19 patients, PP was indicated from a partial pressure of oxygen in arterial blood (PaO2)/fraction of inspired oxygen (FIO2) ratio of <200. Patients were left prone for 16 h each. Pressure levels, FIO2, were adjusted to ensure a PaO2 greater than 60 mmHg. Blood gas analyses were performed before (baseline 0.5 h), during (1/2/5.5/9.5/13 h), and after being in the PP (1 h), the circulatory/ventilation parameters were continuously monitored, and lung compliance (LC) was roughly calculated. Responders were defined compared to the baseline value (PaO2/FIO2 ratio increase of ≥15%; partial pressure of carbon dioxide (PaCO2) decrease of ≥2%). 13 patients were included and 36 PP sessions were conducted. Overall, PaO2/FIO2 increased significantly in the PP (p < 0.001). Most PaO2/FIO2 responders (29/36 PP sessions, 77%) were identified 9.5 h after turning prone (14% slow responders), while most PaCO2 responders (15/36 PP sessions, 42%) were identified 13 h after turning prone. A subgroup of patients (interval intubation to PP ≥3 days) showed less PaO2/FIO2 responders (16% vs. 77%). An increase in PaCO2 and minute ventilation in the PP showed a significant negative correlation (p < 0.001). LC (median before the PP = 38 mL/cm H2O; two patients with LC >80 mL/cm H2O) showed a significant positive correlation with the 28 day survival of patients (p = 0.01). The PP significantly improves oxygenation in COVID-19 ARDS patients. The data suggest that they also benefit most from an early PP. A decrease in minute ventilation may result in fewer PaCO2 responders. LC may be a predictive outcome parameter in COVID-19 patients. Retrospectively registered.

Project description:ObjectivesTherapies for patients with respiratory failure from coronavirus disease 2019 are urgently needed. Early implementation of prone positioning ventilation improves survival in patients with acute respiratory distress syndrome, but studies examining the effect of proning on survival in patients with coronavirus disease 2019 are lacking. Our objective was to estimate the effect of early proning initiation on survival in patients with coronavirus disease 2019-associated respiratory failure.DesignData were derived from the Study of the Treatment and Outcomes in Critically Ill Patients with coronavirus disease 2019, a multicenter cohort study of critically ill adults with coronavirus disease 2019 admitted to 68 U.S. hospitals. Using these data, we emulated a target trial of prone positioning ventilation by categorizing mechanically ventilated hypoxemic (ratio of Pao2 over the corresponding Fio2 ≤ 200 mm Hg) patients as having been initiated on proning or not within 2 days of ICU admission. We fit an inverse probability-weighted Cox model to estimate the mortality hazard ratio for early proning versus no early proning. Patients were followed until death, hospital discharge, or end of follow-up.SettingICUs at 68 U.S. sites.PatientsCritically ill adults with laboratory-confirmed coronavirus disease 2019 receiving invasive mechanical ventilation with ratio of Pao2 over the corresponding Fio2 less than or equal to 200 mm Hg.InterventionsNone.Measurements and main resultsAmong 2,338 eligible patients, 702 (30.0%) were proned within the first 2 days of ICU admission. After inverse probability weighting, baseline and severity of illness characteristics were well-balanced between groups. A total of 1,017 (43.5%) of the 2,338 patients were discharged alive, 1,101 (47.1%) died, and 220 (9.4%) were still hospitalized at last follow-up. Patients proned within the first 2 days of ICU admission had a lower adjusted risk of death compared with nonproned patients (hazard ratio, 0.84; 95% CI, 0.73-0.97).ConclusionsIn-hospital mortality was lower in mechanically ventilated hypoxemic patients with coronavirus disease 2019 treated with early proning compared with patients whose treatment did not include early proning.

Dataset Information

Predicting responders to prone positioning in mechanically ventilated patients with COVID-19 using machine learning.

Background

Methods

Results

Conclusions

Publications

Predicting responders to prone positioning in mechanically ventilated patients with COVID-19 using machine learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets