eARDS: A multi-center validation of an interpretable machine learning algorithm of early onset Acute Respiratory Distress Syndrome (ARDS) among critically ill adults with COVID-19

A common data model for the standardization of intensive care unit medication features

Objective

Common data models provide a standard means of describing data for artificial intelligence (AI) applications, but this process has never been undertaken for medications used in the intensive care unit (ICU). We sought to develop a common data model (CDM) for ICU medications to standardize the medication features needed to support future ICU AI efforts.

Materials and Methods

A 9-member, multi-professional team of ICU clinicians and AI experts conducted a 5-round modified Delphi process employing conference calls, web-based communication, and electronic surveys to define the most important medication features for AI efforts. Candidate ICU medication features were generated through group discussion and then independently scored by each team member based on relevance to ICU clinical decision-making and feasibility for collection and coding. A key consideration was to ensure the final ontology both distinguished unique medications and met Findable, Accessible, Interoperable, and Reusable (FAIR) guiding principles.

Results

Using a list of 889 ICU medications, the team initially generated 106 different medication features, and 71 were ranked as being core features for the CDM. Through this process, 106 medication features were assigned to 2 key feature domains: drug product-related (n = 43) and clinical practice-related (n = 63). Each feature included a standardized definition and suggested response values housed in the electronic data library. This CDM for ICU medications is available online.

Conclusion

The CDM for ICU medications represents an important first step for the research community focused on exploring how AI can improve patient outcomes and will require ongoing engagement and refinement.

A pretrain-finetune approach for improving model generalizability in outcome prediction of acute respiratory distress syndrome patients

BACKGROUND

Early prediction of acute respiratory distress syndrome (ARDS) of critically ill patients in intensive care units (ICUs) has been intensively studied in the past years. Yet a prediction model trained on data from one hospital might not be well generalized to other hospitals. It is therefore essential to develop an accurate and generalizable ARDS prediction model adaptive to different hospital or medical centers.

METHODS

We analyzed electronic medical records of 200,859 and 50,920 hospitalized patients within 24 h after being diagnosed with ARDS from the Philips eICU Institute (eICU-CRD) and the Medical Information Mart for Intensive Care (MIMIC-IV) dataset, respectively. Patients were sorted into three groups, including rapid death, long stay, and recovery, based on their condition or outcome between 24 and 72 h after ARDS diagnosis. To improve prediction performance and generalizability, a "pretrain-finetune" approach was applied, where we pretrained models on the eICU-CRD dataset and performed model finetuning using only a part (35%) of the MIMIC-IV dataset, and then tested the finetuned models on the remaining data from the MIMIC-IV dataset. Well-known machine-learning algorithms, including logistic regression, random forest, extreme gradient boosting, and multilayer perceptron neural networks, were employed to predict ARDS outcomes. Prediction performance was evaluated using the area under the receiver-operating characteristic curve (AUC).

RESULTS

Results show that, in general, multilayer perceptron neural networks outperformed the other models. The use of pretrain-finetune yielded improved performance in predicting ARDS outcomes achieving a micro-AUC of 0.870 for the MIMIC-IV dataset, an improvement of 0.046 over the pretrain model.

CONCLUSIONS

The proposed pretrain-finetune approach can effectively improve model generalizability from one to another dataset in ARDS prediction.

Acute Respiratory Distress Syndrome: Definition, Diagnosis, and Routine Management

Acute respiratory distress syndrome (ARDS) is an acute inflammatory lung injury characterized by severe hypoxemic respiratory failure, bilateral opacities on chest imaging, and low lung compliance. ARDS is a heterogeneous syndrome that is the common end point of a wide variety of predisposing conditions, with complex pathophysiology and underlying mechanisms. Routine management of ARDS is centered on lung-protective ventilation strategies such as low tidal volume ventilation and targeting low airway pressures to avoid exacerbation of lung injury, as well as a conservative fluid management strategy.

Machine learning for the early prediction of acute respiratory distress syndrome (ARDS) in patients with sepsis in the ICU based on clinical data

Background

Acute respiratory distress syndrome (ARDS) is a fatal outcome of severe sepsis. Machine learning models are helpful for accurately predicting ARDS in patients with sepsis at an early stage.

Objective

We aim to develop a machine-learning model for predicting ARDS in patients with sepsis in the intensive care unit (ICU).

Methods

The initial clinical data of patients with sepsis admitted to the hospital (including population characteristics, clinical diagnosis, complications, and laboratory tests) were used to predict ARDS, and screen out the crucial variables. After comparing eight different algorithms, namely, XG boost, logistic regression, light GBM, random forest, GaussianNB, complement NB, support vector machine (SVM), and K nearest neighbors (KNN), rebuilding a prediction model with the best one. When remodeling with the best algorithm, 10% was randomly selected to test, and the remaining was trained for cross-validation. Using the area under the curve (AUC), sensitivity, accuracy, specificity, positive and negative predictive value, F1 score, kappa value, and clinical decision curve to evaluate the model's performance. Eventually, the application in the model illustrated by the SHAP package.

Results

Ten critical features were screened utilizing the lasso method, namely, PaO2/PAO2, A-aDO2, PO2(T), CRP, gender, PO2, RDW, MCH, SG, and chlorine. The prior ranking of variables demonstrated that PaO2/PAO2 was the most significant variable. Among the eight algorithms, the performance of the Gaussian NB algorithm was significantly better than that of the others. After remodeling with the best algorithm, the AUC in the training and validation sets were 0.777 and 0.770, respectively, and the algorithm performed well in the test set (AUC = 0.781, accuracy = 78.6%, sensitivity = 82.4%, F1 score = 0.824). A comparison of the overlap factors with those of previous models revealed that the model we developed performs better.

Conclusion

Sepsis-associated ARDS can be accurately predicted early via a machine learning model based on existing clinical data. These findings are helpful for accurate identification and improvement of the prognosis in patients with sepsis-associated ARDS.

Optimizing the Synergistic Potential of Pseudo-Labels from Radiology Notes and Annotated Ground Truth in Identifying Pulmonary Opacities on Chest Radiographs for Early Detection of Acute Respiratory Distress Syndrome

Acute Respiratory Distress Syndrome (ARDS) is a life-threatening lung injury, hallmarks of which are bilateral radiographic opacities. Studies have shown that early recognition of ARDS could reduce severity and lethal clinical sequela. A Convolutional Neural Network (CNN) model that can identify bilateral pulmonary opacities on chest x-ray (CXR) images can aid early ARDS recognition. Obtaining large datasets with ground truth labels to train CNNs is challenging, as medical image annotation requires clinical expertise and meticulous consideration. In this work, we implement a natural language processing pipeline that extracts pseudo-labels CXR images by parsing radiology notes for abnormal findings. We obtain ground-truth annotations from clinicians for the presence of pulmonary opacities for a subset of these images. A knowledge distillation-based teacher-student training framework is implemented to leverage the larger dataset with noisy pseudo-labels. Our results show an AUC of 0.93 (95%CI 0.92-0.94) for the prediction of bilateral opacities on chest radiographs.

Critical evaluation of established risk prediction models for acute respiratory distress syndrome in adult patients: A systematic review and meta-analysis

AIM

To assess the performance of validated prediction models for acute respiratory distress syndrome (ARDS) by systematic review and meta-analysis.

METHODS

Eight databases (Medline, CINAHL, Embase, The Cochrane Library, CNKI, WanFang Data, Sinomed, and VIP) were searched up to March 26, 2023. Studies developed and validated a prediction model for ARDS in adult patients were included. Items on study design, incidence, derivation methods, predictors, discrimination, and calibration were collected. The risk of bias was assessed by the Prediction model Risk of Bias Assessment Tool. Models with a reported area under the curve of the receiver operating characteristic (AUC) metric were analyzed.

RESULTS

A total of 25 studies were retrieved, including 48 unique prediction models. Discrimination was reported in all studies, with AUC ranging from 0.701 to 0.95. Emerged AUC value of the logistic regression model was 0.837 (95% CI: 0.814 to 0.859). Besides, the value in the ICU group was 0.856 (95% CI: 0.812 to 0.899), the acute pancreatitis group was 0.863 (95% CI: 0.844 to 0.882), and the postoperation group was 0.835 (95% CI: 0.808 to 0.861). In total, 24 of the included studies had a high risk of bias, which was mostly due to the improper methods in predictor screening (13/24), model calibration assessment (9/24), and dichotomization of continuous predictors (6/24).

CONCLUSIONS

This study shows that most prediction models for ARDS are at high risk of bias, and the discrimination ability of the model is excellent. Adherence to standardized guidelines for model development is necessary to derive a prediction model of value to clinicians.

Cluster analysis driven by unsupervised latent feature learning of medications to identify novel pharmacophenotypes of critically ill patients

Unsupervised clustering of intensive care unit (ICU) medications may identify unique medication clusters (i.e., pharmacophenotypes) in critically ill adults. We performed an unsupervised analysis with Restricted Boltzmann Machine of 991 medications profiles of patients managed in the ICU to explore pharmacophenotypes that correlated with ICU complications (e.g., mechanical ventilation) and patient-centered outcomes (e.g., length of stay, mortality). Six unique pharmacophenotypes were observed, with unique medication profiles and clinically relevant differences in ICU complications and patient-centered outcomes. While pharmacophenotypes 2 and 4 had no statistically significant difference in ICU length of stay, duration of mechanical ventilation, or duration of vasopressor use, their mortality differed significantly (9.0% vs. 21.9%, p < 0.0001). Pharmacophenotype 4 had a mortality rate of 21.9%, compared with the rest of the pharmacophenotypes ranging from 2.5 to 9%. Phenotyping approaches have shown promise in classifying the heterogenous syndromes of critical illness to predict treatment response and guide clinical decision support systems but have never included comprehensive medication information. This first-ever machine learning approach revealed differences among empirically-derived subgroups of ICU patients that are not typically revealed by traditional classifiers. Identification of pharmacophenotypes may enable enhanced decision making to optimize treatment decisions.

A Transformer-Based Model Trained on Large Scale Claims Data for Prediction of Severe COVID-19 Disease Progression

In situations like the COVID-19 pandemic, healthcare systems are under enormous pressure as they can rapidly collapse under the burden of the crisis. Machine learning (ML) based risk models could lift the burden by identifying patients with a high risk of severe disease progression. Electronic Health Records (EHRs) provide crucial sources of information to develop these models because they rely on routinely collected healthcare data. However, EHR data is challenging for training ML models because it contains irregularly timestamped diagnosis, prescription, and procedure codes. For such data, transformer-based models are promising. We extended the previously published Med-BERT model by including age, sex, medications, quantitative clinical measures, and state information. After pre-training on approximately 988 million EHRs from 3.5 million patients, we developed models to predict Acute Respiratory Manifestations (ARM) risk using the medical history of 80,211 COVID-19 patients. Compared to Random Forests, XGBoost, and RETAIN, our transformer-based models more accurately forecast the risk of developing ARM after COVID-19 infection. We used Integrated Gradients and Bayesian networks to understand the link between the essential features of our model. Finally, we evaluated adapting our model to Austrian in-patient data. Our study highlights the promise of predictive transformer-based models for precision medicine.

Uncertainty-Aware Convolutional Neural Network for Identifying Bilateral Opacities on Chest X-rays: A Tool to Aid Diagnosis of Acute Respiratory Distress Syndrome

Acute Respiratory Distress Syndrome (ARDS) is a severe lung injury with high mortality, primarily characterized by bilateral pulmonary opacities on chest radiographs and hypoxemia. In this work, we trained a convolutional neural network (CNN) model that can reliably identify bilateral opacities on routine chest X-ray images of critically ill patients. We propose this model as a tool to generate predictive alerts for possible ARDS cases, enabling early diagnosis. Our team created a unique dataset of 7800 single-view chest-X-ray images labeled for the presence of bilateral or unilateral pulmonary opacities, or 'equivocal' images, by three blinded clinicians. We used a novel training technique that enables the CNN to explicitly predict the 'equivocal' class using an uncertainty-aware label smoothing loss. We achieved an Area under the Receiver Operating Characteristic Curve (AUROC) of 0.82 (95% CI: 0.80, 0.85), a precision of 0.75 (95% CI: 0.73, 0.78), and a sensitivity of 0.76 (95% CI: 0.73, 0.78) on the internal test set while achieving an (AUROC) of 0.84 (95% CI: 0.81, 0.86), a precision of 0.73 (95% CI: 0.63, 0.69), and a sensitivity of 0.73 (95% CI: 0.70, 0.75) on an external validation set. Further, our results show that this approach improves the model calibration and diagnostic odds ratio of the hypothesized alert tool, making it ideal for clinical decision support systems.

Pharmacophenotype identification of intensive care unit medications using unsupervised cluster analysis of the ICURx common data model

BACKGROUND

Identifying patterns within ICU medication regimens may help artificial intelligence algorithms to better predict patient outcomes; however, machine learning methods incorporating medications require further development, including standardized terminology. The Common Data Model for Intensive Care Unit (ICU) Medications (CDM-ICURx) may provide important infrastructure to clinicians and researchers to support artificial intelligence analysis of medication-related outcomes and healthcare costs. Using an unsupervised cluster analysis approach in combination with this common data model, the objective of this evaluation was to identify novel patterns of medication clusters (termed 'pharmacophenotypes') correlated with ICU adverse events (e.g., fluid overload) and patient-centered outcomes (e.g., mortality).

METHODS

This was a retrospective, observational cohort study of 991 critically ill adults. To identify pharmacophenotypes, unsupervised machine learning analysis with automated feature learning using restricted Boltzmann machine and hierarchical clustering was performed on the medication administration records of each patient during the first 24 h of their ICU stay. Hierarchical agglomerative clustering was applied to identify unique patient clusters. Distributions of medications across pharmacophenotypes were described, and differences among patient clusters were compared using signed rank tests and Fisher's exact tests, as appropriate.

RESULTS

A total of 30,550 medication orders for the 991 patients were analyzed; five unique patient clusters and six unique pharmacophenotypes were identified. For patient outcomes, compared to patients in Clusters 1 and 3, patients in Cluster 5 had a significantly shorter duration of mechanical ventilation and ICU length of stay (p < 0.05); for medications, Cluster 5 had a higher distribution of Pharmacophenotype 1 and a smaller distribution of Pharmacophenotype 2, compared to Clusters 1 and 3. For outcomes, patients in Cluster 2, despite having the highest severity of illness and greatest medication regimen complexity, had the lowest overall mortality; for medications, Cluster 2 also had a comparably higher distribution of Pharmacophenotype 6.

CONCLUSION

The results of this evaluation suggest that patterns among patient clusters and medication regimens may be observed using empiric methods of unsupervised machine learning in combination with a common data model. These results have potential because while phenotyping approaches have been used to classify heterogenous syndromes in critical illness to better define treatment response, the entire medication administration record has not been incorporated in those analyses. Applying knowledge of these patterns at the bedside requires further algorithm development and clinical application but may have the future potential to be leveraged in guiding medication-related decision making to improve treatment outcomes.

Machine learning algorithm for early-stage prediction of severe morbidity in COVID-19 pneumonia patients based on bio-signals

BACKGROUND

Paralysis of medical systems has emerged as a major problem not only in Korea but also globally because of the COVID-19 pandemic. Therefore, early identification and treatment of COVID-19 are crucial. This study aims to develop a machine-learning algorithm based on bio-signals that predicts the infection three days in advance before it progresses from mild to severe, which may necessitate high-flow oxygen therapy or mechanical ventilation.

METHODS

The study included 2758 hospitalized patients with mild severity COVID-19 between July 2020 and October 2021. Bio-signals, clinical information, and laboratory findings were retrospectively collected from the electronic medical records of patients. Machine learning methods included random forest, random forest ranger, gradient boosting machine, and support vector machine (SVM).

RESULTS

SVM showed the best performance in terms of accuracy, kappa, sensitivity, detection rate, balanced accuracy, and run-time; the area under the receiver operating characteristic curve was also quite high at 0.96. Body temperature and SpO₂ three and four days before discharge or exacerbation were ranked high among SVM features.

CONCLUSIONS

The proposed algorithm can predict the exacerbation of severity three days in advance in patients with mild COVID-19. This prediction can help effectively manage the reallocation of appropriate medical resources in clinical settings. Therefore, this algorithm can facilitate adequate oxygen therapy and mechanical ventilator preparation, thereby improving patient prognosis, increasing the efficiency of medical systems, and mitigating the damage caused by a global pandemic.

Overt and Occult Hypoxemia in Patients Hospitalized With COVID-19

Progressive hypoxemia is the predominant mode of deterioration in COVID-19. Among hypoxemia measures, the ratio of the Pao₂ to the Fio₂ (P/F ratio) has optimal construct validity but poor availability because it requires arterial blood sampling. Pulse oximetry reports oxygenation continuously (ratio of the Spo₂ to the Fio₂ [S/F ratio]), but it is affected by skin color and occult hypoxemia can occur in Black patients. Oxygen dissociation curves allow noninvasive estimation of P/F ratios (ePFRs) but remain unproven.

OBJECTIVES

Measure overt and occult hypoxemia using ePFR.

DESIGN SETTING AND PARTICIPANTS

We retrospectively studied COVID-19 hospital encounters (n = 5,319) at two academic centers (University of Virginia [UVA] and Emory University).

MAIN OUTCOMES AND MEASURES

We measured primary outcomes (death or ICU transfer within 24 hr), ePFR, conventional hypoxemia measures, baseline predictors (age, sex, race, comorbidity), and acute predictors (National Early Warning Score [NEWS] and Sequential Organ Failure Assessment [SOFA]). We updated predictors every 15 minutes. We assessed predictive validity using adjusted odds ratios (AORs) and area under the receiver operating characteristic curves (AUROCs). We quantified disparities (Black vs non-Black) in empirical cumulative distributions using the Kolmogorov-Smirnov (K-S) two-sample test.

RESULTS

Overt hypoxemia (low ePFR) predicted bad outcomes (AOR for a 100-point ePFR drop: 2.7 [UVA]; 1.7 [Emory]; p < 0.01) with better discrimination (AUROC: 0.76 [UVA]; 0.71 [Emory]) than NEWS (0.70 [both sites]) or SOFA (0.68 [UVA]; 0.65 [Emory]) and similar to S/F ratio (0.76 [UVA]; 0.70 [Emory]). We found racial differences consistent with occult hypoxemia. Black patients had better apparent oxygenation (K-S distance: 0.17 [both sites]; p < 0.01) but, for comparable ePFRs, worse outcomes than other patients (AOR: 2.2 [UVA]; 1.2 [Emory]; p < 0.01).

CONCLUSIONS AND RELEVANCE

The ePFR was a valid measure of overt hypoxemia. In COVID-19, it may outperform multi-organ dysfunction models. By accounting for biased oximetry as well as clinicians' real-time responses to it (supplemental oxygen adjustment), ePFRs may reveal racial disparities attributable to occult hypoxemia.

Higher Risk of Acute Respiratory Distress Syndrome and Risk Factors among Patients with COVID-19: A Systematic Review, Meta-Analysis and Meta-Regression

OBJECTIVE

To estimate the global risk and risk factors associated with acute respiratory distress syndrome (ARDS) among patients with COVID-19: Design: A systematic review, meta-analysis and meta-regression.

SETTING AND PARTICIPANTS

Hospitals or nursing homes and patients with acute respiratory distress syndrome after COVID-19.

METHODS

The literature review was systematically conducted on Embase, MEDLINE, CINAHL, and Web of Science, in addition to manual searches and reference list checking from 1 January 2019 to 2 March 2022. The search terms included coronavirus, acute respiratory syndrome, acute respiratory distress syndrome and observational studies. Three reviewers independently appraised the quality of the studies and extracted the relevant data using the Joanna Briggs Institute abstraction form and critical appraisal tools. A study protocol was registered in PROSPERO (CRD42022311957). Eligible studies were meta-analyzed and underwent meta-regression.

RESULTS

A total of 12 studies were included, with 148,080 participants. The risk ratio (RR) of ARDS was 23%. Risk factors were age ≥ 41-64 years old (RR = 15.3%, 95% CI =0.14-2.92, p = 0.03); fever (RR = 10.3%, 95% CI = 0.03-2.03, p = 0.04); multilobe involvement of the chest (RR = 33.5%, 95% CI = 0.35-6.36, p = 0.02); lymphopenia (RR = 25.9%, 95% CI = 1.11-4.08, p = 0.01); mechanical ventilation with oxygen therapy (RR = 31.7%, 95% CI = 1.10-5.25, p = 0.002); European region (RR = 16.3%, 95% CI = 0.09-3.17, p = 0.03); sample size ≤ 500 (RR = 18.0%, 95% CI = 0.70-2.89, p = 0.001).

CONCLUSIONS AND IMPLICATIONS

One in four patients experienced ARDS after having COVID-19. The age group 41-64 years old and the European region were high-risk groups. These findings can be used by policymakers to allocate resources for respiratory care facilities and can also provide scientific evidence in the design of protocols to manage COVID-19 worldwide.

A Machine Learning Model for Predicting Hospitalization in Patients with Respiratory Symptoms during the COVID-19 Pandemic

A machine learning approach is a useful tool for risk-stratifying patients with respiratory symptoms during the COVID-19 pandemic, as it is still evolving. We aimed to verify the predictive capacity of a gradient boosting decision trees (XGboost) algorithm to select the most important predictors including clinical and demographic parameters in patients who sought medical support due to respiratory signs and symptoms (RAPID RISK COVID-19). A total of 7336 patients were enrolled in the study, including 6596 patients that did not require hospitalization and 740 that required hospitalization. We identified that patients with respiratory signs and symptoms, in particular, lower oxyhemoglobin saturation by pulse oximetry (SpO₂) and higher respiratory rate, fever, higher heart rate, and lower levels of blood pressure, associated with age, male sex, and the underlying conditions of diabetes mellitus and hypertension, required hospitalization more often. The predictive model yielded a ROC curve with an area under the curve (AUC) of 0.9181 (95% CI, 0.9001 to 0.9361). In conclusion, our model had a high discriminatory value which enabled the identification of a clinical and demographic profile predictive, preventive, and personalized of COVID-19 severity symptoms.

Overt and occult hypoxemia in patients hospitalized with novel coronavirus disease 2019

Background

Progressive hypoxemia is the predominant mode of deterioration in COVID-19. Among hypoxemia measures, the ratio of the partial pressure of arterial oxygen to the fraction of inspired oxygen (P/F ratio) has optimal construct validity but poor availability because it requires arterial blood sampling. Pulse oximetry reports oxygenation continuously, but occult hypoxemia can occur in Black patients because the technique is affected by skin color. Oxygen dissociation curves allow non-invasive estimation of P/F ratios (ePFR) but this approach remains unproven.

Research Question

Can ePFRs measure overt and occult hypoxemia?

Study Design and methods

We retrospectively studied COVID-19 hospital encounters (n=5319) at two academic centers (University of Virginia [UVA] and Emory University). We measured primary outcomes (death or ICU transfer within 24 hours), ePFR, conventional hypoxemia measures, baseline predictors (age, sex, race, comorbidity), and acute predictors (National Early Warning Score (NEWS) and Sepsis-3). We updated predictors every 15 minutes. We assessed predictive validity using adjusted odds ratios (AOR) and area under receiver operating characteristics curves (AUROC). We quantified disparities (Black vs non-Black) in empirical cumulative distributions using the Kolmogorov-Smirnov (K-S) two-sample test.

Results

Overt hypoxemia (low ePFR) predicted bad outcomes (AOR for a 100-point ePFR drop: 2.7 [UVA]; 1.7 [Emory]; p<0.01) with better discrimination (AUROC: 0.76 [UVA]; 0.71 [Emory]) than NEWS (AUROC: 0.70 [UVA]; 0.70 [Emory]) or Sepsis-3 (AUROC: 0.68 [UVA]; 0.65 [Emory]). We found racial differences consistent with occult hypoxemia. Black patients had better apparent oxygenation (K-S distance: 0.17 [both sites]; p<0.01) but, for comparable ePFRs, worse outcomes than other patients (AOR: 2.2 [UVA]; 1.2 [Emory], p<0.01).

Interpretation

The ePFR was a valid measure of overt hypoxemia. In COVID-19, it may outperform multi-organ dysfunction models like NEWS and Sepsis-3. By accounting for biased oximetry as well as clinicians’ real-time responses to it (supplemental oxygen adjustment), ePFRs may enable statistical modelling of racial disparities in outcomes attributable to occult hypoxemia.

Development of Machine-Learning Model to Predict COVID-19 Mortality: Application of Ensemble Model and Regarding Feature Impacts

This study was designed to develop machine-learning models to predict COVID-19 mortality and identify its key features based on clinical characteristics and laboratory tests. For this, deep-learning (DL) and machine-learning (ML) models were developed using receiver operating characteristic (ROC) area under the curve (AUC) and F1 score optimization of 87 parameters. Of the two, the DL model exhibited better performance (AUC 0.8721, accuracy 0.84, and F1 score 0.76). However, we also blended DL with ML, and the ensemble model performed the best (AUC 0.8811, accuracy 0.85, and F1 score 0.77). The DL model is generally unable to extract feature importance; however, we succeeded by using the Shapley Additive exPlanations method for each model. This study demonstrated both the applicability of DL and ML models for classifying COVID-19 mortality using hospital-structured data and that the ensemble model had the best predictive ability.

Insights Regarding the Berlin Definition of ARDS from Prospective Observational Studies

The definition of acute respiratory distress syndrome (ARDS), has evolved since it was first described in 1967 by Ashbaugh and Petty to the current "Berlin" definition of ARDS developed in 2012 by an expert panel, that provided clarification on the definition of "acute," and on the cardiac failure criteria. It expanded the definition to include patients receiving non-invasive ventilation, and removed the term "acute lung injury" and added a requirement of patients to be receiving a minimum 5 cmH₂O expiratory pressure.Since 2012, a series of observational cohort studies have generated insights into the utility and robustness of this definition. This review will examine novel insights into the epidemiology of ARDS, failures in ARDS diagnosis, the role of lung imaging in ARDS, the novel ARDS cohort that is not invasively ventilated, lung compliance profiles in patients with ARDS, sex differences that exist in ARDS management and outcomes, the progression of ARDS following initial diagnosis, and the clinical profile and outcomes of confirmed versus resolved ARDS. Furthermore, we will discuss studies that challenge the utility of distinguishing ARDS from other causes of acute hypoxemic respiratory failure (AHRF) and identify issues that may need to be addressed in a revised definition.

Machine Learning Approaches to Identify Discriminative Signatures of Volatile Organic Compounds (VOCs) from Bacteria and Fungi Using SPME-DART-MS

Point-of-care screening tools are essential to expedite patient care and decrease reliance on slow diagnostic tools (e.g., microbial cultures) to identify pathogens and their associated antibiotic resistance. Analysis of volatile organic compounds (VOC) emitted from biological media has seen increased attention in recent years as a potential non-invasive diagnostic procedure. This work explores the use of solid phase micro-extraction (SPME) and ambient plasma ionization mass spectrometry (MS) to rapidly acquire VOC signatures of bacteria and fungi. The MS spectrum of each pathogen goes through a preprocessing and feature extraction pipeline. Various supervised and unsupervised machine learning (ML) classification algorithms are trained and evaluated on the extracted feature set. These are able to classify the type of pathogen as bacteria or fungi with high accuracy, while marked progress is also made in identifying specific strains of bacteria. This study presents a new approach for the identification of pathogens from VOC signatures collected using SPME and ambient ionization MS by training classifiers on just a few samples of data. This ambient plasma ionization and ML approach is robust, rapid, precise, and can potentially be used as a non-invasive clinical diagnostic tool for point-of-care applications.

Multi-Task Learning with Recurrent Neural Networks for ARDS Prediction using only EHR Data: Model Development and Validation Study (Preprint)

Background

Acute respiratory distress syndrome (ARDS) is a condition that is often considered to have broad and subjective diagnostic criteria and is associated with significant mortality and morbidity. Early and accurate prediction of ARDS and related conditions such as hypoxemia and sepsis could allow timely administration of therapies, leading to improved patient outcomes.

Objective

The aim of this study is to perform an exploration of how multilabel classification in the clinical setting can take advantage of the underlying dependencies between ARDS and related conditions to improve early prediction of ARDS in patients.

Methods

The electronic health record data set included 40,703 patient encounters from 7 hospitals from April 20, 2018, to March 17, 2021. A recurrent neural network (RNN) was trained using data from 5 hospitals, and external validation was conducted on data from 2 hospitals. In addition to ARDS, 12 target labels for related conditions such as sepsis, hypoxemia, and COVID-19 were used to train the model to classify a total of 13 outputs. As a comparator, XGBoost models were developed for each of the 13 target labels. Model performance was assessed using the area under the receiver operating characteristic curve. Heat maps to visualize attention scores were generated to provide interpretability to the neural networks. Finally, cluster analysis was performed to identify potential phenotypic subgroups of patients with ARDS.

Results

The single RNN model trained to classify 13 outputs outperformed the individual XGBoost models for ARDS prediction, achieving an area under the receiver operating characteristic curve of 0.842 on the external test sets. Models trained on an increasing number of tasks resulted in improved performance. Earlier prediction of ARDS nearly doubled the rate of in-hospital survival. Cluster analysis revealed distinct ARDS subgroups, some of which had similar mortality rates but different clinical presentations.

Conclusions

The RNN model presented in this paper can be used as an early warning system to stratify patients who are at risk of developing one of the multiple risk outcomes, hence providing practitioners with the means to take early action.

Artificial intelligence–aided diagnosis model for acute respiratory distress syndrome combining clinical data and chest radiographs

Objective

The aim of this study was to develop an artificial intelligence–based model to detect the presence of acute respiratory distress syndrome (ARDS) using clinical data and chest X-ray (CXR) data.

Method

The transfer learning method was used to train a convolutional neural network (CNN) model with an external image dataset to extract the image features. Then, the last layer of the model was fine-tuned to determine the probability of ARDS. The clinical data were trained using three machine learning algorithms—eXtreme Gradient Boosting (XGB), random forest (RF), and logistic regression (LR)—to estimate the probability of ARDS. Finally, ensemble-weighted methods were proposed that combined the image model and the clinical data model to estimate the probability of ARDS. An analysis of the importance of clinical features was performed to explore the most important features in detecting ARDS. A gradient-weighted class activation mapping (Grad-CAM) model was used to explain what our CNN sees and understands when making a decision.

Results

The proposed ensemble-weighted methods improved the performances of the ARDS classifiers (XGB + CNN, area under the curve [AUC] = 0.916; RF + CNN, AUC = 0.920; LR + CNN, AUC = 0.920; XGB + RF + LR + CNN, AUC = 0.925). In addition, the ML model using clinical data to present the top 15 important features to identify the risk factors of ARDS.

Conclusion

This study developed combined machine learning models with clinical data and CXR images to detect ARDS. According to the results of the Shapley Additive exPlanations values and the Grad-CAM techniques, an explicable ARDS diagnosis model is suitable for a real-life scenario.

Clinical prediction system of complications among patients with COVID-19: A development and validation retrospective multicentre study during first wave of the pandemic

Clinical evidence suggests that some patients diagnosed with coronavirus disease 2019 (COVID-19) experience a variety of complications associated with significant morbidity, especially in severe cases during the initial spread of the pandemic. To support early interventions, we propose a machine learning system that predicts the risk of developing multiple complications. We processed data collected from 3,352 patient encounters admitted to 18 facilities between April 1 and April 30, 2020, in Abu Dhabi (AD), United Arab Emirates. Using data collected during the first 24 h of admission, we trained machine learning models to predict the risk of developing any of three complications after 24 h of admission. The complications include Secondary Bacterial Infection (SBI), Acute Kidney Injury (AKI), and Acute Respiratory Distress Syndrome (ARDS). The hospitals were grouped based on geographical proximity to assess the proposed system's learning generalizability, AD Middle region and AD Western & Eastern regions, A and B, respectively. The overall system includes a data filtering criterion, hyperparameter tuning, and model selection. In test set A, consisting of 587 patient encounters (mean age: 45.5), the system achieved a good area under the receiver operating curve (AUROC) for the prediction of SBI (0.902 AUROC), AKI (0.906 AUROC), and ARDS (0.854 AUROC). Similarly, in test set B, consisting of 225 patient encounters (mean age: 42.7), the system performed well for the prediction of SBI (0.859 AUROC), AKI (0.891 AUROC), and ARDS (0.827 AUROC). The performance results and feature importance analysis highlight the system's generalizability and interpretability. The findings illustrate how machine learning models can achieve a strong performance even when using a limited set of routine input variables. Since our proposed system is data-driven, we believe it can be easily repurposed for different outcomes considering the changes in COVID-19 variants over time.

Management of COVID-19-Associated Acute Respiratory Failure with Alternatives to Invasive Mechanical Ventilation: High-Flow Oxygen, Continuous Positive Airway Pressure, and Noninvasive Ventilation

Patients admitted to hospital with coronavirus disease 2019 (COVID-19) may develop acute respiratory failure (ARF) with compromised gas exchange. These patients require oxygen and possibly ventilatory support, which can be delivered via different devices. Initially, oxygen therapy will often be administered through a conventional binasal oxygen catheter or air-entrainment mask. However, when higher rates of oxygen flow are needed, patients are often stepped up to high-flow nasal cannula oxygen therapy (HFNC), continuous positive airway pressure (CPAP), bilevel positive airway pressure (BiPAP), or invasive mechanical ventilation (IMV). BiPAP, CPAP, and HFNC may be beneficial alternatives to IMV for COVID-19-associated ARF. Current evidence suggests that when nasal catheter oxygen therapy is insufficient for adequate oxygenation of patients with COVID-19-associated ARF, CPAP should be provided for prolonged periods. Subsequent escalation to IMV may be implemented if necessary.

Management of COVID-19-Associated Acute Respiratory Failure with Alternatives to Invasive Mechanical Ventilation: High-Flow Oxygen, Continuous Positive Airway Pressure, and Noninvasive Ventilation

Patients admitted to hospital with coronavirus disease 2019 (COVID-19) may develop acute respiratory failure (ARF) with compromised gas exchange. These patients require oxygen and possibly ventilatory support, which can be delivered via different devices. Initially, oxygen therapy will often be administered through a conventional binasal oxygen catheter or air-entrainment mask. However, when higher rates of oxygen flow are needed, patients are often stepped up to high-flow nasal cannula oxygen therapy (HFNC), continuous positive airway pressure (CPAP), bilevel positive airway pressure (BiPAP), or invasive mechanical ventilation (IMV). BiPAP, CPAP, and HFNC may be beneficial alternatives to IMV for COVID-19-associated ARF. Current evidence suggests that when nasal catheter oxygen therapy is insufficient for adequate oxygenation of patients with COVID-19-associated ARF, CPAP should be provided for prolonged periods. Subsequent escalation to IMV may be implemented if necessary.