Prediction Of Acute Respiratory Distress Syndrome Research Articles

The mechanical ventilation guided by electrical impedance tomography in acute respiratory distress syndrome

Acute respiratory distress syndrome (ARDS) is a common and critical clinical condition characterized by diffuse damage to the lung interstitium, alveoli, and increased permeability of pulmonary blood vessels. CT can be used to assess the imaging features, severity, and prediction of ARDS, but it requires patient transportation to the CT room and is only a static examination. Electrical impedance tomography (EIT) is an increasingly widely used monitoring tool in clinical applications in recent years. It enables continuous real-time assessment of lung ventilation distribution at the bedside and has high clinical value in optimizing mechanical ventilation parameters for critically ill patients. This article introduces the basic principles of EIT and how to better utilize EIT technology to guide mechanical ventilation treatment for ARDS patients.

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

BackgroundEarly 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. MethodsWe 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). ResultsResults 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.

A prediction model for acute respiratory distress syndrome in immunocompetent adults with adenovirus-associated Pneumonia: a multicenter retrospective analysis

BackgroundIn recent years, the number of human adenovirus (HAdV)-related pneumonia cases has increased in immunocompetent adults. Acute respiratory distress syndrome (ARDS) in these patients is the predominant cause of HADV-associated fatality rates. This study aimed to identify early risk factors to predict early HAdV-related ARDS.MethodsData from immunocompetent adults with HAdV pneumonia between June 2018 and May 2022 in ten tertiary general hospitals in central China was analyzed retrospectively. Patients were categorized into the ARDS group based on the Berlin definition. The prediction model of HAdV-related ARDS was developed using multivariate stepwise logistic regression and visualized using a nomogram.ResultsOf 102 patients with adenovirus pneumonia, 41 (40.2%) developed ARDS. Overall, most patients were male (94.1%), the median age was 38.0 years. Multivariate logistic regression showed that dyspnea, SOFA (Sequential Organ Failure Assessment) score, lactate dehydrogenase (LDH) and mechanical ventilation status were independent risk factors for this development, which has a high mortality rate (41.5%). Incorporating these factors, we established a nomogram with good concordance statistics of 0.904 (95% CI 0.844–0.963) which may help to predict early HAdV-related ARDS.ConclusionA nomogram with good accuracy in the early prediction of ARDS in patients with HAdV-associated pneumonia may could contribute to the early management and effective treatment of severe HAdV infection.

Vascular Endothelial Growth Factor as Potential Biomarker for COVID-19 Severity.

COVID-19 is characterized by immunological responses to viral replication and coherent with endothelitis, microvascular disturbance of lung vasculature and coagulopathy. Vascular Endothelial Growth Factor (VEGF) is a proangiogenic mediator regulating endothelial changes. It is induced by proinflammatory signaling and hypoxia. We sought to determine whether VEGF levels differ between SARS-CoV-2-positive patients of different disease severity and whether VEGF might be useful in risk stratification. After retrospective screening of all SARS-CoV-2-positive patients treated in Unfallkrankenhaus Berlin in 2020, we included those with documented VEGF measurement. We extracted laboratory values and clinical parameters. An exploratory data analysis was performed to detect possible relations between VEGF level and clinical disease features. We included 167 SARS-CoV-2-positive patients of which 139 suffered from COVID-19. Seventy-one of the COVID-19 patients had to be treated in the intensive care unit (ICU), those patients exhibited higher VEGF levels than those being admitted to normal wards (535 vs 279 pg/L, P < .001). APACHE-2 (Acute Physiology And Chronic Health Evaluation Score) correlated with mortality and patients with high values showed higher VEGF concentrations on admission (456 vs 875 pg/L, p = 0.006). Receiver operating characteristic analytic revealed that the occurrence of organ dysfunctions like acute respiratory distress syndrome (ARDS), shock, or acute kidney injury could be predicted by VEGF. It was significantly higher in patients who later died compared to survivors (637 vs 389 pg/mL, P = 0.041) and predicted mortality with same accuracy as established markers. In our cohort, association of VEGF above 277 pg/L on admission with risk of ARDS could be confirmed in logistic regression adjusting for possible confounding factors (odds ratio 3.1, 95% confidence interval: 1.34-7.7). Even though there are several limitations to this retrospective study it revealed that in COVID-19 patients VEGF can contribute to the prediction of necessity of ICU, mortality and the prediction of ARDS, kidney injury or shock. Its use in risk stratification and potential pathogenetic involvement should be further investigated.

Value of plasma AGEs and sRAGE expression in predicting the occurrence of ARDS in elderly COPD patients.

This study aimed to explore the diagnostic value of advanced glycation end products (AGEs) and soluble receptors for advanced glycation end products (sRAGE) by detecting the expression levels of AGEs and sRAGE in the plasma of elderly patients with chronic obstructive pulmonary disease (COPD) combined with acute respiratory distress syndrome (ARDS). For this purpose, 110 COPD patients were divided into the elderly COPD group (n=95) and the elderly COPD combined with the ARDS group (n=15). An additional 100 healthy people were recruited as the control group. All patients were assessed with Acute Physiology and Chronic Health Evaluation (APACHE II score) after admission. The levels of AGEs and sRAGE in the plasma were measured by enzyme-linked immunosorbent assay. Results showed that compared with the elderly COPD group, the APACHE II score was significantly higher in the elderly COPD combined with ARDS group (P < 0.05); the level of plasma AGEs in the control group, the elderly COPD group, and the elderly COPD combined with ARDS group decreased in turn (P < 0.05), and the level of sRAGE increased in turn (P < 0.05). Pearson analysis exhibited that the plasma AGEs level was negatively correlated with the APACHE II score (r = -0.681, P < 0.05), and plasma sRAGE level was positively correlated with the APACHE II score (r = 0.653, P < 0.05). Binary Logistic analysis demonstrated that AGEs were the protective factor for ARDS in elderly COPD patients (P < 0.05), and sRAGE was the risk factor for ARDS in elderly COPD patients (P < 0.05). The areas under the curve of plasma AGEs, sRAGE, and their combination in the prediction of ARDS in elderly COPD patients were 0.860 (95%CI: 0.785-0.935), 0.756 (95%CI: 0.659-0.853), and 0.882 (95%CI: 0.813-0.951), respectively. The decreased level of AGEs and the increased level of sRAGE in the plasma of COPD patients with ARDS are associated with the disease severity and have a certain diagnostic value for ARDS in COPD patients, which may be potential markers for the clinical diagnosis of COPD combined with ARDS.

Deep Learning With Chest Radiographs for Making Prognoses in Patients With COVID-19: Retrospective Cohort Study.

An artificial intelligence (AI) model using chest radiography (CXR) may provide good performance in making prognoses for COVID-19. We aimed to develop and validate a prediction model using CXR based on an AI model and clinical variables to predict clinical outcomes in patients with COVID-19. This retrospective longitudinal study included patients hospitalized for COVID-19 at multiple COVID-19 medical centers between February 2020 and October 2020. Patients at Boramae Medical Center were randomly classified into training, validation, and internal testing sets (at a ratio of 8:1:1, respectively). An AI model using initial CXR images as input, a logistic regression model using clinical information, and a combined model using the output of the AI model (as CXR score) and clinical information were developed and trained to predict hospital length of stay (LOS) ≤2 weeks, need for oxygen supplementation, and acute respiratory distress syndrome (ARDS). The models were externally validated in the Korean Imaging Cohort of COVID-19 data set for discrimination and calibration. The AI model using CXR and the logistic regression model using clinical variables were suboptimal to predict hospital LOS ≤2 weeks or the need for oxygen supplementation but performed acceptably in the prediction of ARDS (AI model area under the curve [AUC] 0.782, 95% CI 0.720-0.845; logistic regression model AUC 0.878, 95% CI 0.838-0.919). The combined model performed better in predicting the need for oxygen supplementation (AUC 0.704, 95% CI 0.646-0.762) and ARDS (AUC 0.890, 95% CI 0.853-0.928) compared to the CXR score alone. Both the AI and combined models showed good calibration for predicting ARDS (P=.079 and P=.859). The combined prediction model, comprising the CXR score and clinical information, was externally validated as having acceptable performance in predicting severe illness and excellent performance in predicting ARDS in patients with COVID-19.

Chest computed tomography radiomics to predict the outcome for patients with COVID-19 at an early stage

Early monitoring and intervention for patients with novel coronavirus disease-2019 (COVID-19) will benefit both patients and the medical system. Chest computed tomography (CT) radiomics provide more information regarding the prognosis of COVID-19. A total of 833 quantitative features of 157 COVID-19 patients in the hospital were extracted. By filtering unstable features using the least absolute shrinkage and selection operator algorithm, a radiomic signature was built to predict the prognosis of COVID-19 pneumonia. The main outcomes were the area under the curve (AUC) of the prediction models for death, clinical stage, and complications. Internal validation was performed using the bootstrapping validation technique. The AUC of each model demonstrated good predictive accuracy [death, 0.846; stage, 0.918; complication, 0.919; acute respiratory distress syndrome (ARDS), 0.852]. After finding the optimal cut-off for each outcome, the respective accuracy, sensitivity, and specificity were 0.854, 0.700, and 0.864 for the prediction of the death of COVID-19 patients; 0.814, 0.949, and 0.732 for the prediction of a higher stage of COVID-19; 0.846, 0.920, and 0.832 for the prediction of complications of COVID-19 patients; and 0.814, 0.818, and 0.814 for ARDS of COVID-19 patients. The AUCs after bootstrapping were 0.846 [95% confidence interval (CI): 0.844-0.848] for the death prediction model, 0.919 (95% CI: 0.917-0.922) for the stage prediction model, 0.919 (95% CI: 0.916-0.921) for the complication prediction model, and 0.853 (95% CI: 0.852-0.0.855) for the ARDS prediction model in the internal validation. Based on the decision curve analysis, the radiomics nomogram was clinically significant and useful. The radiomic signature from the chest CT was significantly associated with the prognosis of COVID-19. A radiomic signature model achieved maximum accuracy in the prognosis prediction. Although our results provide vital insights into the prognosis of COVID-19, they need to be verified by large samples in multiple centers.

Combination of transcriptional biomarkers and clinical parameters for early prediction of sepsis indued acute respiratory distress syndrome.

As a common yet intractable complication of severe sepsis, acute respiratory distress syndrome (ARDS) is closely associated with poor clinical outcomes and elevated medical expenses. The aim of the current study is to generate a model combining transcriptional biomarkers and clinical parameters to alarm the development of ARDS in septic patients. Gene expression profile (GSE66890) was downloaded from the Gene Expression Omnibus database and clinical data were extracted. Differentially expressed genes (DEGs) from whole blood leukocytes were identified between patients with sepsis alone and septic patients who develop ARDS. ARDS prediction model was constructed using backward stepwise regression and Akaike Information Criterion (AIC). Meanwhile, a nomogram based on this model was established, with subsequent internal validation. A total of 57 severe septic patients were enrolled in this study, and 28 (49.1%) developed ARDS. Based on the differential expression analysis, six DEGs (BPI, OLFM4, LCN2, CD24, MMP8 and MME) were screened. According to the outcome prediction model, six valuable risk factors (direct lung injury, shock, tumor, BPI, MME and MMP8) were incorporated into a nomogram, which was used to predict the onset of ARDS in septic patients. The calibration curves of the nomogram showed good consistency between the probabilities and observed values. The decision curve analysis also revealed the potential clinical usefulness of the nomogram. The area under the receiver operating characteristic (AUROC) for the prediction of ARDS occurrence in septic patients by the nomogram was 0.86 (95% CI = 0.767-0.952). A sensitivity analysis showed that the AUROC for the prediction of ARDS development in septic patients without direct lung injury was 0.967 (95% CI = 0.896-1.0). The nomogram based on transcriptional biomarkers and clinical parameters showed a good performance for the prediction of ARDS occurrence in septic patients.

Development and Validation of an Acute Respiratory Distress Syndrome Prediction Model in Coronavirus Disease 2019: Updated Lung Injury Prediction Score

ObjectiveTo develop and validate an updated lung injury prediction score for coronavirus disease 2019 (COVID-19) (c-LIPS) tailored for predicting acute respiratory distress syndrome (ARDS) in COVID-19. Patients and MethodsThis was a registry-based cohort study using the Viral Infection and Respiratory Illness Universal Study. Hospitalized adult patients between January 2020 and January 2022 were screened. Patients who qualified for ARDS within the first day of admission were excluded. Development cohort consisted of patients enrolled from participating Mayo Clinic sites. The validation analyses were performed on remaining patients enrolled from more than 120 hospitals in 15 countries. The original lung injury prediction score (LIPS) was calculated and enhanced using reported COVID-19–specific laboratory risk factors, constituting c-LIPS. The main outcome was ARDS development and secondary outcomes included hospital mortality, invasive mechanical ventilation, and progression in WHO ordinal scale. ResultsThe derivation cohort consisted of 3710 patients, of whom 1041 (28.1%) developed ARDS. The c-LIPS discriminated COVID-19 patients who developed ARDS with an area under the curve (AUC) of 0.79 compared with original LIPS (AUC, 0.74; P<.001) with good calibration accuracy (Hosmer-Lemeshow P=.50). Despite different characteristics of the two cohorts, the c-LIPS’s performance was comparable in the validation cohort of 5426 patients (15.9% ARDS), with an AUC of 0.74; and its discriminatory performance was significantly higher than the LIPS (AUC, 0.68; P<.001). The c-LIPS’s performance in predicting the requirement for invasive mechanical ventilation in derivation and validation cohorts had an AUC of 0.74 and 0.72, respectively. ConclusionIn this large patient sample c-LIPS was successfully tailored to predict ARDS in COVID-19 patients.

Developing and evaluating a machine-learning-based algorithm to predict the incidence and severity of ARDS with continuous non-invasive parameters from ordinary monitors and ventilators

ObjectivesMajor observational studies report that the mortality rate of acute respiratory distress syndrome (ARDS) is close to 40%. Different treatment strategies are required for each patient, according to the degree of ARDS. Early prediction of ARDS is helpful to implement targeted drug therapy and mechanical ventilation strategies for patients with different degrees of potential ARDS. In this paper, a new dynamic prediction machine learning model for ARDS incidence and severity is established and evaluated based on 28 parameters from ordinary monitors and ventilators, capable of dynamic prediction of the incidence and severity of ARDS. This new method is expected to meet the clinical practice requirements of user-friendliness and timeliness for wider application. MethodsA total of 4738 hospitalized patients who required ICU care from 159 hospitals are employed in this study. The models are trained by standardized data from electronic medical records. There are 28 structured, continuous non-invasive parameters that are recorded every hour. Seven machine learning models using only continuous, non-invasive parameters are developed for dynamic prediction and compared with methods trained by complete parameters and the traditional risk adjustment method (i.e., oxygenation saturation index method). ResultsThe optimal prediction performance (area under the curve) of the ARDS incidence and severity prediction models built using continuous noninvasive parameters reached0.8691 and 0.7765, respectively. In terms of mild and severe ARDS prediction, the AUC values are both above 0.85. The performance of the model using only continuous non-invasive parameters have an AUC of 0.0133 lower, in comparison with that employing a complete feature set, including continuous non-invasive parameters, demographic information, laboratory parameters and clinical natural language text. ConclusionsA machine learning method was developed in this study using only continuous non-invasive parameters for ARDS incidence and severity prediction. Because the continuous non-invasive parameters can be easily obtained from ordinary monitors and ventilators, the method presented in this study is friendly and convenient to use. It is expected to be applied in pre-hospital setting for early ARDS warning.

Using machine learning for the early prediction of sepsis-associated ARDS in the ICU and identification of clinical phenotypes with differential responses to treatment.

Background: An early diagnosis model with clinical phenotype classification is key for the early identification and precise treatment of sepsis-associated acute respiratory distress syndrome (ARDS). This study aimed to: 1) build a machine learning diagnostic model for patients with sepsis-associated ARDS using easily accessible early clinical indicators, 2) conduct rapid classification of clinical phenotypes in this population, and 3) explore the differences in clinical characteristics, outcomes, and treatment responses of different phenotypes. Methods: This study is based on data from the Telehealth Intensive Care Unit (eICU) and Medical Information Mart for Intensive Care IV (MIMIC-IV). We trained and tested the early diagnostic model of sepsis-associated ARDS patients in the eICU. We used key predictive indicators to cluster sepsis-associated ARDS patients and determine the characteristics and clinical outcomes of different phenotypes, as well to explore the differences of in-hospital mortality among different the positive end-expiratory pressure (PEEP) levels in different phenotypes. These results are verified in MIMIC-IV to evaluate whether they are repeatable. Results: Among the diagnostic models constructed in 19,249 sepsis patients and 5,947 sepsis-associated ARDS patients, the AdaBoost (Decision Tree) model achieved the best performance with an area under the receiver operating characteristic curve (AUC) of 0.895, which is higher than that of the traditional Logistic Regression model (Z = -2.40,p = 0.013), and the accuracy of 70.06%, sensitivity of 78.11% and specificity of 78.74%. We simultaneously identified three sepsis-associated ARDS phenotypes. Cluster 0 (n = 3,669) had the lowest in-hospital mortality rate (6.51%) and fewer laboratory abnormalities (lower WBC (median:15.000K/mcL), lower blood glucose (median:158.000mg/dl), lower creatinine (median:1.200mg/dl), lower lactic acid (median:3.000mmol/L); p < 0.001). Cluster 1 (n = 1,554) had the highest in-hospital mortality rate (75.29%) and the most laboratory abnormalities (higher WBC (median:18.300K/mcL), higher blood glucose (median:188.000mg/dl), higher creatinine (median:2.300mg/dl), higher lactic acid (median:3.900mmol/L); p < 0.001). Cluster 2 (n = 724) had the most complex condition, with a moderate in-hospital mortality rate (29.7%) and the longest intensive care unit stay. In Clusters 0 and 1, patients with high PEEP had higher in-hospital mortality rate than those with low PEEP, but the opposite trend was seen in Cluster 2. These results were repeatable in 11,935 patients with sepsis and 2,699 patients with sepsis-associated ARDS patients in the MIMIC-IV cohort. Conclusion: A machine learning diagnostic model of sepsis-associated ARDS patients was established. Three phenotypes with different clinical features and outcomes were clustered, and these had different therapeutic responses. These findings are helpful for the early and rapid identification of sepsis-associated ARDS patients and their precise individualized treatment.

Deep Learning Chest CT for Clinically Precise Prediction of Sepsis-Induced Acute Respiratory Distress Syndrome: A Protocol for an Observational Ambispective Cohort Study.

Background: Sepsis commonly causes acute respiratory distress syndrome (ARDS), and ARDS contributes to poor prognosis in sepsis patients. Early prediction of ARDS for sepsis patients remains a clinical challenge. This study aims to develop and validate chest computed tomography (CT) radiomic-based signatures for early prediction of ARDS and assessment of individual severity in sepsis patients. Methods: In this ambispective observational cohort study, a deep learning model, a sepsis-induced acute respiratory distress syndrome (SI-ARDS) prediction neural network, will be developed to extract radiomics features of chest CT from sepsis patients. The datasets will be collected from these retrospective and prospective cohorts, including 400 patients diagnosed with sepsis-3 definition during a period from 1 May 2015 to 30 May 2022. 160 patients of the retrospective cohort will be selected as a discovering group to reconstruct the model and 40 patients of the retrospective cohort will be selected as a testing group for internal validation. Additionally, 200 patients of the prospective cohort from two hospitals will be selected as a validating group for external validation. Data pertaining to chest CT, clinical information, immune-associated inflammatory indicators and follow-up will be collected. The primary outcome is to develop and validate the model, predicting in-hospital incidence of SI-ARDS. Finally, model performance will be evaluated using the area under the curve (AUC) of receiver operating characteristic (ROC), sensitivity and specificity, using internal and external validations. Discussion: Present studies reveal that early identification and classification of the SI-ARDS is essential to improve prognosis and disease management. Chest CT has been sought as a useful diagnostic tool to identify ARDS. However, when characteristic imaging findings were clearly presented, delays in diagnosis and treatment were impossible to avoid. In this ambispective cohort study, we hope to develop a novel model incorporating radiomic signatures and clinical signatures to provide an easy-to-use and individualized prediction of SI-ARDS occurrence and severe degree in patients at early stage.

Establishment and verification of LIPS score combined with APACHE II score and oxygenation index to predict the occurrence model of ARDS

To construct and verify the occurrence model of acute respiratory distress syndrome (ARDS) using lung injury prediction score (LIPS) combined with acute physiology and chronic health evaluation II (APACHE II) score and oxygenation index (PaO2/FiO2). Using a prospective cohort study method, 244 patients with complete medical records who were admitted to the intensive care unit (ICU) of Peking University Third Hospital from December 2020 to July 2022 were selected as research objects according to the inclusion and exclusion criteria. They were divided into training set (173 cases) and validation set (71 cases). Patients' gender, age, body mass index (BMI), various causes (shock, sepsis, craniocerebral injury, pulmonary contusion, multiple trauma, aspiration, pneumonia, acute abdomen, hypoproteinemia, acidosis, major surgery, etc.), underlying diseases (diabetes, malignant tumor, cerebrovascular disease, liver disease, kidney disease) and laboratory test indicators were collected. According to the above data, the LIPS score, APACHE II score, sequential organ failure assessment (SOFA) and PaO2/FiO2, etc within 24 hours after admission to the ICU were calculated. Univariate analysis was used to screen the influencing factors for the occurrence of ARDS, and the factors with P < 0.2 were included in the multivariate Logistic regression analysis to screen out the independent predictive factors for the occurrence of ARDS. According to the results of multivariate Logistic regression analysis, the risk score of patients with ARDS was obtained to construct the risk prediction model of ARDS, the receiver operator characteristic curve (ROC curve) was drawn, and the area under the ROC curve (AUC) was calculated. The established ARDS prediction model was externally validated, and ROC curves were drawn to evaluate the predictive accuracy of the prediction model for the occurrence of ARDS in critically ill patients, and the AUC of the validation set was calculated to analyze the predictive performance of each risk factor on the occurrence of ARDS. A total of 173 patients were enrolled in the training set, including 121 patients without ARDS and 52 patients with ARDS; 77 cases of acute abdomen, 64 cases of sepsis, 60 cases of shock, 51 cases of acidosis, 40 cases of hypoproteinemia, 37 cases of diabetes, 34 cases of craniocerebral injury, 34 cases of abnormal liver function, 28 cases of multiple trauma, 23 cases of malignant tumor, 23 cases of spinal orthopedic surgery, 17 cases of obesity, 12 cases of pneumonia, 11 cases of pulmonary contusion, and 7 cases of chronic kidney disease, chemotherapy in 6 cases, and aspiration in 2 cases. The rates of shock, sepsis, acute abdomen, acidosis, abnormal liver function, lung contusion, pneumonia and aspiration, gender, age, LIPS score, APACHE II score, and SOFA score in the ARDS group were significantly higher than those in the non-ARDS group (all P < 0.05), moreover, PaO2/FiO2 ratio was significantly lower than that of non-ARDS group (P < 0.01). Multivariate Logistic regression analysis showed that LIPS score, APACHE II score, and PaO2/FiO2 ratio were independent risk factors for ARDS in ICU patients with high risk factors for ARDS, and the odds ratio (OR) was 1.768 [95% confidence interval (95%CI) was 1.380-2.266], 1.242 (95%CI was 1.089-1.417), 0.985 (95%CI was 0.978-0.991), all P < 0.05. ROC curve analysis showed that the AUC of the ARDS prediction model training set was 0.920, the sensitivity was 86.5%, and the specificity was 86.8%; the AUC of the verification set was 0.896, the sensitivity was 96.8%, and the specificity was 76.6%. LIPS score, APACHE II score and PaO2/FiO2 are independent risk factors for the occurrence of ARDS in ICU patients with high risk factors for ARDS. The ARDS risk prediction model established based on these three indicators has a good predictive ability for the occurrence of ARDS in critically ill patients, wihich needs to be verified by multicenter cohort studies.

Multitask Learning With Recurrent Neural Networks for Acute Respiratory Distress Syndrome Prediction Using Only Electronic Health Record Data: Model Development and Validation Study

BackgroundAcute 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.ObjectiveThe 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.MethodsThe 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.ResultsThe 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.ConclusionsThe 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.

Establishing prediction model of community-acquired pneumonia complicated with acute respiratory distress syndrome based on artificial neural network

To investigate the independent risk factors of community-acquired pneumonia (CAP) complicated with acute respiratory distress syndrome (ARDS), and the accuracy and prevention value of ARDS prediction based on artificial neural network model in CAP patients. A case-control study was conducted. Clinical data of 414 patients with CAP who met the inclusion criteria and were admitted to the comprehensive intensive care unit and respiratory department of Changzhou Second People's Hospital Affiliated to Nanjing Medical University from February 2020 to February 2021 were analyzed. They were divided into two groups according to whether they had complicated with ARDS. The clinical data of the two groups were collected within 24 hours after admission, the influencing factors of ARDS were screened out by univariate analysis, and the artificial neural network model was constructed. Through the artificial neural network model, the importance of input layer independent variables (that was, the influence factors obtained from univariate analysis) on the output layer dependent variables (whether ARDS occurred) was drawn. The artificial neural network modeling data pairs were randomly divided into training group (n = 290) and verification group (n = 124) in a ratio of 7:3. The overall prediction accuracy of the training group and the verification group was calculated respectively. At the same time, the receiver operator characteristic curve (ROC curve) was drawn, and the area under the ROC curve (AUC) was calculated. All 414 patients were enrolled in the analysis, including 82 patients with ARDS and 332 patients without ARDS. Univariate analysis showed that gender, age, heart rate (HR), maximum systolic blood pressure (MSBP), maximum respiratory rate (MRR), source of admission, C-reactive protein (CRP), procalcitonin (PCT), erythrocyte sedimentation rate (ESR), neutrophil count (NEUT), eosinophil count (EOS), fibrinogen equivalent unit (FEU), activated partial thromboplastin time (APTT), total bilirubin (TBil), albumin (ALB), lactate dehydrogenase (LDH), serum creatinine (SCr), hemoglobin (Hb) and blood glucose (GLU) were significantly different between the two groups, which might be the risk factors of CAP patients complicated with ARDS. Taking the above 19 risk factors as the input layer and whether ARDS occurred as the output layer, the artificial neural network model was constructed. Among the input layer independent variables, the top five indicators with the largest influence weight on the neural network model were LDH (100.0%), PCT (74.4%), FEU (61.5%), MRR (56.9%), and APTT (51.6%), indicating that that these five indicators had a greater impact on the occurrence of ARDS in patients with CAP. The overall prediction accuracy of the artificial neural network model in the training group was 94.1% (273/290), and that of the verification group was 89.5% (111/124). The AUC predicted by the aforementioned artificial neural network model for ARDS in CAP patients was 0.977 (95% confidence interval was 0.956-1.000). The prediction model of ARDS in CAP patients based on artificial neural network model has good prediction ability, which can be used to calculate the accuracy of ARDS in CAP patients, and specific preventive measures can be given.

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.

Accuracy of conventional and novel scoring systems in predicting severity and outcomes of acute pancreatitis: a retrospective study

BackgroundRecently, several novel scoring systems have been developed to evaluate the severity and outcomes of acute pancreatitis. This study aimed to compare the effectiveness of novel and conventional scoring systems in predicting the severity and outcomes of acute pancreatitis.MethodsPatients treated between January 2003 and August 2020 were reviewed. The Ranson score (RS), Glasgow score (GS), bedside index of severity in acute pancreatitis (BISAP), pancreatic activity scoring system (PASS), and Chinese simple scoring system (CSSS) were determined within 48 h after admission. Multivariate logistic regression was used for severity, mortality, and organ failure prediction. Optimum cutoffs were identified using receiver operating characteristic curve analysis.ResultsA total of 1848 patients were included. The areas under the curve (AUCs) of RS, GS, BISAP, PASS, and CSSS for severity prediction were 0.861, 0.865, 0.829, 0.778, and 0.816, respectively. The corresponding AUCs for mortality prediction were 0.693, 0.736, 0.789, 0.858, and 0.759. The corresponding AUCs for acute respiratory distress syndrome prediction were 0.745, 0.784, 0.834, 0.936, and 0.820. Finally, the corresponding AUCs for acute renal failure prediction were 0.707, 0.734, 0.781, 0.868, and 0.816.ConclusionsRS and GS predicted severity better than they predicted mortality and organ failure, while PASS predicted mortality and organ failure better. BISAP and CSSS performed equally well in severity and outcome predictions.

Establishment of a predictive model for acute respiratory distress syndrome and analysis of its predictive value in critical burn patients

Objective: To establish a predictive model for acute respiratory distress syndrome (ARDS) in critical burn patients with the screened independent risk factors, and to validate its predictive value. Methods: Totally 131 critical burn patients (101 males and 30 females, aged 18-84 years) who met the inclusion criteria were admitted to the Department of Burns of the Second Affiliated Hospital of Zhejiang University School of Medicine from January 2018 to December 2019. A retrospective case-control study was conducted. The patients were divided into ARDS group (54 cases) and non-ARDS group (77 cases) according to whether ARDS occurred or not. The statistics of patients in the two groups were recorded including the gender, age, burn index, combination of inhalation injury, smoking history, delayed resuscitation, indwelling nasogastric tube, and complication of sepsis, and the data were statistically analyzed with independent sample t test, chi-square test, and Fisher's exact probability test. The multivariate logistic regression analysis was performed on the indicators with statistically significant differences between the two groups to screen the independent risk factors for developing ARDS in critical burn patients, and the corresponding nomograph prediction model for the risk of ARDS in critical burn patients was established. The risk scores for patients developing ARDS were therefore obtained based on the above-mentioned nomograph, and the corresponding receiver operating characteristic (ROC) curve was drawn to calculate the area under the curve. The internal validation of the above-mentioned ARDS prediction model was performed using the Bootstrap method, and the area under the ROC curve was calculated for modeling group (79 cases) and validation group (52 cases), respectively. A calibration curve was drawn to assess the predictive conformity of the above-mentioned ARDS prediction model for the occurrence of ARDS in critical burn patients. Results: The burn index, proportion of combination of inhalation injury, and proportion of complication of sepsis of patients were significantly higher in ARDS group than in non-ARDS group (t=0.36, χ2=33.78, 49.92, P<0.01). The gender, age, smoking history, delayed resuscitation, and indwelling nasogastric tube of patients in ARDS group were close to those in non-ARDS group (P>0.05). The multivariate logistic regression analysis showed that the burn index, combination of inhalation injury, and complication of sepsis were the independent risk factors for developing ARDS in critical burn patients (odds ratio=1.05, 15.33, 5.02, 95% confidence interval=1.01-1.10, 2.65-88.42, 1.28-19.71, P<0.05 or P<0.01). The overall area under the ROC curve of the above-mentioned ARDS prediction model was 0.92 (95% confidence interval=0.88-0.97), and the area under the ROC curve was 0.95 and 0.91 (95% confidence interval=0.90-1.00, 0.86-0.97) for validation group and modeling group, respectively. When applying the above-mentioned ARDS prediction model for ARDS incidence prediction, there might be some risk of overestimating ARDS incidence when the prediction probability was <35.0% or >85.0%, and some risk of underestimating ARDS incidence when the prediction probability was 35.0%-85.0%. Conclusions: The burn index, inhalation injury, and sepsis are the independent risk factors for the occurrence of ARDS in critical burn patients. The risk prediction model for ARDS based on these three indicators has good predictive ability for ARDS in critical burn patients.

Risk Factors and a Scoring System to Predict ARDS in Patients with COVID-19 Pneumonia in Korea: A Multicenter Cohort Study.

Predictive studies of acute respiratory distress syndrome (ARDS) in patients with coronavirus disease 2019 (COVID-19) are limited. In this study, the predictors of ARDS were investigated and a score that can predict progression to ARDS in patients with COVID-19 pneumonia was developed. All patients who were diagnosed with COVID-19 pneumonia between February 1, 2020, and May 15, 2020, at five university hospitals in Korea were enrolled. Their demographic, clinical, and epidemiological characteristics and the outcomes were collected using the World Health Organization COVID-19 Case Report Form. A logistic regression analysis was performed to determine the predictors for ARDS. The receiver operating characteristic (ROC) curves were constructed for the scoring model. Of the 166 patients with COVID-19 pneumonia, 37 (22.3%) patients developed ARDS. The areas under the curves for the infiltration on a chest X-ray, C-reactive protein, neutrophil/lymphocyte ratio, and age, for prediction of ARDS were 0.91, 0.90, 0.87, and 0.80, respectively (all P < 0.001). The COVID-19 ARDS Prediction Score (CAPS) was constructed using age (≥60 years old), C-reactive protein (≥5 mg/dL), and the infiltration on a chest X-ray (≥22%), with each predictor allocated 1 point. The area under the curve of COVID-19 ARDS prediction score (CAPS) for prediction of ARDS was 0.90 (95% CI 0.86–0.95; P < 0.001). It provided 100% sensitivity and 75% specificity when the CAPS score cutoff value was 2 points. CAPS, which consists of age, C-reactive protein, and the area of infiltration on a chest X-ray, was predictive of the development of ARDS in patients with COVID-19 pneumonia.

Acute Respiratory Distress Syndrome Prediction Score: Derivation and Validation

Despite advances in treatment strategies, acute respiratory distress syndrome (ARDS) after cardiac surgery remains associated with high morbidity and mortality. A method of screening patients for risk of ARDS after cardiac surgery is needed. To develop and validate an ARDS prediction score designed to identify patients at high risk of ARDS after cardiac or aortic surgery. An ARDS prediction score was derived from a retrospective derivation cohort and validated in a prospective cohort. Discrimination and calibration of the score were assessed with area under the receiver operating characteristic curve and the Hosmer-Lemeshow goodness-of-fit test, respectively. A sensitivity analysis was conducted to assess model performance at different cutoff points. The retrospective derivation cohort consisted of 201 patients with and 602 patients without ARDS who had undergone cardiac or aortic surgery. Nine routinely available clinical variables were included in the ARDS prediction score. In the derivation cohort, the score distinguished patients with versus without ARDS with area under the curve of 0.84 (95% CI, 0.81-0.88; Hosmer-Lemeshow P = .55). In the validation cohort, 46 of 1834 patients (2.5%) had ARDS develop within 7 days after cardiac or aortic surgery. Area under the curve was 0.78 (95% CI, 0.71-0.85), and the score was well calibrated (Hosmer-Lemeshow P = .53). The ARDS prediction score can be used to identify high-risk patients from the first day after cardiac or aortic surgery. Patients with a score of 3 or greater should be closely monitored. The score requires external validation before clinical use.