Validation and utility of ARDS subphenotypes identified by machine-learning models using clinical data: an observational, multicohort, retrospective analysis

Phenotyping COVID-19 respiratory failure in spontaneously breathing patients with AI on lung CT-scan

BACKGROUND

Automated analysis of lung computed tomography (CT) scans may help characterize subphenotypes of acute respiratory illness. We integrated lung CT features measured via deep learning with clinical and laboratory data in spontaneously breathing subjects to enhance the identification of COVID-19 subphenotypes.

METHODS

This is a multicenter observational cohort study in spontaneously breathing patients with COVID-19 respiratory failure exposed to early lung CT within 7 days of admission. We explored lung CT images using deep learning approaches to quantitative and qualitative analyses; latent class analysis (LCA) by using clinical, laboratory and lung CT variables; regional differences between subphenotypes following 3D spatial trajectories.

RESULTS

Complete datasets were available in 559 patients. LCA identified two subphenotypes (subphenotype 1 and 2). As compared with subphenotype 2 (n = 403), subphenotype 1 patients (n = 156) were older, had higher inflammatory biomarkers, and were more hypoxemic. Lungs in subphenotype 1 had a higher density gravitational gradient with a greater proportion of consolidated lungs as compared with subphenotype 2. In contrast, subphenotype 2 had a higher density submantellar-hilar gradient with a greater proportion of ground glass opacities as compared with subphenotype 1. Subphenotype 1 showed higher prevalence of comorbidities associated with endothelial dysfunction and higher 90-day mortality than subphenotype 2, even after adjustment for clinically meaningful variables.

CONCLUSIONS

Integrating lung-CT data in a LCA allowed us to identify two subphenotypes of COVID-19, with different clinical trajectories. These exploratory findings suggest a role of automated imaging characterization guided by machine learning in subphenotyping patients with respiratory failure.

TRIAL REGISTRATION

ClinicalTrials.gov Identifier: NCT04395482. Registration date: 19/05/2020.

Algor-ethics: charting the ethical path for AI in critical care

The integration of Clinical Decision Support Systems (CDSS) based on artificial intelligence (AI) in healthcare is groundbreaking evolution with enormous potential, but its development and ethical implementation, presents unique challenges, particularly in critical care, where physicians often deal with life-threating conditions requiring rapid actions and patients unable to participate in the decisional process. Moreover, development of AI-based CDSS is complex and should address different sources of bias, including data acquisition, health disparities, domain shifts during clinical use, and cognitive biases in decision-making. In this scenario algor-ethics is mandatory and emphasizes the integration of 'Human-in-the-Loop' and 'Algorithmic Stewardship' principles, and the benefits of advanced data engineering. The establishment of Clinical AI Departments (CAID) is necessary to lead AI innovation in healthcare, ensuring ethical integrity and human-centered development in this rapidly evolving field.

From ICU Syndromes to ICU Subphenotypes: Consensus Report and Recommendations for Developing Precision Medicine in the ICU

Critical care uses syndromic definitions to describe patient groups for clinical practice and research. There is growing recognition that a "precision medicine" approach is required and that integrated biologic and physiologic data identify reproducible subpopulations that may respond differently to treatment. This article reviews the current state of the field and considers how to successfully transition to a precision medicine approach. To impact clinical care, identification of subpopulations must do more than differentiate prognosis. It must differentiate response to treatment, ideally by defining subgroups with distinct functional or pathobiological mechanisms (endotypes). There are now multiple examples of reproducible subpopulations of sepsis, acute respiratory distress syndrome, and acute kidney or brain injury described using clinical, physiological, and/or biological data. Many of these subpopulations have demonstrated the potential to define differential treatment response, largely in retrospective studies, and that the same treatment-responsive subpopulations may cross multiple clinical syndromes (treatable traits). To bring about a change in clinical practice, a precision medicine approach must be evaluated in prospective clinical studies requiring novel adaptive trial designs. Several such studies are underway, but there are multiple challenges to be tackled. Such subpopulations must be readily identifiable and be applicable to all critically ill populations around the world. Subdividing clinical syndromes into subpopulations will require large patient numbers. Global collaboration of investigators, clinicians, industry, and patients over many years will therefore be required to transition to a precision medicine approach and ultimately realize treatment advances seen in other medical fields.

Association between sub-phenotypes identified using latent class analysis and neurological outcomes in patients with out-of-hospital cardiac arrest in Japan

BACKGROUND

In patients who experience out-of-hospital cardiac arrest (OHCA), it is important to assess the association of sub-phenotypes identified by latent class analysis (LCA) using pre-hospital prognostic factors and factors measurable immediately after hospital arrival with neurological outcomes at 30 days, which would aid in making treatment decisions.

METHODS

This study retrospectively analyzed data obtained from the Japanese OHCA registry between June 2014 and December 2019. The registry included a complete set of data on adult patients with OHCA, which was used in the LCA. The association between the sub-phenotypes and 30-day survival with favorable neurological outcomes was investigated. Furthermore, adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were estimated by multivariate logistic regression analysis using in-hospital data as covariates.

RESULTS

A total of, 22,261 adult patients who experienced OHCA were classified into three sub-phenotypes. The factor with the highest discriminative power upon patient's arrival was Glasgow Coma Scale followed by partial pressure of oxygen. Thirty-day survival with favorable neurological outcome as the primary outcome was evident in 66.0% participants in Group 1, 5.2% in Group 2, and 0.5% in Group 3. The 30-day survival rates were 80.6%, 11.8%, and 1.3% in groups 1, 2, and 3, respectively. Logistic regression analysis revealed that the ORs (95% CI) for 30-day survival with favorable neurological outcomes were 137.1 (99.4-192.2) for Group 1 and 4.59 (3.46-6.23) for Group 2 in comparison to Group 3. For 30-day survival, the ORs (95%CI) were 161.7 (124.2-212.1) for Group 1 and 5.78 (4.78-7.04) for Group 2, compared to Group 3.

CONCLUSIONS

This study identified three sub-phenotypes based on the prognostic factors available immediately after hospital arrival that could predict neurological outcomes and be useful in determining the treatment strategy of patients experiencing OHCA upon their arrival at the hospital.

A systematic review of machine learning models for management, prediction and classification of ARDS

AIM

Acute respiratory distress syndrome or ARDS is an acute, severe form of respiratory failure characterised by poor oxygenation and bilateral pulmonary infiltrates. Advancements in signal processing and machine learning have led to promising solutions for classification, event detection and predictive models in the management of ARDS.

METHOD

In this review, we provide systematic description of different studies in the application of Machine Learning (ML) and artificial intelligence for management, prediction, and classification of ARDS. We searched the following databases: Google Scholar, PubMed, and EBSCO from 2009 to 2023. A total of 243 studies was screened, in which, 52 studies were included for review and analysis. We integrated knowledge of previous work providing the state of art and overview of explainable decision models in machine learning and have identified areas for future research.

RESULTS

Gradient boosting is the most common and successful method utilised in 12 (23.1%) of the studies. Due to limitation of data size available, neural network and its variation is used by only 8 (15.4%) studies. Whilst all studies used cross validating technique or separated database for validation, only 1 study validated the model with clinician input. Explainability methods were presented in 15 (28.8%) of studies with the most common method is feature importance which used 14 times.

CONCLUSION

For databases of 5000 or fewer samples, extreme gradient boosting has the highest probability of success. A large, multi-region, multi centre database is required to reduce bias and take advantage of neural network method. A framework for validating with and explaining ML model to clinicians involved in the management of ARDS would be very helpful for development and deployment of the ML model.

Acute lung injury: a view from the perspective of necroptosis

BACKGROUND

ALI/ARDS is a syndrome of acute onset characterized by progressive hypoxemia and noncardiogenic pulmonary edema as the primary clinical manifestations. Necroptosis is a form of programmed cell necrosis that is precisely regulated by molecular signals. This process is characterized by organelle swelling and membrane rupture, is highly immunogenic, involves extensive crosstalk with various cellular stress mechanisms, and is significantly implicated in the onset and progression of ALI/ARDS.

METHODS

The current body of literature on necroptosis and ALI/ARDS was thoroughly reviewed. Initially, an overview of the molecular mechanism of necroptosis was provided, followed by an examination of its interactions with apoptosis, pyroptosis, autophagy, ferroptosis, PANOptosis, and NETosis. Subsequently, the involvement of necroptosis in various stages of ALI/ARDS progression was delineated. Lastly, drugs targeting necroptosis, biomarkers, and current obstacles were presented.

CONCLUSION

Necroptosis plays an important role in the progression of ALI/ARDS. However, since ALI/ARDS is a clinical syndrome caused by a variety of mechanisms, we emphasize that while focusing on necroptosis, it may be more beneficial to treat ALI/ARDS by collaborating with other mechanisms.

Differential Effect of Positive End-Expiratory Pressure Strategies in Patients With ARDS: A Bayesian Analysis of Clinical Subphenotypes

BACKGROUND

ARDS is a heterogeneous condition with two subphenotypes identified by different methodologies. Our group similarly identified two ARDS subphenotypes using nine routinely available clinical variables. However, whether these are associated with differential response to treatment has yet to be explored.

RESEARCH QUESTION

Are there differential responses to positive end-expiratory pressure (PEEP) strategies on 28-day mortality according to subphenotypes in adult patients with ARDS?

STUDY DESIGN AND METHODS

We evaluated data from two prior ARDS trials (Higher vs Lower Positive End-Expiratory Pressures in Patients With the ARDS [ALVEOLI] and the Alveolar Recruitment in ARDS Trial [ART]) that compared different PEEP strategies. We classified patients into one of two subphenotypes as described previously. We assessed the differential effect of PEEP with a Bayesian hierarchical logistic model for the primary outcome of 28-day mortality.

RESULTS

We analyzed data from 1,559 patients with ARDS. Compared with lower PEEP, a higher PEEP strategy resulted in higher 28-day mortality in patients with subphenotype A disease in the ALVEOLI study (OR, 1.61; 95% credible interval [CrI], 0.90-2.94) and ART (OR, 1.73; 95% CrI, 1.01-2.98), with a probability of harm resulting from higher PEEP in this subphenotype of 94.3% and 97.7% in the ALVEOLI and ART studies, respectively. Higher PEEP was not associated with mortality in patients with subphenotype B disease in each trial (OR, 0.95 [95% CrI, 0.51-1.73] and 1.00 [95% CrI, 0.63-1.55], respectively), with probability of benefit of 56.4% and 50.7% in the ALVEOLI and ART studies, respectively. These effects were not modified by Pao2 to Fio2 ratio, driving pressure, or the severity of illness for the cohorts.

INTERPRETATION

We found evidence of differential response to PEEP strategies across two ARDS subphenotypes, suggesting possible harm with a higher PEEP strategy in one subphenotype. These observations may assist with predictive enrichment in future clinical trials.

Derivation and Validation of Generalized Sepsis-induced Acute Respiratory Failure Phenotypes Among Critically Ill Patients: A Retrospective Study

Background

Septic patients who develop acute respiratory failure (ARF) requiring mechanical ventilation represent a heterogenous subgroup of critically ill patients with widely variable clinical characteristics. Identifying distinct phenotypes of these patients may reveal insights about the broader heterogeneity in the clinical course of sepsis. We aimed to derive novel phenotypes of sepsis-induced ARF using observational clinical data and investigate their generalizability across multi-ICU specialties, considering multi-organ dynamics.

Methods

We performed a multi-center retrospective study of ICU patients with sepsis who required mechanical ventilation for ≥24 hours. Data from two different high-volume academic hospital systems were used as a derivation set with N=3,225 medical ICU (MICU) patients and a validation set with N=848 MICU patients. For the multi-ICU validation, we utilized retrospective data from two surgical ICUs at the same hospitals (N=1,577). Clinical data from 24 hours preceding intubation was used to derive distinct phenotypes using an explainable machine learning-based clustering model interpreted by clinical experts.

Results

Four distinct ARF phenotypes were identified: A (severe multi-organ dysfunction (MOD) with a high likelihood of kidney injury and heart failure), B (severe hypoxemic respiratory failure [median P/F=123]), C (mild hypoxia [median P/F=240]), and D (severe MOD with a high likelihood of hepatic injury, coagulopathy, and lactic acidosis). Patients in each phenotype showed differences in clinical course and mortality rates despite similarities in demographics and admission co-morbidities. The phenotypes were reproduced in external validation utilizing an external MICU from second hospital and SICUs from both centers. Kaplan-Meier analysis showed significant difference in 28-day mortality across the phenotypes (p<0.01) and consistent across both centers. The phenotypes demonstrated differences in treatment effects associated with high positive end-expiratory pressure (PEEP) strategy.

Conclusion

The phenotypes demonstrated unique patterns of organ injury and differences in clinical outcomes, which may help inform future research and clinical trial design for tailored management strategies.

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.

ARDS Mortality Prediction Model Using Evolving Clinical Data and Chest Radiograph Analysis

INTRODUCTION

Within primary ARDS, SARS-CoV-2-associated ARDS (C-ARDS) emerged in late 2019, reaching its peak during the subsequent two years. Recent efforts in ARDS research have concentrated on phenotyping this heterogeneous syndrome to enhance comprehension of its pathophysiology.

METHODS AND RESULTS

A retrospective study was conducted on C-ARDS patients from April 2020 to February 2021, encompassing 110 participants with a mean age of 63.2 ± 11.92 (26-83 years). Of these, 61.2% (68) were male, and 25% (17) experienced severe ARDS, resulting in a mortality rate of 47.3% (52). Ventilation settings, arterial blood gases, and chest X-ray (CXR) were evaluated on the first day of invasive mechanical ventilation and between days two and three. CXR images were scrutinized using a convolutional neural network (CNN). A binary logistic regression model for predicting C-ARDS mortality was developed based on the most influential variables: age, PaO2/FiO2 ratio (P/F) on days one and three, CNN-extracted CXR features, and age. Initial performance assessment on test data (23 patients out of the 110) revealed an area under the receiver operating characteristic (ROC) curve of 0.862 with a 95% confidence interval (0.654-0.969).

CONCLUSION

Integrating data available in all intensive care units enables the prediction of C-ARDS mortality by utilizing evolving P/F ratios and CXR. This approach can assist in tailoring treatment plans and initiating early discussions to escalate care and extracorporeal life support. Machine learning algorithms for imaging classification can uncover otherwise inaccessible patterns, potentially evolving into another form of ARDS phenotyping. The combined features of these algorithms and clinical variables demonstrate superior performance compared to either element alone.

Forecasting disease trajectories in critical illness: comparison of probabilistic dynamic systems to static models to predict patient status in the intensive care unit

OBJECTIVE

Conventional prediction models fail to integrate the constantly evolving nature of critical illness. Alternative modelling approaches to study dynamic changes in critical illness progression are needed. We compare static risk prediction models to dynamic probabilistic models in early critical illness.

DESIGN

We developed models to simulate disease trajectories of critically ill COVID-19 patients across different disease states. Eighty per cent of cases were randomly assigned to a training and 20% of the cases were used as a validation cohort. Conventional risk prediction models were developed to analyse different disease states for critically ill patients for the first 7 days of intensive care unit (ICU) stay. Daily disease state transitions were modelled using a series of multivariable, multinomial logistic regression models. A probabilistic dynamic systems modelling approach was used to predict disease trajectory over the first 7 days of an ICU admission. Forecast accuracy was assessed and simulated patient clinical trajectories were developed through our algorithm.

SETTING AND PARTICIPANTS

We retrospectively studied patients admitted to a Cleveland Clinic Healthcare System in Ohio, for the treatment of COVID-19 from March 2020 to December 2022.

RESULTS

5241 patients were included in the analysis. For ICU days 2-7, the static (conventional) modelling approach, the accuracy of the models steadily decreased as a function of time, with area under the curve (AUC) for each health state below 0.8. But the dynamic forecasting approach improved its ability to predict as a function of time. AUC for the dynamic forecasting approach were all above 0.90 for ICU days 4-7 for all states.

CONCLUSION

We demonstrated that modelling critical care outcomes as a dynamic system improved the forecasting accuracy of the disease state. Our model accurately identified different disease conditions and trajectories, with a <10% misclassification rate over the first week of critical illness.

High protein intake and nitrogen balance in patients receiving venovenous extracorporeal membrane oxygenation: A descriptive cohort study

BACKGROUND

This retrospective cohort study sought to describe the ability of high protein regimens to achieve nitrogen equilibrium in patients receiving venovenous extracorporeal membrane oxygenation (VV ECMO).

METHODS

Patients aged ≥18 years with a documented nitrogen balance study (NB) on VV ECMO between February 2018 and December 2021 were included. Studies with incomplete 24-h urine collections or changes in blood urea nitrogen ≥10 mg/dl were excluded. Data were summarized, correlation between first NB and potentially contributing variables was assessed with Kendall tau. Subanalysis described findings after stratifying for weight class (obese vs nonobese) and duration of VV ECMO at the time of NB.

RESULTS

A total of 68 NBs in 30 patients were included; 47% of the cohort had obesity. The number of NBs per patient was 2.2 ± 1.1, which were completed on a median of 31.5 (interquartile range: 16, 53.8) days receiving ECMO. Nitrogen equilibrium or positive balance was achieved in 72% of studies despite elevated nitrogen excretion. Patients received 87.9 ± 16.8% of prescribed protein on NB days for average intakes of 2.4 ± 0.4 g/kg of actual weight per day and 2.4 ± 0.5 g/kg of ideal weight per day in patients without and with obesity. Median NB in patients without obesity was -1.46 (-8.96, 2.98) g/day and -0.21 (-10.58, 4.04) g/day in patients with obesity. A difference in median NB after stratification for timing was observed (P = 0.029).

CONCLUSION

Nitrogen equilibrium can be achieved with high protein intake in adults receiving VV ECMO. NB monitoring is one tool to individualize protein prescriptions throughout the course of VV ECMO.

Construction of a novel lower-extremity peripheral artery disease subtype prediction model using unsupervised machine learning and neutrophil-related biomarkers

Lower-extremity peripheral artery disease (LE-PAD) is a prevalent circulatory disorder with risks of critical limb ischemia and amputation. This study aimed to develop a prediction model for a novel LE-PAD subtype to predict the severity of the disease and guide personalized interventions. Additionally, LE-PAD pathogenesis involves altered immune microenvironment, we examined the immune differences to elucidate LE-PAD pathogenesis. A total of 460 patients with LE-PAD were enrolled and clustered using unsupervised machine learning algorithms (UMLAs). Logistic regression analyses were performed to screen and identify predictive factors for the novel subtype of LE-PAD and a prediction model was built. We performed a comparative analysis regarding neutrophil levels in different subgroups of patients and an immune cell infiltration analysis to explore the associations between neutrophil levels and LE-PAD. Through hematoxylin and eosin (H&E) staining of lower-extremity arteries, neutrophil infiltration in patients with and without LE-PAD was compared. We found that UMLAs can helped in constructing a prediction model for patients with novel LE-PAD subtypes which enabled risk stratification for patients with LE-PAD using routinely available clinical data to assist clinical decision-making and improve personalized management for patients with LE-PAD. Additionally, the results indicated the critical role of neutrophil infiltration in LE-PAD pathogenesis.

Chinese experts' consensus on the application of intensive care big data

The development of intensive care medicine is inseparable from the diversified monitoring data. Intensive care medicine has been closely integrated with data since its birth. Critical care research requires an integrative approach that embraces the complexity of critical illness and the computational technology and algorithms that can make it possible. Considering the need of standardization of application of big data in intensive care, Intensive Care Medicine Branch of China Health Information and Health Care Big Data Society, Standard Committee has convened expert group, secretary group and the external audit expert group to formulate Chinese Experts' Consensus on the Application of Intensive Care Big Data (2022). This consensus makes 29 recommendations on the following five parts: Concept of intensive care big data, Important scientific issues, Standards and principles of database, Methodology in solving big data problems, Clinical application and safety consideration of intensive care big data. The consensus group believes this consensus is the starting step of application big data in the field of intensive care. More explorations and big data based retrospective research should be carried out in order to enhance safety and reliability of big data based models of critical care field.

Identification of Clinical Heterogeneity and Construction of Prediction Models for Novel Subtypes in Patients with Abdominal Aortic Aneurysm: An Unsupervised Machine Learning Study

BACKGROUND

Abdominal aortic aneurysm (AAA) is one of the most common diseases in vascular surgery. Endovascular aneurysm repair (EVAR) can effectively treat AAA. It is essential to accurately classify patients with AAA who need EVAR.

METHODS

We enrolled 266 patients with AAA who underwent EVAR. Unsupervised machine learning algorithms (UMLAs) were used to cluster subjects according to similar clinical characteristics. To verify UMLA's accuracy, the operative and postoperative results of the 2 clusters were analyzed. Finally, a prediction model was developed using binary logistic regression analysis.

RESULTS

UMLAs could correctly classify patients based on their clinical characteristics. Patients in Cluster 1 were older, had a higher BMI, and were more likely than patients in Cluster 2 to develop pneumonia, chronic obstructive pulmonary disease, and cerebrovascular disease. The aneurysm diameter, neck angulation, diameter and angulation of bilateral common iliac arteries, and incidence of iliac artery aneurysm were significantly higher in cluster 1 patients than in cluster 2. Cluster 1 had a longer operative time, a longer length of stay in the intensive care unit and hospital, a higher medical expense, and a higher incidence of reintervention. A nomogram was established based on the BMI, neck angulation, left common iliac artery (LCIA) diameter and angulation, and right common iliac artery (RCIA) diameter and angulation. The nomogram was evaluated using receiver operating characteristic curve analysis, with an area under the curve of 0.933 (95% confidence interval, 0.902-0.963) and a C-index of 0.927.

CONCLUSIONS

Our findings demonstrate that UMLAs can be used to rationally classify a heterogeneous cohort of patients with AAA effectively, and the analysis of postoperative variables also verified the accuracy of UMLAs. We established a prediction model for new subtypes of AAA, which can improve the quality of management of patients with AAA.

A New Global Definition of Acute Respiratory Distress Syndrome

Background: Since publication of the 2012 Berlin definition of acute respiratory distress syndrome (ARDS), several developments have supported the need for an expansion of the definition, including the use of high-flow nasal oxygen, the expansion of the use of pulse oximetry in place of arterial blood gases, the use of ultrasound for chest imaging, and the need for applicability in resource-limited settings. Methods: A consensus conference of 32 critical care ARDS experts was convened, had six virtual meetings (June 2021 to March 2022), and subsequently obtained input from members of several critical care societies. The goal was to develop a definition that would 1) identify patients with the currently accepted conceptual framework for ARDS, 2) facilitate rapid ARDS diagnosis for clinical care and research, 3) be applicable in resource-limited settings, 4) be useful for testing specific therapies, and 5) be practical for communication to patients and caregivers. Results: The committee made four main recommendations: 1) include high-flow nasal oxygen with a minimum flow rate of ⩾30 L/min; 2) use PaO2:FiO2 ⩽ 300 mm Hg or oxygen saturation as measured by pulse oximetry SpO2:FiO2 ⩽ 315 (if oxygen saturation as measured by pulse oximetry is ⩽97%) to identify hypoxemia; 3) retain bilateral opacities for imaging criteria but add ultrasound as an imaging modality, especially in resource-limited areas; and 4) in resource-limited settings, do not require positive end-expiratory pressure, oxygen flow rate, or specific respiratory support devices. Conclusions: We propose a new global definition of ARDS that builds on the Berlin definition. The recommendations also identify areas for future research, including the need for prospective assessments of the feasibility, reliability, and prognostic validity of the proposed global definition.

Pneumomediastinum and pneumothorax in acute respiratory distress syndrome (ARDS) patients: a narrative review

Background and Objective

Acute respiratory distress syndrome (ARDS) is a severe, life-threatening medical condition characterized by poor oxygenation due to non-compliant lungs secondary diffuse alveolar damage. Encouragingly, the incidence of ARDS has declined steadily recently, attributed mainly to implementation of keystone guidelines and continuous research efforts. Mechanical ventilation is the cornerstone of supportive care for ARDS patients. This review aims to consolidate the current knowledge on pneumothorax (PNX) and pneumomediastinum (PMD) and to enhance the understanding of the readers. The objectives are to (I) explore the etiology and risk factors of PNX and PMD, (II) discuss the various diagnostic modalities available, (III) evaluate management options, and (IV) recent advancements.

Methods

A search of the literature was conducted using PubMed, MEDLINE, and Google Scholar for relevant articles pertaining to PNX and PMD in ARDS population. The clinical presentation, diagnostic and management strategies of PNX, PMD, and ARDS were summarized, and all authors reviewed the selection and decide which studies to include.

Key Content and Findings

The adoption of lung-protective ventilation strategies, based on the review of literature from the recent years, shows that it has played a significant role in reducing the occurrence of barotrauma, such as PNX and PMD. However, PNX and PMD remains to be a challenging complication to manage. With a specific focus on PNX and PMD, this review provides valuable insights into effectively managing and understanding these critical complications among ARDS patients.

Conclusions

ARDS, with its evolving definition, continues to pose a life-threatening threat. Despite the widespread adoption of lung-protective ventilation strategies, PNX and PMD present persistent challenges in management. Further research is imperative to enhance the risk assessment of ARDS patients prone to developing PNX and PMD and to institute more effective prevention and treatment measures.

Mitigation of Sepsis-Induced Acute Lung Injury by BMSC-Derived Exosomal miR-125b-5p Through STAT3-Mediated Suppression of Macrophage Pyroptosis

Introduction

Sepsis is a syndrome characterized by high morbidity and mortality rates. One of its most severe complications is acute lung injury, which exhibits a multitude of clinical and biological features, including macrophage pyroptosis. This study investigates the regulatory effects of exosomes derived from Bone Marrow-Derived Mesenchymal Stem Cells (BMSCs) on sepsis-associated acute lung injury (ALI) and explores the potential mechanisms mediated by exosomal miRNAs.

Methods

Exosomes were isolated from primary BMSCs of adult C57BL/6J mice using differential centrifugation. Their uptake and distribution in both in vitro and in vivo contexts were validated. Key sepsis-associated hub gene signal transducer and activator of transcription 3 (STAT3) and its upstream non-coding miR-125b-5p were elucidated through a combination of bioinformatics, machine learning, and miRNA sequencing. Subsequently, the therapeutic potential of BMSC-derived exosomes in alleviating sepsis-induced acute lung injury was substantiated. Moreover, the functionalities of miR-125b-5p and STAT3 were corroborated through miR-125b-5p inhibitor and STAT3 agonist interventions, employing gain and loss-of-function strategies both in vitro and in vivo. Finally, a dual-luciferase reporter assay reaffirmed the interaction between miR-125b-5p and STAT3.

Results

We isolated exosomes from primary BMSCs and confirmed their accumulation in the mouse lung as well as their uptake by macrophages in vitro. This study identified the pivotal sepsis-associated hub gene STAT3 and demonstrated that exosomes derived from BMSCs can target STAT3, thereby inhibiting macrophage pyroptosis. MiR-125b-5p inhibition experiments showed that exosomes mitigate macrophage pyroptosis and lung injury by delivering miR-125b-5p. STAT3 overexpression experiments validated that miR-125b-5p reduces macrophage pyroptosis and lung injury by suppressing STAT3. Furthermore, a dual-luciferase reporter assay confirmed the binding interaction between miR-125b-5p and STAT3.

Conclusion

Exosomes derived from BMSCs, serving as carriers for delivering miR-125b-5p, can downregulate STAT3, thereby inhibiting macrophage pyroptosis and alleviating sepsis-associated ALI. These significant findings provide valuable insights into the potential development of ALI therapies centred around exosomes derived from BMSC.

Dynamic changes in lactate levels within the first 24 hours in septic patients as a prognostic indicator: A retrospective cohort study utilizing latent class growth analysis

Elevated lactate levels are common in sepsis patients. This study aimed to assess the effect of dynamic changes in lactate levels within the first 24 hours following admission on patient prognosis. We extracted data from the Medical Information Mart for Intensive Care (MIMIC)-IV database and classified patients using latent class growth analysis (LCGA). This analysis classified sepsis patients into different groups based on dynamic changes in lactate levels during the initial 24 hours post-admission, dividing this time frame into four periods (0-3 h, 3-6 h, 6-12 h, and 12-24 h). The highest lactate level recorded in each period was then used for patient classification. We subsequently compared the baseline characteristics and outcomes between these different groups. Our study encompassed 7,830 patients, whom LCGA successfully divided into two classes: class 1 (steady lactate class) and class 2 (increasing lactate class). Class 2 demonstrated a worse clinical status at baseline, as indicated by vital signs, disease severity scores, and laboratory results. Importantly, class 2 also had a significantly higher 28-day mortality rate than class 1 (55.6% vs 13.5%, P < 0.001). In conclusion, LCGA effectively categorized sepsis patients into two distinct groups based on their dynamic changes in lactate levels during the first 24 hours post-admission. This methodology has potential utility in clinical practice for managing sepsis patients.

Identifying Cardiovascular Disease Risk Factors in Adults with Explainable Artificial Intelligence

BACKGROUND

The aim of this study was to evaluate the relationship between risk factors causing cardiovascular diseases and their importance with explainable machine learning models.

METHODS

In this retrospective study, multiple databases were searched, and data on 11 risk factors of 70 000 patients were obtained. Data included risk factors highly associated with cardiovascular disease and having/not having any cardiovascular disease. The explainable prediction model was constructed using 7 machine learning algorithms: Random Forest Classifier, Extreme Gradient Boost Classifier, Decision Tree Classifier, KNeighbors Classifier, Support Vector Machine Classifier, and GaussianNB. Receiver operating characteristic curve, Brier scores, and mean accuracy were used to assess the model's performance. The interpretability of the predicted results was examined using Shapley additive description values.

RESULTS

The accuracy, area under the curve values, and Brier scores of the Extreme Gradient Boost model (the best prediction model for cardiovascular disease risk factors) were calculated as 0.739, 0.803, and 0.260, respectively. The most important risk factors in the permutation feature importance method and explainable artificial intelligence-Shapley's explanations method are systolic blood pressure (ap_hi) [0.1335 ± 0.0045 w (weight)], cholesterol (0.0341 ± 0.0022 w), and age (0.0211 ± 0.0036 w).

CONCLUSION

The created explainable machine learning model has become a successful clinical model that can predict cardiovascular patients and explain the impact of risk factors. Especially in the clinical setting, this model, which has an accurate, explainable, and transparent algorithm, will help encourage early diagnosis of patients with cardiovascular diseases, risk factors, and possible treatment options.

Identifying molecular phenotypes in sepsis: an analysis of two prospective observational cohorts and secondary analysis of two randomised controlled trials

BACKGROUND

In sepsis and acute respiratory distress syndrome (ARDS), heterogeneity has contributed to difficulty identifying effective pharmacotherapies. In ARDS, two molecular phenotypes (hypoinflammatory and hyperinflammatory) have consistently been identified, with divergent outcomes and treatment responses. In this study, we sought to derive molecular phenotypes in critically ill adults with sepsis, determine their overlap with previous ARDS phenotypes, and evaluate whether they respond differently to treatment in completed sepsis trials.

METHODS

We used clinical data and plasma biomarkers from two prospective sepsis cohorts, the Validating Acute Lung Injury biomarkers for Diagnosis (VALID) study (N=1140) and the Early Assessment of Renal and Lung Injury (EARLI) study (N=818), in latent class analysis (LCA) to identify the optimal number of classes in each cohort independently. We used validated models trained to classify ARDS phenotypes to evaluate concordance of sepsis and ARDS phenotypes. We applied these models retrospectively to the previously published Prospective Recombinant Human Activated Protein C Worldwide Evaluation in Severe Sepsis and Septic Shock (PROWESS-SHOCK) trial and Vasopressin and Septic Shock Trial (VASST) to assign phenotypes and evaluate heterogeneity of treatment effect.

FINDINGS

A two-class model best fit both VALID and EARLI (p<0·0001). In VALID, 804 (70·5%) of the 1140 patients were classified as hypoinflammatory and 336 (29·5%) as hyperinflammatory; in EARLI, 530 (64·8%) of 818 were hypoinflammatory and 288 (35·2%) hyperinflammatory. We observed higher plasma pro-inflammatory cytokines, more vasopressor use, more bacteraemia, lower protein C, and higher mortality in the hyperinflammatory than in the hypoinflammatory phenotype (p<0·0001 for all). Classifier models indicated strong concordance between sepsis phenotypes and previously identified ARDS phenotypes (area under the curve 0·87-0·96, depending on the model). Findings were similar excluding participants with both sepsis and ARDS. In PROWESS-SHOCK, 1142 (68·0%) of 1680 patients had the hypoinflammatory phenotype and 538 (32·0%) had the hyperinflammatory phenotype, and response to activated protein C differed by phenotype (p=0·0043). In VASST, phenotype proportions were similar to other cohorts; however, no treatment interaction with the type of vasopressor was observed (p=0·72).

INTERPRETATION

Molecular phenotypes previously identified in ARDS are also identifiable in multiple sepsis cohorts and respond differently to activated protein C. Molecular phenotypes could represent a treatable trait in critical illness beyond the patient's syndromic diagnosis.

FUNDING

US National Institutes of Health.

Development of genomic phenotype and immunophenotype of acute respiratory distress syndrome using autophagy and metabolism-related genes

Background

Distinguishing ARDS phenotypes is of great importance for its precise treatment. In the study, we attempted to ascertain its phenotypes based on metabolic and autophagy-related genes and infiltrated immune cells.

Methods

Transcription datasets of ARDS patients were obtained from Gene expression omnibus (GEO), autophagy and metabolic-related genes were from the Human Autophagy Database and the GeneCards Database, respectively. Autophagy and metabolism-related differentially expressed genes (AMRDEGs) were further identified by machine learning and processed for constructing the nomogram and the risk prediction model. Functional enrichment analyses of differentially expressed genes were performed between high- and low-risk groups. According to the protein-protein interaction network, these hub genes closely linked to increased risk of ARDS were identified with CytoHubba. ssGSEA and CIBERSORT was applied to analyze the infiltration pattern of immune cells in ARDS. Afterwards, immunologically characterized and molecular phenotypes were constructed according to infiltrated immune cells and hub genes.

Results

A total of 26 AMRDEGs were obtained, and CTSB and EEF2 were identified as crucial AMRDEGs. The predictive capability of the risk score, calculated based on the expression levels of CTSB and EEF2, was robust for ARDS in both the discovery cohort (AUC = 1) and the validation cohort (AUC = 0.826). The mean risk score was determined to be 2.231332, and based on this score, patients were classified into high-risk and low-risk groups. 371 differential genes in high- and low-risk groups were analyzed. ITGAM, TYROBP, ITGB2, SPI1, PLEK, FGR, MPO, S100A12, HCK, and MYC were identified as hub genes. A total of 12 infiltrated immune cells were differentially expressed and have correlations with hub genes. According to hub genes and implanted immune cells, ARDS patients were divided into two different molecular phenotypes (Group 1: n = 38; Group 2: n = 19) and two immune phenotypes (Cluster1: n = 22; Cluster2: n = 35), respectively.

Conclusion

This study picked up hub genes of ARDS related to autophagy and metabolism and clustered ARDS patients into different molecular phenotypes and immunophenotypes, providing insights into the precision medicine of treating patients with ARDS.

Identification and Prediction of Clinical Phenotypes in Hospitalized Patients With COVID-19: Machine Learning From Medical Records

BACKGROUND

There is significant heterogeneity in disease progression among hospitalized patients with COVID-19. The pathogenesis of SARS-CoV-2 infection is attributed to a complex interplay between virus and host immune response that in some patients unpredictably and rapidly leads to "hyperinflammation" associated with increased risk of mortality. The early identification of patients at risk of progression to hyperinflammation may help inform timely therapeutic decisions and lead to improved outcomes.

OBJECTIVE

The primary objective of this study was to use machine learning to reproducibly identify specific risk-stratifying clinical phenotypes across hospitalized patients with COVID-19 and compare treatment response characteristics and outcomes. A secondary objective was to derive a predictive phenotype classification model using routinely available early encounter data that may be useful in informing optimal COVID-19 bedside clinical management.

METHODS

This was a retrospective analysis of electronic health record data of adult patients (N=4379) who were admitted to a Johns Hopkins Health System hospital for COVID-19 treatment from 2020 to 2021. Phenotypes were identified by clustering 38 routine clinical observations recorded during inpatient care. To examine the reproducibility and validity of the derived phenotypes, patient data were randomly divided into 2 cohorts, and clustering analysis was performed independently for each cohort. A predictive phenotype classifier using the gradient-boosting machine method was derived using routine clinical observations recorded during the first 6 hours following admission.

RESULTS

A total of 2 phenotypes (designated as phenotype 1 and phenotype 2) were identified in patients admitted for COVID-19 in both the training and validation cohorts with similar distributions of features, correlations with biomarkers, treatments, comorbidities, and outcomes. In both the training and validation cohorts, phenotype-2 patients were older; had elevated markers of inflammation; and were at an increased risk of requiring intensive care unit-level care, developing sepsis, and mortality compared with phenotype-1 patients. The gradient-boosting machine phenotype prediction model yielded an area under the curve of 0.89 and a positive predictive value of 0.83.

CONCLUSIONS

Using machine learning clustering, we identified and internally validated 2 clinical COVID-19 phenotypes with distinct treatment or response characteristics consistent with similar 2-phenotype models derived from other hospitalized populations with COVID-19, supporting the reliability and generalizability of these findings. COVID-19 phenotypes can be accurately identified using machine learning models based on readily available early encounter clinical data. A phenotype prediction model based on early encounter data may be clinically useful for timely bedside risk stratification and treatment personalization.

Derivation, Validation, and Clinical Relevance of a Pediatric Sepsis Phenotype With Persistent Hypoxemia, Encephalopathy, and Shock

OBJECTIVES

Untangling the heterogeneity of sepsis in children and identifying clinically relevant phenotypes could lead to the development of targeted therapies. Our aim was to analyze the organ dysfunction trajectories of children with sepsis-associated multiple organ dysfunction syndrome (MODS) to identify reproducible and clinically relevant sepsis phenotypes and determine if they are associated with heterogeneity of treatment effect (HTE) to common therapies.

DESIGN

Multicenter observational cohort study.

SETTING

Thirteen PICUs in the United States.

PATIENTS

Patients admitted with suspected infections to the PICU between 2012 and 2018.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We used subgraph-augmented nonnegative matrix factorization to identify candidate trajectory-based phenotypes based on the type, severity, and progression of organ dysfunction in the first 72 hours. We analyzed the candidate phenotypes to determine reproducibility as well as prognostic, therapeutic, and biological relevance. Overall, 38,732 children had suspected infection, of which 15,246 (39.4%) had sepsis-associated MODS with an in-hospital mortality of 10.1%. We identified an organ dysfunction trajectory-based phenotype (which we termed persistent hypoxemia, encephalopathy, and shock) that was highly reproducible, had features of systemic inflammation and coagulopathy, and was independently associated with higher mortality. In a propensity score-matched analysis, patients with persistent hypoxemia, encephalopathy, and shock phenotype appeared to have HTE and benefit from adjuvant therapy with hydrocortisone and albumin. When compared with other high-risk clinical syndromes, the persistent hypoxemia, encephalopathy, and shock phenotype only overlapped with 50%-60% of patients with septic shock, moderate-to-severe pediatric acute respiratory distress syndrome, or those in the top tier of organ dysfunction burden, suggesting that it represents a nonsynonymous clinical phenotype of sepsis-associated MODS.

CONCLUSIONS

We derived and validated the persistent hypoxemia, encephalopathy, and shock phenotype, which is highly reproducible, clinically relevant, and associated with HTE to common adjuvant therapies in children with sepsis.

The application of artificial intelligence in the management of sepsis

Sepsis is a complex and heterogeneous syndrome that remains a serious challenge to healthcare worldwide. Patients afflicted by severe sepsis or septic shock are customarily placed under intensive care unit (ICU) supervision, where a multitude of apparatus is poised to produce high-granularity data. This reservoir of high-quality data forms the cornerstone for the integration of AI into clinical practice. However, existing reviews currently lack the inclusion of the latest advancements. This review examines the evolving integration of artificial intelligence (AI) in sepsis management. Applications of artificial intelligence include early detection, subtyping analysis, precise treatment and prognosis assessment. AI-driven early warning systems provide enhanced recognition and intervention capabilities, while profiling analyzes elucidate distinct sepsis manifestations for targeted therapy. Precision medicine harnesses the potential of artificial intelligence for pathogen identification, antibiotic selection, and fluid optimization. In conclusion, the seamless amalgamation of artificial intelligence into the domain of sepsis management heralds a transformative shift, ushering in novel prospects to elevate diagnostic precision, therapeutic efficacy, and prognostic acumen. As AI technologies develop, their impact on shaping the future of sepsis care warrants ongoing research and thoughtful implementation.

Leveraging Data Science and Novel Technologies to Develop and Implement Precision Medicine Strategies in Critical Care

Precision medicine aims to identify treatments that are most likely to result in favorable outcomes for subgroups of patients with similar clinical and biological characteristics. The gaps for the development and implementation of precision medicine strategies in the critical care setting are many, but the advent of data science and multi-omics approaches, combined with the rich data ecosystem in the intensive care unit, offer unprecedented opportunities to realize the promise of precision critical care. In this article, the authors review the data-driven and technology-based approaches being leveraged to discover and implement precision medicine strategies in the critical care setting.

Sepsis and acute respiratory failure in patients with cancer: how can we improve care and outcomes even further?

PURPOSE OF REVIEW

Care and outcomes of critically ill patients with cancer have improved over the past decade. This selective review will discuss recent updates in sepsis and acute respiratory failure among patients with cancer, with particular focus on important opportunities to improve outcomes further through attention to phenotyping, predictive analytics, and improved outcome measures.

RECENT FINDINGS

The prevalence of cancer diagnoses in intensive care units (ICUs) is nontrivial and increasing. Sepsis and acute respiratory failure remain the most common critical illness syndromes affecting these patients, although other complications are also frequent. Recent research in oncologic sepsis has described outcome variation - including ICU, hospital, and 28-day mortality - across different types of cancer (e.g., solid vs. hematologic malignancies) and different sepsis definitions (e.g., Sepsis-3 vs. prior definitions). Research in acute respiratory failure in oncology patients has highlighted continued uncertainty in the value of diagnostic bronchoscopy for some patients and in the optimal respiratory support strategy. For both of these syndromes, specific challenges include multifactorial heterogeneity (e.g. in etiology and/or underlying cancer), delayed recognition of clinical deterioration, and complex outcomes measurement.

SUMMARY

Improving outcomes in oncologic critical care requires attention to the heterogeneity of cancer diagnoses, timely recognition and management of critical illness, and defining appropriate ICU outcomes.

Estimating the attributable fraction of mortality from acute respiratory distress syndrome to inform enrichment in future randomised clinical trials

BACKGROUND

Efficiency of randomised clinical trials of acute respiratory distress syndrome (ARDS) depends on the fraction of deaths attributable to ARDS (AFARDS) to which interventions are targeted. Estimates of AFARDS in subpopulations of ARDS could improve design of ARDS trials.

METHODS

We performed a matched case-control study using the Large observational study to UNderstand the Global impact of Severe Acute respiratory FailurE cohort. Primary outcome was intensive care unit mortality. We used nearest neighbour propensity score matching without replacement to match ARDS to non-ARDS populations. We derived two separate AFARDS estimates by matching patients with ARDS to patients with non-acute hypoxaemic respiratory failure (non-AHRF) and to patients with AHRF with unilateral infiltrates only (AHRF-UL). We also estimated AFARDS in subgroups based on severity of hypoxaemia, number of lung quadrants involved and hyperinflammatory versus hypoinflammatory phenotypes. Additionally, we derived AFAHRF estimates by matching patients with AHRF to non-AHRF controls, and AFAHRF-UL estimates by matching patients with AHRF-UL to non-AHRF controls.

RESULTS

Estimated AFARDS was 20.9% (95% CI 10.5% to 31.4%) when compared with AHRF-UL controls and 38.0% (95% CI 34.4% to 41.6%) compared with non-AHRF controls. Within subgroups, estimates for AFARDS compared with AHRF-UL controls were highest in patients with severe hypoxaemia (41.1% (95% CI 25.2% to 57.1%)), in those with four quadrant involvement on chest radiography (28.9% (95% CI 13.4% to 44.3%)) and in the hyperinflammatory subphenotype (26.8% (95% CI 6.9% to 46.7%)). Estimated AFAHRF was 33.8% (95% CI 30.5% to 37.1%) compared with non-AHRF controls. Estimated AFAHRF-UL was 21.3% (95% CI 312.8% to 29.7%) compared with non-AHRF controls.

CONCLUSIONS

Overall AFARDS mean values were between 20.9% and 38.0%, with higher AFARDS seen with severe hypoxaemia, four quadrant involvement on chest radiography and hyperinflammatory ARDS.

Precision management of acute kidney injury in the intensive care unit: current state of the art

Acute kidney injury (AKI) is a prototypical example of a common syndrome in critical illness defined by consensus. The consensus definition for AKI, traditionally defined using only serum creatinine and urine output, was needed to standardize the description for epidemiology and to harmonize eligibility for clinical trials. However, AKI is not a simple disease, but rather a complex and multi-factorial syndrome characterized by a wide spectrum of pathobiology. AKI is now recognized to be comprised of numerous sub-phenotypes that can be discriminated through shared features such as etiology, prognosis, or common pathobiological mechanisms of injury and damage. The characterization of sub-phenotypes can serve to enable prognostic enrichment (i.e., identify subsets of patients more likely to share an outcome of interest) and predictive enrichment (identify subsets of patients more likely to respond favorably to a given therapy). Existing and emerging biomarkers will aid in discriminating sub-phenotypes of AKI, facilitate expansion of diagnostic criteria, and be leveraged to realize personalized approaches to management, particularly for recognizing treatment-responsive mechanisms (i.e., endotypes) and targets for intervention (i.e., treatable traits). Specific biomarkers (e.g., serum renin; olfactomedin 4 (OLFM4); interleukin (IL)-9) may further enable identification of pathobiological mechanisms to serve as treatment targets. However, even non-specific biomarkers of kidney injury (e.g., neutrophil gelatinase-associated lipocalin, NGAL; [tissue inhibitor of metalloproteinases 2, TIMP2]·[insulin like growth factor binding protein 7, IGFBP7]; kidney injury molecule 1, KIM-1) can direct greater precision management for specific sub-phenotypes of AKI. This review will summarize these evolving concepts and recent innovations in precision medicine approaches to the syndrome of AKI in critical illness, along with providing examples of how they can be leveraged to guide patient care.

Clinical Sepsis Phenotypes in Critically Ill Patients

Sepsis, defined as the life-threatening dysregulated host response to an infection leading to organ dysfunction, is considered as one of the leading causes of mortality worldwide, especially in intensive care units (ICU). Moreover, sepsis remains an enigmatic clinical syndrome, with complex pathophysiology incompletely understood and a great heterogeneity both in terms of clinical expression, patient response to currently available therapeutic interventions and outcomes. This heterogeneity proves to be a major obstacle in our quest to deliver improved treatment in septic critical care patients; thus, identification of clinical phenotypes is absolutely necessary. Although this might be seen as an extremely difficult task, nowadays, artificial intelligence and machine learning techniques can be recruited to quantify similarities between individuals within sepsis population and differentiate them into distinct phenotypes regarding not only temperature, hemodynamics or type of organ dysfunction, but also fluid status/responsiveness, trajectories in ICU and outcome. Hopefully, we will eventually manage to determine both the subgroup of septic patients that will benefit from a therapeutic intervention and the correct timing of applying the intervention during the disease process.

Machine learning-based prediction of in-ICU mortality in pneumonia patients

Conventional severity-of-illness scoring systems have shown suboptimal performance for predicting in-intensive care unit (ICU) mortality in patients with severe pneumonia. This study aimed to develop and validate machine learning (ML) models for mortality prediction in patients with severe pneumonia. This retrospective study evaluated patients admitted to the ICU for severe pneumonia between January 2016 and December 2021. The predictive performance was analyzed by comparing the area under the receiver operating characteristic curve (AU-ROC) of ML models to that of conventional severity-of-illness scoring systems. Three ML models were evaluated: (1) logistic regression with L2 regularization, (2) gradient-boosted decision tree (LightGBM), and (3) multilayer perceptron (MLP). Among the 816 pneumonia patients included, 223 (27.3%) patients died. All ML models significantly outperformed the Simplified Acute Physiology Score II (AU-ROC: 0.650 [0.584-0.716] vs 0.820 [0.771-0.869] for logistic regression vs 0.827 [0.777-0.876] for LightGBM 0.838 [0.791-0.884] for MLP; P < 0.001). In the analysis for NRI, the LightGBM and MLP models showed superior reclassification compared with the logistic regression model in predicting in-ICU mortality in all length of stay in the ICU subgroups; all age subgroups; all subgroups with any APACHE II score, PaO₂/FiO₂ ratio < 200; all subgroups with or without history of respiratory disease; with or without history of CVA or dementia; treatment with mechanical ventilation, and use of inotropic agents. In conclusion, the ML models have excellent performance in predicting in-ICU mortality in patients with severe pneumonia. Moreover, this study highlights the potential advantages of selecting individual ML models for predicting in-ICU mortality in different subgroups.

Machine learning predicts lung recruitment in acute respiratory distress syndrome using single lung CT scan

BACKGROUND

To develop and validate classifier models that could be used to identify patients with a high percentage of potentially recruitable lung from readily available clinical data and from single CT scan quantitative analysis at intensive care unit admission. 221 retrospectively enrolled mechanically ventilated, sedated and paralyzed patients with acute respiratory distress syndrome (ARDS) underwent a PEEP trial at 5 and 15 cmH2O of PEEP and two lung CT scans performed at 5 and 45 cmH2O of airway pressure. Lung recruitability was defined at first as percent change in not aerated tissue between 5 and 45 cmH2O (radiologically defined; recruiters: Δ45-5non-aerated tissue  > 15%) and secondly as change in PaO2 between 5 and 15 cmH2O (gas exchange-defined; recruiters: Δ15-5PaO2  > 24 mmHg). Four machine learning (ML) algorithms were evaluated as classifiers of radiologically defined and gas exchange-defined lung recruiters using different models including different variables, separately or combined, of lung mechanics, gas exchange and CT data.

RESULTS

ML algorithms based on CT scan data at 5 cmH2O classified radiologically defined lung recruiters with similar AUC as ML based on the combination of lung mechanics, gas exchange and CT data. ML algorithm based on CT scan data classified gas exchange-defined lung recruiters with the highest AUC.

CONCLUSIONS

ML based on a single CT data at 5 cmH2O represented an easy-to-apply tool to classify ARDS patients in recruiters and non-recruiters according to both radiologically defined and gas exchange-defined lung recruitment within the first 48 h from the start of mechanical ventilation.

ESICM guidelines on acute respiratory distress syndrome: definition, phenotyping and respiratory support strategies

The aim of these guidelines is to update the 2017 clinical practice guideline (CPG) of the European Society of Intensive Care Medicine (ESICM). The scope of this CPG is limited to adult patients and to non-pharmacological respiratory support strategies across different aspects of acute respiratory distress syndrome (ARDS), including ARDS due to coronavirus disease 2019 (COVID-19). These guidelines were formulated by an international panel of clinical experts, one methodologist and patients' representatives on behalf of the ESICM. The review was conducted in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement recommendations. We followed the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach to assess the certainty of evidence and grade recommendations and the quality of reporting of each study based on the EQUATOR (Enhancing the QUAlity and Transparency Of health Research) network guidelines. The CPG addressed 21 questions and formulates 21 recommendations on the following domains: (1) definition; (2) phenotyping, and respiratory support strategies including (3) high-flow nasal cannula oxygen (HFNO); (4) non-invasive ventilation (NIV); (5) tidal volume setting; (6) positive end-expiratory pressure (PEEP) and recruitment maneuvers (RM); (7) prone positioning; (8) neuromuscular blockade, and (9) extracorporeal life support (ECLS). In addition, the CPG includes expert opinion on clinical practice and identifies the areas of future research.

Latent class analysis-derived classification improves the cancer-specific death stratification of molecular subtyping in colorectal cancer

The molecular subtypes of colorectal cancer (CRC) represent a comprehensive dissection of CRC heterogeneity. However, molecular feature-based classification systems have limitations in accurately prognosticating stratification due to the inability to distinguish cancer-specific deaths. This study aims to establish a classification system that bridges clinical characteristics, cause-specific deaths, and molecular features. We adopted latent class analysis (LCA) on 491,107 first primary CRC patients from the Surveillance, Epidemiology, and End Results (SEER) database to reveal hidden profiles of CRC. The LCA-derived classification scheme was further applied to The Cancer Genome Atlas (TCGA) to assess its effectiveness in improving the accurate stratification of molecular-based subtypes of CRC. Four classes were identified based on latent class analysis integrating demographic and clinicopathological information of CRC patients. The LCA-derived Class 1 (LCAC1) and the LCAC2 showed a high risk of dying from non-CRC, while patients in LCAC3 had a risk of dying from CRC 1.41 times that of LCAC1 (95% confidence interval [CI] = 1.39-1.43). LCAC4 had the lowest probability to die from non-CRC (hazard ratio [HR] = 0.22, 95% CI = 0.21-0.24) compared with LCAC1. Since the LCA-derived classification can identify patients susceptible to CRC-specific death, adjusting for this classification allows molecular-based subtypes to achieve more accurate survival stratification. We provided a classification system capable of distinguish CRC-specific death, which will improve the accuracy of consensus molecular subtypes for CRC patients' survival stratification. Further studies are warranted to confirm the molecular features of LCA-derived classification to inform potential therapeutic strategies and treatment recommendations.

Personalized Respiratory Support in ARDS: A Physiology-to-Bedside Review

Acute respiratory distress syndrome (ARDS) is a leading cause of disability and mortality worldwide, and while no specific etiologic interventions have been shown to improve outcomes, noninvasive and invasive respiratory support strategies are life-saving interventions that allow time for lung recovery. However, the inappropriate management of these strategies, which neglects the unique features of respiratory, lung, and chest wall mechanics may result in disease progression, such as patient self-inflicted lung injury during spontaneous breathing or by ventilator-induced lung injury during invasive mechanical ventilation. ARDS characteristics are highly heterogeneous; therefore, a physiology-based approach is strongly advocated to titrate the delivery and management of respiratory support strategies to match patient characteristics and needs to limit ARDS progression. Several tools have been implemented in clinical practice to aid the clinician in identifying the ARDS sub-phenotypes based on physiological peculiarities (inspiratory effort, respiratory mechanics, and recruitability), thus allowing for the appropriate application of personalized supportive care. In this narrative review, we provide an overview of noninvasive and invasive respiratory support strategies, as well as discuss how identifying ARDS sub-phenotypes in daily practice can help clinicians to deliver personalized respiratory support and potentially improve patient outcomes.

Bottom-up and top-down paradigms of artificial intelligence research approaches to healthcare data science using growing real-world big data

OBJECTIVES

As the real-world electronic health record (EHR) data continue to grow exponentially, novel methodologies involving artificial intelligence (AI) are becoming increasingly applied to enable efficient data-driven learning and, ultimately, to advance healthcare. Our objective is to provide readers with an understanding of evolving computational methods and help in deciding on methods to pursue.

TARGET AUDIENCE

The sheer diversity of existing methods presents a challenge for health scientists who are beginning to apply computational methods to their research. Therefore, this tutorial is aimed at scientists working with EHR data who are early entrants into the field of applying AI methodologies.

SCOPE

This manuscript describes the diverse and growing AI research approaches in healthcare data science and categorizes them into 2 distinct paradigms, the bottom-up and top-down paradigms to provide health scientists venturing into artificial intelligent research with an understanding of the evolving computational methods and help in deciding on methods to pursue through the lens of real-world healthcare data.

Inhaled Nitric Oxide in Acute Respiratory Distress Syndrome Subsets: Rationale and Clinical Applications

Acute respiratory distress syndrome (ARDS) is a life-threatening condition, characterized by diffuse inflammatory lung injury. Since the coronavirus disease 2019 (COVID-19) pandemic spread worldwide, the most common cause of ARDS has been the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. Both the COVID-19-associated ARDS and the ARDS related to other causes-also defined as classical ARDS-are burdened by high mortality and morbidity. For these reasons, effective therapeutic interventions are urgently needed. Among them, inhaled nitric oxide (iNO) has been studied in patients with ARDS since 1993 and it is currently under investigation. In this review, we aim at describing the biological and pharmacological rationale of iNO treatment in ARDS by elucidating similarities and differences between classical and COVID-19 ARDS. Thereafter, we present the available evidence on the use of iNO in clinical practice in both types of respiratory failure. Overall, iNO seems a promising agent as it could improve the ventilation/perfusion mismatch, gas exchange impairment, and right ventricular failure, which are reported in ARDS. In addition, iNO may act as a viricidal agent and prevent lung hyperinflammation and thrombosis of the pulmonary vasculature in the specific setting of COVID-19 ARDS. However, the current evidence on the effects of iNO on outcomes is limited and clinical studies are yet to demonstrate any survival benefit by administering iNO in ARDS.

Conditional Treatment Effect Analysis of Two Infusion Rates for Fluid Challenges in Critically Ill Patients: A Secondary Analysis of Balanced Solution versus Saline in Intensive Care Study (BaSICS) Trial

Rationale: Optimal infusion rate for fluid challenges in critically ill patients is unknown. A large clinical trial comparing two different infusion rates yielded neutral results. Conditional average treatment effect (CATE) assessment may aid in tailoring therapy. Objectives: To estimate CATE in patients enrolled in the BaSICS trial and to assess the effects of receiving CATE model-recommended treatment in terms of hospital mortality. Methods: Post hoc analysis of the BaSICS trial assessing the effect of two infusion rates for the fluid challenge (fast, 999 ml/h, control group; vs. slow, 333 ml/h, intervention group) on hospital mortality. CATE was estimated as the difference in outcome for treatment arms in counterfactuals obtained from a Bayesian model trained in the first half of the trial adjusted for predictors hypothesized to interact with the intervention. The model recommended slow or fast infusion or made no recommendation in the second half. A threshold greater than 0.90 probability of benefit was considered. Results: A total of 10,465 patients were analyzed. The model was trained in 5,230 patients and tested in 5,235 patients. A recommendation could be made in the test set in 19% of patients (14% were recommended the control group and 5% the treatment group); for 81% of patients, no recommendation could be made. Slow infusion was more frequently recommended in cases of planned admissions in younger patients; fast infusion was recommended for older patients with sepsis. Slow infusion rate in the subgroup of patients in the test set in which slow infusion was recommended by the model was associated with an odds ratio of 0.58 (95% credible interval of 0.32-0.90; 0.99 posterior probability of benefit) for hospital mortality. Fast infusion in the subgroup in which the model recommended fast infusion was associated with an odds ratio of 0.72 (credible intervals from 0.54 to 0.91; probability of benefit >0.99). Conclusions: Estimation of CATEs from counterfactual probabilities in data from BaSICS provided additional information on trial data. Agreement between treatment recommendation and actual treatment was associated with lower hospital mortality. Clinical trial registered with clinicaltrials.gov (NCT02875873).

Cardiovascular Subphenotypes in ARDS: Diagnostic and Therapeutic Implications and Overlap with Other ARDS Subphenotypes

Acute respiratory distress syndrome (ARDS) is a highly heterogeneous clinical condition. Shock is a poor prognostic sign in ARDS, and heterogeneity in its pathophysiology may be a barrier to its effective treatment. Although right ventricular dysfunction is commonly implicated, there is no consensus definition for its diagnosis, and left ventricular function is neglected. There is a need to identify the homogenous subgroups within ARDS, that have a similar pathobiology, which can then be treated with targeted therapies. Haemodynamic clustering analyses in patients with ARDS have identified two subphenotypes of increasingly severe right ventricular injury, and a further subphenotype of hyperdynamic left ventricular function. In this review, we discuss how phenotyping the cardiovascular system in ARDS may align with haemodynamic pathophysiology, can aid in optimally defining right ventricular dysfunction and can identify tailored therapeutic targets for shock in ARDS. Additionally, clustering analyses of inflammatory, clinical and radiographic data describe other subphenotypes in ARDS. We detail the potential overlap between these and the cardiovascular phenotypes.

miR-125b-5p in adipose derived stem cells exosome alleviates pulmonary microvascular endothelial cells ferroptosis via Keap1/Nrf2/GPX4 in sepsis lung injury

BACKGROUND

Sepsis is a fatal disease with a high rate of morbidity and mortality, during which acute lung injury is the earliest and most serious complication. Injury of pulmonary microvascular endothelial cells (PMVECs) induced by excessive inflammation plays an important role in sepsis acute lung injury. This study is meant to explore the protective effect and mechanism of ADSCs exosomes on excessive inflammation PMVECs injury.

RESULTS

We successfully isolated ADSCs exosomes, the characteristic of which were confirmed. ADSCs exosomes reduced excessive inflammatory response induced ROS accumulation and cell injury in PMVECs. Besides, ADSCs exosomes inhibited excessive inflammatory response induced ferroptosis while upregulated expression of GPX4 in PMVECs. And further GPX4 inhibition experiments revealed that ADSCs exosomes alleviated inflammatory response induced ferroptosis via upregulating GPX4. Meanwhile, ADSCs exosomes could increase the expression and nucleus translocation of Nrf2, while decrease the expression of Keap1. miRNA analysis and further inhibition experiments verified that specific delivery of miR-125b-5p by ADSCs exosomes inhibited Keap1 and alleviated ferroptosis. In CLP induced sepsis model, ADSCs exosomes could relieve the lung tissue injury and reduced the death rate. Besides, ADSCs exosomes alleviated oxidative stress injury and ferroptosis of lung tissue, while remarkably increase expression of Nrf2 and GPX4.

CONCLUSION

Collectively, we illustrated a novel potentially therapeutic mechanism that miR-125b-5p in ADSCs exosomes could alleviate the inflammation induced PMVECs ferroptosis in sepsis induced acute lung injury via regulating Keap1/Nrf2/GPX4 expression, hence improve the acute lung injury in sepsis.

Identifying two distinct subphenotypes of patent ductus arteriosus in preterm infants using machine learning

To use unsupervised machine learning to identify potential subphenotypes of preterm infants with patent ductus arteriosus (PDA). The study was conducted retrospectively at a neonatal intensive care unit in Brazil. Patients with a gestational age < 28 weeks who had undergone at least one echocardiogram within the first two weeks of life and had PDA size > 1.5 or LA/AO ratio > 1.5 were included. Agglomerative hierarchical clustering on principal components was used to divide the data into different clusters based on common characteristics. Two distinct subphenotypes of preterm infants with hemodynamically significant PDA were identified: "inflamed," characterized by high leukocyte, neutrophil, and neutrophil-to-lymphocyte ratio, and "respiratory acidosis," characterized by low pH and high pCO2 levels.    Conclusions: This study suggests that there may be two distinct subphenotypes of preterm infants with hemodynamically significant PDA: "inflamed" and "respiratory acidosis." By dividing the population into different subgroups based on common characteristics, it is possible to get a more nuanced understanding of the effectiveness of PDA interventions. What is Known: • Treatment of PDA in preterm infants has been controversial. • Stratification of preterm infants with PDA into subgroups is important in order to determine the best treatment. What is New: • Unsupervised machine learning was used to identify two subphenotypes of preterm infants with hemodynamically significant PDA. • The 'inflamed' cluster was characterized by higher values of leukocyte, neutrophil, and neutrophil-to-lymphocyte ratio. The 'respiratory acidosis' cluster was characterized by lower pH values and higher pCO2 values.

Identification of clinical heterogeneity and construction of a novel subtype predictive model in patients with ankylosing spondylitis: An unsupervised machine learning study

BACKGROUND

Accurate classification of patients with ankylosing spondylitis (AS) is the premise of precision medicine so as to perform different medical interventions for different patient types. AS pathology is closely related to the changes in the immune microenvironment. In this study, we used unsupervised machine learning (UML) to classify patients with AS based on clinical characteristics. We then constructed a novel subtype predictive model for AS based on the clinical classification, after which we investigated the difference in the immune microenvironment to unravel the AS pathogenesis.

METHODS

Overall, 196 patients with AS were enrolled. UML was used to cluster AS patients by similar clinical characteristics. Functional ability, disease status, and grading of radiologic features were assessed to verify the accuracy and heterogeneity of UML clustering. Least Absolute Shrinkage and Selection Operator (LASSO) regression and Random Forest algorithm were used to screen and identify predictive factors for the novel subtype of AS. Logistic regression was also performed to construct a predictive model of this novel subtype. Datasets were downloaded from the Gene Expression Omnibus database to assess immune cell infiltration, and the results were validated using data of routine blood tests from 3671 AS patients and 5720 non-AS patients. The differential expression of Fat Mass and Obesity-Associated Protein (FTO), an m⁶A regulator, between AS patients and healthy control subjects was confirmed using immunohistochemistry.

RESULTS

UML clustering identified two clusters. The clinical characteristics of the two clusters were significantly heterogeneous. For the novel subtype of AS identified in UML clustering, a predictive model was built using three predictive factors, namely, C-reactive protein (CRP), absolute value of neutrophils (NEU), and absolute value of monocytes (MONO). The area under the curve of the predictive model was 0.983. Heterogeneity in the neutrophil and monocyte counts in AS was verified through immune cell infiltration analysis. Data from routine blood tests revealed that NEU and MONO were significantly higher in AS patients than in non-AS patients (p < 0.001). FTO expression was negatively correlated with both NEU and MONO. Immunohistochemistry analysis confirmed the downregulated expression of FTO.

CONCLUSIONS

UML provides an explicable and remarkable classification of a heterogeneous cohort of AS patients. A novel subtype of AS was identified in UML clustering. CRP, NEU, and MONO were the independent predictive factors for the novel subtype of AS. FTO expression was correlated with immune cell infiltration in AS patients.

Early Restrictive or Liberal Fluid Management for Sepsis-Induced Hypotension

BACKGROUND

Intravenous fluids and vasopressor agents are commonly used in early resuscitation of patients with sepsis, but comparative data for prioritizing their delivery are limited.

METHODS

In an unblinded superiority trial conducted at 60 U.S. centers, we randomly assigned patients to either a restrictive fluid strategy (prioritizing vasopressors and lower intravenous fluid volumes) or a liberal fluid strategy (prioritizing higher volumes of intravenous fluids before vasopressor use) for a 24-hour period. Randomization occurred within 4 hours after a patient met the criteria for sepsis-induced hypotension refractory to initial treatment with 1 to 3 liters of intravenous fluid. We hypothesized that all-cause mortality before discharge home by day 90 (primary outcome) would be lower with a restrictive fluid strategy than with a liberal fluid strategy. Safety was also assessed.

RESULTS

A total of 1563 patients were enrolled, with 782 assigned to the restrictive fluid group and 781 to the liberal fluid group. Resuscitation therapies that were administered during the 24-hour protocol period differed between the two groups; less intravenous fluid was administered in the restrictive fluid group than in the liberal fluid group (difference of medians, -2134 ml; 95% confidence interval [CI], -2318 to -1949), whereas the restrictive fluid group had earlier, more prevalent, and longer duration of vasopressor use. Death from any cause before discharge home by day 90 occurred in 109 patients (14.0%) in the restrictive fluid group and in 116 patients (14.9%) in the liberal fluid group (estimated difference, -0.9 percentage points; 95% CI, -4.4 to 2.6; P = 0.61); 5 patients in the restrictive fluid group and 4 patients in the liberal fluid group had their data censored (lost to follow-up). The number of reported serious adverse events was similar in the two groups.

CONCLUSIONS

Among patients with sepsis-induced hypotension, the restrictive fluid strategy that was used in this trial did not result in significantly lower (or higher) mortality before discharge home by day 90 than the liberal fluid strategy. (Funded by the National Heart, Lung, and Blood Institute; CLOVERS ClinicalTrials.gov number, NCT03434028.).

Integrating biology into clinical trial design

PURPOSE OF REVIEW

Critical care medicine revolves around syndromes, such as acute respiratory distress syndrome (ARDS), sepsis and acute kidney injury. Few interventions have shown to be effective in large clinical trials, likely because of between-patient heterogeneity. Translational evidence suggests that more homogeneous biological subgroups can be identified and that differential treatment effects exist. Integrating biological considerations into clinical trial design is therefore an important frontier of critical care research.

RECENT FINDINGS

The pathophysiology of critical care syndromes involves a multiplicity of processes, which emphasizes the difficulty of integrating biology into clinical trial design. Biological assessment can be integrated into clinical trials using predictive enrichment at trial inclusion, time-dependent variation to better understand treatment effects and biological markers as surrogate outcomes.

SUMMARY

Integrating our knowledge on biological heterogeneity into clinical trial design, which has revolutionized other medical fields, could serve as a solution to implement personalized treatment in critical care syndromes. Changing the trial design by using predictive enrichment, incorporation of the evaluation of time-dependent changes and biological markers as surrogate outcomes may improve the likelihood of detecting a beneficial effect from targeted therapeutic interventions and the opportunity to test multiple lines of treatment per patient.

A narrative review of progress in the application of artificial intelligence in acute respiratory distress syndrome: subtypes and predictive models

Background and Objective

Acute respiratory distress syndrome (ARDS) occurs in different populations, and it is very challenging to manage heterogeneous patient groups. Artificial intelligence (AI) aids in interpreting complex data of patients with ARDS and can be used to detect adverse events as it can automatically capture complex relationships. This review aimed to explore the application and progress of AI in ARDS (e.g., subgroup classification of patients with ARDS via unsupervised clustering and supervised predictive models for early detection) and identify the current ARDS-related problems that can be solved using AI.

Methods

This comprehensive and narrative review was performed to obtain information about the application of AI in ARDS and summarize its subtypes and predictive models.

Key Content and Findings

The current applications of AI and machine learning in ARDS include ARDS subgroup classification, diagnosis, and survival prediction. In this review, the current problems that should be addressed by AI in ARDS were identified, and our findings may serve as a useful reference for its translational use in the ARDS field.

Conclusions

Owing to the discovery of hyper- and hypoinflammatory subtypes, individualized treatment of ARDS is possible, and diagnosis and survival prediction are essential in disease management and planning. However, prospective studies should clarify the reliability and generalizability of the results using AI and machine learning and performing bedside testing in larger populations to establish a more stable and time-resilient model. Therefore, a consensus on conducting and reporting machine learning studies in medicine should be urgently established.

Biological Phenotyping in Sepsis and Acute Respiratory Distress Syndrome

Heterogeneity in sepsis and acute respiratory distress syndrome (ARDS) is increasingly being recognized as one of the principal barriers to finding efficacious targeted therapies. The advent of multiple high-throughput biological data ("omics"), coupled with the widespread access to increased computational power, has led to the emergence of phenotyping in critical care. Phenotyping aims to use a multitude of data to identify homogenous subgroups within an otherwise heterogenous population. Increasingly, phenotyping schemas are being applied to sepsis and ARDS to increase understanding of these clinical conditions and identify potential therapies. Here we present a selective review of the biological phenotyping schemas applied to sepsis and ARDS. Further, we outline some of the challenges involved in translating these conceptual findings to bedside clinical decision-making tools.

Machine learning model successfully identifies important clinical features for predicting outpatients with rotator cuff tears

PURPOSE

The aim of this study is to develop a machine learning model to identify important clinical features related to rotator cuff tears (RCTs) using explainable artificial intelligence (XAI) for efficiently predicting outpatients with RCTs.

METHODS

A retrospective review of a local clinical registry dataset was performed to include patients with shoulder pain and dysfunction who underwent questionnaires and physical examinations between 2019 and 2022. RCTs were diagnosed by shoulder arthroscopy. Six machine-learning algorithms (Stacking, Gradient Boosting Machine, Bagging, Random Forest, Extreme Gradient Boost (XGBoost), and Adaptive Boosting) were developed for the prediction. The performance of the models was assessed by the area under the receiver operating characteristic curve (AUC), Brier scores, and Decision curve. The interpretability of the predicted outcomes was evaluated using Shapley additive explanation (SHAP) values.

RESULTS

A total of 1684 patients who completed questionnaires and clinical tests were included, and 417 patients with RCTs underwent shoulder arthroscopy. In six machining learning algorithms for predicting RCTs, the accuracy, AUC values, and Brier scores were in the range of 0.81-0.86, 0.75-0.92, and 0.15-0.19, respectively. The XGBoost model showed superior performance with accuracy, AUC, and Brier scores of 0.85(95% confidence interval, 0.82-0.87), 0.92 (95% confidence interval,0.90-0.94), and 0.15 (95% confidence interval,0.14-0.16), respectively. The Shapley plot showed the impact of the clinical features on predicting RCTs. The most important variables were Jobe test, Bear hug test, and age for prediction, with mean SHAP values of 1.458, 0.950, and 0.790, respectively.

CONCLUSION

The machine learning model successfully identified important clinical variables for predicting patients with RCTs. In addition, the best algorithm was also integrated into a digital application to provide predictions in outpatient settings. This tool may assist patients in reducing their pain experience and providing prompt treatments.

LEVEL OF EVIDENCE

Level III.

Identification of spinal tuberculosis subphenotypes using routine clinical data: a study based on unsupervised machine learning

OBJECTIVE

The identification of spinal tuberculosis subphenotypes is an integral component of precision medicine. However, we lack proper study models to identify subphenotypes in patients with spinal tuberculosis. Here we identified possible subphenotypes of spinal tuberculosis and compared their clinical results.

METHODS

A total of 422 patients with spinal tuberculosis who received surgical treatment were enrolled. Clustering analysis was performed using the K-means clustering algorithm and the routinely available clinical data collected from patients within 24 h after admission. Finally, the differences in clinical characteristics, surgical efficacy, and postoperative complications among the subphenotypes were compared.

RESULTS

Two subphenotypes of spinal tuberculosis were identified. Laboratory examination results revealed that the levels of more than one inflammatory index in cluster 2 were higher than those in cluster 1. In terms of disease severity, Cluster 2 showed a higher Oswestry Disability Index (ODI), a higher visual analysis scale (VAS) score, and a lower Japanese Orthopedic Association (JOA) score. In addition, in terms of postoperative outcomes, cluster 2 patients were more prone to complications, especially wound infections, and had a longer hospital stay.

CONCLUSION

K-means clustering analysis based on conventional available clinical data can rapidly identify two subtypes of spinal tuberculosis with different clinical results. We believe this finding will help clinicians to rapidly and easily identify the subtypes of spinal tuberculosis at the bedside and become the cornerstone of individualized treatment strategies.