A Multimodal Video-Based AI Biomarker for Aortic Stenosis Development and Progression

Importance

Aortic stenosis (AS) is a major public health challenge with a growing therapeutic landscape, but current biomarkers do not inform personalized screening and follow-up. A video-based artificial intelligence (AI) biomarker (Digital AS Severity index [DASSi]) can detect severe AS using single-view long-axis echocardiography without Doppler characterization.

Objective

To deploy DASSi to patients with no AS or with mild or moderate AS at baseline to identify AS development and progression.

Design, Setting, and Participants

This is a cohort study that examined 2 cohorts of patients without severe AS undergoing echocardiography in the Yale New Haven Health System (YNHHS; 2015-2021) and Cedars-Sinai Medical Center (CSMC; 2018-2019). A novel computational pipeline for the cross-modal translation of DASSi into cardiac magnetic resonance (CMR) imaging was further developed in the UK Biobank. Analyses were performed between August 2023 and February 2024.

Exposure

DASSi (range, 0-1) derived from AI applied to echocardiography and CMR videos.

Main Outcomes and Measures

Annualized change in peak aortic valve velocity (AV-Vmax) and late (>6 months) aortic valve replacement (AVR).

Results

A total of 12 599 participants were included in the echocardiographic study (YNHHS: n = 8798; median [IQR] age, 71 [60-80] years; 4250 [48.3%] women; median [IQR] follow-up, 4.1 [2.4-5.4] years; and CSMC: n = 3801; median [IQR] age, 67 [54-78] years; 1685 [44.3%] women; median [IQR] follow-up, 3.4 [2.8-3.9] years). Higher baseline DASSi was associated with faster progression in AV-Vmax (per 0.1 DASSi increment: YNHHS, 0.033 m/s per year [95% CI, 0.028-0.038] among 5483 participants; CSMC, 0.082 m/s per year [95% CI, 0.053-0.111] among 1292 participants), with values of 0.2 or greater associated with a 4- to 5-fold higher AVR risk than values less than 0.2 (YNHHS: 715 events; adjusted hazard ratio [HR], 4.97 [95% CI, 2.71-5.82]; CSMC: 56 events; adjusted HR, 4.04 [95% CI, 0.92-17.70]), independent of age, sex, race, ethnicity, ejection fraction, and AV-Vmax. This was reproduced across 45 474 participants (median [IQR] age, 65 [59-71] years; 23 559 [51.8%] women; median [IQR] follow-up, 2.5 [1.6-3.9] years) undergoing CMR imaging in the UK Biobank (for participants with DASSi ≥0.2 vs those with DASSi <.02, adjusted HR, 11.38 [95% CI, 2.56-50.57]). Saliency maps and phenome-wide association studies supported associations with cardiac structure and function and traditional cardiovascular risk factors.

Conclusions and Relevance

In this cohort study of patients without severe AS undergoing echocardiography or CMR imaging, a new AI-based video biomarker was independently associated with AS development and progression, enabling opportunistic risk stratification across cardiovascular imaging modalities as well as potential application on handheld devices.

Utilizing large language models in breast cancer management: systematic review

PURPOSE

Despite advanced technologies in breast cancer management, challenges remain in efficiently interpreting vast clinical data for patient-specific insights. We reviewed the literature on how large language models (LLMs) such as ChatGPT might offer solutions in this field.

METHODS

We searched MEDLINE for relevant studies published before December 22, 2023. Keywords included: "large language models", "LLM", "GPT", "ChatGPT", "OpenAI", and "breast". The risk bias was evaluated using the QUADAS-2 tool.

RESULTS

Six studies evaluating either ChatGPT-3.5 or GPT-4, met our inclusion criteria. They explored clinical notes analysis, guideline-based question-answering, and patient management recommendations. Accuracy varied between studies, ranging from 50 to 98%. Higher accuracy was seen in structured tasks like information retrieval. Half of the studies used real patient data, adding practical clinical value. Challenges included inconsistent accuracy, dependency on the way questions are posed (prompt-dependency), and in some cases, missing critical clinical information.

CONCLUSION

LLMs hold potential in breast cancer care, especially in textual information extraction and guideline-driven clinical question-answering. Yet, their inconsistent accuracy underscores the need for careful validation of these models, and the importance of ongoing supervision.

Integrated multiomics implicates dysregulation of ECM and cell adhesion pathways as drivers of severe COVID-associated kidney injury

COVID-19 has been a significant public health concern for the last four years; however, little is known about the mechanisms that lead to severe COVID-associated kidney injury. In this multicenter study, we combined quantitative deep urinary proteomics and machine learning to predict severe acute outcomes in hospitalized COVID-19 patients. Using a 10-fold cross-validated random forest algorithm, we identified a set of urinary proteins that demonstrated predictive power for both discovery and validation set with 87% and 79% accuracy, respectively. These predictive urinary biomarkers were recapitulated in non-COVID acute kidney injury revealing overlapping injury mechanisms. We further combined orthogonal multiomics datasets to understand the mechanisms that drive severe COVID-associated kidney injury. Functional overlap and network analysis of urinary proteomics, plasma proteomics and urine sediment single-cell RNA sequencing showed that extracellular matrix and autophagy-associated pathways were uniquely impacted in severe COVID-19. Differentially abundant proteins associated with these pathways exhibited high expression in cells in the juxtamedullary nephron, endothelial cells, and podocytes, indicating that these kidney cell types could be potential targets. Further, single-cell transcriptomic analysis of kidney organoids infected with SARS-CoV-2 revealed dysregulation of extracellular matrix organization in multiple nephron segments, recapitulating the clinically observed fibrotic response across multiomics datasets. Ligand-receptor interaction analysis of the podocyte and tubule organoid clusters showed significant reduction and loss of interaction between integrins and basement membrane receptors in the infected kidney organoids. Collectively, these data suggest that extracellular matrix degradation and adhesion-associated mechanisms could be a main driver of COVID-associated kidney injury and severe outcomes.

Pediatric ECG-Based Deep Learning to Predict Left Ventricular Dysfunction and Remodeling

BACKGROUND

Artificial intelligence-enhanced ECG analysis shows promise to detect ventricular dysfunction and remodeling in adult populations. However, its application to pediatric populations remains underexplored.

METHODS

A convolutional neural network was trained on paired ECG-echocardiograms (≤2 days apart) from patients ≤18 years of age without major congenital heart disease to detect human expert-classified greater than mild left ventricular (LV) dysfunction, hypertrophy, and dilation (individually and as a composite outcome). Model performance was evaluated on single ECG-echocardiogram pairs per patient at Boston Children's Hospital and externally at Mount Sinai Hospital using area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC).

RESULTS

The training cohort comprised 92 377 ECG-echocardiogram pairs (46 261 patients; median age, 8.2 years). Test groups included internal testing (12 631 patients; median age, 8.8 years; 4.6% composite outcomes), emergency department (2830 patients; median age, 7.7 years; 10.0% composite outcomes), and external validation (5088 patients; median age, 4.3 years; 6.1% composite outcomes) cohorts. Model performance was similar on internal test and emergency department cohorts, with model predictions of LV hypertrophy outperforming the pediatric cardiologist expert benchmark. Adding age and sex to the model added no benefit to model performance. When using quantitative outcome cutoffs, model performance was similar between internal testing (composite outcome: AUROC, 0.88, AUPRC, 0.43; LV dysfunction: AUROC, 0.92, AUPRC, 0.23; LV hypertrophy: AUROC, 0.88, AUPRC, 0.28; LV dilation: AUROC, 0.91, AUPRC, 0.47) and external validation (composite outcome: AUROC, 0.86, AUPRC, 0.39; LV dysfunction: AUROC, 0.94, AUPRC, 0.32; LV hypertrophy: AUROC, 0.84, AUPRC, 0.25; LV dilation: AUROC, 0.87, AUPRC, 0.33), with composite outcome negative predictive values of 99.0% and 99.2%, respectively. Saliency mapping highlighted ECG components that influenced model predictions (precordial QRS complexes for all outcomes; T waves for LV dysfunction). High-risk ECG features include lateral T-wave inversion (LV dysfunction), deep S waves in V1 and V2 and tall R waves in V6 (LV hypertrophy), and tall R waves in V4 through V6 (LV dilation).

CONCLUSIONS

This externally validated algorithm shows promise to inexpensively screen for LV dysfunction and remodeling in children, which may facilitate improved access to care by democratizing the expertise of pediatric cardiologists.

Artificial intelligence-guided detection of under-recognized cardiomyopathies on point-of-care cardiac ultrasound

Background

Point-of-care ultrasonography (POCUS) enables access to cardiac imaging directly at the bedside but is limited by brief acquisition, variation in acquisition quality, and lack of advanced protocols.

Objective

To develop and validate deep learning models for detecting underdiagnosed cardiomyopathies on cardiac POCUS, leveraging a novel acquisition quality-adapted modeling strategy.

Methods

To develop the models, we identified transthoracic echocardiograms (TTEs) of patients across five hospitals in a large U.S. health system with transthyretin amyloid cardiomyopathy (ATTR-CM, confirmed by Tc99m-pyrophosphate imaging), hypertrophic cardiomyopathy (HCM, confirmed by cardiac magnetic resonance), and controls enriched for the presence of severe AS. In a sample of 290,245 TTE videos, we used novel augmentation approaches and a customized loss function to weigh image and view quality to train a multi-label, view agnostic video-based convolutional neural network (CNN) to discriminate the presence of ATTR-CM, HCM, and/or AS. Models were tested across 3,758 real-world POCUS videos from 1,879 studies in 1,330 independent emergency department (ED) patients from 2011 through 2023.

Results

Our multi-label, view-agnostic classifier demonstrated state-of-the-art performance in discriminating ATTR-CM (AUROC 0.98 [95%CI: 0.96-0.99]) and HCM (AUROC 0.95 [95% CI: 0.94-0.96]) on standard TTE studies. Automated metrics of anatomical view correctness confirmed significantly lower quality in POCUS vs TTE videos (median view classifier confidence of 0.63 [IQR: 0.44-0.88] vs 0.93 [IQR: 0.69-1.00], p<0.001). When deployed to POCUS videos, our algorithm effectively discriminated ATTR-CM and HCM with AUROC of up to 0.94 (parasternal long-axis (PLAX)), and 0.85 (apical 4 chamber), corresponding to positive diagnostic odds ratios of 46.7 and 25.5, respectively. In total, 18/35 (51.4%) of ATTR-CM and 32/57 (41.1%) of HCM patients in the POCUS cohort had an AI-positive screen in the year before their eventual confirmatory imaging.

Conclusions

We define and validate an AI framework that enables scalable, opportunistic screening of under-diagnosed cardiomyopathies using POCUS.

A Multimodality Video-Based AI Biomarker For Aortic Stenosis Development And Progression

Importance

Aortic stenosis (AS) is a major public health challenge with a growing therapeutic landscape, but current biomarkers do not inform personalized screening and follow-up.

Objective

A video-based artificial intelligence (AI) biomarker (Digital AS Severity index [DASSi]) can detect severe AS using single-view long-axis echocardiography without Doppler. Here, we deploy DASSi to patients with no or mild/moderate AS at baseline to identify AS development and progression.

Design Setting and Participants

We defined two cohorts of patients without severe AS undergoing echocardiography in the Yale-New Haven Health System (YNHHS) (2015-2021, 4.1[IQR:2.4-5.4] follow-up years) and Cedars-Sinai Medical Center (CSMC) (2018-2019, 3.4[IQR:2.8-3.9] follow-up years). We further developed a novel computational pipeline for the cross-modality translation of DASSi into cardiac magnetic resonance (CMR) imaging in the UK Biobank (2.5[IQR:1.6-3.9] follow-up years). Analyses were performed between August 2023-February 2024.

Exposure

DASSi (range: 0-1) derived from AI applied to echocardiography and CMR videos.

Main Outcomes and Measures

Annualized change in peak aortic valve velocity (AV-Vmax) and late (>6 months) aortic valve replacement (AVR).

Results

A total of 12,599 participants were included in the echocardiographic study (YNHHS: n=8,798, median age of 71 [IQR (interquartile range):60-80] years, 4250 [48.3%] women, and CSMC: n=3,801, 67 [IQR:54-78] years, 1685 [44.3%] women). Higher baseline DASSi was associated with faster progression in AV-Vmax (per 0.1 DASSi increments: YNHHS: +0.033 m/s/year [95%CI:0.028-0.038], n=5,483, and CSMC: +0.082 m/s/year [0.053-0.111], n=1,292), with levels ≥ vs <0.2 linked to a 4-to-5-fold higher AVR risk (715 events in YNHHS; adj.HR 4.97 [95%CI: 2.71-5.82], 56 events in CSMC: 4.04 [0.92-17.7]), independent of age, sex, ethnicity/race, ejection fraction and AV-Vmax. This was reproduced across 45,474 participants (median age 65 [IQR:59-71] years, 23,559 [51.8%] women) undergoing CMR in the UK Biobank (adj.HR 11.4 [95%CI:2.56-50.60] for DASSi ≥vs<0.2). Saliency maps and phenome-wide association studies supported links with traditional cardiovascular risk factors and diastolic dysfunction.

Conclusions and Relevance

In this cohort study of patients without severe AS undergoing echocardiography or CMR imaging, a new AI-based video biomarker is independently associated with AS development and progression, enabling opportunistic risk stratification across cardiovascular imaging modalities as well as potential application on handheld devices.

Automated Diagnostic Reports from Images of Electrocardiograms at the Point-of-Care

Timely and accurate assessment of electrocardiograms (ECGs) is crucial for diagnosing, triaging, and clinically managing patients. Current workflows rely on a computerized ECG interpretation using rule-based tools built into the ECG signal acquisition systems with limited accuracy and flexibility. In low-resource settings, specialists must review every single ECG for such decisions, as these computerized interpretations are not available. Additionally, high-quality interpretations are even more essential in such low-resource settings as there is a higher burden of accuracy for automated reads when access to experts is limited. Artificial Intelligence (AI)-based systems have the prospect of greater accuracy yet are frequently limited to a narrow range of conditions and do not replicate the full diagnostic range. Moreover, these models often require raw signal data, which are unavailable to physicians and necessitate costly technical integrations that are currently limited. To overcome these challenges, we developed and validated a format-independent vision encoder-decoder model - ECG-GPT - that can generate free-text, expert-level diagnosis statements directly from ECG images. The model shows robust performance, validated on 2.6 million ECGs across 6 geographically distinct health settings: (1) 2 large and diverse US health systems- Yale-New Haven and Mount Sinai Health Systems, (2) a consecutive ECG dataset from a central ECG repository from Minas Gerais, Brazil, (3) the prospective cohort study, UK Biobank, (4) a Germany-based, publicly available repository, PTB-XL, and (5) a community hospital in Missouri. The model demonstrated consistently high performance (AUROC≥0.81) across a wide range of rhythm and conduction disorders. This can be easily accessed via a web-based application capable of receiving ECG images and represents a scalable and accessible strategy for generating accurate, expert-level reports from images of ECGs, enabling accurate triage of patients globally, especially in low-resource settings.

Comparison of predicting cardiovascular disease hospitalization using individual, ZIP code-derived, and machine learning model-predicted educational attainment in New York City

BACKGROUND

Area-level social determinants of health (SDOH) based on patients' ZIP codes or census tracts have been commonly used in research instead of individual SDOHs. To our knowledge, whether machine learning (ML) could be used to derive individual SDOH measures, specifically individual educational attainment, is unknown.

METHODS

This is a retrospective study using data from the Mount Sinai BioMe Biobank. We included participants that completed a validated questionnaire on educational attainment and had home addresses in New York City. ZIP code-level education was derived from the American Community Survey matched for the participant's gender and race/ethnicity. We tested several algorithms to predict individual educational attainment from routinely collected clinical and demographic data. To evaluate how using different measures of educational attainment will impact model performance, we developed three distinct models for predicting cardiovascular (CVD) hospitalization. Educational attainment was imputed into models as either survey-derived, ZIP code-derived, or ML-predicted educational attainment.

RESULTS

A total of 20,805 participants met inclusion criteria. Concordance between survey and ZIP code-derived education was 47%, while the concordance between survey and ML model-predicted education was 67%. A total of 13,715 patients from the cohort were included into our CVD hospitalization prediction models, of which 1,538 (11.2%) had a history of CVD hospitalization. The AUROC of the model predicting CVD hospitalization using survey-derived education was significantly higher than the model using ZIP code-level education (0.77 versus 0.72; p < 0.001) and the model using ML model-predicted education (0.77 versus 0.75; p < 0.001). The AUROC for the model using ML model-predicted education was also significantly higher than that using ZIP code-level education (p = 0.003).

CONCLUSION

The concordance of survey and ZIP code-level educational attainment in NYC was low. As expected, the model utilizing survey-derived education achieved the highest performance. The model incorporating our ML model-predicted education outperformed the model relying on ZIP code-derived education. Implementing ML techniques can improve the accuracy of SDOH data and consequently increase the predictive performance of outcome models.

A novel method leveraging time series data to improve subphenotyping and application in critically ill patients with COVID-19

Computational subphenotyping, a data-driven approach to understanding disease subtypes, is a prominent topic in medical research. Numerous ongoing studies are dedicated to developing advanced computational subphenotyping methods for cross-sectional data. However, the potential of time-series data has been underexplored until now. Here, we propose a Multivariate Levenshtein Distance (MLD) that can account for address correlation in multiple discrete features over time-series data. Our algorithm has two distinct components: it integrates an optimal threshold score to enhance the sensitivity in discriminating between pairs of instances, and the MLD itself. We have applied the proposed distance metrics on the k-means clustering algorithm to derive temporal subphenotypes from time-series data of biomarkers and treatment administrations from 1039 critically ill patients with COVID-19 and compare its effectiveness to standard methods. In conclusion, the Multivariate Levenshtein Distance metric is a novel method to quantify the distance from multiple discrete features over time-series data and demonstrates superior clustering performance among competing time-series distance metrics.

Derivation, External Validation and Clinical Implications of a deep learning approach for intracranial pressure estimation using non-cranial waveform measurements

Importance

Increased intracranial pressure (ICP) is associated with adverse neurological outcomes, but needs invasive monitoring.

Objective

Development and validation of an AI approach for detecting increased ICP (aICP) using only non-invasive extracranial physiological waveform data.

Design

Retrospective diagnostic study of AI-assisted detection of increased ICP. We developed an AI model using exclusively extracranial waveforms, externally validated it and assessed associations with clinical outcomes.

Setting

MIMIC-III Waveform Database (2000-2013), a database derived from patients admitted to an ICU in an academic Boston hospital, was used for development of the aICP model, and to report association with neurologic outcomes. Data from Mount Sinai Hospital (2020-2022) in New York City was used for external validation.

Participants

Patients were included if they were older than 18 years, and were monitored with electrocardiograms, arterial blood pressure, respiratory impedance plethysmography and pulse oximetry. Patients who additionally had intracranial pressure monitoring were used for development (N=157) and external validation (N=56). Patients without intracranial monitors were used for association with outcomes (N=1694).

Exposures

Extracranial waveforms including electrocardiogram, arterial blood pressure, plethysmography and SpO2.

Main Outcomes and Measures

Intracranial pressure > 15 mmHg. Measures were Area under receiver operating characteristic curves (AUROCs), sensitivity, specificity, and accuracy at threshold of 0.5. We calculated odds ratios and p-values for phenotype association.

Results

The AUROC was 0.91 (95% CI, 0.90-0.91) on testing and 0.80 (95% CI, 0.80-0.80) on external validation. aICP had accuracy, sensitivity, and specificity of 73.8% (95% CI, 72.0%-75.6%), 99.5% (95% CI 99.3%-99.6%), and 76.9% (95% CI, 74.0-79.8%) on external validation. A ten-percentile increment was associated with stroke (OR=2.12; 95% CI, 1.27-3.13), brain malignancy (OR=1.68; 95% CI, 1.09-2.60), subdural hemorrhage (OR=1.66; 95% CI, 1.07-2.57), intracerebral hemorrhage (OR=1.18; 95% CI, 1.07-1.32), and procedures like percutaneous brain biopsy (OR=1.58; 95% CI, 1.15-2.18) and craniotomy (OR = 1.43; 95% CI, 1.12-1.84; P < 0.05 for all).

Conclusions and Relevance

aICP provides accurate, non-invasive estimation of increased ICP, and is associated with neurological outcomes and neurosurgical procedures in patients without intracranial monitoring.

Quantitative Prediction of Right Ventricular Size and Function From the ECG

BACKGROUND

Right ventricular ejection fraction (RVEF) and end-diastolic volume (RVEDV) are not readily assessed through traditional modalities. Deep learning-enabled ECG analysis for estimation of right ventricular (RV) size or function is unexplored.

METHODS AND RESULTS

We trained a deep learning-ECG model to predict RV dilation (RVEDV >120 mL/m2), RV dysfunction (RVEF ≤40%), and numerical RVEDV and RVEF from a 12-lead ECG paired with reference-standard cardiac magnetic resonance imaging volumetric measurements in UK Biobank (UKBB; n=42 938). We fine-tuned in a multicenter health system (MSHoriginal [Mount Sinai Hospital]; n=3019) with prospective validation over 4 months (MSHvalidation; n=115). We evaluated performance with area under the receiver operating characteristic curve for categorical and mean absolute error for continuous measures overall and in key subgroups. We assessed the association of RVEF prediction with transplant-free survival with Cox proportional hazards models. The prevalence of RV dysfunction for UKBB/MSHoriginal/MSHvalidation cohorts was 1.0%/18.0%/15.7%, respectively. RV dysfunction model area under the receiver operating characteristic curve for UKBB/MSHoriginal/MSHvalidation cohorts was 0.86/0.81/0.77, respectively. The prevalence of RV dilation for UKBB/MSHoriginal/MSHvalidation cohorts was 1.6%/10.6%/4.3%. RV dilation model area under the receiver operating characteristic curve for UKBB/MSHoriginal/MSHvalidation cohorts was 0.91/0.81/0.92, respectively. MSHoriginal mean absolute error was RVEF=7.8% and RVEDV=17.6 mL/m2. The performance of the RVEF model was similar in key subgroups including with and without left ventricular dysfunction. Over a median follow-up of 2.3 years, predicted RVEF was associated with adjusted transplant-free survival (hazard ratio, 1.40 for each 10% decrease; P=0.031).

CONCLUSIONS

Deep learning-ECG analysis can identify significant cardiac magnetic resonance imaging RV dysfunction and dilation with good performance. Predicted RVEF is associated with clinical outcome.

A Real-World Precision Medicine Program Including the KidneyIntelX Test Effectively Changes Management Decisions and Outcomes for Patients With Early-Stage Diabetic Kidney Disease

INTRODUCTION/OBJECTIVE

The KidneyIntelX is a multiplex, bioprognostic, immunoassay consisting of 3 plasma biomarkers and clinical variables that uses machine learning to predict a patient's risk for a progressive decline in kidney function over 5 years. We report the 1-year pre- and post-test clinical impact on care management, eGFR slope, and A1C along with engagement of population health clinical pharmacists and patient coordinators to promote a program of sustainable kidney, metabolic, and cardiac health.

METHODS

The KidneyIntelX in vitro prognostic test was previously validated for patients with type 2 diabetes and diabetic kidney disease (DKD) to predict kidney function decline within 5 years was introduced into the RWE study (NCT04802395) across the Health System as part of a population health chronic disease management program from [November 2020 to April 2023]. Pre- and post-test patients with a minimum of 12 months of follow-up post KidneyIntelX were assessed across all aspects of the program.

RESULTS

A total of 5348 patients with DKD had a KidneyIntelX assay. The median age was 68 years old, 52% were female, 27% self-identified as Black, and 89% had hypertension. The median baseline eGFR was 62 ml/min/1.73 m2, urine albumin-creatinine ratio was 54 mg/g, and A1C was 7.3%. The KidneyIntelX risk level was low in 49%, intermediate in 40%, and high in 11% of cases. New prescriptions for SGLT2i, GLP-1 RA, or referral to a specialist were noted in 19%, 33%, and 43% among low-, intermediate-, and high-risk patients, respectively. The median A1C decreased from 8.2% pre-test to 7.5% post-test in the high-risk group (P < .001). UACR levels in the intermediate-risk patients with albuminuria were reduced by 20%, and in a subgroup treated with new scripts for SGLT2i, UACR levels were lowered by approximately 50%. The median eGFR slope improved from -7.08 ml/min/1.73 m2/year to -4.27 ml/min/1.73 m2/year in high-risk patients (P = .0003), -2.65 to -1.04 in intermediate risk, and -3.26 ml/min/1.73 m2/year to +0.45 ml/min/1.73 m2/year in patients with low-risk (P < .001).

CONCLUSIONS

Deployment and risk stratification by KidneyIntelX was associated with an escalation in action taken to optimize cardio-kidney-metabolic health including medications and specialist referrals. Glycemic control and kidney function trajectories improved post-KidneyIntelX testing, with the greatest improvements observed in those scored as high-risk.

Evaluating the role of ChatGPT in gastroenterology: a comprehensive systematic review of applications, benefits, and limitations

Background

The integration of artificial intelligence (AI) into healthcare has opened new avenues for enhancing patient care and clinical research. In gastroenterology, the potential of AI tools, specifically large language models like ChatGPT, is being explored to understand their utility and effectiveness.

Objectives

The primary goal of this systematic review is to assess the various applications, ascertain the benefits, and identify the limitations of utilizing ChatGPT within the realm of gastroenterology.

Design

Through a systematic approach, this review aggregates findings from multiple studies to evaluate the impact of ChatGPT on the field.

Data sources and methods

The review was based on a detailed literature search of the PubMed database, targeting research that delves into the use of ChatGPT for gastroenterological purposes. It incorporated six selected studies, which were meticulously evaluated for quality using the Joanna Briggs Institute critical appraisal instruments. The data were then synthesized narratively to encapsulate the roles, advantages, and drawbacks of ChatGPT in gastroenterology.

Results

The investigation unearthed various roles of ChatGPT, including its use in patient education, diagnostic self-assessment, disease management, and the formulation of research queries. Notable benefits were its capability to provide pertinent recommendations, enhance communication between patients and physicians, and prompt valuable research inquiries. Nonetheless, it encountered obstacles in decoding intricate medical questions, yielded inconsistent responses at times, and exhibited limitations in generating novel content. The review also considered ethical implications.

Conclusion

ChatGPT has demonstrated significant potential in the field of gastroenterology, especially in facilitating patient-physician interactions and managing diseases. Despite these advancements, the review underscores the necessity for ongoing refinement, customization, and ethical regulation of AI tools. These findings serve to enrich the dialog concerning AI's role in healthcare, with a specific focus on ChatGPT's application in gastroenterology.

Machine learning models to predict end-stage kidney disease in chronic kidney disease stage 4

INTRODUCTION

End-stage kidney disease (ESKD) is associated with increased morbidity and mortality. Identifying patients with stage 4 CKD (CKD4) at risk of rapid progression to ESKD remains challenging. Accurate prediction of CKD4 progression can improve patient outcomes by improving advanced care planning and optimizing healthcare resource allocation.

METHODS

We obtained electronic health record data from patients with CKD4 in a large health system between January 1, 2006, and December 31, 2016. We developed and validated four models, including Least Absolute Shrinkage and Selection Operator (LASSO) regression, random forest, eXtreme Gradient Boosting (XGBoost), and artificial neural network (ANN), to predict ESKD at 3 years. We utilized area under the receiver operating characteristic curve (AUROC) to evaluate model performances and utilized Shapley additive explanation (SHAP) values and plots to define feature dependence of the best performance model.

RESULTS

We included 3,160 patients with CKD4. ESKD was observed in 538 patients (21%). All approaches had similar AUROCs; ANN yielded the highest AUROC (0.77; 95%CI 0.75 to 0.79) and LASSO regression (0.77; 95%CI 0.75 to 0.79), followed by random forest (0.76; 95% CI 0.74 to 0.79), and XGBoost (0.76; 95% CI 0.74 to 0.78).

CONCLUSIONS

We developed and validated several models for near-term prediction of kidney failure in CKD4. ANN, random forest, and XGBoost demonstrated similar predictive performances. Using this suite of models, interventions can be customized based on risk, and population health and resources appropriately allocated.

Joint Modeling of Social Determinants and Clinical Factors to Define Subphenotypes in Out-of-Hospital Cardiac Arrest Survival: Cluster Analysis

Background

Machine learning clustering offers an unbiased approach to better understand the interactions of complex social and clinical variables via integrative subphenotypes, an approach not studied in out-of-hospital cardiac arrest (OHCA).

Objective

We conducted a cluster analysis for a cohort of OHCA survivors to examine the association of clinical and social factors for mortality at 1 year.

Methods

We used a retrospective observational OHCA cohort identified from Medicare claims data, including area-level social determinants of health (SDOH) features and hospital-level data sets. We applied k-means clustering algorithms to identify subphenotypes of beneficiaries who had survived an OHCA and examined associations of outcomes by subphenotype.

Results

We identified 27,028 unique beneficiaries who survived to discharge after OHCA. We derived 4 distinct subphenotypes. Subphenotype 1 included a distribution of more urban, female, and Black beneficiaries with the least robust area-level SDOH measures and the highest 1-year mortality (2375/4417, 53.8%). Subphenotype 2 was characterized by a greater distribution of male, White beneficiaries and had the strongest zip code-level SDOH measures, with 1-year mortality at 49.9% (4577/9165). Subphenotype 3 had the highest rates of cardiac catheterization at 34.7% (1342/3866) and the greatest distribution with a driving distance to the index OHCA hospital from their primary residence >16.1 km at 85.4% (8179/9580); more were also discharged to a skilled nursing facility after index hospitalization. Subphenotype 4 had moderate median household income at US $51,659.50 (IQR US $41,295 to $67,081) and moderate to high median unemployment at 5.5% (IQR 4.2%-7.1%), with the lowest 1-year mortality (1207/3866, 31.2%). Joint modeling of these features demonstrated an increased hazard of death for subphenotypes 1 to 3 but not for subphenotype 4 when compared to reference.

Conclusions

We identified 4 distinct subphenotypes with differences in outcomes by clinical and area-level SDOH features for OHCA. Further work is needed to determine if individual or other SDOH domains are specifically tied to long-term survival after OHCA.

Deep Learning and Gastric Cancer: Systematic Review of AI-Assisted Endoscopy

BACKGROUND

Gastric cancer (GC), a significant health burden worldwide, is typically diagnosed in the advanced stages due to its non-specific symptoms and complex morphological features. Deep learning (DL) has shown potential for improving and standardizing early GC detection. This systematic review aims to evaluate the current status of DL in pre-malignant, early-stage, and gastric neoplasia analysis.

METHODS

A comprehensive literature search was conducted in PubMed/MEDLINE for original studies implementing DL algorithms for gastric neoplasia detection using endoscopic images. We adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The focus was on studies providing quantitative diagnostic performance measures and those comparing AI performance with human endoscopists.

RESULTS

Our review encompasses 42 studies that utilize a variety of DL techniques. The findings demonstrate the utility of DL in GC classification, detection, tumor invasion depth assessment, cancer margin delineation, lesion segmentation, and detection of early-stage and pre-malignant lesions. Notably, DL models frequently matched or outperformed human endoscopists in diagnostic accuracy. However, heterogeneity in DL algorithms, imaging techniques, and study designs precluded a definitive conclusion about the best algorithmic approach.

CONCLUSIONS

The promise of artificial intelligence in improving and standardizing gastric neoplasia detection, diagnosis, and segmentation is significant. This review is limited by predominantly single-center studies and undisclosed datasets used in AI training, impacting generalizability and demographic representation. Further, retrospective algorithm training may not reflect actual clinical performance, and a lack of model details hinders replication efforts. More research is needed to substantiate these findings, including larger-scale multi-center studies, prospective clinical trials, and comprehensive technical reporting of DL algorithms and datasets, particularly regarding the heterogeneity in DL algorithms and study designs.

Derivation and independent validation of kidneyintelX.dkd: A prognostic test for the assessment of diabetic kidney disease progression

AIMS

To develop and validate an updated version of KidneyIntelX (kidneyintelX.dkd) to stratify patients for risk of progression of diabetic kidney disease (DKD) stages 1 to 3, to simplify the test for clinical adoption and support an application to the US Food and Drug Administration regulatory pathway.

METHODS

We used plasma biomarkers and clinical data from the Penn Medicine Biobank (PMBB) for training, and independent cohorts (BioMe and CANVAS) for validation. The primary outcome was progressive decline in kidney function (PDKF), defined by a ≥40% sustained decline in estimated glomerular filtration rate or end-stage kidney disease within 5 years of follow-up.

RESULTS

In 573 PMBB participants with DKD, 15.4% experienced PDKF over a median of 3.7 years. We trained a random forest model using biomarkers and clinical variables. Among 657 BioMe participants and 1197 CANVAS participants, 11.7% and 7.5%, respectively, experienced PDKF. Based on training cut-offs, 57%, 35% and 8% of BioMe participants, and 56%, 38% and 6% of CANVAS participants were classified as having low-, moderate- and high-risk levels, respectively. The cumulative incidence at these risk levels was 5.9%, 21.2% and 66.9% in BioMe and 6.7%, 13.1% and 59.6% in CANVAS. After clinical risk factor adjustment, the adjusted hazard ratios were 7.7 (95% confidence interval [CI] 3.0-19.6) and 3.7 (95% CI 2.0-6.8) in BioMe, and 5.4 (95% CI 2.5-11.9) and 2.3 (95% CI 1.4-3.9) in CANVAS, for high- versus low-risk and moderate- versus low-risk levels, respectively.

CONCLUSIONS

Using two independent cohorts and a clinical trial population, we validated an updated KidneyIntelX test (named kidneyintelX.dkd), which significantly enhanced risk stratification in patients with DKD for PDKF, independently from known risk factors for progression.

Digital health and acute kidney injury: consensus report of the 27th Acute Disease Quality Initiative workgroup

Acute kidney injury (AKI), which is a common complication of acute illnesses, affects the health of individuals in community, acute care and post-acute care settings. Although the recognition, prevention and management of AKI has advanced over the past decades, its incidence and related morbidity, mortality and health care burden remain overwhelming. The rapid growth of digital technologies has provided a new platform to improve patient care, and reports show demonstrable benefits in care processes and, in some instances, in patient outcomes. However, despite great progress, the potential benefits of using digital technology to manage AKI has not yet been fully explored or implemented in clinical practice. Digital health studies in AKI have shown variable evidence of benefits, and the digital divide means that access to digital technologies is not equitable. Upstream research and development costs, limited stakeholder participation and acceptance, and poor scalability of digital health solutions have hindered their widespread implementation and use. Here, we provide recommendations from the Acute Disease Quality Initiative consensus meeting, which involved experts in adult and paediatric nephrology, critical care, pharmacy and data science, at which the use of digital health for risk prediction, prevention, identification and management of AKI and its consequences was discussed.

Development of the ehive Digital Health App: Protocol for a Centralized Research Platform

BACKGROUND

The increasing use of smartphones, wearables, and connected devices has enabled the increasing application of digital technologies for research. Remote digital study platforms comprise a patient-interfacing digital application that enables multimodal data collection from a mobile app and connected sources. They offer an opportunity to recruit at scale, acquire data longitudinally at a high frequency, and engage study participants at any time of the day in any place. Few published descriptions of centralized digital research platforms provide a framework for their development.

OBJECTIVE

This study aims to serve as a road map for those seeking to develop a centralized digital research platform. We describe the technical and functional aspects of the ehive app, the centralized digital research platform of the Hasso Plattner Institute for Digital Health at Mount Sinai Hospital, New York, New York. We then provide information about ongoing studies hosted on ehive, including usership statistics and data infrastructure. Finally, we discuss our experience with ehive in the broader context of the current landscape of digital health research platforms.

METHODS

The ehive app is a multifaceted and patient-facing central digital research platform that permits the collection of e-consent for digital health studies. An overview of its development, its e-consent process, and the tools it uses for participant recruitment and retention are provided. Data integration with the platform and the infrastructure supporting its operations are discussed; furthermore, a description of its participant- and researcher-facing dashboard interfaces and the e-consent architecture is provided.

RESULTS

The ehive platform was launched in 2020 and has successfully hosted 8 studies, namely 6 observational studies and 2 clinical trials. Approximately 1484 participants downloaded the app across 36 states in the United States. The use of recruitment methods such as bulk messaging through the EPIC electronic health records and standard email portals enables broad recruitment. Light-touch engagement methods, used in an automated fashion through the platform, maintain high degrees of engagement and retention. The ehive platform demonstrates the successful deployment of a central digital research platform that can be modified across study designs.

CONCLUSIONS

Centralized digital research platforms such as ehive provide a novel tool that allows investigators to expand their research beyond their institution, engage in large-scale longitudinal studies, and combine multimodal data streams. The ehive platform serves as a model for groups seeking to develop similar digital health research programs.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

DERR1-10.2196/49204.

Per- and polyfluoroalkyl substances (PFAS) exposure and thyroid cancer risk

BACKGROUND

Although per- and polyfluoroalkyl substances (PFAS) exposure is a potential contributor to the increasing thyroid cancer trend, limited studies have investigated the association between PFAS exposure and thyroid cancer in human populations. We therefore investigated associations between plasma PFAS levels and thyroid cancer diagnosis using a nested case-control study of patients with thyroid cancer with plasma samples collected at/before cancer diagnosis.

METHODS

88 patients with thyroid cancer using diagnosis codes and 88 healthy (non-cancer) controls pair-matched on sex, age (±5 years), race/ethnicity, body mass index, smoking status, and year of sample collection were identified in the BioMe population (a medical record-linked biobank at the Icahn School of Medicine at Mount Sinai in New York); 74 patients had papillary thyroid cancer. Eight plasma PFAS were measured using untargeted analysis with liquid chromatography-high resolution mass spectrometry and suspect screening. Associations between individual PFAS levels and thyroid cancer were evaluated using unconditional logistic regression models to estimate adjusted odds ratios (ORadj) and 95% confidence intervals (CI).

FINDINGS

There was a 56% increased rate of thyroid cancer diagnosis per doubling of linear perfluorooctanesulfonic acid (n-PFOS) intensity (ORadj, 1.56, 95% CI: 1.17-2.15, P = 0.004); results were similar when including patients with papillary thyroid cancer only (ORadj, 1.56, 95% CI: 1.13-2.21, P = 0.009). This positive association remained in subset analysis investigating exposure timing including 31 thyroid cancer cases diagnosed ≥1 year after plasma sample collection (ORadj, 2.67, 95% CI: 1.59-4.88, P < 0.001).

INTERPRETATION

This study reports associations between exposure to PFAS and increased rate of (papillary) thyroid cancer. Thyroid cancer risk from PFAS exposure is a global concern given the prevalence of PFAS exposure. Individual PFAS studied here are a small proportion of the total number of PFAS supporting additional large-scale prospective studies investigating thyroid cancer risk associated with exposure to PFAS chemicals.

FUNDING

National Institutes of Health grants and The Andrea and Charles Bronfman Philanthropies.

Discrepancy between predicted and measured exercise intensity for eliciting the maximal rate of lipid oxidation

BACKGROUND AND AIMS

Ectopic lipid storage is implicated in type 2 diabetes pathogenesis; hence, exercise to deplete stores (i.e., at the intensity that allows for maximal rate of lipid oxidation; MLO) might be optimal for restoring metabolic health. This intensity ("Fatmax") is estimated during incremental exercise ("Fatmax test"). However, in "the field" general recommendations exist regarding a range of percentages of maximal heart rate (HR) to elicit MLO. The degree to which this range is aligned with measured Fatmax has not been investigated. We compared measured HR at Fatmax, with maximal HR percentages within the typically recommended range in a sample of 26 individuals (Female: n = 11, European ancestry: n = 17).

METHODS AND RESULTS

Subjects completed a modified Fatmax test with a 5-min warmup, followed by incremental stages starting at 15 W with work rate increased by 15 W every 5 min until termination criteria were reached. Pulmonary gas exchange was recorded and average values for V˙ o2 and V˙ co2 for the final minute of each stage were used to estimate substrate-oxidation rates. We modeled lipid-oxidation kinetics using a sinusoidal model and expressed MLO relative to peak V˙ o2 and HR. Bland-Altman analysis demonstrated lack of concordance between HR at Fatmax and at 50%, 70%, and 80% of age-predicted maximum with a mean difference of 23 b·min-1.

CONCLUSION

Our results indicate that estimated "fat-burning" heart rate zones are inappropriate for prescribing exercise to elicit MLO and we recommend direct individual exercise lipid oxidation measurements to elicit these values.

Peripheral Transcriptomics in Acute and Long-Term Kidney Dysfunction in SARS-CoV2 Infection

Background

Acute kidney injury (AKI) is common in hospitalized patients with SARS-CoV2 infection despite vaccination and leads to long-term kidney dysfunction. However, peripheral blood molecular signatures in AKI from COVID-19 and their association with long-term kidney dysfunction are yet unexplored.

Methods

In patients hospitalized with SARS-CoV2, we performed bulk RNA sequencing using peripheral blood mononuclear cells(PBMCs). We applied linear models accounting for technical and biological variability on RNA-Seq data accounting for false discovery rate (FDR) and compared functional enrichment and pathway results to a historical sepsis-AKI cohort. Finally, we evaluated the association of these signatures with long-term trends in kidney function.

Results

Of 283 patients, 106 had AKI. After adjustment for sex, age, mechanical ventilation, and chronic kidney disease (CKD), we identified 2635 significant differential gene expressions at FDR<0.05. Top canonical pathways were EIF2 signaling, oxidative phosphorylation, mTOR signaling, and Th17 signaling, indicating mitochondrial dysfunction and endoplasmic reticulum (ER) stress. Comparison with sepsis associated AKI showed considerable overlap of key pathways (48.14%). Using follow-up estimated glomerular filtration rate (eGFR) measurements from 115 patients, we identified 164/2635 (6.2%) of the significantly differentiated genes associated with overall decrease in long-term kidney function. The strongest associations were 'autophagy', 'renal impairment via fibrosis', and 'cardiac structure and function'.

Conclusions

We show that AKI in SARS-CoV2 is a multifactorial process with mitochondrial dysfunction driven by ER stress whereas long-term kidney function decline is associated with cardiac structure and function and immune dysregulation. Functional overlap with sepsis-AKI also highlights common signatures, indicating generalizability in therapeutic approaches.

SIGNIFICANCE STATEMENT

Peripheral transcriptomic findings in acute and long-term kidney dysfunction after hospitalization for SARS-CoV2 infection are unclear. We evaluated peripheral blood molecular signatures in AKI from COVID-19 (COVID-AKI) and their association with long-term kidney dysfunction using the largest hospitalized cohort with transcriptomic data. Analysis of 283 hospitalized patients of whom 37% had AKI, highlighted the contribution of mitochondrial dysfunction driven by endoplasmic reticulum stress in the acute stages. Subsequently, long-term kidney function decline exhibits significant associations with markers of cardiac structure and function and immune mediated dysregulation. There were similar biomolecular signatures in other inflammatory states, such as sepsis. This enhances the potential for repurposing and generalizability in therapeutic approaches.

A clustering approach to improve our understanding of the genetic and phenotypic complexity of chronic kidney disease

Chronic kidney disease (CKD) is a complex disorder that causes a gradual loss of kidney function, affecting approximately 9.1% of the world's population. Here, we use a soft-clustering algorithm to deconstruct its genetic heterogeneity. First, we selected 322 CKD-associated independent genetic variants from published genome-wide association studies (GWAS) and added association results for 229 traits from the GWAS catalog. We then applied nonnegative matrix factorization (NMF) to discover overlapping clusters of related traits and variants. We computed cluster-specific polygenic scores and validated each cluster with a phenome-wide association study (PheWAS) on the BioMe biobank (n=31,701). NMF identified nine clusters that reflect different aspects of CKD, with the top-weighted traits signifying areas such as kidney function, type 2 diabetes (T2D), and body weight. For most clusters, the top-weighted traits were confirmed in the PheWAS analysis. Results were found to be more significant in the cross-ancestry analysis, although significant ancestry-specific associations were also identified. While all alleles were associated with a decreased kidney function, associations with CKD-related diseases (e.g., T2D) were found only for a smaller subset of variants and differed across genetic ancestry groups. Our findings leverage genetics to gain insights into the underlying biology of CKD and investigate population-specific associations.

Implications of the Use of Artificial Intelligence Predictive Models in Health Care Settings : A Simulation Study

BACKGROUND

Substantial effort has been directed toward demonstrating uses of predictive models in health care. However, implementation of these models into clinical practice may influence patient outcomes, which in turn are captured in electronic health record data. As a result, deployed models may affect the predictive ability of current and future models.

OBJECTIVE

To estimate changes in predictive model performance with use through 3 common scenarios: model retraining, sequentially implementing 1 model after another, and intervening in response to a model when 2 are simultaneously implemented.

DESIGN

Simulation of model implementation and use in critical care settings at various levels of intervention effectiveness and clinician adherence. Models were either trained or retrained after simulated implementation.

SETTING

Admissions to the intensive care unit (ICU) at Mount Sinai Health System (New York, New York) and Beth Israel Deaconess Medical Center (Boston, Massachusetts).

PATIENTS

130 000 critical care admissions across both health systems.

INTERVENTION

Across 3 scenarios, interventions were simulated at varying levels of clinician adherence and effectiveness.

MEASUREMENTS

Statistical measures of performance, including threshold-independent (area under the curve) and threshold-dependent measures.

RESULTS

At fixed 90% sensitivity, in scenario 1 a mortality prediction model lost 9% to 39% specificity after retraining once and in scenario 2 a mortality prediction model lost 8% to 15% specificity when created after the implementation of an acute kidney injury (AKI) prediction model; in scenario 3, models for AKI and mortality prediction implemented simultaneously, each led to reduced effective accuracy of the other by 1% to 28%.

LIMITATIONS

In real-world practice, the effectiveness of and adherence to model-based recommendations are rarely known in advance. Only binary classifiers for tabular ICU admissions data were simulated.

CONCLUSION

In simulated ICU settings, a universally effective model-updating approach for maintaining model performance does not seem to exist. Model use may have to be recorded to maintain viability of predictive modeling.

PRIMARY FUNDING SOURCE

National Center for Advancing Translational Sciences.

Automated Identification of Heart Failure with Reduced Ejection Fraction using Deep Learning-based Natural Language Processing

Background

The lack of automated tools for measuring care quality has limited the implementation of a national program to assess and improve guideline-directed care in heart failure with reduced ejection fraction (HFrEF). A key challenge for constructing such a tool has been an accurate, accessible approach for identifying patients with HFrEF at hospital discharge, an opportunity to evaluate and improve the quality of care.

Methods

We developed a novel deep learning-based language model for identifying patients with HFrEF from discharge summaries using a semi-supervised learning framework. For this purpose, hospitalizations with heart failure at Yale New Haven Hospital (YNHH) between 2015 to 2019 were labeled as HFrEF if the left ventricular ejection fraction was under 40% on antecedent echocardiography. The model was internally validated with model-based net reclassification improvement (NRI) assessed against chart-based diagnosis codes. We externally validated the model on discharge summaries from hospitalizations with heart failure at Northwestern Medicine, community hospitals of Yale New Haven Health in Connecticut and Rhode Island, and the publicly accessible MIMIC-III database, confirmed with chart abstraction.

Results

A total of 13,251 notes from 5,392 unique individuals (mean age 73 ± 14 years, 48% female), including 2,487 patients with HFrEF (46.1%), were used for model development (train/held-out test: 70/30%). The deep learning model achieved an area under receiving operating characteristic (AUROC) of 0.97 and an area under precision-recall curve (AUPRC) of 0.97 in detecting HFrEF on the held-out set. In external validation, the model had high performance in identifying HFrEF from discharge summaries with AUROC 0.94 and AUPRC 0.91 on 19,242 notes from Northwestern Medicine, AUROC 0.95 and AUPRC 0.96 on 139 manually abstracted notes from Yale community hospitals, and AUROC 0.91 and AUPRC 0.92 on 146 manually reviewed notes at MIMIC-III. Model-based prediction of HFrEF corresponded to an overall NRI of 60.2 ± 1.9% compared with the chart diagnosis codes (p-value < 0.001) and an increase in AUROC from 0.61 [95% CI: 060-0.63] to 0.91 [95% CI 0.90-0.92].

Conclusions

We developed and externally validated a deep learning language model that automatically identifies HFrEF from clinical notes with high precision and accuracy, representing a key element in automating quality assessment and improvement for individuals with HFrEF.

Deep learning on electronic medical records identifies distinct subphenotypes of diabetic kidney disease driven by genetic variations in the Rho pathway

Kidney disease affects 50% of all diabetic patients; however, prediction of disease progression has been challenging due to inherent disease heterogeneity. We use deep learning to identify novel genetic signatures prognostically associated with outcomes. Using autoencoders and unsupervised clustering of electronic health record data on 1,372 diabetic kidney disease patients, we establish two clusters with differential prevalence of end-stage kidney disease. Exome-wide associations identify a novel variant in ARHGEF18, a Rho guanine exchange factor specifically expressed in glomeruli. Overexpression of ARHGEF18 in human podocytes leads to impairments in focal adhesion architecture, cytoskeletal dynamics, cellular motility, and RhoA/Rac1 activation. Mutant GEF18 is resistant to ubiquitin mediated degradation leading to pathologically increased protein levels. Our findings uncover the first known disease-causing genetic variant that affects protein stability of a cytoskeletal regulator through impaired degradation, a potentially novel class of expression quantitative trait loci that can be therapeutically targeted.

Deep Learning on Electrocardiograms for Prediction of In-hospital Intradialytic Hypotension in Patients with ESKD

Intradialytic hypotension is common in patients who are on hemodialysis. We applied deep learning techniques to ECGs to predict patients at risk of IDH. The performance of the model was good with an AUC of 0.763 and AUPRC of 0.35.

Clinical Informatics in Critical Care Medicine

Continuous monitoring and treatment of patients in intensive care units generates vast amounts of data. Critical Care Medicine clinicians incorporate this continuously evolving data to make split-second, life or death decisions for management of these patients. Despite the abundance of data, it can be challenging to consider every accessible data point when making the quick decisions necessary at the point of care. Consequently, Clinical Informatics offers a natural partnership to improve the care for critically ill patients. The last two decades have seen a significant evolution in the role of Clinical Informatics in Critical Care Medicine. In this review, we will discuss how Clinical Informatics improves the care of critically ill patients by enhancing not only data collection and visualization but also bedside medical decision making. We will further discuss the evolving role of machine learning algorithms in Clinical Informatics as it pertains to Critical Care Medicine.

Development of a decentralized cohort for studying post-acute sequelae of COVID-19 in India in the Data4life Study

BACKGROUND

Decentralized, digital health studies can provide real-world evidence of the lasting effects of COVID-19 on physical, socioeconomic, psychological, and social determinant factors of health in India. Existing research cohorts, however, are small and were not designed for longitudinal collection of comprehensive data from India's diverse population. Data4Life is a nationwide, digitally enabled, health research initiative to examine the post-acute sequelae of COVID-19 across individuals, communities, and regions. Data4Life seeks to build an ethnically and geographically diverse population of at least 100,000 participants in India.

METHODS

Here we discuss the feasibility of developing a completely decentralized COVID-19 cohort in India through qualitative analysis of data collection procedures, participant characteristics, participant perspectives on recruitment and reported study motivation.

RESULTS

As of June 13th, 2022, more than 6,000 participants from 17 Indian states completed baseline surveys. Friend and family referral were identified as the most common recruitment method (64.8%) across all demographic groups. Helping family and friends was the primary reason reported for joining the study (61.5%).

CONCLUSIONS

Preliminary findings support the use of digital technology for rapid enrollment and data collection to develop large health research cohorts in India. This demonstrates the potential for expansion of digitally enabled health research in India. These findings also outline the value of person-to-person recruitment strategies when conducting digital health research in modern-day India. Qualitative analysis reveals opportunities to increase diversity and retention in real time. It also informs strategies for improving participant experiences in the current Data4Life initiative and future studies.

A Novel ECG-Based Deep Learning Algorithm to Predict Cardiomyopathy in Patients With Premature Ventricular Complexes

BACKGROUND

Premature ventricular complexes (PVCs) are prevalent and, although often benign, they may lead to PVC-induced cardiomyopathy. We created a deep-learning algorithm to predict left ventricular ejection fraction (LVEF) reduction in patients with PVCs from a 12-lead electrocardiogram (ECG).

OBJECTIVES

This study aims to assess a deep-learning model to predict cardiomyopathy among patients with PVCs.

METHODS

We used electronic medical records from 5 hospitals and identified ECGs from adults with documented PVCs. Internal training and testing were performed at one hospital. External validation was performed with the others. The primary outcome was first diagnosis of LVEF ≤40% within 6 months. The dataset included 383,514 ECGs, of which 14,241 remained for analysis. We analyzed area under the receiver operating curves and explainability plots for representative patients, algorithm prediction, PVC burden, and demographics in a multivariable Cox model to assess independent predictors for cardiomyopathy.

RESULTS

Among the 14,241-patient cohort (age 67.6 ± 14.8 years; female 43.8%; White 29.5%, Black 8.6%, Hispanic 6.5%, Asian 2.2%), 22.9% experienced reductions in LVEF to ≤40% within 6 months. The model predicted reductions in LVEF to ≤40% with area under the receiver operating curve of 0.79 (95% CI: 0.77-0.81). The gradient weighted class activation map explainability framework highlighted the sinus rhythm QRS complex-ST segment. In patients who underwent successful PVC ablation there was a post-ablation improvement in LVEF with resolution of cardiomyopathy in most (89%) patients.

CONCLUSIONS

Deep-learning on the 12-lead ECG alone can accurately predict new-onset cardiomyopathy in patients with PVCs independent of PVC burden. Model prediction performed well across sex and race, relying on the QRS complex/ST-segment in sinus rhythm, not PVC morphology.

A study of gene expression in the living human brain

A goal of medical research is to determine the molecular basis of human brain health and illness. One way to achieve this goal is through observational studies of gene expression in human brain tissue. Due to the unavailability of brain tissue from living people, most such studies are performed using tissue from postmortem brain donors. An assumption underlying this practice is that gene expression in the postmortem human brain is an accurate representation of gene expression in the living human brain. Here, this assumption - which, until now, had not been adequately tested - is tested by comparing human prefrontal cortex gene expression between 275 living samples and 243 postmortem samples. Expression levels differed significantly for nearly 80% of genes, and a systematic examination of alternative explanations for this observation determined that these differences are not a consequence of cell type composition, RNA quality, postmortem interval, age, medication, morbidity, symptom severity, tissue pathology, sample handling, batch effects, or computational methods utilized. Analyses integrating the data generated for this study with data from earlier landmark studies that used tissue from postmortem brain donors showed that postmortem brain gene expression signatures of neurological and mental illnesses, as well as of normal traits such as aging, may not be accurate representations of these gene expression signatures in the living brain. By using tissue from large cohorts living people, future observational studies of human brain biology have the potential to (1) determine the medical research questions that can be addressed using postmortem tissue as a proxy for living tissue and (2) expand the scope of medical research to include questions about the molecular basis of human brain health and illness that can only be addressed in living people (e.g., "What happens at the molecular level in the brain as a person experiences an emotion?").

Time-to-Event Genome-Wide Association Study for Incident Cardiovascular Disease in People with Type 2 Diabetes Mellitus

BACKGROUND

Type 2 diabetes mellitus (T2D) confers a two- to three-fold increased risk of cardiovascular disease (CVD). However, the mechanisms underlying increased CVD risk among people with T2D are only partially understood. We hypothesized that a genetic association study among people with T2D at risk for developing incident cardiovascular complications could provide insights into molecular genetic aspects underlying CVD.

METHODS

From 16 studies of the Cohorts for Heart & Aging Research in Genomic Epidemiology (CHARGE) Consortium, we conducted a multi-ancestry time-to-event genome-wide association study (GWAS) for incident CVD among people with T2D using Cox proportional hazards models. Incident CVD was defined based on a composite of coronary artery disease (CAD), stroke, and cardiovascular death that occurred at least one year after the diagnosis of T2D. Cohort-level estimated effect sizes were combined using inverse variance weighted fixed effects meta-analysis. We also tested 204 known CAD variants for association with incident CVD among patients with T2D.

RESULTS

A total of 49,230 participants with T2D were included in the analyses (31,118 European ancestries and 18,112 non-European ancestries) which consisted of 8,956 incident CVD cases over a range of mean follow-up duration between 3.2 and 33.7 years (event rate 18.2%). We identified three novel, distinct genetic loci for incident CVD among individuals with T2D that reached the threshold for genome-wide significance (P<5.0×10-8): rs147138607 (intergenic variant between CACNA1E and ZNF648) with a hazard ratio (HR) 1.23, 95% confidence interval (CI) 1.15 - 1.32, P=3.6×10-9, rs11444867 (intergenic variant near HS3ST1) with HR 1.89, 95% CI 1.52 - 2.35, P=9.9×10-9, and rs335407 (intergenic variant between TFB1M and NOX3) HR 1.25, 95% CI 1.16 - 1.35, P=1.5×10-8. Among 204 known CAD loci, 32 were associated with incident CVD in people with T2D with P<0.05, and 5 were significant after Bonferroni correction (P<0.00024, 0.05/204). A polygenic score of these 204 variants was significantly associated with incident CVD with HR 1.14 (95% CI 1.12 - 1.16) per 1 standard deviation increase (P=1.0×10-16).

CONCLUSIONS

The data point to novel and known genomic regions associated with incident CVD among individuals with T2D.

A machine learning approach to determine resilience utilizing wearable device data: analysis of an observational cohort

Objective

To assess whether an individual's degree of psychological resilience can be determined from physiological metrics passively collected from a wearable device.

Materials and Methods

Data were analyzed in this secondary analysis of the Warrior Watch Study dataset, a prospective cohort of healthcare workers enrolled across 7 hospitals in New York City. Subjects wore an Apple Watch for the duration of their participation. Surveys were collected measuring resilience, optimism, and emotional support at baseline.

Results

We evaluated data from 329 subjects (mean age 37.4 years, 37.1% male). Across all testing sets, gradient-boosting machines (GBM) and extreme gradient-boosting models performed best for high- versus low-resilience prediction, stratified on a median Connor-Davidson Resilience Scale-2 score of 6 (interquartile range = 5-7), with an AUC of 0.60. When predicting resilience as a continuous variable, multivariate linear models had a correlation of 0.24 (P = .029) and RMSE of 1.37 in the testing data. A positive psychological construct, comprised of resilience, optimism, and emotional support was also evaluated. The oblique random forest method performed best in estimating high- versus low-composite scores stratified on a median of 32.5, with an AUC of 0.65, a sensitivity of 0.60, and a specificity of 0.70.

Discussion

In a post hoc analysis, machine learning models applied to physiological metrics collected from wearable devices had some predictive ability in identifying resilience states and a positive psychological construct.

Conclusions

These findings support the further assessment of psychological characteristics from passively collected wearable data in dedicated studies.

Genome-wide association study of thoracic aortic aneurysm and dissection in the Million Veteran Program

The current understanding of the genetic determinants of thoracic aortic aneurysms and dissections (TAAD) has largely been informed through studies of rare, Mendelian forms of disease. Here, we conducted a genome-wide association study (GWAS) of TAAD, testing ~25 million DNA sequence variants in 8,626 participants with and 453,043 participants without TAAD in the Million Veteran Program, with replication in an independent sample of 4,459 individuals with and 512,463 without TAAD from six cohorts. We identified 21 TAAD risk loci, 17 of which have not been previously reported. We leverage multiple downstream analytic methods to identify causal TAAD risk genes and cell types and provide human genetic evidence that TAAD is a non-atherosclerotic aortic disorder distinct from other forms of vascular disease. Our results demonstrate that the genetic architecture of TAAD mirrors that of other complex traits and that it is not solely inherited through protein-altering variants of large effect size.

Researching COVID to Enhance Recovery (RECOVER) adult study protocol: Rationale, objectives, and design

IMPORTANCE

SARS-CoV-2 infection can result in ongoing, relapsing, or new symptoms or other health effects after the acute phase of infection; termed post-acute sequelae of SARS-CoV-2 infection (PASC), or long COVID. The characteristics, prevalence, trajectory and mechanisms of PASC are ill-defined. The objectives of the Researching COVID to Enhance Recovery (RECOVER) Multi-site Observational Study of PASC in Adults (RECOVER-Adult) are to: (1) characterize PASC prevalence; (2) characterize the symptoms, organ dysfunction, natural history, and distinct phenotypes of PASC; (3) identify demographic, social and clinical risk factors for PASC onset and recovery; and (4) define the biological mechanisms underlying PASC pathogenesis.

METHODS

RECOVER-Adult is a combined prospective/retrospective cohort currently planned to enroll 14,880 adults aged ≥18 years. Eligible participants either must meet WHO criteria for suspected, probable, or confirmed infection; or must have evidence of no prior infection. Recruitment occurs at 86 sites in 33 U.S. states, Washington, DC and Puerto Rico, via facility- and community-based outreach. Participants complete quarterly questionnaires about symptoms, social determinants, vaccination status, and interim SARS-CoV-2 infections. In addition, participants contribute biospecimens and undergo physical and laboratory examinations at approximately 0, 90 and 180 days from infection or negative test date, and yearly thereafter. Some participants undergo additional testing based on specific criteria or random sampling. Patient representatives provide input on all study processes. The primary study outcome is onset of PASC, measured by signs and symptoms. A paradigm for identifying PASC cases will be defined and updated using supervised and unsupervised learning approaches with cross-validation. Logistic regression and proportional hazards regression will be conducted to investigate associations between risk factors, onset, and resolution of PASC symptoms.

DISCUSSION

RECOVER-Adult is the first national, prospective, longitudinal cohort of PASC among US adults. Results of this study are intended to inform public health, spur clinical trials, and expand treatment options.

REGISTRATION

NCT05172024.

Development of a Definition of Postacute Sequelae of SARS-CoV-2 Infection

Importance

SARS-CoV-2 infection is associated with persistent, relapsing, or new symptoms or other health effects occurring after acute infection, termed postacute sequelae of SARS-CoV-2 infection (PASC), also known as long COVID. Characterizing PASC requires analysis of prospectively and uniformly collected data from diverse uninfected and infected individuals.

Objective

To develop a definition of PASC using self-reported symptoms and describe PASC frequencies across cohorts, vaccination status, and number of infections.

Design, Setting, and Participants

Prospective observational cohort study of adults with and without SARS-CoV-2 infection at 85 enrolling sites (hospitals, health centers, community organizations) located in 33 states plus Washington, DC, and Puerto Rico. Participants who were enrolled in the RECOVER adult cohort before April 10, 2023, completed a symptom survey 6 months or more after acute symptom onset or test date. Selection included population-based, volunteer, and convenience sampling.

Exposure

SARS-CoV-2 infection.

Main Outcomes and Measures

PASC and 44 participant-reported symptoms (with severity thresholds).

Results

A total of 9764 participants (89% SARS-CoV-2 infected; 71% female; 16% Hispanic/Latino; 15% non-Hispanic Black; median age, 47 years [IQR, 35-60]) met selection criteria. Adjusted odds ratios were 1.5 or greater (infected vs uninfected participants) for 37 symptoms. Symptoms contributing to PASC score included postexertional malaise, fatigue, brain fog, dizziness, gastrointestinal symptoms, palpitations, changes in sexual desire or capacity, loss of or change in smell or taste, thirst, chronic cough, chest pain, and abnormal movements. Among 2231 participants first infected on or after December 1, 2021, and enrolled within 30 days of infection, 224 (10% [95% CI, 8.8%-11%]) were PASC positive at 6 months.

Conclusions and Relevance

A definition of PASC was developed based on symptoms in a prospective cohort study. As a first step to providing a framework for other investigations, iterative refinement that further incorporates other clinical features is needed to support actionable definitions of PASC.

Proteomic characterization of acute kidney injury in patients hospitalized with SARS-CoV2 infection

BACKGROUND

Acute kidney injury (AKI) is a known complication of COVID-19 and is associated with an increased risk of in-hospital mortality. Unbiased proteomics using biological specimens can lead to improved risk stratification and discover pathophysiological mechanisms.

METHODS

Using measurements of ~4000 plasma proteins in two cohorts of patients hospitalized with COVID-19, we discovered and validated markers of COVID-associated AKI (stage 2 or 3) and long-term kidney dysfunction. In the discovery cohort (N = 437), we identified 413 higher plasma abundances of protein targets and 30 lower plasma abundances of protein targets associated with COVID-AKI (adjusted p < 0.05). Of these, 62 proteins were validated in an external cohort (p < 0.05, N = 261).

RESULTS

We demonstrate that COVID-AKI is associated with increased markers of tubular injury (NGAL) and myocardial injury. Using estimated glomerular filtration (eGFR) measurements taken after discharge, we also find that 25 of the 62 AKI-associated proteins are significantly associated with decreased post-discharge eGFR (adjusted p < 0.05). Proteins most strongly associated with decreased post-discharge eGFR included desmocollin-2, trefoil factor 3, transmembrane emp24 domain-containing protein 10, and cystatin-C indicating tubular dysfunction and injury.

CONCLUSIONS

Using clinical and proteomic data, our results suggest that while both acute and long-term COVID-associated kidney dysfunction are associated with markers of tubular dysfunction, AKI is driven by a largely multifactorial process involving hemodynamic instability and myocardial damage.

A foundational vision transformer improves diagnostic performance for electrocardiograms

The electrocardiogram (ECG) is a ubiquitous diagnostic modality. Convolutional neural networks (CNNs) applied towards ECG analysis require large sample sizes, and transfer learning approaches for biomedical problems may result in suboptimal performance when pre-training is done on natural images. We leveraged masked image modeling to create a vision-based transformer model, HeartBEiT, for electrocardiogram waveform analysis. We pre-trained this model on 8.5 million ECGs and then compared performance vs. standard CNN architectures for diagnosis of hypertrophic cardiomyopathy, low left ventricular ejection fraction and ST elevation myocardial infarction using differing training sample sizes and independent validation datasets. We find that HeartBEiT has significantly higher performance at lower sample sizes compared to other models. We also find that HeartBEiT improves explainability of diagnosis by highlighting biologically relevant regions of the EKG vs. standard CNNs. Domain specific pre-trained transformer models may exceed the classification performance of models trained on natural images especially in very low data regimes. The combination of the architecture and such pre-training allows for more accurate, granular explainability of model predictions.

Joint Modeling of Clinical and Biomarker Data in Acute Kidney Injury Defines Unique Subphenotypes with Differing Outcomes

BACKGROUND

AKI is a heterogeneous syndrome. Current subphenotyping approaches have only used limited laboratory data to understand a much more complex condition.

METHODS

We focused on patients with AKI from the Assessment, Serial Evaluation, and Subsequent Sequelae in AKI (ASSESS-AKI). We used hierarchical clustering with Ward linkage on biomarkers of inflammation, injury, and repair/health. We then evaluated clinical differences between subphenotypes and examined their associations with cardiorenal events and death using Cox proportional hazard models.

RESULTS

We included 748 patients with AKI: 543 (73%) of them had AKI stage 1, 112 (15%) had AKI stage 2, and 93 (12%) had AKI stage 3. The mean age (±SD) was 64 (13) years; 508 (68%) were men; and the median follow-up was 4.7 (Q1: 2.9, Q3: 5.7) years. Patients with AKI subphenotype 1 ( N =181) had the highest kidney injury molecule (KIM-1) and troponin T levels. Subphenotype 2 ( N =250) had the highest levels of uromodulin. AKI subphenotype 3 ( N =159) comprised patients with markedly high pro-brain natriuretic peptide and plasma tumor necrosis factor receptor-1 and -2 and low concentrations of KIM-1 and neutrophil gelatinase-associated lipocalin. Finally, patients with subphenotype 4 ( N =158) predominantly had sepsis-AKI and the highest levels of vascular/kidney inflammation (YKL-40, MCP-1) and injury (neutrophil gelatinase-associated lipocalin, KIM-1). AKI subphenotypes 3 and 4 were independently associated with a higher risk of death compared with subphenotype 2 and had adjusted hazard ratios of 2.9 (95% confidence interval, 1.8 to 4.6) and 1.6 (95% confidence interval, 1.01 to 2.6, P = 0.04), respectively. Subphenotype 3 was also independently associated with a three-fold risk of CKD and cardiovascular events.

CONCLUSIONS

We discovered four AKI subphenotypes with differing clinical features and biomarker profiles that are associated with longitudinal clinical outcomes.

Polygenic prediction of preeclampsia and gestational hypertension

Preeclampsia and gestational hypertension are common pregnancy complications associated with adverse maternal and child outcomes. Current tools for prediction, prevention and treatment are limited. Here we tested the association of maternal DNA sequence variants with preeclampsia in 20,064 cases and 703,117 control individuals and with gestational hypertension in 11,027 cases and 412,788 control individuals across discovery and follow-up cohorts using multi-ancestry meta-analysis. Altogether, we identified 18 independent loci associated with preeclampsia/eclampsia and/or gestational hypertension, 12 of which are new (for example, MTHFR-CLCN6, WNT3A, NPR3, PGR and RGL3), including two loci (PLCE1 and FURIN) identified in the multitrait analysis. Identified loci highlight the role of natriuretic peptide signaling, angiogenesis, renal glomerular function, trophoblast development and immune dysregulation. We derived genome-wide polygenic risk scores that predicted preeclampsia/eclampsia and gestational hypertension in external cohorts, independent of clinical risk factors, and reclassified eligibility for low-dose aspirin to prevent preeclampsia. Collectively, these findings provide mechanistic insights into the hypertensive disorders of pregnancy and have the potential to advance pregnancy risk stratification.

Randomized Trial of Anticoagulation Strategies for Noncritically Ill Patients Hospitalized With COVID-19

BACKGROUND

Prior studies of therapeutic-dose anticoagulation in patients with COVID-19 have reported conflicting results.

OBJECTIVES

We sought to determine the safety and effectiveness of therapeutic-dose anticoagulation in noncritically ill patients with COVID-19.

METHODS

Patients hospitalized with COVID-19 not requiring intensive care unit treatment were randomized to prophylactic-dose enoxaparin, therapeutic-dose enoxaparin, or therapeutic-dose apixaban. The primary outcome was the 30-day composite of all-cause mortality, requirement for intensive care unit-level of care, systemic thromboembolism, or ischemic stroke assessed in the combined therapeutic-dose groups compared with the prophylactic-dose group.

RESULTS

Between August 26, 2020, and September 19, 2022, 3,398 noncritically ill patients hospitalized with COVID-19 were randomized to prophylactic-dose enoxaparin (n = 1,141), therapeutic-dose enoxaparin (n = 1,136), or therapeutic-dose apixaban (n = 1,121) at 76 centers in 10 countries. The 30-day primary outcome occurred in 13.2% of patients in the prophylactic-dose group and 11.3% of patients in the combined therapeutic-dose groups (HR: 0.85; 95% CI: 0.69-1.04; P = 0.11). All-cause mortality occurred in 7.0% of patients treated with prophylactic-dose enoxaparin and 4.9% of patients treated with therapeutic-dose anticoagulation (HR: 0.70; 95% CI: 0.52-0.93; P = 0.01), and intubation was required in 8.4% vs 6.4% of patients, respectively (HR: 0.75; 95% CI: 0.58-0.98; P = 0.03). Results were similar in the 2 therapeutic-dose groups, and major bleeding in all 3 groups was infrequent.

CONCLUSIONS

Among noncritically ill patients hospitalized with COVID-19, the 30-day primary composite outcome was not significantly reduced with therapeutic-dose anticoagulation compared with prophylactic-dose anticoagulation. However, fewer patients who were treated with therapeutic-dose anticoagulation required intubation and fewer died (FREEDOM COVID [FREEDOM COVID Anticoagulation Strategy]; NCT04512079).

A machine learning model identifies patients in need of autoimmune disease testing using electronic health records

Systemic autoimmune rheumatic diseases (SARDs) can lead to irreversible damage if left untreated, yet these patients often endure long diagnostic journeys before being diagnosed and treated. Machine learning may help overcome the challenges of diagnosing SARDs and inform clinical decision-making. Here, we developed and tested a machine learning model to identify patients who should receive rheumatological evaluation for SARDs using longitudinal electronic health records of 161,584 individuals from two institutions. The model demonstrated high performance for predicting cases of autoantibody-tested individuals in a validation set, an external test set, and an independent cohort with a broader case definition. This approach identified more individuals for autoantibody testing compared with current clinical standards and a greater proportion of autoantibody carriers among those tested. Diagnoses of SARDs and other autoimmune conditions increased with higher model probabilities. The model detected a need for autoantibody testing and rheumatology encounters up to five years before the test date and assessment date, respectively. Altogether, these findings illustrate that the clinical manifestations of a diverse array of autoimmune conditions are detectable in electronic health records using machine learning, which may help systematize and accelerate autoimmune testing.

Major cardiovascular events and subsequent risk of kidney failure with replacement therapy: a CKD Prognosis Consortium study

AIMS

Chronic kidney disease (CKD) increases risk of cardiovascular disease (CVD). Less is known about how CVD associates with future risk of kidney failure with replacement therapy (KFRT).

METHODS AND RESULTS

The study included 25 903 761 individuals from the CKD Prognosis Consortium with known baseline estimated glomerular filtration rate (eGFR) and evaluated the impact of prevalent and incident coronary heart disease (CHD), stroke, heart failure (HF), and atrial fibrillation (AF) events as time-varying exposures on KFRT outcomes. Mean age was 53 (standard deviation 17) years and mean eGFR was 89 mL/min/1.73 m2, 15% had diabetes and 8.4% had urinary albumin-to-creatinine ratio (ACR) available (median 13 mg/g); 9.5% had prevalent CHD, 3.2% prior stroke, 3.3% HF, and 4.4% prior AF. During follow-up, there were 269 142 CHD, 311 021 stroke, 712 556 HF, and 605 596 AF incident events and 101 044 (0.4%) patients experienced KFRT. Both prevalent and incident CVD were associated with subsequent KFRT with adjusted hazard ratios (HRs) of 3.1 [95% confidence interval (CI): 2.9-3.3], 2.0 (1.9-2.1), 4.5 (4.2-4.9), 2.8 (2.7-3.1) after incident CHD, stroke, HF and AF, respectively. HRs were highest in first 3 months post-CVD incidence declining to baseline after 3 years. Incident HF hospitalizations showed the strongest association with KFRT [HR 46 (95% CI: 43-50) within 3 months] after adjustment for other CVD subtype incidence.

CONCLUSION

Incident CVD events strongly and independently associate with future KFRT risk, most notably after HF, then CHD, stroke, and AF. Optimal strategies for addressing the dramatic risk of KFRT following CVD events are needed.

The Kidney Failure Risk Equation: Evaluation of Novel Input Variables including eGFR Estimated Using the CKD-EPI 2021 Equation in 59 Cohorts

SIGNIFICANCE STATEMENT

The kidney failure risk equation (KFRE) uses age, sex, GFR, and urine albumin-to-creatinine ratio (ACR) to predict 2- and 5-year risk of kidney failure in populations with eGFR <60 ml/min per 1.73 m 2 . However, the CKD-EPI 2021 creatinine equation for eGFR is now recommended for use but has not been fully tested in the context of KFRE. In 59 cohorts comprising 312,424 patients with CKD, the authors assessed the predictive performance and calibration associated with the use of the CKD-EPI 2021 equation and whether additional variables and accounting for the competing risk of death improves the KFRE's performance. The KFRE generally performed well using the CKD-EPI 2021 eGFR in populations with eGFR <45 ml/min per 1.73 m 2 and was not improved by adding the 2-year prior eGFR slope and cardiovascular comorbidities.

BACKGROUND

The kidney failure risk equation (KFRE) uses age, sex, GFR, and urine albumin-to-creatinine ratio (ACR) to predict kidney failure risk in people with GFR <60 ml/min per 1.73 m 2 .

METHODS

Using 59 cohorts with 312,424 patients with CKD, we tested several modifications to the KFRE for their potential to improve the KFRE: using the CKD-EPI 2021 creatinine equation for eGFR, substituting 1-year average ACR for single-measure ACR and 1-year average eGFR in participants with high eGFR variability, and adding 2-year prior eGFR slope and cardiovascular comorbidities. We also assessed calibration of the KFRE in subgroups of eGFR and age before and after accounting for the competing risk of death.

RESULTS

The KFRE remained accurate and well calibrated overall using the CKD-EPI 2021 eGFR equation. The other modifications did not improve KFRE performance. In subgroups of eGFR 45-59 ml/min per 1.73 m 2 and in older adults using the 5-year time horizon, the KFRE demonstrated systematic underprediction and overprediction, respectively. We developed and tested a new model with a spline term in eGFR and incorporating the competing risk of mortality, resulting in more accurate calibration in those specific subgroups but not overall.

CONCLUSIONS

The original KFRE is generally accurate for eGFR <45 ml/min per 1.73 m 2 when using the CKD-EPI 2021 equation. Incorporating competing risk methodology and splines for eGFR may improve calibration in low-risk settings with longer time horizons. Including historical averages, eGFR slopes, or a competing risk design did not meaningfully alter KFRE performance in most circumstances.

Multi-center retrospective cohort study applying deep learning to electrocardiograms to identify left heart valvular dysfunction

BACKGROUND

Aortic Stenosis and Mitral Regurgitation are common valvular conditions representing a hidden burden of disease within the population. The aim of this study was to develop and validate deep learning-based screening and diagnostic tools that can help guide clinical decision making.

METHODS

In this multi-center retrospective cohort study, we acquired Transthoracic Echocardiogram reports from five Mount Sinai hospitals within New York City representing a demographically diverse cohort of patients. We developed a Natural Language Processing pipeline to extract ground-truth labels about valvular status and paired these to Electrocardiograms (ECGs). We developed and externally validated deep learning models capable of detecting valvular disease, in addition to considering scenarios of clinical deployment.

RESULTS

We use 617,338 ECGs paired to transthoracic echocardiograms from 123,096 patients to develop a deep learning model for detection of Mitral Regurgitation. Area Under Receiver Operating Characteristic curve (AUROC) is 0.88 (95% CI:0.88-0.89) in internal testing, and 0.81 (95% CI:0.80-0.82) in external validation. To develop a model for detection of Aortic Stenosis, we use 617,338 Echo-ECG pairs for 128,628 patients. AUROC is 0.89 (95% CI: 0.88-0.89) in internal testing, going to 0.86 (95% CI: 0.85-0.87) in external validation. The model's performance increases leading up to the time of the diagnostic echo, and it performs well in validation against requirement of Transcatheter Aortic Valve Replacement procedures.

CONCLUSIONS

Deep learning based tools can increase the amount of information extracted from ubiquitous investigations such as the ECG. Such tools are inexpensive, can help in earlier disease detection, and potentially improve prognosis.

Evaluation of the portability of computable phenotypes with natural language processing in the eMERGE network

The electronic Medical Records and Genomics (eMERGE) Network assessed the feasibility of deploying portable phenotype rule-based algorithms with natural language processing (NLP) components added to improve performance of existing algorithms using electronic health records (EHRs). Based on scientific merit and predicted difficulty, eMERGE selected six existing phenotypes to enhance with NLP. We assessed performance, portability, and ease of use. We summarized lessons learned by: (1) challenges; (2) best practices to address challenges based on existing evidence and/or eMERGE experience; and (3) opportunities for future research. Adding NLP resulted in improved, or the same, precision and/or recall for all but one algorithm. Portability, phenotyping workflow/process, and technology were major themes. With NLP, development and validation took longer. Besides portability of NLP technology and algorithm replicability, factors to ensure success include privacy protection, technical infrastructure setup, intellectual property agreement, and efficient communication. Workflow improvements can improve communication and reduce implementation time. NLP performance varied mainly due to clinical document heterogeneity; therefore, we suggest using semi-structured notes, comprehensive documentation, and customization options. NLP portability is possible with improved phenotype algorithm performance, but careful planning and architecture of the algorithms is essential to support local customizations.