Cohort design and natural language processing to reduce bias in electronic health records research

The AORTA Gene score for detection and risk stratification of ascending aortic dilation

BACKGROUND AND AIMS

This study assessed whether a model incorporating clinical features and a polygenic score for ascending aortic diameter would improve diameter estimation and prediction of adverse thoracic aortic events over clinical features alone.

METHODS

Aortic diameter estimation models were built with a 1.1 million-variant polygenic score (AORTA Gene) and without it. Models were validated internally in 4394 UK Biobank participants and externally in 5469 individuals from Mass General Brigham (MGB) Biobank, 1298 from the Framingham Heart Study (FHS), and 610 from All of Us. Model fit for adverse thoracic aortic events was compared in 401 453 UK Biobank and 164 789 All of Us participants.

RESULTS

AORTA Gene explained more of the variance in thoracic aortic diameter compared to clinical factors alone: 39.5% (95% confidence interval 37.3%-41.8%) vs. 29.3% (27.0%-31.5%) in UK Biobank, 36.5% (34.4%-38.5%) vs. 32.5% (30.4%-34.5%) in MGB, 41.8% (37.7%-45.9%) vs. 33.0% (28.9%-37.2%) in FHS, and 34.9% (28.8%-41.0%) vs. 28.9% (22.9%-35.0%) in All of Us. AORTA Gene had a greater area under the receiver operating characteristic curve for identifying diameter ≥ 4 cm: 0.836 vs. 0.776 (P < .0001) in UK Biobank, 0.808 vs. 0.767 in MGB (P < .0001), 0.856 vs. 0.818 in FHS (P < .0001), and 0.827 vs. 0.791 (P = .0078) in All of Us. AORTA Gene was more informative for adverse thoracic aortic events in UK Biobank (P = .0042) and All of Us (P = .049).

CONCLUSIONS

A comprehensive model incorporating polygenic information and clinical risk factors explained 34.9%-41.8% of the variation in ascending aortic diameter, improving the identification of ascending aortic dilation and adverse thoracic aortic events compared to clinical risk factors.

Frequency of Electrocardiogram-Defined Cardiac Conduction Disorders in a Multi-Institutional Primary Care Cohort

Background

Disorders affecting cardiac conduction are associated with substantial morbidity. Understanding the epidemiology and risk factors for conduction disorders may enable earlier diagnosis and preventive efforts.

Objectives

The purpose of this study was to quantify contemporary frequency and risk factors for electrocardiogram (ECG)-defined cardiac conduction disorders in a large multi-institutional primary care sample.

Methods

We quantified prevalence and incidence of conduction disorders among adults receiving longitudinal primary care between 2001 and 2019, each with at least one 12-lead ECG performed prior to the start of follow-up and at least one ECG during follow-up. We defined conduction disorders using curated terms extracted from ECG diagnostic statements by cardiologists. We grouped conduction disorders by inferred anatomic location of abnormal conduction. We tested associations between clinical factors and incident conduction disease using multivariable proportional hazards regression.

Results

We analyzed 189,163 individuals (median age 55 years; 58% female). The overall prevalence of conduction disorders was 27% among men and 15% among women. Among 119,926 individuals (median age 55 years; 51% female), 6,802 developed an incident conduction system abnormality over a median of 10 years (Q1, Q3: 6, 15 years) of follow-up. Incident conduction disorders were more common in men (8.78 events/1,000 person-years) vs women (4.34 events/1,000 person-years, P < 0.05). In multivariable models, clinical factors including older age (HR: 1.25 per 5-year increase [95% CI: 1.24-1.26]) and myocardial infarction (HR: 1.39 [95% CI: 1.26-1.54]) were associated with incident conduction disorders.

Conclusions

Cardiac conduction disorders are common in a primary care population, especially among older individuals with cardiovascular risk factors.

The emerging role of generative artificial intelligence in transplant medicine

Generative artificial intelligence (AI), a subset of machine learning that creates new content based on training data, has witnessed tremendous advances in recent years. Practical applications have been identified in health care in general, and there is significant opportunity in transplant medicine for generative AI to simplify tasks in research, medical education, and clinical practice. In addition, patients stand to benefit from patient education that is more readily provided by generative AI applications. This review aims to catalyze the development and adoption of generative AI in transplantation by introducing basic AI and generative AI concepts to the transplant clinician and summarizing its current and potential applications within the field. We provide an overview of applications to the clinician, researcher, educator, and patient. We also highlight the challenges involved in bringing these applications to the bedside and need for ongoing refinement of generative AI applications to sustainably augment the transplantation field.

Consider this a WARNing

Atrial fibrillation (AF) prediction can be valuable at many timescales and in many populations. In this issue of Patterns, Gavidia et al. train a model called WARN for short-term prediction of AF in the timescale of minutes in patients wearing 24-h continuous Holter electrocardiograms. The ability to predict near-term (e.g., 30 min) AF has the potential to enable preventive therapies with rapid mechanisms of action (e.g., oral anticoagulation, anti-arrhythmic drugs). In this way, efficient, continuous, and algorithmic monitoring of AF risk could reduce burden on healthcare workers and represents a valuable clinical pursuit.

Recent advancements and applications of deep learning in heart failure: Α systematic review

BACKGROUND

Heart failure (HF), a global health challenge, requires innovative diagnostic and management approaches. The rapid evolution of deep learning (DL) in healthcare necessitates a comprehensive review to evaluate these developments and their potential to enhance HF evaluation, aligning clinical practices with technological advancements.

OBJECTIVE

This review aims to systematically explore the contributions of DL technologies in the assessment of HF, focusing on their potential to improve diagnostic accuracy, personalize treatment strategies, and address the impact of comorbidities.

METHODS

A thorough literature search was conducted across four major electronic databases: PubMed, Scopus, Web of Science and IEEE Xplore, yielding 137 articles that were subsequently categorized into five primary application areas: cardiovascular disease (CVD) classification, HF detection, image analysis, risk assessment, and other clinical analyses. The selection criteria focused on studies utilizing DL algorithms for HF assessment, not limited to HF detection but extending to any attempt in analyzing and interpreting HF-related data.

RESULTS

The analysis revealed a notable emphasis on CVD classification and HF detection, with DL algorithms showing significant promise in distinguishing between affected individuals and healthy subjects. Furthermore, the review highlights DL's capacity to identify underlying cardiomyopathies and other comorbidities, underscoring its utility in refining diagnostic processes and tailoring treatment plans to individual patient needs.

CONCLUSIONS

This review establishes DL as a key innovation in HF management, highlighting its role in advancing diagnostic accuracy and personalized care. The insights provided advocate for the integration of DL in clinical settings and suggest directions for future research to enhance patient outcomes in HF care.

Assessing Real-World Data From Electronic Health Records for Health Technology Assessment: The SUITABILITY Checklist: A Good Practices Report of an ISPOR Task Force

This ISPOR Good Practices report provides a framework for assessing the suitability of electronic health records data for use in health technology assessments (HTAs). Although electronic health record (EHR) data can fill evidence gaps and improve decisions, several important limitations can affect its validity and relevance. The ISPOR framework includes 2 components: data delineation and data fitness for purpose. Data delineation provides a complete understanding of the data and an assessment of its trustworthiness by describing (1) data characteristics; (2) data provenance; and (3) data governance. Fitness for purpose comprises (1) data reliability items, ie, how accurate and complete the estimates are for answering the question at hand and (2) data relevance items, which assess how well the data are suited to answer the particular question from a decision-making perspective. The report includes a checklist specific to EHR data reporting: the ISPOR SUITABILITY Checklist. It also provides recommendations for HTA agencies and policy makers to improve the use of EHR-derived data over time. The report concludes with a discussion of limitations and future directions in the field, including the potential impact from the substantial and rapid advances in the diffusion and capabilities of large language models and generative artificial intelligence. The report's immediate audiences are HTA evidence developers and users. We anticipate that it will also be useful to other stakeholders, particularly regulators and manufacturers, in the future.

Risk factors for pediatric ischemic stroke and intracranial hemorrhage: A national electronic health record based study

Background

Stroke is an important cause of morbidity in pediatrics. Large studies are needed to better understand the epidemiology, pathogenesis and risk factors associated with pediatric stroke. Large administrative datasets can provide information on risk factors in perinatal and childhood stroke at low cost. The aim of this hypothesis-generating study was to use a large administrative dataset to assess for prevalence and odds-ratios of rare exposures associated with pediatric stroke.

Methods

The data for patients aged 0-18 with a diagnosis of either ischemic stroke or intracranial hemorrhage were extracted from the Cerner Health Facts EMR Database from 2000 to 2018. Prevalence of various possible risk factors for pediatric and adult stroke was assessed using ICD 9 and 10 codes. Odds ratios were calculated using a control group of patients without stroke.

Results

10,688 children were identified with stroke. 6339 (59 %) were ischemic and 4349 (41 %) were hemorrhagic. The most frequently identified risk factors for ischemic stroke across age groups were hypertension (29-44 %), trauma (19-33 %), and malignancy (11-24 %). The most common risk factors seen with hemorrhagic stroke were trauma (32-64 %), malignancy (5-19 %) and arrhythmia (9-12 %). Odds ratios across all age groups for dyslipidemia (17-64), hypertension (20-63), and tobacco exposure (3-59) were high in the ischemic stroke cohort.

Conclusion

This is the largest retrospective study of pediatric stroke of its kind from hospitals across the US in both academic and non-academic clinical settings. Much of our data was consistent with prior studies. ICD codes for tobacco exposure, hyperlipidemia, diabetes, and hypertension all had high odds ratios for stroke in children, which suggest a relationship between these conditions and pediatric stroke. However, ascertainment bias is a major concern with electronic health record-based studies. More focused study is needed into the role of these exposures into the pathogenesis of pediatric stroke.

Challenges in and Opportunities for Electronic Health Record-Based Data Analysis and Interpretation

Electronic health records (EHRs) have been increasingly adopted in clinical practices across the United States, providing a primary source of data for clinical research, particularly observational cohort studies. EHRs are a high-yield, low-maintenance source of longitudinal real-world data for large patient populations and provide a wealth of information and clinical contexts that are useful for clinical research and translation into practice. Despite these strengths, it is important to recognize the multiple limitations and challenges related to the use of EHR data in clinical research. Missing data are a major source of error and biases and can affect the representativeness of the cohort of interest, as well as the accuracy of the outcomes and exposures. Here, we aim to provide a critical understanding of the types of data available in EHRs and describe the impact of data heterogeneity, quality, and generalizability, which should be evaluated prior to and during the analysis of EHR data. We also identify challenges pertaining to data quality, including errors and biases, and examine potential sources of such biases and errors. Finally, we discuss approaches to mitigate and remediate these limitations. A proactive approach to addressing these issues can help ensure the integrity and quality of EHR data and the appropriateness of their use in clinical studies.

Making causal inferences from transactional data: A narrative review of opportunities and challenges when implementing the target trial framework

The target trial framework has emerged as a powerful tool for addressing causal questions in clinical practice and in public health. In the healthcare sector, where decision-making is increasingly data-driven, transactional databases, such as electronic health records (EHR) and insurance claims, present an untapped potential for answering complex causal questions. This narrative review explores the potential of the integration of the target trial framework with real-world data to enhance healthcare decision-making processes. We outline essential elements of the target trial framework, and identify pertinent challenges in data quality, privacy concerns, and methodological limitations, proposing solutions to overcome these obstacles and optimize the framework's application.

Biases in Electronic Health Records Data for Generating Real-World Evidence: An Overview

Electronic Health Records (EHR) are increasingly being perceived as a unique source of data for clinical research as they provide unprecedentedly large volumes of real-time data from real-world settings. In this review of the secondary uses of EHR, we identify the anticipated breadth of opportunities, pointing out the data deficiencies and potential biases that are likely to limit the search for true causal relationships. This paper provides a comprehensive overview of the types of biases that arise along the pathways that generate real-world evidence and the sources of these biases. We distinguish between two levels in the production of EHR data where biases are likely to arise: (i) at the healthcare system level, where the principal source of bias resides in access to, and provision of, medical care, and in the acquisition and documentation of medical and administrative data; and (ii) at the research level, where biases arise from the processes of extracting, analyzing, and interpreting these data. Due to the plethora of biases, mainly in the form of selection and information bias, we conclude with advising extreme caution about making causal inferences based on secondary uses of EHRs.

Artificial intelligence: revolutionizing cardiology with large language models

Natural language processing techniques are having an increasing impact on clinical care from patient, clinician, administrator, and research perspective. Among others are automated generation of clinical notes and discharge letters, medical term coding for billing, medical chatbots both for patients and clinicians, data enrichment in the identification of disease symptoms or diagnosis, cohort selection for clinical trial, and auditing purposes. In the review, an overview of the history in natural language processing techniques developed with brief technical background is presented. Subsequently, the review will discuss implementation strategies of natural language processing tools, thereby specifically focusing on large language models, and conclude with future opportunities in the application of such techniques in the field of cardiology.

Natural Language Processing for Adjudication of Heart Failure in a Multicenter Clinical Trial: A Secondary Analysis of a Randomized Clinical Trial

Importance

The gold standard for outcome adjudication in clinical trials is medical record review by a physician clinical events committee (CEC), which requires substantial time and expertise. Automated adjudication of medical records by natural language processing (NLP) may offer a more resource-efficient alternative but this approach has not been validated in a multicenter setting.

Objective

To externally validate the Community Care Cohort Project (C3PO) NLP model for heart failure (HF) hospitalization adjudication, which was previously developed and tested within one health care system, compared to gold-standard CEC adjudication in a multicenter clinical trial.

Design, Setting, and Participants

This was a retrospective analysis of the Influenza Vaccine to Effectively Stop Cardio Thoracic Events and Decompensated Heart Failure (INVESTED) trial, which compared 2 influenza vaccines in 5260 participants with cardiovascular disease at 157 sites in the US and Canada between September 2016 and January 2019. Analysis was performed from November 2022 to October 2023.

Exposures

Individual sites submitted medical records for each hospitalization. The central INVESTED CEC and the C3PO NLP model independently adjudicated whether the cause of hospitalization was HF using the prepared hospitalization dossier. The C3PO NLP model was fine-tuned (C3PO + INVESTED) and a de novo NLP model was trained using half the INVESTED hospitalizations.

Main Outcomes and Measures

Concordance between the C3PO NLP model HF adjudication and the gold-standard INVESTED CEC adjudication was measured by raw agreement, κ, sensitivity, and specificity. The fine-tuned and de novo INVESTED NLP models were evaluated in an internal validation cohort not used for training.

Results

Among 4060 hospitalizations in 1973 patients (mean [SD] age, 66.4 [13.2] years; 514 [27.4%] female and 1432 [72.6%] male]), 1074 hospitalizations (26%) were adjudicated as HF by the CEC. There was good agreement between the C3PO NLP and CEC HF adjudications (raw agreement, 87% [95% CI, 86-88]; κ, 0.69 [95% CI, 0.66-0.72]). C3PO NLP model sensitivity was 94% (95% CI, 92-95) and specificity was 84% (95% CI, 83-85). The fine-tuned C3PO and de novo NLP models demonstrated agreement of 93% (95% CI, 92-94) and κ of 0.82 (95% CI, 0.77-0.86) and 0.83 (95% CI, 0.79-0.87), respectively, vs the CEC. CEC reviewer interrater reproducibility was 94% (95% CI, 93-95; κ, 0.85 [95% CI, 0.80-0.89]).

Conclusions and Relevance

The C3PO NLP model developed within 1 health care system identified HF events with good agreement relative to the gold-standard CEC in an external multicenter clinical trial. Fine-tuning the model improved agreement and approximated human reproducibility. Further study is needed to determine whether NLP will improve the efficiency of future multicenter clinical trials by identifying clinical events at scale.

Deep learned representations of the resting 12-lead electrocardiogram to predict at peak exercise

AIMS

To leverage deep learning on the resting 12-lead electrocardiogram (ECG) to estimate peak oxygen consumption (V˙O2peak) without cardiopulmonary exercise testing (CPET).

METHODS AND RESULTS

V ˙ O 2 peak estimation models were developed in 1891 individuals undergoing CPET at Massachusetts General Hospital (age 45 ± 19 years, 38% female) and validated in a separate test set (MGH Test, n = 448) and external sample (BWH Test, n = 1076). Three penalized linear models were compared: (i) age, sex, and body mass index ('Basic'), (ii) Basic plus standard ECG measurements ('Basic + ECG Parameters'), and (iii) basic plus 320 deep learning-derived ECG variables instead of ECG measurements ('Deep ECG-V˙O2'). Associations between estimated V˙O2peak and incident disease were assessed using proportional hazards models within 84 718 primary care patients without CPET. Inference ECGs preceded CPET by 7 days (median, interquartile range 27-0 days). Among models, Deep ECG-V˙O2 was most accurate in MGH Test [r = 0.845, 95% confidence interval (CI) 0.817-0.870; mean absolute error (MAE) 5.84, 95% CI 5.39-6.29] and BWH Test (r = 0.552, 95% CI 0.509-0.592, MAE 6.49, 95% CI 6.21-6.67). Deep ECG-V˙O2 also outperformed the Wasserman, Jones, and FRIEND reference equations (P < 0.01 for comparisons of correlation). Performance was higher in BWH Test when individuals with heart failure (HF) were excluded (r = 0.628, 95% CI 0.567-0.682; MAE 5.97, 95% CI 5.57-6.37). Deep ECG-V˙O2 estimated V˙O2peak <14 mL/kg/min was associated with increased risks of incident atrial fibrillation [hazard ratio 1.36 (95% CI 1.21-1.54)], myocardial infarction [1.21 (1.02-1.45)], HF [1.67 (1.49-1.88)], and death [1.84 (1.68-2.03)].

CONCLUSION

Deep learning-enabled analysis of the resting 12-lead ECG can estimate exercise capacity (V˙O2peak) at scale to enable efficient cardiovascular risk stratification.

Characteristics and Attitudes of Wearable Device Users and Nonusers in a Large Health Care System

BACKGROUND

Consumer wearable devices with health and wellness features are increasingly common and may enhance disease detection and management. Yet studies informing relationships between wearable device use, attitudes toward device data, and comprehensive clinical profiles are lacking.

METHODS AND RESULTS

WATCH-IT (Wearable Activity Tracking for Comprehensive Healthcare-Integrated Technology) studied adults receiving longitudinal primary or ambulatory cardiovascular care in the Mass General Brigham health care system from January 2010 to July 2021. Participants completed a 20-question electronic survey about perceptions and use of consumer wearable devices, with responses linked to electronic health records. Multivariable logistic regression was used to identify factors associated with device use. Among 214 992 individuals receiving longitudinal primary or cardiovascular care with an active electronic portal, 11 121 responded (5.2%). Most respondents (55.8%) currently used a wearable device, and most nonusers (95.3%) would use a wearable if provided at no cost. Although most users (70.2%) had not shared device data with their doctor previously, most believed it would be very (20.4%) or moderately (34.4%) important to share device-related health information with providers. In multivariable models, older age (odds ratio [OR], 0.80 per 10-year increase [95% CI, 0.77-0.82]), male sex (OR, 0.87 [95% CI, 0.80-0.95]), and heart failure (OR, 0.75 [95% CI, 0.63-0.89]) were associated with lower odds of wearable device use, whereas higher median income (OR, 1.08 per 1-quartile increase [95% CI, 1.04-1.12]) and care in a cardiovascular medicine clinic (OR, 1.17 [95% CI, 1.05-1.30]) were associated with greater odds of device use.

CONCLUSIONS

Among patients in primary and cardiovascular medicine clinics, consumer wearable device use is common, and most users perceive value in wearable health data.

Deep Learning-Enabled Assessment of Left Heart Structure and Function Predicts Cardiovascular Outcomes

BACKGROUND

Deep learning interpretation of echocardiographic images may facilitate automated assessment of cardiac structure and function.

OBJECTIVES

We developed a deep learning model to interpret echocardiograms and examined the association of deep learning-derived echocardiographic measures with incident outcomes.

METHODS

We trained and validated a 3-dimensional convolutional neural network model for echocardiographic view classification and quantification of left atrial dimension, left ventricular wall thickness, chamber diameter, and ejection fraction. The training sample comprised 64,028 echocardiograms (n = 27,135) from a retrospective multi-institutional ambulatory cardiology electronic health record sample. Validation was performed in a separate longitudinal primary care sample and an external health care system data set. Cox models evaluated the association of model-derived left heart measures with incident outcomes.

RESULTS

Deep learning discriminated echocardiographic views (area under the receiver operating curve >0.97 for parasternal long axis, apical 4-chamber, and apical 2-chamber views vs human expert annotation) and quantified standard left heart measures (R2 range = 0.53 to 0.91 vs study report values). Model performance was similar in 2 external validation samples. Model-derived left heart measures predicted incident heart failure, atrial fibrillation, myocardial infarction, and death. A 1-SD lower model-left ventricular ejection fraction was associated with 43% greater risk of heart failure (HR: 1.43; 95% CI: 1.23-1.66) and 17% greater risk of death (HR: 1.17; 95% CI: 1.06-1.30). Similar results were observed for other model-derived left heart measures.

CONCLUSIONS

Deep learning echocardiographic interpretation accurately quantified standard measures of left heart structure and function, which in turn were associated with future clinical outcomes. Deep learning may enable automated echocardiogram interpretation and disease prediction at scale.

AORTA Gene: Polygenic prediction improves detection of thoracic aortic aneurysm

Background

Thoracic aortic disease is an important cause of morbidity and mortality in the US, and aortic diameter is a heritable contributor to risk. Could a polygenic prediction of ascending aortic diameter improve detection of aortic aneurysm?

Methods

Deep learning was used to measure ascending thoracic aortic diameter in 49,939 UK Biobank participants. A genome-wide association study (GWAS) was conducted in 39,524 participants and leveraged to build a 1.1 million-variant polygenic score with PRScs-auto. Aortic diameter prediction models were built with the polygenic score ("AORTA Gene") and without it. The models were tested in a held-out set of 4,962 UK Biobank participants and externally validated in 5,469 participants from Mass General Brigham Biobank (MGB), 1,298 from the Framingham Heart Study (FHS), and 610 participants from All of Us.

Results

In each test set, the AORTA Gene model explained more of the variance in thoracic aortic diameter compared to clinical factors alone: 39.9% (95% CI 37.8-42.0%) vs 29.2% (95% CI 27.1-31.4%) in UK Biobank, 36.5% (95% CI 34.4-38.5%) vs 32.5% (95% CI 30.4-34.5%) in MGB, 41.8% (95% CI 37.7-45.9%) vs 33.0% (95% CI 28.9-37.2%) in FHS, and 34.9% (95% CI 28.8-41.0%) vs 28.9% (95% CI 22.9-35.0%) in All of Us. AORTA Gene had a greater AUROC for identifying diameter ≥4cm in each test set: 0.834 vs 0.765 (P=7.3E-10) in UK Biobank, 0.808 vs 0.767 in MGB (P=4.5E-12), 0.856 vs 0.818 in FHS (P=8.5E-05), and 0.827 vs 0.791 (P=7.8E-03) in All of Us.

Conclusions

Genetic information improved estimation of thoracic aortic diameter when added to clinical risk factors. Larger and more diverse cohorts will be needed to develop more powerful and equitable scores.

Natural Language Processing for Adjudication of Heart Failure Hospitalizations in a Multi-Center Clinical Trial

Background

The gold standard for outcome adjudication in clinical trials is chart review by a physician clinical events committee (CEC), which requires substantial time and expertise. Automated adjudication by natural language processing (NLP) may offer a more resource-efficient alternative. We previously showed that the Community Care Cohort Project (C3PO) NLP model adjudicates heart failure (HF) hospitalizations accurately within one healthcare system.

Methods

This study externally validated the C3PO NLP model against CEC adjudication in the INVESTED trial. INVESTED compared influenza vaccination formulations in 5260 patients with cardiovascular disease at 157 North American sites. A central CEC adjudicated the cause of hospitalizations from medical records. We applied the C3PO NLP model to medical records from 4060 INVESTED hospitalizations and evaluated agreement between the NLP and final consensus CEC HF adjudications. We then fine-tuned the C3PO NLP model (C3PO+INVESTED) and trained a de novo model using half the INVESTED hospitalizations, and evaluated these models in the other half. NLP performance was benchmarked to CEC reviewer inter-rater reproducibility.

Results

1074 hospitalizations (26%) were adjudicated as HF by the CEC. There was high agreement between the C3PO NLP and CEC HF adjudications (agreement 87%, kappa statistic 0.69). C3PO NLP model sensitivity was 94% and specificity was 84%. The fine-tuned C3PO and de novo NLP models demonstrated agreement of 93% and kappa of 0.82 and 0.83, respectively. CEC reviewer inter-rater reproducibility was 94% (kappa 0.85).

Conclusion

Our NLP model developed within a single healthcare system accurately identified HF events relative to the gold-standard CEC in an external multi-center clinical trial. Fine-tuning the model improved agreement and approximated human reproducibility. NLP may improve the efficiency of future multi-center clinical trials by accurately identifying clinical events at scale.

Characteristics and Attitudes of Wearable Device Users and Non-Users in a Large Healthcare System

Introduction

Consumer wearable devices with health and wellness features are increasingly common and may enhance prevention and management of cardiovascular disease. However, the characteristics and attitudes of wearable device users versus non-users are poorly understood.

Methods

Wearable Activity Tracking for Comprehensive Healthcare-Integrated Technology (WATCH-IT) was a prospective study of adults aged ≥18 years receiving longitudinal primary or ambulatory cardiovascular care at one of eleven hospitals within the Mass General Brigham multi-institutional healthcare system between January 2010-July 2021. We invited patients, including wearable users and non-users, to participate via an electronic patient portal. Participants were asked to complete a 20-question survey regarding perceptions and use of consumer wearable devices. Responses were linked to electronic health record data. Multivariable logistic regression was used to identify factors associated with device use.

Results

Among 280,834 individuals receiving longitudinal primary or cardiovascular care, 65,842 did not have an active electronic portal or opted out of research contact. Of the 214,992 individuals sent a survey link, 11,121 responded (5.2%), comprising the WATCH-IT patient sample. Most respondents (55.8%) reported current use of a wearable device, and most non-users (95.3%) reported they would use a wearable device if provided at no cost. Although most users (70.2%) had not shared device data with their doctor previously, the majority believed it would be very (20.4%) or moderately (34.4%) important to share device-related health information with providers. In multivariable models, older age (odds ratio [OR] 0.80 per 10-year increase, 95% CI 0.77-0.82), male sex (0.87, 95% CI 0.80-0.95), and heart failure (0.75, 95% CI 0.63-0.89) were associated with lower odds of wearable device use, whereas higher median zip code income (1.08 per 1-quartile increase, 95% CI 1.04-1.12) and care in a cardiovascular medicine clinic (1.17, 95% CI 1.05-1.30) were associated with greater odds of device use. Nearly all respondents (98%) stated they would share device data with researchers studying health outcomes.

Conclusions

Within an electronically assembled cohort of patients in primary and cardiovascular medicine clinics with linkage to detailed health records, wearable device use is common. Most users perceive value in wearable data. Our platform may enable future study of the relationships between wearable technology and resource utilization, clinical outcomes, and health disparities.

Genetic Susceptibility to Atrial Fibrillation Identified via Deep Learning of 12-Lead Electrocardiograms

BACKGROUND

Artificial intelligence (AI) models applied to 12-lead ECG waveforms can predict atrial fibrillation (AF), a heritable and morbid arrhythmia. However, the factors forming the basis of risk predictions from AI models are usually not well understood. We hypothesized that there might be a genetic basis for an AI algorithm for predicting the 5-year risk of new-onset AF using 12-lead ECGs (ECG-AI)-based risk estimates.

METHODS

We applied a validated ECG-AI model for predicting incident AF to ECGs from 39 986 UK Biobank participants without AF. We then performed a genome-wide association study (GWAS) of the predicted AF risk and compared it with an AF GWAS and a GWAS of risk estimates from a clinical variable model.

RESULTS

In the ECG-AI GWAS, we identified 3 signals (P<5×10-8) at established AF susceptibility loci marked by the sarcomeric gene TTN and sodium channel genes SCN5A and SCN10A. We also identified 2 novel loci near the genes VGLL2 and EXT1. In contrast, the clinical variable model prediction GWAS indicated a different genetic profile. In genetic correlation analysis, the prediction from the ECG-AI model was estimated to have a higher correlation with AF than that from the clinical variable model.

CONCLUSIONS

Predicted AF risk from an ECG-AI model is influenced by genetic variation implicating sarcomeric, ion channel and body height pathways. ECG-AI models may identify individuals at risk for disease via specific biological pathways.

Leveraging electronic health record data for endometriosis research

Endometriosis is a chronic, complex disease for which there are vast disparities in diagnosis and treatment between sociodemographic groups. Clinical presentation of endometriosis can vary from asymptomatic disease-often identified during (in)fertility consultations-to dysmenorrhea and debilitating pelvic pain. Because of this complexity, delayed diagnosis (mean time to diagnosis is 1.7-3.6 years) and misdiagnosis is common. Early and accurate diagnosis of endometriosis remains a research priority for patient advocates and healthcare providers. Electronic health records (EHRs) have been widely adopted as a data source in biomedical research. However, they remain a largely untapped source of data for endometriosis research. EHRs capture diverse, real-world patient populations and care trajectories and can be used to learn patterns of underlying risk factors for endometriosis which, in turn, can be used to inform screening guidelines to help clinicians efficiently and effectively recognize and diagnose the disease in all patient populations reducing inequities in care. Here, we provide an overview of the advantages and limitations of using EHR data to study endometriosis. We describe the prevalence of endometriosis observed in diverse populations from multiple healthcare institutions, examples of variables that can be extracted from EHRs to enhance the accuracy of endometriosis prediction, and opportunities to leverage longitudinal EHR data to improve our understanding of long-term health consequences for all patients.

Information extraction from German radiological reports for general clinical text and language understanding

Recent advances in deep learning and natural language processing (NLP) have opened many new opportunities for automatic text understanding and text processing in the medical field. This is of great benefit as many clinical downstream tasks rely on information from unstructured clinical documents. However, for low-resource languages like German, the use of modern text processing applications that require a large amount of training data proves to be difficult, as only few data sets are available mainly due to legal restrictions. In this study, we present an information extraction framework that was initially pre-trained on real-world computed tomographic (CT) reports of head examinations, followed by domain adaptive fine-tuning on reports from different imaging examinations. We show that in the pre-training phase, the semantic and contextual meaning of one clinical reporting domain can be captured and effectively transferred to foreign clinical imaging examinations. Moreover, we introduce an active learning approach with an intrinsic strategic sampling method to generate highly informative training data with low human annotation cost. We see that the model performance can be significantly improved by an appropriate selection of the data to be annotated, without the need to train the model on a specific downstream task. With a general annotation scheme that can be used not only in the radiology field but also in a broader clinical setting, we contribute to a more consistent labeling and annotation process that also facilitates the verification and evaluation of language models in the German clinical setting.

Validation of a Deep Learning-Based Model to Predict Lung Cancer Risk Using Chest Radiographs and Electronic Medical Record Data

IMPORTANCE

Lung cancer screening with chest computed tomography (CT) prevents lung cancer death; however, fewer than 5% of eligible Americans are screened. CXR-LC, an open-source deep learning tool that estimates lung cancer risk from existing chest radiograph images and commonly available electronic medical record (EMR) data, may enable automated identification of high-risk patients as a step toward improving lung cancer screening participation.

OBJECTIVE

To validate CXR-LC using EMR data to identify individuals at high-risk for lung cancer to complement 2022 US Centers for Medicare & Medicaid Services (CMS) lung cancer screening eligibility guidelines.

DESIGN, SETTING, AND PARTICIPANTS

This prognostic study compared CXR-LC estimates with CMS screening guidelines using patient data from a large US hospital system. Included participants were persons who currently or formerly smoked cigarettes with an outpatient posterior-anterior chest radiograph between January 1, 2013, and December 31, 2014, with no history of lung cancer or screening CT. Data analysis was performed between May 2021 and June 2022.

EXPOSURES

CXR-LC lung cancer screening eligibility (previously defined as having a 3.297% or greater 12-year risk) based on inputs (chest radiograph image, age, sex, and whether currently smoking) extracted from the EMR.

MAIN OUTCOMES AND MEASURES

6-year incident lung cancer.

RESULTS

A total of 14 737 persons were included in the study population (mean [SD] age, 62.6 [6.8] years; 7154 [48.5%] male; 204 [1.4%] Asian, 1051 [7.3%] Black, 432 [2.9%] Hispanic, 12 330 [85.2%] White) with a 2.4% rate of incident lung cancer over 6 years (361 patients with cancer). CMS eligibility could be determined in 6277 patients (42.6%) using smoking pack-year and quit-date from the EMR. Patients eligible by both CXR-LC and 2022 CMS criteria had a high rate of lung cancer (83 of 974 patients [8.5%]), higher than those eligible by 2022 CMS criteria alone (5 of 177 patients [2.8%]; P < .001). Patients eligible by CXR-LC but not 2022 CMS criteria also had a high 6-year incidence of lung cancer (121 of 3703 [3.3%]). In the 8460 cases (57.4%) where CMS eligibility was unknown, CXR-LC eligible patients had a 5-fold higher rate of lung cancer than ineligible (127 of 5177 [2.5%] vs 18 of 2283 [0.5%]; P < .001). Similar results were found in subgroups, including female patients and Black persons.

CONCLUSIONS AND RELEVANCE

Using routine chest radiographs and other data automatically extracted from the EMR, CXR-LC identified high-risk individuals who may benefit from lung cancer screening CT.

Development of a Prediction Model for Ascending Aortic Diameter Among Asymptomatic Individuals

IMPORTANCE

Ascending thoracic aortic disease is an important cause of sudden death in the US, yet most aortic aneurysms are identified incidentally.

OBJECTIVE

To develop and validate a clinical score to estimate ascending aortic diameter.

DESIGN, SETTING, AND PARTICIPANTS

Using an ongoing magnetic resonance imaging substudy of the UK Biobank cohort study, which had enrolled participants from 2006 through 2010, score derivation was performed in 30 018 participants and internal validation in an additional 6681. External validation was performed in 1367 participants from the Framingham Heart Study (FHS) offspring cohort who had undergone computed tomography from 2002 through 2005, and in 50 768 individuals who had undergone transthoracic echocardiography in the Community Care Cohort Project, a retrospective hospital-based cohort of longitudinal primary care patients in the Mass General Brigham (MGB) network between 2001-2018.

EXPOSURES

Demographic and clinical variables (11 covariates that would not independently prompt thoracic imaging).

MAIN OUTCOMES AND MEASURES

Ascending aortic diameter was modeled with hierarchical group least absolute shrinkage and selection operator (LASSO) regression. Correlation between estimated and measured diameter and performance for identifying diameter 4.0 cm or greater were assessed.

RESULTS

The 30 018-participant training cohort (52% women), were a median age of 65.1 years (IQR, 58.6-70.6 years). The mean (SD) ascending aortic diameter was 3.04 (0.31) cm for women and 3.32 (0.34) cm for men. A score to estimate ascending aortic diameter explained 28.2% of the variance in aortic diameter in the UK Biobank validation cohort (95% CI, 26.4%-30.0%), 30.8% in the FHS cohort (95% CI, 26.8%-34.9%), and 32.6% in the MGB cohort (95% CI, 31.9%-33.2%). For detecting individuals with an ascending aortic diameter of 4 cm or greater, the score had an area under the receiver operator characteristic curve of 0.770 (95% CI, 0.737-0.803) in the UK Biobank, 0.813 (95% CI, 0.772-0.854) in the FHS, and 0.766 (95% CI, 0.757-0.774) in the MGB cohorts, although the model significantly overestimated or underestimated aortic diameter in external validation. Using a fixed-score threshold of 3.537, 9.7 people in UK Biobank, 1.8 in the FHS, and 4.6 in the MGB cohorts would need imaging to confirm 1 individual with an ascending aortic diameter of 4 cm or greater. The sensitivity at that threshold was 8.9% in the UK Biobank, 11.3% in the FHS, and 18.8% in the MGB cohorts, with specificities of 98.1%, 99.2%, and 96.2%, respectively.

CONCLUSIONS AND RELEVANCE

A prediction model based on common clinically available data was derived and validated to predict ascending aortic diameter. Further research is needed to optimize the prediction model and to determine whether its use is associated with improved outcomes.

Validation of a Natural Language Processing Algorithm for the Extraction of the Sleep Parameters from the Polysomnography Reports

Background: There is a need to better understand the association between sleep and chronic diseases. In this study we developed a natural language processing (NLP) algorithm to mine polysomnography (PSG) free-text notes from electronic medical records (EMR) and evaluated the performance. Methods: Using the Veterans Health Administration EMR, we identified 46,093 PSG studies using CPT code 95,810 from 1 October 2000−30 September 2019. We randomly selected 200 notes to compare the accuracy of the NLP algorithm in mining sleep parameters including total sleep time (TST), sleep efficiency (SE) and sleep onset latency (SOL), wake after sleep onset (WASO), and apnea-hypopnea index (AHI) compared to visual inspection by raters masked to the NLP output. Results: The NLP performance on the training phase was >0.90 for precision, recall, and F-1 score for TST, SOL, SE, WASO, and AHI. The NLP performance on the test phase was >0.90 for precision, recall, and F-1 score for TST, SOL, SE, WASO, and AHI. Conclusions: This study showed that NLP is an accurate technique to extract sleep parameters from PSG reports in the EMR. Thus, NLP can serve as an effective tool in large health care systems to evaluate and improve patient care.

One Clinician Is All You Need-Cardiac Magnetic Resonance Imaging Measurement Extraction: Deep Learning Algorithm Development

BACKGROUND

Cardiac magnetic resonance imaging (CMR) is a powerful diagnostic modality that provides detailed quantitative assessment of cardiac anatomy and function. Automated extraction of CMR measurements from clinical reports that are typically stored as unstructured text in electronic health record systems would facilitate their use in research. Existing machine learning approaches either rely on large quantities of expert annotation or require the development of engineered rules that are time-consuming and are specific to the setting in which they were developed.

OBJECTIVE

We hypothesize that the use of pretrained transformer-based language models may enable label-efficient numerical extraction from clinical text without the need for heuristics or large quantities of expert annotations. Here, we fine-tuned pretrained transformer-based language models on a small quantity of CMR annotations to extract 21 CMR measurements. We assessed the effect of clinical pretraining to reduce labeling needs and explored alternative representations of numerical inputs to improve performance.

METHODS

Our study sample comprised 99,252 patients that received longitudinal cardiology care in a multi-institutional health care system. There were 12,720 available CMR reports from 9280 patients. We adapted PRAnCER (Platform Enabling Rapid Annotation for Clinical Entity Recognition), an annotation tool for clinical text, to collect annotations from a study clinician on 370 reports. We experimented with 5 different representations of numerical quantities and several model weight initializations. We evaluated extraction performance using macroaveraged F1-scores across the measurements of interest. We applied the best-performing model to extract measurements from the remaining CMR reports in the study sample and evaluated established associations between selected extracted measures with clinical outcomes to demonstrate validity.

RESULTS

All combinations of weight initializations and numerical representations obtained excellent performance on the gold-standard test set, suggesting that transformer models fine-tuned on a small set of annotations can effectively extract numerical quantities. Our results further indicate that custom numerical representations did not appear to have a significant impact on extraction performance. The best-performing model achieved a macroaveraged F1-score of 0.957 across the evaluated CMR measurements (range 0.92 for the lowest-performing measure of left atrial anterior-posterior dimension to 1.0 for the highest-performing measures of left ventricular end systolic volume index and left ventricular end systolic diameter). Application of the best-performing model to the study cohort yielded 136,407 measurements from all available reports in the study sample. We observed expected associations between extracted left ventricular mass index, left ventricular ejection fraction, and right ventricular ejection fraction with clinical outcomes like atrial fibrillation, heart failure, and mortality.

CONCLUSIONS

This study demonstrated that a domain-agnostic pretrained transformer model is able to effectively extract quantitative clinical measurements from diagnostic reports with a relatively small number of gold-standard annotations. The proposed workflow may serve as a roadmap for other quantitative entity extraction.

Establishment of a Simple Pediatric Lower Respiratory Tract Infections Database Based on the Structured Electronic Medical Records

OBJECTIVE

This study aimed to establish a pediatric lower respiratory tract infections (PLRTIs) database based on the structured electronic medical records (SEMRs), to provide a brief overview and the usage process of the SEMRs and the database.

METHODS

All the medical information is recorded by a clinical information system developed by Eureka Systems Company. A plugin of the software was used to set the properties of items of the SEMR. Children with lower respiratory tract infections (LRTIs) who were admitted to the department of respiratory medicine of our hospital from May 2020 were included. PostgreSQL 13.1 software was used to construct the PLRTIs database.

RESULTS

Seven kinds of SEMRs were established, and the admission record was the most important and complex among them. It was mainly composed of 10 parts, i.e., basic information, history of present illness, past history (without respiratory disease), past history of respiratory diseases, personal history, family history, physical examination, the score of LRTIs, auxiliary examination, and diagnosis. With the three-level doctor ward round, the recorded information of the SEMR would be checked repeatedly, thus guaranteeing the correctness of the information. The data of the SEMR and laboratory tests could be extracted directly from the hospital information system (HIS) by PostgreSQL 13.1 software with the specific structured query language (SQL) code. After manually checking the original records, the datasets were imported into PostgreSQL 13.1 software, and a simple PLRTIs database was constructed. According to the inclusion criteria of this study, a total of 1,184 children with LRTIs were included in this database from 1 May 2020 to 30 April 2021.

CONCLUSION

A series of SEMRs for PLRTIs were designed and used during the hospitalization. A simple PLRTIs database was established based on the SEMR. The SEMRs will provide complete and high-quality data on LRTIs in children.