Performance of risk models to predict mortality risk for patients with heart failure: evaluation in an integrated health system.

BACKGROUND: Referral of patients with heart failure (HF) who are at high mortality risk for specialist evaluation is recommended. Yet, most tools for identifying such patients are difficult to implement in electronic health record (EHR) systems. OBJECTIVE: To assess the performance and ease of implementation of Machine learning Assessment of RisK and EaRly mortality in Heart Failure (MARKER-HF), a machine-learning model that uses structured data that is readily available in the EHR, and compare it with two commonly used risk scores: the Seattle Heart Failure Model (SHFM) and Meta-Analysis Global Group in Chronic (MAGGIC) Heart Failure Risk Score. DESIGN: Retrospective, cohort study. PARTICIPANTS: Data from 6764 adults with HF were abstracted from EHRs at a large integrated health system from 1/1/10 to 12/31/19. MAIN MEASURES: One-year survival from time of first cardiology or primary care visit was estimated using MARKER-HF, SHFM, and MAGGIC. Discrimination was measured by the area under the receiver operating curve (AUC). Calibration was assessed graphically. KEY RESULTS: Compared to MARKER-HF, both SHFM and MAGGIC required a considerably larger amount of data engineering and imputation to generate risk score estimates. MARKER-HF, SHFM, and MAGGIC exhibited similar discriminations with AUCs of 0.70 (0.69-0.73), 0.71 (0.69-0.72), and 0.71 (95% CI 0.70-0.73), respectively. All three scores showed good calibration across the full risk spectrum. CONCLUSIONS: These findings suggest that MARKER-HF, which uses readily available clinical and lab measurements in the EHR and required less imputation and data engineering than SHFM and MAGGIC, is an easier tool to identify high-risk patients in ambulatory clinics who could benefit from referral to a HF specialist.

Comparative Effectiveness of Aspirin Dosing in Cardiovascular Disease and Diabetes Mellitus: A Subgroup Analysis of the ADAPTABLE Trial.

OBJECTIVE: Patients with diabetes mellitus (DM) and concomitant atherosclerotic cardiovascular disease (ASCVD) must be on the most effective dose of aspirin to mitigate risk of future adverse cardiovascular events. RESEARCH DESIGN AND METHODS: ADAPTABLE, an open-label, pragmatic study, randomized patients with stable, chronic ASCVD to 81 mg or 325 mg of daily aspirin. The effects of aspirin dosing was assessed on the primary effectiveness outcome, a composite of all-cause death, hospitalization for myocardial infarction, or hospitalization for stroke, and the primary safety outcome of hospitalization for major bleeding. In this prespecified analysis, we used Cox proportional hazards models to compare aspirin dosing in patients with and without DM for the primary effectiveness and safety outcome. RESULTS: Of 15,076 patients, 5,676 (39%) had DM of whom 2,820 (49.7%) were assigned to 81 mg aspirin and 2,856 (50.3%) to 325 mg aspirin. Patients with versus without DM had higher rates of the composite cardiovascular outcome (9.6% vs. 5.9%; P < 0.001) and bleeding events (0.78% vs. 0.50%; P < 0.001). When comparing 81 mg vs. 325 mg of aspirin, patients with DM had no difference in the primary effectiveness outcome (9.3% vs. 10.0%; hazard ratio [HR] 0.98 [95% CI 0.83-1.16]; P = 0.265) or safety outcome (0.87% vs. 0.69%; subdistribution HR 1.25 [95% CI 0.72-2.16]; P = 0.772). CONCLUSIONS: This study confirms the inherently higher risk of patients with DM irrespective of aspirin dosing. Our findings suggest that a higher dose of aspirin yields no added clinical benefit, even in a more vulnerable population.

Comparative antimicrobial use in coronavirus disease 2019 (COVID-19) and non-COVID-19 inpatients from 2019 to 2020: A multicenter ecological study.

OBJECTIVE: We sought to determine whether increased antimicrobial use (AU) at the onset of the coronavirus disease 2019 (COVID-19) pandemic was driven by greater AU in COVID-19 patients only, or whether AU also increased in non-COVID-19 patients. DESIGN: In this retrospective observational ecological study from 2019 to 2020, we stratified inpatients by COVID-19 status and determined relative percentage differences in median monthly AU in COVID-19 patients versus non-COVID-19 patients during the COVID-19 period (March-December 2020) and the pre-COVID-19 period (March-December 2019). We also determined relative percentage differences in median monthly AU in non-COVID-19 patients during the COVID-19 period versus the pre-COVID-19 period. Statistical significance was assessed using Wilcoxon signed-rank tests. SETTING: The study was conducted in 3 acute-care hospitals in Chicago, Illinois. PATIENTS: Hospitalized patients. RESULTS: Facility-wide AU for broad-spectrum antibacterial agents predominantly used for hospital-onset infections was significantly greater in COVID-19 patients versus non-COVID-19 patients during the COVID-19 period (with relative increases of 73%, 66%, and 91% for hospitals A, B, and C, respectively), and during the pre-COVID-19 period (with relative increases of 52%, 64%, and 66% for hospitals A, B, and C, respectively). In contrast, facility-wide AU for all antibacterial agents was significantly lower in non-COVID-19 patients during the COVID-19 period versus the pre-COVID-19 period (with relative decreases of 8%, 7%, and 8% in hospitals A, B, and C, respectively). CONCLUSIONS: AU for broad-spectrum antimicrobials was greater in COVID-19 patients compared to non-COVID-19 patients at the onset of the pandemic. AU for all antibacterial agents in non-COVID-19 patients decreased in the COVID-19 period compared to the pre-COVID-19 period.

Health Status in Heart Failure and Cancer: Analysis of the Medicare Health Outcomes Survey 2016-2020.

BACKGROUND: People with heart failure (HF) and cancer experience impaired physical and mental health status. However, health-related quality of life (HRQOL) has not been directly compared between these conditions in a contemporary population of older people. OBJECTIVES: The authors sought to compare HRQOL in people with HF vs those with lung, colorectal, breast, and prostate cancers. METHODS: The authors performed a pooled analysis of Medicare Health Outcomes Survey data from 2016 to 2020 in participants ≥65 years of age with a self-reported history of HF or active treatment for lung, colon, breast, or prostate cancer. They used the Veterans RAND-12 physical component score (PCS) and mental component score (MCS), which range from 0-100 with a mean score of 50 (based on the U.S. general population) and an SD of 10. The authors used pairwise Student's t-tests to evaluate for differences in PCS and MCS between groups. RESULTS: Among participants with HF (n = 71,025; 54% female, 16% Black), mean PCS was 29.5 and mean MCS 47.9. Mean PCS was lower in people with HF compared with lung (31.2; n = 4,165), colorectal (35.6; n = 4,270), breast (37.7; n = 14,542), and prostate (39.6; n = 17,670) cancer (all P < 0.001). Participants with HF had a significantly lower mean MCS than those with lung (31.2), colon (50.0), breast (52.0), and prostate (53.0) cancer (all P < 0.001). CONCLUSIONS: People with HF experience worse HRQOL than those with cancer actively receiving treatment. The pervasiveness of low HRQOL in HF underscores the need to implement evidence-based interventions that target physical and mental health status and scale multidisciplinary clinics.

Synchronized wearables for the detection of haemodynamic states via electrocardiography and multispectral photoplethysmography.

Cardiovascular health is typically monitored by measuring blood pressure. Here we describe a wireless on-skin system consisting of synchronized sensors for chest electrocardiography and peripheral multispectral photoplethysmography for the continuous monitoring of metrics related to vascular resistance, cardiac output and blood-pressure regulation. We used data from the sensors to train a support-vector-machine model for the classification of haemodynamic states (resulting from exposure to heat or cold, physical exercise, breath holding, performing the Valsalva manoeuvre or from vasopressor administration during post-operative hypotension) that independently affect blood pressure, cardiac output and vascular resistance. The model classified the haemodynamic states on the basis of an unseen subset of sensor data for 10 healthy individuals, 20 patients with hypertension undergoing haemodynamic stimuli and 15 patients recovering from cardiac surgery, with an average precision of 0.878 and an overall area under the receiver operating characteristic curve of 0.958. The multinodal sensor system may provide clinically actionable insights into haemodynamic states for use in the management of cardiovascular disease.

Aspirin Dosing for Secondary Prevention of Atherosclerotic Cardiovascular Disease in Patients Treated With P2Y12 Inhibitors.

Background The ADAPTABLE (Aspirin Dosing: A Patient-Centric Trial Assessing Benefits and Long-Term Effectiveness) was a large, pragmatic, randomized controlled trial that found no difference between high- versus low-dose aspirin for secondary prevention of atherosclerotic cardiovascular disease. Whether concomitant P2Y12 inhibitor therapy modifies the effect of aspirin dose on clinical events remains unclear. Methods and Results Participants in ADAPTABLE were stratified according to baseline use of clopidogrel or prasugrel (P2Y12 group). The primary effectiveness end point was a composite of death, myocardial infarction, or stroke; and the primary safety end point was major bleeding requiring blood transfusions. We used multivariable Cox regression to compare the relative effectiveness and safety of aspirin dose within P2Y12 and non-P2Y12 groups. Of 13 815 (91.6%) participants with available data, 3051 (22.1%) were receiving clopidogrel (2849 [93.4%]) or prasugrel (203 [6.7%]) at baseline. P2Y12 inhibitor use was associated with higher risk of the primary effectiveness end point (10.86% versus 6.31%; adjusted hazard ratio [HR], 1.40 [95% CI, 1.22-1.62]) but was not associated with bleeding (0.95% versus 0.53%; adjusted HR, 1.42 [95% CI, 0.91-2.22]). We found no interaction in the relative effectiveness and safety of high- versus low-dose aspirin by P2Y12 inhibitor use. Overall, dose switching or discontinuation was more common in the high-dose compared with low-dose aspirin group, but the pattern was not modified by P2Y12 inhibitor use. Conclusions In this prespecified analysis of ADAPTABLE, we found that the relative effectiveness and safety of high- versus low-dose aspirin was not modified by baseline P2Y12 inhibitor use. Registration https://www.clinical.trials.gov. Unique identifier: NCT02697916.

An exploratory survey about using ChatGPT in education, healthcare, and research.

OBJECTIVE: ChatGPT is the first large language model (LLM) to reach a large, mainstream audience. Its rapid adoption and exploration by the population at large has sparked a wide range of discussions regarding its acceptable and optimal integration in different areas. In a hybrid (virtual and in-person) panel discussion event, we examined various perspectives regarding the use of ChatGPT in education, research, and healthcare. MATERIALS AND METHODS: We surveyed in-person and online attendees using an audience interaction platform (Slido). We quantitatively analyzed received responses on questions about the use of ChatGPT in various contexts. We compared pairwise categorical groups with a Fisher's Exact. Furthermore, we used qualitative methods to analyze and code discussions. RESULTS: We received 420 responses from an estimated 844 participants (response rate 49.7%). Only 40% of the audience had tried ChatGPT. More trainees had tried ChatGPT compared with faculty. Those who had used ChatGPT were more interested in using it in a wider range of contexts going forwards. Of the three discussed contexts, the greatest uncertainty was shown about using ChatGPT in education. Pros and cons were raised during discussion for the use of this technology in education, research, and healthcare. DISCUSSION: There was a range of perspectives around the uses of ChatGPT in education, research, and healthcare, with still much uncertainty around its acceptability and optimal uses. There were different perspectives from respondents of different roles (trainee vs faculty vs staff). More discussion is needed to explore perceptions around the use of LLMs such as ChatGPT in vital sectors such as education, healthcare and research. Given involved risks and unforeseen challenges, taking a thoughtful and measured approach in adoption would reduce the likelihood of harm.

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 for Cardiovascular Imaging: A Review.

Importance: Artificial intelligence (AI), driven by advances in deep learning (DL), has the potential to reshape the field of cardiovascular imaging (CVI). While DL for CVI is still in its infancy, research is accelerating to aid in the acquisition, processing, and/or interpretation of CVI across various modalities, with several commercial products already in clinical use. It is imperative that cardiovascular imagers are familiar with DL systems, including a basic understanding of how they work, their relative strengths compared with other automated systems, and possible pitfalls in their implementation. The goal of this article is to review the methodology and application of DL to CVI in a simple, digestible fashion toward demystifying this emerging technology. Observations: At its core, DL is simply the application of a series of tunable mathematical operations that translate input data into a desired output. Based on artificial neural networks that are inspired by the human nervous system, there are several types of DL architectures suited to different tasks; convolutional neural networks are particularly adept at extracting valuable information from CVI data. We survey some of the notable applications of DL to tasks across the spectrum of CVI modalities. We also discuss challenges in the development and implementation of DL systems, including avoiding overfitting, preventing systematic bias, improving explainability, and fostering a human-machine partnership. Finally, we conclude with a vision of the future of DL for CVI. Conclusions and Relevance: Deep learning has the potential to meaningfully affect the field of CVI. Rather than a threat, DL could be seen as a partner to cardiovascular imagers in reducing technical burden and improving efficiency and quality of care. High-quality prospective evidence is still needed to demonstrate how the benefits of DL CVI systems may outweigh the risks.

Clinical Inertia Among Outpatients With Heart Failure: Application of Treatment Nonintensification Taxonomy to EPIC-HF Trial.

BACKGROUND: The contribution of clinical inertia to suboptimal guideline-directed medical therapy (GDMT) for patients with heart failure with reduced ejection fraction (HFrEF) remains unclear. OBJECTIVES: This study examined reasons for GDMT nonintensification and characterized clinical inertia. METHODS: In this secondary analysis of EPIC-HF (Electronically Delivered, Patient-Activation Tool for Intensification of Medications for Chronic Heart Failure with Reduced Ejection Fraction), a randomized clinical trial evaluating a patient-activation tool on GDMT utilization, we performed a sequential, explanatory mixed-methods study. Reasons for nonintensification among 4 medication classes were assigned according to an expanded published taxonomy using structured chart reviews. Audio transcripts of clinic encounters were analyzed to further characterize nonintensification reasons. Integration occurred during the interpretation phase. RESULTS: Among 292 HFrEF patients who completed a cardiology visit, 185 (63.4%) experienced no treatment intensification, of whom 90 (48.6%) had at least 1 opportunity for intensification of a medication class with no documented contraindication or barriers (ie, clinical inertia). Nonintensification reasons varied by medication class, and included heightened risk of adverse effects (range 18.2%-31.6%), patient nonadherence (range 0.8%-1.1%), patient preferences and beliefs (range 0.6%-0.9%), comanagement with other providers (range 4.6%-5.6%), prioritization of other issues (range 15.6%-31.8%), multiple categories (range 16.5%-22.7%), and clinical inertia (range 22.7%-31.6%). A qualitative analysis of 32 clinic audio recordings demonstrated common characteristics of clinical inertia: 1) clinician review of medication regimens without education or intensification discussions; 2) patient stability as justification for nonintensification; and 3) shorter encounters for nonintensification vs intensification. CONCLUSIONS: In this comprehensive study exploring HFrEF prescribing, clinical inertia is a main contributor to nonintensification within an updated taxonomy classification for suboptimal GDMT prescribing. This approach should help target strategies overcoming GDMT underuse.

"I Just Wanted Nothing More Than to Get in a Real Shower": Patient Experience of the Inpatient Wait for a Heart Transplant.

BACKGROUND: Patients waiting for heart transplant may be hospitalized for weeks to months before undergoing transplantation. This high-stress period is further complicated by restrictions of daily privileges including diet, rooming, access to the outdoors, and hygiene (eg, limited in ability to shower). However, there is a paucity of research on the experience of this waiting period. We sought to describe the inpatient experience among patients awaiting heart transplantation and to better understand the needs of inpatients waiting for heart transplant. METHODS AND RESULTS: We conducted in-depth, semistructured phone interviews with a purposeful sample of patients who received a heart transplant in the past 10 years and waited in the hospital for at least 2 weeks before surgery. Using the prior literature, the lived experience of the lead author, and input from qualitative experts, we developed an interview guide. Interviews were recorded, transcribed, and analyzed in an iterative process until theoretical saturation was achieved. A 3-person coding team identified, discussed, and reconciled emergent themes. We conducted interviews with 15 patients. Overarching themes included food, hygiene, relationship with health care professionals, living environment, and stressors. Patients reported that strong bonds were formed between the patients and the staff, and the overwhelming majority only had positive comments about these relationships. However, many expressed negative comments about the experience of the food and limitations in personal hygiene. Other stressors included the unknown length of the waiting period, lack of communication about position on the transplant list, worry about family, and concerns that their life must be saved by the death of another. Many participants described that they would benefit from more interaction with recent heart transplant recipients. CONCLUSIONS: Hospitals and care units have the opportunity to make small changes that could greatly benefit the experience of waiting for a heart transplant, as well as the experience of hospitalization more generally.

Artificial intelligence-enabled tools in cardiovascular medicine: A survey of current use, perceptions, and challenges.

Background: Numerous artificial intelligence (AI)-enabled tools for cardiovascular diseases have been published, with a high impact on public health. However, few have been adopted into, or have meaningfully affected, routine clinical care. Objective: To evaluate current awareness, perceptions, and clinical use of AI-enabled digital health tools for patients with cardiovascular disease, and challenges to adoption. Methods: This mixed-methods study included interviews with 12 cardiologists and 8 health information technology (IT) administrators, and a follow-on survey of 90 cardiologists and 30 IT administrators. Results: We identified 5 major challenges: (1) limited knowledge, (2) insufficient usability, (3) cost constraints, (4) poor electronic health record interoperability, and (5) lack of trust. A minority of cardiologists were using AI tools; more were prepared to implement AI tools, but their sophistication level varied greatly. Conclusion: Most respondents believe in the potential of AI-enabled tools to improve care quality and efficiency, but they identified several fundamental barriers to wide-scale adoption.

Racial and Ethnic Differences in Cardiac Surveillance Evaluation of Patients Treated With Anthracycline-Based Chemotherapy.

Background Anthracyclines remain a key treatment for many malignancies but can increase the risk of heart failure or cardiomyopathy. Specific guidelines recommend echocardiography and serum cardiac biomarkers such as BNP (B-type natriuretic peptide) or NT-proBNP (N-terminal proBNP) evaluation before and 6 to 12 months after treatment. Our objective was to evaluate associations between racial and ethnic groups in cardiac surveillance of survivors of cancer after exposure to anthracyclines. Methods and Results Adult patients in the OneFlorida Consortium without prior cardiovascular disease who received at least 2 cycles of anthracyclines were included in the analysis. Multivariable logistic regression was performed to estimate the odds ratios (ORs) and 95% CIs for receiving cardiac surveillance at baseline before anthracycline therapy, 6 months after, and 12 months after anthracycline exposure among different racial and ethnic groups. Among the entire cohort of 5430 patients, 63.4% had a baseline echocardiogram, with 22.3% receiving an echocardiogram at 6 months and 25% at 12 months. Non-Hispanic Black (NHB) patients had a lower likelihood of receiving a baseline echocardiogram than Non-Hispanic White (NHW) patients (OR, 0.75 [95% CI, 0.63-0.88]; P=0.0006) or any baseline cardiac surveillance (OR, 0.76 [95% CI, 0.64-0.89]; P=0.001). Compared with NHW patients, Hispanic patients received significantly less cardiac surveillance at the 6-month (OR, 0.84 [95% CI, 0.72-0.98]; P=0.03) and 12-month (OR, 0.85 [95% CI, 0.74-0.98]; P=0.03) time points, respectively. Conclusions There were significant racial and ethnic differences in cardiac surveillance among survivors of cancer at baseline and following anthracycline-based treatment in NHB and Hispanic cohorts. Health care providers need to be cognizant of these social inequities and initiate efforts to ensure recommended cardiac surveillance occurs following anthracyclines.

The Emerging Role of Artificial Intelligence in Valvular Heart Disease.

Valvular heart disease (VHD) is a morbid condition in which timely identification and evidence-based treatments can lead to improved outcomes. Artificial intelligence broadly refers to the ability for computers to perform tasks and problem solve like the human mind. Studies applying AI to VHD have used a variety of structured (eg, sociodemographic, clinical) and unstructured (eg, electrocardiogram, phonocardiogram, and echocardiograms) and machine learning modeling approaches. Additional researches in diverse populations, including prospective clinical trials, are needed to evaluate the effectiveness and value of AI-enabled medical technologies in clinical care for patients with VHD.

Disruption in Blood Pressure Control With the COVID-19 Pandemic: The PCORnet Blood Pressure Control Laboratory.

OBJECTIVE: To explore trends in blood pressure (BP) control before and during the COVID-19 pandemic. PATIENTS AND METHODS: Health systems participating in the National Patient-Centered Clinical Research Network (PCORnet) Blood Pressure Control Laboratory Surveillance System responded to data queries, producing 9 BP control metrics. Averages of the BP control metrics (weighted by numbers of observations in each health system) were calculated and compared between two 1-year measurement periods (January 1, 2019, through December 31, 2019, and January 1, 2020, through December 31, 2020). RESULTS: Among 1,770,547 hypertensive persons in 2019, BP control to <140/<90 mm Hg varied across 24 health systems (range, 46%-74%). Reduced BP control occurred in most health systems with onset of the COVID-19 pandemic; the weighted average BP control was 60.5% in 2019 and 53.3% in 2020. Reductions were also evident for BP control to <130/<80 mm Hg (29.9% in 2019 and 25.4% in 2020) and improvement in BP (reduction of 10 mm Hg in systolic BP or achievement of systolic BP <140 mm Hg; 29.7% in 2019 and 23.8% in 2020). Two BP control process metrics exhibited pandemic-associated disruption: repeat visit in 4 weeks after a visit with uncontrolled hypertension (36.7% in 2019 and 31.7% in 2020) and prescription of fixed-dose combination medications among those with 2 or more drug classes (24.6% in 2019 and 21.5% in 2020). CONCLUSION: BP control decreased substantially during the COVID-19 pandemic, with a corresponding reduction in follow-up health care visits among persons with uncontrolled hypertension. It is unclear whether the observed decline in BP control during the pandemic will contribute to future cardiovascular events.

An exploratory survey about using ChatGPT in education, healthcare, and research.

Objective ChatGPT is the first large language model (LLM) to reach a large, mainstream audience. Its rapid adoption and exploration by the population at large has sparked a wide range of discussions regarding its acceptable and optimal integration in different areas. In a hybrid (virtual and in-person) panel discussion event, we examined various perspectives regarding the use of ChatGPT in education, research, and healthcare. Materials and Methods We surveyed in-person and online attendees using an audience interaction platform (Slido). We quantitatively analyzed received responses on questions about the use of ChatGPT in various contexts. We compared pairwise categorical groups with Fisher's Exact. Furthermore, we used qualitative methods to analyze and code discussions. Results We received 420 responses from an estimated 844 participants (response rate 49.7%). Only 40% of the audience had tried ChatGPT. More trainees had tried ChatGPT compared with faculty. Those who had used ChatGPT were more interested in using it in a wider range of contexts going forwards. Of the three discussed contexts, the greatest uncertainty was shown about using ChatGPT in education. Pros and cons were raised during discussion for the use of this technology in education, research, and healthcare. Discussion There was a range of perspectives around the uses of ChatGPT in education, research, and healthcare, with still much uncertainty around its acceptability and optimal uses. There were different perspectives from respondents of different roles (trainee vs faculty vs staff). More discussion is needed to explore perceptions around the use of LLMs such as ChatGPT in vital sectors such as education, healthcare and research. Given involved risks and unforeseen challenges, taking a thoughtful and measured approach in adoption would reduce the likelihood of harm.

Association Between End-Stage Liver Disease and Incident Heart Failure in an Integrated Health System.

BACKGROUND: End-stage liver disease (ESLD) and heart failure (HF) often coexist and are associated with significant morbidity and mortality. However, the true incidence of HF among patients with ESLD remains understudied. OBJECTIVE: This study aims to evaluate the association between ESLD and incident HF in a real-world clinical cohort. DESIGN AND PARTICIPANTS: A retrospective electronic health records database analysis of individuals with ESLD and frequency-matched controls without ESLD in a large integrated health system. MAIN MEASURES: The primary outcome was incident HF, which was defined by the International Classification of Disease codes and manually adjudicated by physician reviewers. The Kaplan-Meier method was used to estimate the cumulative incidence of HF. Multivariate proportional hazards models adjusted for shared metabolic factors (diabetes, hypertension, chronic kidney disease, coronary heart disease, body mass index) were used to compare the risk of HF in patients with and without ESLD. KEY RESULTS: Of 5004 patients (2502 with ESLD and 2502 without ESLD), the median (Q1-Q3) age was 57.0 (55.0-65.0) years, 59% were male, and 18% had diabetes. Over a median (Q1-Q3) follow-up of 2.3 (0.6-6.0) years, 121 incident HF cases occurred. Risk for incident HF was significantly higher for patients with ESLD compared with the non-ESLD group (adjusted HR: 4.67; 95% CI: 2.82-7.75; p < 0.001), with the majority of the ESLD group (70.7%) having HF with preserved ejection fraction (ejection fraction ≥ 50%). CONCLUSION: ESLD was significantly associated with a higher risk of incident HF, independent of shared metabolic risk factors, with the predominant phenotype being HF with preserved ejection fraction.

Seasonal variation in blood pressure control across US health systems.

OBJECTIVE: We aimed to characterize seasonal variation in US population-based blood pressure (BP) control and BP-related metrics and evaluate the association between outdoor temperature and BP control variation. METHODS: We queried electronic health records (EHRs) from 26 health systems, representing 21 states, to summarize BP metrics by quarters of 12-month periods from January 2017 to March 2020. Patients with at least one ambulatory visit during the measurement period and a hypertension diagnosis during the first 6 months or prior to the measurement period were included. Changes in BP control, BP improvement, medication intensification, average SBP reduction after medication intensification across quarters and association with outdoor temperature were analyzed using weighted generalized linear models with repeated measures. RESULTS: Among 1 818 041 people with hypertension, the majority were more than 65 years of age (52.2%), female (52.1%), white non-Hispanic (69.8%) and had stage 1/2 hypertension (64.8%). Overall, BP control and process metrics were highest in quarters 2 and 3, and lowest in quarters 1 and 4. Quarter 2 had the highest percentage of improved BP (31.95 ±â€Š0.90%) and average SBP reduction after medication intensification (16 ±â€Š0.23 mmHg). Quarter 3 had the highest percentage of BP controlled (62.25 ±â€Š2.55%) and lowest with medication intensification (9.73 ±â€Š0.60%). Results were largely consistent in adjusted models. Average temperature was associated with BP control metrics in unadjusted models, but associations were attenuated following adjustment. CONCLUSION: In this large, national, EHR-based study, BP control and BP-related process metrics improved during spring/summer months, but outdoor temperature was not associated with performance following adjustment for potential confounders.

Prediction of left ventricular ejection fraction changes in heart failure patients using machine learning and electronic health records: a multi-site study.

Left ventricular ejection fraction (EF) is a key measure in the diagnosis and treatment of heart failure (HF) and many patients experience changes in EF overtime. Large-scale analysis of longitudinal changes in EF using electronic health records (EHRs) is limited. In a multi-site retrospective study using EHR data from three academic medical centers, we investigated longitudinal changes in EF measurements in patients diagnosed with HF. We observed significant variations in baseline characteristics and longitudinal EF change behavior of the HF cohorts from a previous study that is based on HF registry data. Data gathered from this longitudinal study were used to develop multiple machine learning models to predict changes in ejection fraction measurements in HF patients. Across all three sites, we observed higher performance in predicting EF increase over a 1-year duration, with similarly higher performance predicting an EF increase of 30% from baseline compared to lower percentage increases. In predicting EF decrease we found moderate to high performance with low confidence for various models. Among various machine learning models, XGBoost was the best performing model for predicting EF changes. Across the three sites, the XGBoost model had an F1-score of 87.2, 89.9, and 88.6 and AUC of 0.83, 0.87, and 0.90 in predicting a 30% increase in EF, and had an F1-score of 95.0, 90.6, 90.1 and AUC of 0.54, 0.56, 0.68 in predicting a 30% decrease in EF. Among features that contribute to predicting EF changes, baseline ejection fraction measurement, age, gender, and heart diseases were found to be statistically significant.