Evaluating AI Proficiency in Nuclear Cardiology: Large Language Models take on the Board Preparation Exam.

Background: Previous studies evaluated the ability of large language models (LLMs) in medical disciplines; however, few have focused on image analysis, and none specifically on cardiovascular imaging or nuclear cardiology. Objectives: This study assesses four LLMs - GPT-4, GPT-4 Turbo, GPT-4omni (GPT-4o) (Open AI), and Gemini (Google Inc.) - in responding to questions from the 2023 American Society of Nuclear Cardiology Board Preparation Exam, reflecting the scope of the Certification Board of Nuclear Cardiology (CBNC) examination. Methods: We used 168 questions: 141 text-only and 27 image-based, categorized into four sections mirroring the CBNC exam. Each LLM was presented with the same standardized prompt and applied to each section 30 times to account for stochasticity. Performance over six weeks was assessed for all models except GPT-4o. McNemar's test compared correct response proportions. Results: GPT-4, Gemini, GPT4-Turbo, and GPT-4o correctly answered median percentiles of 56.8% (95% confidence interval 55.4% - 58.0%), 40.5% (39.9% - 42.9%), 60.7% (59.9% - 61.3%) and 63.1% (62.5 - 64.3%) of questions, respectively. GPT4o significantly outperformed other models (p=0.007 vs. GPT-4Turbo, p<0.001 vs. GPT-4 and Gemini). GPT-4o excelled on text-only questions compared to GPT-4, Gemini, and GPT-4 Turbo (p<0.001, p<0.001, and p=0.001), while Gemini performed worse on image-based questions (p<0.001 for all). Conclusion: GPT-4o demonstrated superior performance among the four LLMs, achieving scores likely within or just outside the range required to pass a test akin to the CBNC examination. Although improvements in medical image interpretation are needed, GPT-4o shows potential to support physicians in answering text-based clinical questions.

Identification and correction for collider bias in a genome-wide association study of diabetes-related heart failure.

Type 2 diabetes (T2D) is a major risk factor for heart failure (HF) and has elevated incidence among individuals with HF. Since genetics and HF can independently influence T2D, collider bias may occur when T2D (i.e., collider) is controlled for by design or analysis. Thus, we conducted a genome-wide association study (GWAS) of diabetes-related HF with correction for collider bias. We first performed a GWAS of HF to identify genetic instrumental variables (GIVs) for HF and to enable bidirectional Mendelian randomization (MR) analysis between T2D and HF. We identified 61 genomic loci, significantly associated with all-cause HF in 114,275 individuals with HF and over 1.5 million controls of European ancestry. Using a two-sample bidirectional MR approach with 59 and 82 GIVs for HF and T2D, respectively, we estimated that T2D increased HF risk (odds ratio [OR] 1.07, 95% confidence interval [CI] 1.04-1.10), while HF also increased T2D risk (OR 1.60, 95% CI 1.36-1.88). Then we performed a GWAS of diabetes-related HF corrected for collider bias due to the study design of index cases. After removing the spurious association of TCF7L2 locus due to collider bias, we identified two genome-wide significant loci close to PITX2 (chromosome 4) and CDKN2B-AS1 (chromosome 9) associated with diabetes-related HF in the Million Veteran Program and replicated the associations in the UK Biobank. Our MR findings provide strong evidence that HF increases T2D risk. As a result, collider bias leads to spurious genetic associations of diabetes-related HF, which can be effectively corrected to identify true positive loci.

Adaptive selection at G6PD and disparities in diabetes complications.

Diabetes complications occur at higher rates in individuals of African ancestry. Glucose-6-phosphate dehydrogenase deficiency (G6PDdef), common in some African populations, confers malaria resistance, and reduces hemoglobin A1c (HbA1c) levels by shortening erythrocyte lifespan. In a combined-ancestry genome-wide association study of diabetic retinopathy, we identified nine loci including a G6PDdef causal variant, rs1050828 -T (Val98Met), which was also associated with increased risk of other diabetes complications. The effect of rs1050828 -T on retinopathy was fully mediated by glucose levels. In the years preceding diabetes diagnosis and insulin prescription, glucose levels were significantly higher and HbA1c significantly lower in those with versus without G6PDdef. In the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial, participants with G6PDdef had significantly higher hazards of incident retinopathy and neuropathy. At the same HbA1c levels, G6PDdef participants in both ACCORD and the Million Veteran Program had significantly increased risk of retinopathy. We estimate that 12% and 9% of diabetic retinopathy and neuropathy cases, respectively, in participants of African ancestry are due to this exposure. Across continentally defined ancestral populations, the differences in frequency of rs1050828 -T and other G6PDdef alleles contribute to disparities in diabetes complications. Diabetes management guided by glucose or potentially genotype-adjusted HbA1c levels could lead to more timely diagnoses and appropriate intensification of therapy, decreasing the risk of diabetes complications in patients with G6PDdef alleles.

Type 1 Diabetes Genetic Risk in 109,954 Veterans With Adult-Onset Diabetes: The Million Veteran Program (MVP).

OBJECTIVE: To characterize high type 1 diabetes (T1D) genetic risk in a population where type 2 diabetes (T2D) predominates. RESEARCH DESIGN AND METHODS: Characteristics typically associated with T1D were assessed in 109,594 Million Veteran Program participants with adult-onset diabetes, 2011-2021, who had T1D genetic risk scores (GRS) defined as low (0 to <45%), medium (45 to <90%), high (90 to <95%), or highest (≥95%). RESULTS: T1D characteristics increased progressively with higher genetic risk (P < 0.001 for trend). A GRS ≥90% was more common with diabetes diagnoses before age 40 years, but 95% of those participants were diagnosed at age ≥40 years, and their characteristics resembled those of individuals with T2D in mean age (64.3 years) and BMI (32.3 kg/m2). Compared with the low-risk group, the highest-risk group was more likely to have diabetic ketoacidosis (low GRS 0.9% vs. highest GRS 3.7%), hypoglycemia prompting emergency visits (3.7% vs. 5.8%), outpatient plasma glucose <50 mg/dL (7.5% vs. 13.4%), a shorter median time to start insulin (3.5 vs. 1.4 years), use of a T1D diagnostic code (16.3% vs. 28.1%), low C-peptide levels if tested (1.8% vs. 32.4%), and glutamic acid decarboxylase antibodies (6.9% vs. 45.2%), all P < 0.001. CONCLUSIONS: Characteristics associated with T1D were increased with higher genetic risk, and especially with the top 10% of risk. However, the age and BMI of those participants resemble those of people with T2D, and a substantial proportion did not have diagnostic testing or use of T1D diagnostic codes. T1D genetic screening could be used to aid identification of adult-onset T1D in settings in which T2D predominates.

The Association of Vaccination for Common Adult Infectious Diseases and Uptake of COVID-19 Vaccines among 5,006,851 Veterans, 20 December 2020-31 October 2021.

Disparities in vaccination coverage for coronavirus disease 2019 (COVID-19) in the United States (U.S.) are consistent barriers limiting our ability to control the spread of disease, particularly those by age and race/ethnicity. This study examines the association between previous vaccination for common adult infectious diseases and vaccination for SARS-CoV-2 among a cohort of veterans in the U.S. Sociodemographic and clinical data were utilized from three databases within the Veterans Health Administration included in the electronic health record. We examined the association of previous vaccination for common adult vaccinations through six separate multivariable logistic regression analyses, one for each previous vaccine exposure, adjusting for demographic and clinical variables. We also examined the association of receiving any one of the six common adult vaccinations and vaccination against SARS-CoV-2. Adjusted models indicate higher odds of vaccination for SARS-CoV-2 among those who received each of the previous vaccinations. Significant differences were also noted by race/ethnicity and age. Veterans who recorded receiving any one of the previous vaccinations for common adult infections had significantly greater odds of receiving any vaccination against SARS-CoV-2. Understanding veterans' previous vaccination status can assist researchers and clinicians in impacting the uptake of novel vaccines, such as vaccination against SARS-CoV-2.

Impact of Visit Volume on the Effectiveness of Electronic Tools to Improve Heart Failure Care.

BACKGROUND: Electronic health record (EHR) tools can improve prescribing of guideline-recommended therapies for heart failure with reduced ejection fraction (HFrEF), but their effectiveness may vary by physician workload. OBJECTIVES: This paper aims to assess whether physician workload modifies the effectiveness of EHR tools for HFrEF. METHODS: This was a prespecified subgroup analysis of the BETTER CARE-HF (Building Electronic Tools to Enhance and Reinforce Cardiovascular Recommendations for Heart Failure) cluster-randomized trial, which compared effectiveness of an alert vs message vs usual care on prescribing of mineralocorticoid antagonists (MRAs). The trial included adults with HFrEF seen in cardiology offices who were eligible for and not prescribed MRAs. Visit volume was defined at the cardiologist-level as number of visits per 6-month study period (high = upper tertile vs non-high = remaining). Analysis at the patient-level used likelihood ratio test for interaction with log-binomial models. RESULTS: Among 2,211 patients seen by 174 cardiologists, 932 (42.2%) were seen by high-volume cardiologists (median: 1,853; Q1-Q3: 1,637-2,225 visits/6 mo; and median: 10; Q1-Q3: 9-12 visits/half-day). MRA was prescribed to 5.5% in the high-volume vs 14.8% in the non-high-volume groups in the usual care arm, 10.3% vs 19.6% in the message arm, and 31.2% vs 28.2% in the alert arm, respectively. Visit volume modified treatment effect (P for interaction = 0.02) such that the alert was more effective in the high-volume group (relative risk: 5.16; 95% CI: 2.57-10.4) than the non-high-volume group (relative risk: 1.93; 95% CI: 1.29-2.90). CONCLUSIONS: An EHR-embedded alert increased prescribing by >5-fold among patients seen by high-volume cardiologists. Our findings support use of EHR alerts, especially in busy practice settings. (Building Electronic Tools to Enhance and Reinforce Cardiovascular Recommendations for Heart Failure [BETTER CARE-HF]; NCT05275920).

Association of type 2 diabetes mellitus with dementia-related and non-dementia-related mortality among postmenopausal women: A secondary competing risks analysis of the women's health initiative.

INTRODUCTION: Alzheimer's disease (AD) and AD-related dementias (ADRD) are leading causes of death among older adults in the United States. Efforts to understand risk factors for prevention are needed. METHODS: Participants (n = 146,166) enrolled in the Women's Health Initiative without AD at baseline were included. Diabetes status was ascertained from self-reported questionnaires and deaths attributed to AD/ADRD from hospital, autopsy, and death records. Competing risk regression models were used to estimate the cause-specific hazard ratios (HRs) and 95% confidence intervals (CIs) for the prospective association of type 2 diabetes mellitus (T2DM) with AD/ADRD and non-AD/ADRD mortality. RESULTS: There were 29,393 treated T2DM cases and 8628 AD/ADRD deaths during 21.6 (14.0-23.5) median (IQR) years of follow-up. Fully adjusted HRs (95% CIs) of the association with T2DM were 2.94 (2.76-3.12) for AD/ADRD and 2.65 (2.60-2.71) for the competing risk of non-AD/ADRD mortality. DISCUSSION: T2DM is associated with AD/ADRD and non-AD/ADRD mortality. HIGHLIGHTS: Type 2 diabetes mellitus is more strongly associated with Alzheimer's disease (AD)/AD and related dementias (ADRD) mortality compared to the competing risk of non-AD/ADRD mortality among postmenopausal women. This relationship was consistent for AD and ADRD, respectively. This association is strongest among participants without obesity or hypertension and with younger age at baseline, higher diet quality, higher physical activity, higher alcohol consumption, and older age at the time of diagnosis of type 2 diabetes mellitus.

Development of Portable Electronic Health Record Based Algorithms to Identify Individuals with Diabetic Retinopathy.

Objectives: To develop, validate and implement algorithms to identify diabetic retinopathy (DR) cases and controls from electronic health care records (EHR)s. Methods : We developed and validated EHR-based algorithms to identify DR cases and individuals with type I or II diabetes without DR (controls) in three independent EHR systems: Vanderbilt University Medical Center Synthetic Derivative (VUMC), the VA Northeast Ohio Healthcare System (VANEOHS), and Massachusetts General Brigham (MGB). Cases were required to meet one of three criteria: 1) two or more dates with any DR ICD-9/10 code documented in the EHR, or 2) at least one affirmative health-factor or EPIC code for DR along with an ICD9/10 code for DR on a different day, or 3) at least one ICD-9/10 code for any DR occurring within 24 hours of an ophthalmology exam. Criteria for controls included affirmative evidence for diabetes as well as an ophthalmology exam. Results: The algorithms, developed and evaluated in VUMC through manual chart review, resulted in a positive predictive value (PPV) of 0.93 for cases and negative predictive value (NPV) of 0.97 for controls. Implementation of algorithms yielded similar metrics in VANEOHS (PPV=0.94; NPV=0.86) and lower in MGB (PPV=0.84; NPV=0.76). In comparison, use of DR definition as implemented in Phenome-wide association study (PheWAS) in VUMC, yielded similar PPV (0.92) but substantially reduced NPV (0.48). Implementation of the algorithms to the Million Veteran Program identified over 62,000 DR cases with genetic data including 14,549 African Americans and 6,209 Hispanics with DR. Conclusions/Discussion: We demonstrate the robustness of the algorithms at three separate health-care centers, with a minimum PPV of 0.84 and substantially improved NPV than existing high-throughput methods. We strongly encourage independent validation and incorporation of features unique to each EHR to enhance algorithm performance for DR cases and controls.

Sleep Characteristics are Associated with Risk of Treated Diabetes Among Postmenopausal Women.

OBJECTIVE: The purpose of this study was to determine whether sleep characteristics are associated with incidence of treated diabetes in postmenopausal individuals. METHODS: Postmenopausal participants ages 50-79 years reported sleep duration, sleep-disordered breathing, or insomnia at baseline and again in a subsample 3 years later. The primary outcome was self-reported new diagnosis of diabetes treated with oral drugs or insulin at any time after baseline. Multivariable Cox proportional hazards models were used. RESULTS: In 135,964 participants followed for 18.1 (± 6.3) years, there was a nonlinear association between sleep duration and risk of treated diabetes. Participants sleeping ≤5 hours at baseline had a 21% increased risk of diabetes compared with those sleeping 7 hours (adjusted hazard ratio [aHR] 1.21; 95% confidence interval [CI], 1.00-1.47). Those who slept for ≥9 hours had a nonsignificant 6% increased risk of diabetes compared with those sleeping 7 hours (aHR 1.06; 95% CI, 0.97-1.16). Participants whose sleep duration had decreased at 3 years had a 9% (aHR 1.09; 95% CI, 1.02-1.16) higher risk of diabetes than participants with unchanged sleep duration. Participants who reported increased sleep duration at 3 years had a risk of diabetes (HR 1.01; 95% CI, 0.95-1.08) similar to those with no sleep duration change. Participants at high risk of sleep-disordered breathing at baseline had a 31% higher risk of diabetes than those without (aHR 1.31; 95% CI, 1.26-1.37). No association was found between self-reported insomnia score and diabetes risk. CONCLUSIONS: Sleep-disordered breathing and short or long sleep duration were associated with higher diabetes risk in a postmenopausal population.

Correction for Collider Bias in the Genome-wide Association Study of Diabetes-Related Heart Failure due to Bidirectional Relationship between Heart Failure and Type 2 Diabetes.

Aims: Type 2 diabetes (T2D) is a major risk factor for heart failure (HF) across demographic groups. On the other hand, metabolic impairment, including elevated T2D incidence is a hallmark of HF pathophysiology. We investigated the bidirectional relationship between T2D and HF, and identified genetic associations with diabetes-related HF after correction for potential collider bias. Methods: We performed a genome-wide association study (GWAS) of HF to identify genetic instrumental variables (GIVs) for HF, and to enable bidirectional Mendelian Randomization (MR) analysis between T2D and HF. Since genetics and HF can independently influence T2D, collider bias may occur when T2D (i.e., collider) is controlled for by design or analysis. Thus, we conducted GWAS of diabetes-related HF with correction for collider bias. Results: We first identified 61 genomic loci, including 24 novel loci, significantly associated with all-cause HF in 114,275 HF cases and over 1.5 million controls of European ancestry. Combined with the summary statistics of a T2D GWAS, we obtained 59 and 82 GIVs for HF and T2D, respectively. Using a two-sample bidirectional MR approach, we estimated that T2D increased HF risk (OR 1.07, 95% CI 1.04-1.10), while HF also increased T2D risk (OR 1.60, 95% CI 1.36-1.88). Then we performed a GWAS of diabetes-related HF corrected for collider bias due to prevalent HF affecting incidence of T2D. After removing the spurious association of TCF7L2 locus due to collider bias, we identified two genome-wide significant loci close to PITX2 (chromosome 4) and CDKN2B-AS1 (chromosome 9) associated with diabetes-related HF in the Million Veteran Program, and replicated the associations in the UK Biobank study. Conclusion: We identified novel HF-associated loci to enable bidirectional MR study of T2D and HF. Our MR findings support T2D as a HF risk factor and provide strong evidence that HF increases T2D risk. As a result, collider bias leads to spurious genetic associations of diabetes-related HF, which can be effectively corrected to identify true positive loci. Evaluation of collider bias should be a critical component when conducting GWAS of complex disease phenotypes such as diabetes-related cardiovascular complications.

Heterogeneous effects on type 2 diabetes and cardiovascular outcomes of genetic variants and traits associated with fasting insulin.

Hyperinsulinemia is a complex and heterogeneous phenotype that characterizes molecular alterations that precede the development of type 2 diabetes (T2D). It results from a complex combination of molecular processes, including insulin secretion and insulin sensitivity, that differ between individuals. To better understand the physiology of hyperinsulinemia and ultimately T2D, we implemented a genetic approach grouping fasting insulin (FI)-associated genetic variants based on their molecular and phenotypic similarities. We identified seven distinctive genetic clusters representing different physiologic mechanisms leading to rising FI levels, ranging from clusters of variants with effects on increased FI, but without increased risk of T2D (non-diabetogenic hyperinsulinemia), to clusters of variants that increase FI and T2D risk with demonstrated strong effects on body fat distribution, liver, lipid, and inflammatory processes (diabetogenic hyperinsulinemia). We generated cluster-specific polygenic scores in 1,104,258 individuals from five multi-ancestry cohorts to show that the clusters differed in associations with cardiometabolic traits. Among clusters characterized by non-diabetogenic hyperinsulinemia, there was both increased and decreased risk of coronary artery disease despite the non-increased risk of T2D. Similarly, the clusters characterized by diabetogenic hyperinsulinemia were associated with an increased risk of T2D, yet had differing risks of cardiovascular conditions, including coronary artery disease, myocardial infarction, and stroke. The strongest cluster-T2D associations were observed with the same direction of effect in non-Hispanic Black, Hispanic, non-Hispanic White, and non-Hispanic East Asian populations. These genetic clusters provide important insights into granular metabolic processes underlying the physiology of hyperinsulinemia, notably highlighting specific processes that decouple increasing FI levels from T2D and cardiovascular risk. Our findings suggest that increasing FI levels are not invariably associated with adverse cardiometabolic outcomes.

Space-time clustering of COVID-19 cases in the United States veteran population.

PURPOSE: To assess the distribution and clustering of coronavirus disease 2019 (COVID-19) testing and incidence over space and time, U.S. Department of Veteran's Affairs (VA) data were used to describe where and when veterans experienced highest proportions of test positivity. METHODS: Data for 6,342,455 veterans who utilized VA services between January 1, 2018, and September 30, 2021, were assessed for COVID-19 testing and test positivity. Testing and positivity proportions by county were mapped and focused-cluster tests identified significant clustering around VA facilities. Spatial cluster analysis also identified where and when veterans experienced highest proportions of test positivity. RESULTS: Within the veterans study population and our time window, 21.3% received at least one COVID-19 test, and 20.4% of those tested had at least one positive test. There was statistically significant clustering of testing around VA facilities, revealing regional variation in testing practices. Veterans experienced highest test positivity proportions between November 2020 and January 2021 in a cluster of states in the Midwest, compared to those who received testing outside of the identified cluster (RR: 3.45). CONCLUSIONS: Findings reflect broad regional trends in COVID-19 positivity which can inform VA policy and resource allocation. Additional analysis is needed to understand patterns during Delta and Omicron variant periods.

The Social, Demographic, and Clinical Predictors of COVID-19 Severity: a Model-based Analysis of United States Veterans.

PURPOSE: This study aims to identify the contributions of individual and community social determinants of health (SDOH), demographic, and clinical factors in COVID-19 disease severity through a model-based analysis. METHODS: This national cross-sectional study focused on hospitalization among those tested for COVID-19 and use of intensive care, analyzing data on 220,848 Veterans tested between February 20, 2020 and October 20, 2021. Multiple logistic regression models were constructed using backwards elimination. The predictive value of each model was assessed with a c-statistic. RESULTS: Those hospitalized were older, more likely to be male, of Black or Asian race, have an income less than $39,999, live in an urban residence, and have medical comorbidities. The strongest predictors for hospitalization included Gini inequality index, race, income, heart failure, chronic kidney disease (CKD), and chronic obstructive pulmonary disease (COPD). For intensive care, Asian race, rural residence, COPD, and CKD were the strongest predictors. C-statistics were c = 0.749 for hospitalization and c = 0.582 for ICU admission. CONCLUSIONS: A combination of clinical, demographic, individual and community SDOH factors predict COVID-19 hospitalization with good predictive ability and can inform risk stratification, discharge planning, and public health interventions. Racial disparities were not explained by social or clinical factors. Intensive care models had low discriminative power and may be better explained by other characteristics.

What interventions should we implement in England's mental health services? The mental health implementation network (MHIN) mixed-methods approach to rapid prioritisation.

Introduction: Setting mental health priorities helps researchers, policy makers, and service funders improve mental health services. In the context of a national mental health implementation programme in England, this study aims to identify implementable evidence-based interventions in key priority areas to improve mental health service delivery. Methods: A mixed-methods research design was used for a three step prioritisation approach involving systematic scoping reviews (additional manuscript under development), expert consultations and data triangulation. Groups with diverse expertise, including experts by experience, worked together to improve decision-making quality by promoting more inclusive and comprehensive discussions. A multi-criteria decision analysis (MCDA) model was used to combine participants' varied opinions, data and judgments about the data's relevance to the issues at hand during a decision conferencing workshop where the priorities were finalised. Results: The study identified mental health interventions in three mental health priority areas: mental health inequalities, child and adolescent mental health, comorbidities with a focus on integration of mental and physical health services and mental health and substance misuse problems. Key interventions in all the priority areas are outlined. The programme is putting some of these evidence-based interventions into action nationwide in each of these three priority mental health priority areas. Conclusion: We report an inclusive attempt to ensure that the list of mental health service priorities agrees with perceived needs on the ground and focuses on evidence-based interventions. Other fields of healthcare may also benefit from this methodological approach if they need to make rapid health-prioritisation decisions.

Genome-wide association analysis and Mendelian randomization proteomics identify drug targets for heart failure.

We conduct a large-scale meta-analysis of heart failure genome-wide association studies (GWAS) consisting of over 90,000 heart failure cases and more than 1 million control individuals of European ancestry to uncover novel genetic determinants for heart failure. Using the GWAS results and blood protein quantitative loci, we perform Mendelian randomization and colocalization analyses on human proteins to provide putative causal evidence for the role of druggable proteins in the genesis of heart failure. We identify 39 genome-wide significant heart failure risk variants, of which 18 are previously unreported. Using a combination of Mendelian randomization proteomics and genetic cis-only colocalization analyses, we identify 10 additional putatively causal genes for heart failure. Findings from GWAS and Mendelian randomization-proteomics identify seven (CAMK2D, PRKD1, PRKD3, MAPK3, TNFSF12, APOC3 and NAE1) proteins as potential targets for interventions to be used in primary prevention of heart failure.