OBJECTIVES:Determine differences in quantitative stenosis severity measurements for ultra-high-resolution (UHR) photon-counting detector (PCD) coronary CT angiography (cCTA) relative to energy-integrating detector (EID) cCTA in a large patient cohort. METHODS:Adult participants seen between November 2022 and March 2023 underwent a clinical dual-source EID-CT cCTA and a research dual-source PCD-CT cCTA on the same day. Percent diameter stenosis (PDS) was measured, and stenosis severity was assigned based on the PDS of the most severe lesion per patient to determine a coronary artery disease reporting and data system (CAD-RADS) score. Agreement between EID-CT and PCD-CT for PDS and CAD-RADS was determined. RESULTS:A total of 112 participants were enrolled, yielding 82 subjects with at least 1 stenosis at PCD-CT (mean age, 68.0 ± 10.8 years; 69.5% [57/82] males). A total of 177 paired stenoses were quantified. The percent decrease in mean PDS from PCD-CT (24.3%) to EID-CT (29.4%) was 17.3% (median difference in PDS: -5.0%, 95% CI: -5.7% to -4.0%). The assigned CAD-RADS score changed in 31/82 subjects. In 11/31 subjects, the most severe stenosis was missed at EID-CT due to partial volume averaging of a small calcification, yet PCD clearly identified stenosis at those locations. In 20/31 subjects, the CAD-RADS score decreased at PCD-CT due to the decreased calcium blooming resulting from the improved spatial resolution. CONCLUSIONS:UHR PCD cCTA decreases quantitative measures of stenosis severity when a stenosis is identified at EID, leading to decreases in CAD-RADS assignments. ADVANCES IN KNOWLEDGE:With its improved spatial resolution, PCD-CT identifies stenoses missed on EID-CT.

OBJECTIVE:To evaluate the documentation of menopause-related International Classification of Diseases-10 (ICD-10) codes in the electronic health record (EHR) among midlife women with moderate or greater menopause symptoms receiving primary care. METHODS:This cross-sectional study from the Hormones and Experiences of Aging (HERA) cohort included women aged 45-60 years receiving primary care at one of 4 Mayo Clinic sites who completed a one-time survey between March 1, 2021 and June 30, 2021. The survey captured demographic data, menopause symptoms using the Menopause Rating Scale (MRS), health care utilization, and treatment. Women with an MRS score ≥12 were included. The primary outcome was documentation of a menopause-related ICD-10 code in the EHR in the 12 months before survey completion. RESULTS:Of 5,254 women with completed surveys, 2,414 (49%) had an MRS score ≥12 and were included. Among these, 1,519 (63%) reported seeking care for their menopause symptoms in the past 12 months, but only 345 (23%) had a menopause-related ICD-10 code. Women with an ICD-10 code had higher MRS scores (18 [IQR: 14-22] vs 17 [IQR: 14-20]; P = 0.002) and were more likely to use systemic hormone therapy (HT; 26% vs 9%; P < 0.001), and vaginal HT (20% vs 6%; P < 0.001). CONCLUSIONS:Menopause-related ICD-10 diagnosis codes were under-documented in women with significant menopause symptom burden. Those with a code were more likely to report menopause treatment. These findings highlight a critical gap between symptom burden and diagnosis coding in the EHR, underscoring the need to improve identification and management of menopause symptoms.

OBJECTIVE:To evaluate the accuracy of patient self-reported menopause stage among midlife US women. PATIENTS AND METHODS:This cross-sectional analysis included women 40 to 65 years of age who presented to women's health clinics at one of three Mayo Clinic sites from December 2016 to September 2019. Patients self-reported their menopause stage (pre-, peri-, postmenopausal, or unsure) while clinician-determined stage (pre-, peri-, postmenopausal, or unknown) was assessed by menopause-trained specialists using a standardized form. Only women with both self- and clinician-reported menopause stages were included. RESULTS:A total of 3411 women (mean age 53.5±6.2 years) were included with the majority being White (91.1%), educated (67.2%, at least some college education) and married/partnered (84.1%). Overall, 323 (9.5%), 640 (18.8%), 1182 (34.7%), and 1266 (37.1%) women self-reported being pre-, peri, postmenopausal, or unsure, respectively. Patient-reported and clinician-determined menopause stages aligned in only 56.6% of cases (n=1930, kappa=0.38). Notably, more than one-third of women (n=1266, 37.1%) reported being unsure of their menopause stage, whereas just 389 (11.4%) were classified as unknown by their clinicians. Of those who self-reported being unsure of their stage, 958 (75.7%) were classified as postmenopausal by clinicians. In a subset of women from 45 to 55 years of age (n=1786), the agreement was 56.8% (n=1015, kappa=0.40), with the highest concordance observed in the postmenopausal stage (41.0%). CONCLUSION:These findings highlight significant discrepancies between patient-reported and clinician-determined menopause stages, underscoring a lack of awareness among women regarding their menopause stage. This gap may result in missed opportunities for timely and appropriate interventions for care.

Background:Spirometry can be performed in an office setting or remotely using portable spirometers. Although basic spirometry is used for diagnosis of obstructive lung disease, clinically relevant information such as restriction, hyperinflation, and air trapping require additional testing, such as body plethysmography, which is not as readily available. We hypothesize that spirometry data contains information that can allow estimation of static lung volumes in certain circumstances by leveraging machine learning techniques. Objective:The aim of the study was to develop artificial intelligence-based algorithms for estimating lung volumes and capacities using spirometry measures. Methods:This study obtained spirometry and lung volume measurements from the Mayo Clinic pulmonary function test database for patient visits between February 19, 2001, and December 16, 2022. Preprocessing was performed, and various machine learning algorithms were applied, including a generalized linear model with regularization, random forests, extremely randomized trees, gradient-boosted trees, and XGBoost for both classification and regression cohorts. Results:A total of 121,498 pulmonary function tests were used in this study, with 85,017 allotted for exploratory data analysis and model development (ie, training dataset) and 36,481 tests reserved for model evaluation (ie, testing dataset). The median age of the cohort was 64.7 years (IQR 18-119.6), with a balanced distribution between genders, consisting 48.2% (n=58,607) female and 51.8% (n=62,889) male patients. The classification models showed a robust performance overall, with relatively low root mean square error and mean absolute error values observed across all predicted lung volumes. Across all lung volume categories, the models demonstrated strong discriminatory capacity, as indicated by the high area under the receiver operating characteristic curve values ranging from 0.85 to 0.99 in the training set and 0.81 to 0.98 in the testing set. Conclusions:Overall, the models demonstrate robust performance across lung volume measurements, underscoring their potential utility in clinical practice for accurate diagnosis and prognosis of respiratory conditions, particularly in settings where access to body plethysmography or other lung volume measurement modalities is limited.

Objective:To evaluate if cardiac troponin values predict poor outcomes in COVID-19 patients across the range of patients of different sex and age. Methods:We examined high-sensitivity cardiac troponin T (hs-cTnT) levels in 1,050 severely ill hospitalized COVID-19 patients who had hs-cTnT data available and participated in the Expanded Access Program for convalescent plasma study during the first wave (April-August 2020) of the COVID-19 pandemic. Results:We observed a continuous relationship between hs-cTnT levels and mortality in hospitalized males and females with COVID-19. This finding was present regardless of sex or age. Conclusion:These data indicate the prognostic ability of hs-cTnT to predict mortality in hospitalized COVID-19 patients across all relevant patient groups. Clinical Trials registration number:NCT04338360.

CONTEXT:Higher gluconeogenesis (GNG) contributes to higher nocturnal endogenous glucose production (EGP) in type 2 diabetes (T2D). Studies using 13C magnetic resonance spectroscopy (MRS) have confirmed lower hepatic glycogen content in subjects with T2D than in subjects with no diabetes (ND). OBJECTIVE:We determined the role of glycogen loading (GL) vs nonglycogen loading (NGL) on the contribution of GNG to nocturnal EGP in T2D. METHODS:In total, 14 subjects with T2D and 15 matched subjects with ND were studied on 2 occasions, with GL (60% carbohydrate) vs NGL (40% carbohydrate) isocaloric meals for 3 days, in random order in the overnight state. [6,6-2H2] glucose was infused to measure EGP, deuterium labelled water was used to measure GNG, and 13C MRS scans were performed in fed and fasted states to measure hepatic glycogen content. RESULTS:Hepatic glycogen content and nocturnal EGP were higher (P < .05) in GL vs NGL in both cohorts. The % GNG to EGP averaged ∼50% in subjects with ND throughout the night after both meals. In contrast, % GNG to nocturnal EGP in T2D was lower with GL vs NGL and matched the pattern observed in subjects with ND with GL lowering overnight rates of GNG in subjects with T2D. CONCLUSION:Selective targeting of GNG at night with appropriate medications could reduce nocturnal and early morning fasting hyperglycemia and hepatic insulin resistance in people with T2D.

Despite concerns about potential side effects, based both on historical experience with plasma products and more recent concerns about contemporary use of plasma, COVID-19 convalescent plasma has been shown to be a very safe product. Research early in the COVID-19 pandemic documented-among the very large population of convalescent plasma recipients in the US Convalescent Plasma Study component of the FDA-authorized Expanded Access Program-that the overall risk profile was no different than that seen for fresh frozen plasma, a product used routinely in medical practice. The safety of CCP was further demonstrated using real-world evidence, pragmatic trials, and formal randomized trials. The rates of all serious adverse events were very low, an especially impressive finding in light of the fact that nearly all safety data came from the use of COVID-19 convalescent plasma in patients who were hospitalized, were older, and/or had significant co-morbid cardiopulmonary and metabolic disorders. The well-known complications of blood and plasma transfusions-transfusion-associated circulatory overload and transfusion-related acute lung injury-were found with no higher incidence than with standard use of blood and plasma, nor was there evidence for antibody-dependent enhancement or increased incidence of thromboembolic events. The comprehensive safety profile derived from studies enrolling hundreds of thousands of recipients of COVID-19 convalescent plasma across the world should allay safety fears about the rapid deployment of convalescent plasma in future pandemics.

PURPOSE:Identifying cardiovascular disease before conception and in early pregnancy can better inform obstetric cardiovascular care. Our main objective was to evaluate the diagnostic performance of artificial intelligence (AI)-enabled digital tools for detecting left ventricular systolic dysfunction (LVSD) among women of reproductive age. METHODS:In a pilot cross-sectional study, we enrolled an initial cohort of 100 consecutive women aged 18-49 years who had a primary care physician and a scheduled echocardiography at Mayo Clinic Florida (Jacksonville) (cohort 1). Twelve-lead electrocardiography (ECG) and digital stethoscope recordings (single-lead ECG + phonocardiography) were performed on the date of echocardiography. We used deep learning to generate prediction probabilities for LVSD (defined as left ventricular ejection fraction <50%) for the 12-lead ECG (AI-ECG) and stethoscope (AI-stethoscope) recordings. In a second cohort of 100 participants, we enrolled consecutive women seen in primary care to estimate the prevalence of positive AI screening results when deployed for routine use (cohort 2). RESULTS:The median age of participants was 38.6 years (quartile 1: 30.3 years, quartile 3: 45.5 years), and 71.9% identified as part of the non-Hispanic White population. Among cohort 1, 5% had LVSD. The AI-ECG had an area under the curve of 0.94, and the AI-stethoscope (maximum prediction across all chest locations) had an area under the curve of 0.98. Among cohort 2, the prevalence of a positive AI screen was 1% and 3.2% for AI-ECG and the AI-stethoscope, respectively. CONCLUSION:We found these AI tools to be effective for the detection of cardiomyopathy associated with LVSD among women of reproductive age. These tools could potentially be useful for preconception cardiovascular evaluations.

Neurofibrillary tangles are dynamic neuropathologic hallmarks of Alzheimer's disease with a hypothesized lifespan morphologically-defined by three maturity levels: pretangles, mature tangles, and ghost tangles. To better understand the progression of tangle pathophysiology, we characterized tangle maturity level predilection of 15 tau antibodies recognizing a broad range of linear, phosphorylation, conformational, and truncation epitopes in the hippocampus of 24 postmortem brains. We developed the tangle maturity scoring system to semi-quantitatively evaluate each tangle maturity level. Based on proportions of tangle maturity levels, we classified antibodies as "early" (mostly pretangles and mature tangles), "middling" (mature tangles with pretangles and ghost tangles), and "advanced" (mostly ghost tangles and mature tangles) tangle maturity markers. To summarize tangle maturity predilection, we developed the tangle maturity scale to integrate individual tangle maturity scores. Correlations showed stronger relationships between tangle maturity scale and subsector thickness for more advanced tangle maturity markers in CA1 and subiculum, whereas Braak tangle stage remained consistently correlated throughout markers of the tangle lifespan. To aid in scoring hippocampi, we used machine learning to recognize tangle maturity levels, which performed comparably to a domain expert and showed similar relationships by Spearman correlation. Pattern recognition software was used to assess tangle and neuritic tau burden separately, which generally correlated with Braak stage and neuronal counts. However, tangle-derived tau burden more consistently correlated with hippocampal subsector thickness. In conclusion, we developed manual and automated scoring systems to evaluate tangle maturity levels, demonstrating early 4R, phosphorylated, and oligomeric tau accumulation preceding more advanced 3R and truncated tau. Our study provides supportive evidence of disease-relevant ordering of tau posttranslational modifications in the brain, which may have implications for theragnostic development. These findings underscore the promise of computerized quantitative analyses (i.e., pathomics) for high-throughput feature extraction from whole-slide images to enhance our understanding of microscopically observed morphologic changes.

Objective: To assess the ability of humans to differentiate human-authored vs artificial intelligence (AI)-generated medical manuscripts. Methods: This is a prospective randomized survey study from October 1, 2023, to December 1, 2023, from a single academic center. Artificial intelligence-generated medical manuscripts were created using ChatGPT 3.5 and were evaluated alongside randomly selected human-authored manuscripts. Participants, who were blinded from manuscript selection and creation, were randomized to receive three manuscripts that were either human-authored or AI-generated and had to fill out a survey questionnaire after review regarding who authored the manuscript. The primary outcome was accuracy of human reviewers in differentiating manuscript authors. Secondary outcomes were to identify factors that influenced prediction accuracy. Results: Fifty-one physicians were included in the study, including 12 post-doctorates, 19 assistant professors, and 20 associate or full professors. The overall specificity of 55.6% (95% CI, 30.8% to 78.5%), sensitivity of 31.2% (95% CI,11.0% to 58.7%), positive predictive value of 38.5% (95% CI,13.9% to 68.4%) and negative predictive value of 47.6% (95% CI, 25.7% to 70.2%). A stratified analysis of human-authored manuscripts indicated that high-impact factor manuscripts were identified with higher accuracy than low-impact factor ones (P1/4.037). For individual-level data, neither academic rank nor prior manuscript review experience significantly predicted the accuracy. The frequency of AI interaction was a significant factor, with occasional (odds ratio [OR], 8.20; P1/4.016), fairly frequent (OR, 7.13; P1/4.033), and very frequent (OR, 8.36; P1/4.030) use associated with correct identification. Further analysis revealed no significant predictors among the papers' qualities. Conclusion: Generative AI such as ChatGPT could create medical manuscripts that could not be differentiated from human-authored manuscripts. (c) 2024 Mayo Foundation for Medical Education and Research. Published by Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies. Mayo Clin Proc. 2025;100(4):622-633

The SPEC-AI Nigeria trial (NCT05438576) was designed to use artificial intelligence (AI) to screen for pregnancy-related cardiomyopathy in the peripartum period. Using data from this study, we evaluated the utility of AI-predicted delta age (adjusted AI-predicted age − chronological age) as a surrogate for biological age. We included 1,187 pregnant and postpartum women enrolled between August 2022 and September 2023 with follow-up through May 2024. Standard 12-lead electrocardiograms (ECGs) were recorded at study entry to generate AI age predictions. Artificial intelligence-predicted age was adjusted using estimated reference ranges obtained from a community-dwelling cohort of 25,144 individuals with AI-ECG age estimated and documented chronological age. Logistic and Cox-proportional hazards regression were used to examine associations with comorbid cardiovascular conditions and maternal mortality, respectively. Adjusted AI-predicted age was significantly higher among women with any cardiovascular condition, peripartum cardiomyopathy, or who died within 18 months (7, 14, and 23 years older, respectively) compared to those without these conditions who had values similar to the normal reference ranges (adjusted AI-predicted age difference less than 1 year). Artificial intelligence-predicted delta age greater than the 75th percentile was associated with an odds ratio (OR) of 2.06 for any cardiovascular condition, OR of 4.98 for left ventricular systolic dysfunction, and a hazard ratio of 32.81 for all-cause mortality; all values of P<.001. Artificial intelligence-ECG-derived biological age appears to be a potentially useful measure of cardiovascular health status and risk among pregnant and postpartum women. However, its role in monitoring maternal health requires further exploration.

Nigeria has the highest reported incidence of peripartum cardiomyopathy worldwide. This open-label, pragmatic clinical trial randomized pregnant and postpartum women to usual care or artificial intelligence (AI)-guided screening to assess its impact on the diagnosis left ventricular systolic dysfunction (LVSD) in the perinatal period. The study intervention included digital stethoscope recordings with point of-care AI predictions and a 12-lead electrocardiogram with asynchronous AI predictions for LVSD. The primary end point was identification of LVSD during the study period. In the intervention arm, the primary end point was defined as the number of identified participants with LVSD as determined by a positive AI screen, confirmed by echocardiography. In the control arm, this was the number of participants with clinical recognition and documentation of LVSD on echocardiography in keeping with current standard of care. Participants in the intervention arm had a confirmatory echocardiogram at baseline for AI model validation. A total of 1,232 (616 in each arm) participants were randomized and 1,195 participants (587 intervention arm and 608 control arm) completed the baseline visit at 6 hospitals in Nigeria between August 2022 and September 2023 with follow-up through May 2024. Using the AI-enabled digital stethoscope, the primary study end point was met with detection of 24 out of 587 (4.1%) versus 12 out of 608 (2.0%) patients with LVSD (intervention versus control odds ratio 2.12, 95% CI 1.05–4.27; P = 0.032). With the 12-lead AI-electrocardiogram model, the primary end point was detected in 20 out of 587 (3.4%) versus 12 out of 608 (2.0%) patients (odds ratio 1.75, 95% CI 0.85–3.62; P = 0.125). A similar direction of effect was observed in prespecified subgroup analysis. There were no serious adverse events related to study participation. In pregnant and postpartum women, AI-guided screening using a digital stethoscope improved the diagnosis of pregnancy-related cardiomyopathy. ClinicalTrials.gov registration: NCT05438576 In this pragmatic, randomized clinical trial involving 1,196 pregnant and postpartum women from 6 hospitals in Nigeria, AI-based electrocardiogram screening proved accurate in detecting cardiomyopathies and suggests that it could improve detection of these conditions.

Nigeria is the most populous country in Africa with the highest gross domestic product (GDP) as of 2022. However, Nigeria is burdened by significant health challenges including an extremely high maternal mortality ratio, inadequate human resources, poor healthcare infrastructure, and population-level poverty rates as high as 40%. Nigeria also has the highest reported prevalence of peripartum cardiomyopathy worldwide which contributes to maternal mortality. Unfortunately, the diagnosis of peripartum cardiomyopathy is often delayed and mortality rates following diagnosis are extremely high (approximately 50%). Thus, there is a huge unmet need for simple, effective, and accessible solutions for cardiomyopathy detection in this population. To address maternal mortality through screening and early diagnosis, we designed and conducted a randomized controlled clinical trial (NCT05438576) of an artificial intelligence (AI) technology in Nigeria. The objective of the study was to evaluate the impact of AI-guided screening on cardiomyopathy detection in obstetric patients. The study findings showed AI-guided screening doubled the detection of cardiomyopathy (defined as left ventricular ejection fraction <50%) when compared to usual care with a number needed to screen of 47. As we explore next steps in relation to deploying this technology for clinical use in Nigeria, we sought to gather contextual information and broadly share lessons learned from the recently completed trial. To that end, we convened a round table discussion with all study site investigators aimed at identifying site-specific contextual challenges related to the development and conduct of the study. The SPEC-AI Nigeria study is the first published randomized controlled clinical trial of a health AI intervention in Nigeria. Insights gained from this study can inform future AI intervention studies in clinical care, guide the development of implementation strategies to ensure effective interventions are successfully incorporated into clinical care, and provide a roadmap for key stakeholders to consider when evaluating AI-technologies for use in low-resource settings.

BACKGROUND/OBJECTIVES:The clinical utility of body composition in predicting the severity of acute pancreatitis (AP) remains unclear. We aimed to measure body composition using artificial intelligence (AI) to predict severe AP in hospitalized patients. METHODS:We performed a retrospective study of patients hospitalized with AP at three tertiary care centers in 2018. Patients with computer tomography (CT) imaging of the abdomen at admission were included. A fully automated and validated abdominal segmentation algorithm was used for body composition analysis. The primary outcome was severe AP, defined as having persistent single- or multi-organ failure as per the revised Atlanta classification. RESULTS:352 patients were included. Severe AP occurred in 35 patients (9.9%). In multivariable analysis, adjusting for male sex and first episode of AP, intermuscular adipose tissue (IMAT) was associated with severe AP, OR = 1.06 per 5 cm2, p = 0.0207. Subcutaneous adipose tissue (SAT) area approached significance, OR = 1.05, p = 0.17. Neither visceral adipose tissue (VAT) nor skeletal muscle (SM) was associated with severe AP. In obese patients, a higher SM was associated with severe AP in unadjusted analysis (86.7 vs 75.1 and 70.3 cm2 in moderate and mild, respectively p = 0.009). CONCLUSION:In this multi-site retrospective study using AI to measure body composition, we found elevated IMAT to be associated with severe AP. Although SAT was non-significant for severe AP, it approached statistical significance. Neither VAT nor SM were significant. Further research in larger prospective studies may be beneficial.

Background: Effective pain management is crucial for both comfort and outcomes, yet predicting and managing this pain is difficult. This study aimed to analyze postoperative pain in patients undergoing hand surgery at the Mayo Clinic Florida, examining how patient characteristics and anxiety affect pain outcomes. Methods: We conducted a single-arm clinical trial at Mayo Clinic Florida, recruiting patients undergoing hand surgery. Preoperative pain and anxiety were assessed using the Pain Catastrophizing Scale (PCS) and State-Trait Anxiety Inventory (STAI). Postoperatively, patients used an iPhone app to record pain levels and medication use every four hours. Devices were collected three days after surgery. We analyzed the relationship between demographics, PCS, STAI scores, and pain levels using linear and logistic regression models. All statistical tests were two-sided with significance set at p < 0.05, analyzed with R4.2.2. Results: Data were collected from 62 patients (62.9% women, 37.1% men) undergoing hand surgery. Participants were mainly White (90.3%), with 50% being in the middle-aged adult group. Most had low anxiety levels (80.6% STAI-S, 82.3% STAI-T) and low catastrophizing (61.3% PCS). Postoperative pain was low, with median scores between 1.0 and 2.0 over three days. Demographics, anxiety, and catastrophizing were not significant predictors of pain levels. Logistic regression showed time as a significant factor, with pain levels peaking on Day 3. Conclusions: Postoperative pain after hand surgery was generally low, with time being a significant predictor of increased pain. Demographic factors, anxiety, and catastrophizing did not significantly affect pain levels. Pain management should emphasize time-sensitive interventions and ongoing monitoring.

Background: Pulmonary hypertension (PH) is a life-threatening disease affecting up to 1% of the global population. Diagnosis can be challenging and is often delayed due to the need for advanced imaging and invasive procedures. The use of artificial intelligence applied to ECGs (ECG-AI) has been shown to detect subtle patterns in voltage-time data and may be a valuable tool for the early detection of PH. Hypothesis and Purpose: To evaluate the performance of a previously trained, ECG-AI algorithm to detect PH (ECG-AI PH) using real-world data (RWD) collected in a multicenter, validation study. A joint primary hypothesis required sensitivity (Sn), specificity (Sp), positive predictive value (PPV), and negative predictive value (NPV) to exceed the null values of 76%, 70%, 10% and 90%. Study Design and Methods: This retrospective validation study was conducted at 5 geographically diverse U.S. health systems. Adult subjects were eligible for inclusion if they had a 12-lead ECG paired with an echocardiogram (Echo) in which tricuspid regurgitation velocity (TRV) was recorded, following presentation with dyspnea. Patients were classified according to echocardiographic criteria as either PH (PH+, TRV >3.4 m/s) or controls (PH-, TRV ≤2.8 m/s) to simulate the real-world use of ECG-AI, where a positive result could lead to a follow-up Echo. The study database was locked before processing the digital ECGs with ECG-AI PH. Performance was also estimated in a subset of subjects that later had a right heart catheterization using mPAP ≥ 20mmHg as the definition of PH. Results: A total of 14281 subjects (53% female, 63% aged 50+ years) met the inclusion criteria, including 3019 PH+ cases and 11262 PH- controls (Figure). Sn and Sp were 84.0% (95% CI: 82.6%, 85.3%) and 71.7% (95% CI: 70.9%, 72.6%), respectively. The positive and negative predictive values were 18.9% (95% CI: 18.4%, 19.4%) and 98.3% (95% CI: 98.1%, 98.4%), respectively, at 7.3% prevalence. Each endpoint met pre-defined performance criteria. In the subset of 1683 subjects with a follow up RHC, performance remained robust (Sn 85% (1155/1358); Sp 46% (151/325)). Conclusion: While ECG-AI PH was first developed as an investigational tool to detect PH, continued development as software as a medical device for clinical use demonstrated that the algorithm retained strong performance to detect PH in diverse, non-overlapping clinical settings and patient populations.

Background: Cardiomyopathy occurring during pregnancy or in the postpartum period is a leading case of maternal death in the US. The most commonly described form of pregnancy related cardiomyopathy (PRCM) associated with left ventricular systolic dysfunction is peripartum cardiomyopathy, defined as left ventricular ejection fraction (LVEF) <45% on echocardiography. PRCM symptoms often overlap with those of normal pregnancy and the lack of validated screening tools limits effective recognition, contributing to diagnostic delays. Recently, several artificial intelligence enabled electrocardiogram (AI-ECG) models have been developed and evaluated for PRCM detection. Aim&Hypothesis: We sought to investigate the pooled diagnostic performance of AI-ECG models for detecting PRCM. We hypothesized that AI-ECG models would be effective for detecting PRCM. Methods: A systematic search of PubMed, Scopus, and Cochrane was performed by combining search terms “artificial intelligence”, “machine learning”, “deep learning”, “algorithm”, “ECG”, “EKG”, “electrocardiogram”, “peripartum cardiomyopathy”, “pregnancy”, “pregnancy related cardiomyopathy” , to identify studies investigating the use of AI-ECG models for PRCM detection. Studies included those utilizing standard 12-lead ECGs as well as single lead ECGs from various devices. For inclusion in the meta-analysis, studies had to provide complete confusion matrix data for the AI-ECG model. Inverse variance random effects model meta-analysis was performed. Pooled performance estimates with corresponding 95% CIs are presented in forest plots and a summary receiver operating characteristics (sROC) curve. Results: Our search identified a total of 201 studies (PubMed: 64, Embase: 117, Cochrane: 20) from which 6 studies involving 11 AI-ECG models and 3,006 patients were included in the meta-analysis (Fig. 1) . Following random effect bivariate meta-analysis, the AI-ECG models yielded a pooled sensitivity of 0.721 (95% CI 0.676–0.762); specificity of 0.931 (95% CI 0.932–0.938); diagnostic odds ratio (DOR) of 44.658 (95% CI 23.843–83.644); and AUC of 0.896 (95% CI 0.87–0.91), for PRCM detection (Fig. 2A-C, 3) . Conclusion: AI-ECG models appear to be effective tools for detecting PRCM and may represent a viable solution to reduce diagnostic delays consequently improving maternal outcomes. Additional studies are needed to evaluate its impact on clinical outcomes in pragmatic settings.

Background: Critical congenital heart disease (CCHD) screening algorithms based on pulse oximetry have up to 27% false positive rate at high altitudes (>2500 m) despite altitude-specific cutoff changes. We examined the added value of an adult-based AI model applied to phonocardiography (AI-PCG) in the digital stethoscope EKO Core 500 in these settings. Research Question: What is the diagnostic performance of AI-PCG model compared to pulse oximetry in detecting neonatal CCHD across different altitudes in Latin America? Methods: This observational, prospective, case-control study included newborns born at different altitudes from 0 to 4380 meters in Peru, Mexico, Colombia and Bolivia. All underwent preductal and postductal oximetry at least 18 hours after birth, electrocardiography and PCG with the digital stethoscope. All CCHD cases were confirmed by echocardiography. Non-CCHD cases were defined by echocardiography or clinically if they were alive for 1 month and had 1) negative pulse oximetry, 2) no hospitalization due to cardiac or pulmonary causes and 3) no cyanosis or pneumonia. Those with abnormal pulse oximetry, genetic syndromes or murmur detection underwent echocardiography. Diagnostic performance metrics [area under the receiving operating characteristic curve (AUROC), sensitivity, specificity and false positive rate] were evaluated per modality. Results: A total of 1152 newborns were enrolled, 725 at <2500 m [13 (1.8%) with CCHD] and 427 at >2500 m [6 (1.4%) with CCHD]. Out of 19 CCHD cases, 17 had positive pulse oximetry and 9 had a positive AI-PCG model. One newborn with severe coarctation of the aorta and atrial septal defect had negative pulse oximetry and AI-PCG model. Pulse oximetry showed superior AUROC (0.932 vs 0.664), sensitivity (89.5% vs 47.4%), specificity (97% vs 85.5%) and lower false positive rate (3% vs 14.5%) compared to the AI-PCG model across altitudes. Diagnostic accuracy varied with altitude (Figure): at <2500 m, pulse oximetry showed higher AUROC (0.999 vs 0.612) and lower false positive rate (0.1% vs 16%) than AI-PCG model, but at >2500 m, pulse oximetry had similar AUROC (0.794 vs 0.774) and lower false positive rate (7.8% vs 11.9%) than the AI-PCG model. Conclusion: Pulse oximetry has superior diagnostic performance than an adult-based AI-PCG model when screening for neonatal CCHD at <2500 m, but similar performance is found >2500 m.