Awasthi, Samir

Abstract Background Artificial intelligence (AI) models trained on 12-lead ECGs effectively detect left ventricular systolic dysfunction (LVSD; left ventricular ejection fraction [LVEF] <=40%). Continuous ECG monitoring via Holter recordings provides an opportunity for opportunistic screening for structural heart disease beyond rhythm disorders. We hypothesized that a lead-invariant version of the 12-lead AI model would enable a Holter monitor to screen for both arrhythmias and ventricular dysfunction. Methods We retrospectively analyzed continuous Holter ECGs from 17,665 patients who underwent a Holter and transthoracic echocardiogram (TTE) within 30 days of each other at Mayo Clinic. From each Holter, a random 20-minute of valid (non-flatline/lead disconnect) ECG segment was extracted and analyzed for LVSD detection using the adapted lead-invariant AI model. To evaluate stability, we examined model performance across different time points of the day, presenting results as area under the receiver operating characteristic curve (AUC) over time. Moreover, we illustrated the model’s robustness to noisy data by comparing its performance on raw ECG signals with that on bandpass-filtered inputs. Results Among 17,665 patients (mean age 59 years, 48.57% female), 4.96% had an LVEF <=40%. The AI model demonstrated strong predictive performance (20-minute segment AUC 0.90, mean prediction of 24-hour AUC 0.92). Analysis of results over time (Figure) revealed temporal patterns in predictive accuracy, with specific time periods showing greater stability. Despite modest variability, model performance remained consistently high throughout the day, confirming robustness across different physiological states. The predictions remained robust with noisy input. We did not observe performance improvement when the baseline wander and high frequency noise are removed by the bandpass filter. Conclusion Applying a 12-lead AI ECG model with a lead-invariant framework to a continuous Holter ECG enables effective screening for left ventricular dysfunction. This suggests that AI-based analysis of Holter-monitors can facilitate opportunistic screening of ventricular dysfunction and may enable assessment of an arrhythmia’s impact on LVEF, as well as the relationship between arrhythmia burden and LVEF.Figure 1.Mean prediction AUC of the day Figure 2.AUC for different time point

Background: We previously developed deep-learning algorithms that identify coronary artery disease (CAD) risk based on (i) coronary artery calcium (CAC), (ii) obstructive CAD by angiography, and (iii) left ventricular akinesis in ≥1 segment by echocardiogram, using a 12-lead electrocardiogram (CAD ECG-AI). We tested the hypothesis that those with increased probability of CAD by CAD ECG-AI algorithms will have an increased risk of atherosclerotic cardiovascular disease (ASCVD) events and would refine the AHA Predicting Risk of Cardiovascular Disease Events (PREVENT TM ) equations’ predictive capabilities. Methods: We assessed a group of consecutive patients who sought primary care in Olmsted County, MN, between 1997 and 2003. Passive follow-up was conducted using the Rochester Epidemiology Project's record linkage system. Patients included met the same criteria of the original PREVENT TM equations. The probability output of each CAD ECG-AI algorithm was used to predict ASCVD (fatal and non-fatal myocardial infarction and ischemic stroke) and ASCVD-Plus [further including percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG), and all-cause mortality]. Events were validated in duplicate. Cox proportional hazard models adjusted for age, sex, and risk factors while modeling each CAD ECG-AI algorithm as an individual and additive predictor (see footnote). PREVENT TM risk categories stratified analysis to evaluate the effect of the CAD ECG-AI models on predicted ASCVD risk. Results: We included 21,193 patients, with mean ± SD age 51.6 ± 12.0, 54% women and 95% white. After 14.2±12.4 years follow-up, 1,351 (6.4%) developed ASCVD, and 2,808 (13.2%) had ASCVD-Plus. The risk of ASCVD and ASCVD-Plus increased with positivity for the CAD ECG-AI algorithms after adjustments for age, sex, and PREVENT TM equations factors, all p for trend <0.001 (data not shown). In additive analyses, the risk of ASCVD and ASCVD-Plus increased with increased CAD ECG-AI factors, all p for trend <0.001 ( Fig. A-B ). Furthermore, the CAD ECG-AI algorithm enhanced the predictive capabilities of PREVENT TM equations across most risk subgroups ( Fig. C-E ). Conclusions: The deep-learning CAD ECG-AI algorithms displayed an independent and additive association with long-term ASCVD and ASCVD-Plus events in the community and improved the PREVENT TM predicted ASCVD risk. The CAD ECG-AI algorithms could help identify individuals at risk in primary prevention of cardiovascular events.

Abstract Study Objectives Narcolepsy has a complex phenotype owing to differences in symptomatology, disease severity, and comorbidities. This is the first study to use aggregate electronic health record (EHR) data and natural language processing (NLP) algorithms to characterize the demographics and comorbidities of a large cohort of patients with narcolepsy. Methods First-time Mayo Clinic patients (2000–2020) who had ≥1 narcolepsy-specific ICD-9/10 code and ≥1 disease-supportive statement in the clinical notes (identified using an NLP algorithm) were identified. A control cohort was propensity matched for birth year, age at first institutional encounter, sex, race, ethnicity, number of diagnosis codes, and mortality. Common comorbidities were compared and ranked between cohorts. Results In the EHR database (N = 6 389 186 patients), 2057 patients with narcolepsy were identified (median age, 32 years; 59.6% female; 92.6% white; and 89.2% non-Hispanic) and propensity matched with a control cohort. Among the top 20 comorbidities occurring more frequently in the narcolepsy cohort compared with the control cohort (odds ratio [OR], 1.67–3.94; p &lt; .001]) were sleep disorders (restless legs syndrome, obstructive sleep apnea, and insomnia), mood disorders (depression, dysthymia, and anxiety), and pain disorders (chronic pain syndrome, migraine, fibromyalgia, carpal tunnel syndrome, and myalgia). Other comorbidities significantly associated with narcolepsy (OR, 1.33–1.95) were irritable bowel syndrome (p &lt; .001), asthma (p &lt; .001), cervical spondylosis (p &lt; .01), syncope (p &lt; .01), and hypothyroidism (p &lt; .05). Conclusions This propensity-matched cohort study demonstrates increased psychiatric, sleep, and pain disorders in patients living with narcolepsy and challenges beyond narcolepsy-specific symptoms in this population. Understanding common narcolepsy-associated comorbidities may assist in tailoring treatment modalities.

Introduction: Artificial intelligence applied to electrocardiograms (ECG-AI) enables early and efficient detection of cardiovascular disease. The technical and clinical impact of different approaches to designing ECG-AI remains under-explored. We consider left ventricular hypertrophy (LVH) and examine the impact of disease-specific labeling of minority etiologies of left ventricular wall thickening- hypertrophic cardiomyopathy (HCM) and cardiac amyloidosis (CA) - on ECG-AI performance. Hypothesis: ECG-AI designed to detect HCM or CA outperforms ECG-AI broadly trained to identify LVH due to differences in attention to key ECG features associated with HCM or CA. Methods: Independent neural networks were trained to identify HCM, CA, and LVH. HCM and CA patients were identified using the full depth of EHR data and validated via clinician adjudication (N = 3,025 and 4,251 patients, respectively). Moderate and severe LVH patients (N = 98,144) were identified using echocardiogram results and international LVH criteria. We compared the ability of disease-specific ECG-AI vs. LVH ECG-AI to detect HCM (or CA) by examining model performance within each of the three test sets. Results: 52.6% of HCM patients and 49.5% of CA patients met LVH criteria. Within the HCM test set, the AUROC, sensitivity, specificity, and positive predictive value (PPV) of the HCM ECG-AI were 0.94, 0.88, 0.85, and 60%, respectively, whereas the same for LVH ECG-AI were 0.83, 0.83, 0.64, and 37%. Within the CA test set, CA ECG-AI achieved AUROC, sensitivity, specificity, and PPV of 0.94, 0.81, 0.91, and 69%, respectively, whereas the same for LVH ECG-AI were 0.75, 0.73, 0.68, and 31%. ECGs positive for HCM exhibited higher P wave area, QRS deflection and R wave amplitude compared to LVH-positive ECGs (Cohen's D > 0.75). ECGs positive for CA exhibited lower R wave amplitude/area, QRS deflection and QRS areas (Cohen's D > 0.6). Conclusion: Models trained using carefully labeled HCM or CA identify distinct signals associated with the diseases, resulting in improved disease discrimination compared to a model trained to identify echocardiographic left ventricular thickening. Labeling choices should be communicated transparently to clinicians to ensure safe and effective use of ECG-AI.

Background: Early diagnosis of heart failure with reduced ejection fraction (HFrEF) can lead to reduced healthcare costs and improved outcomes. In this study, we evaluated the economic effectiveness of an AI algorithm for detecting low ejection fraction using 12-lead ECG data. Aims: The primary aim was to estimate the health economic impact of the AI-ECG algorithm intervention to detect low ejection fraction. Methods: The study population consisted of 113,719 individuals without a prior history of HFrEF who visited a multistate health system between 2010-2018 with 12-lead ECG data and an echocardiogram within 1 year. The AI-ECG algorithm was run retrospectively, and projected costs were estimated assuming each patient incurred the average cost based on their age group and AI-detected heart failure stage. Hospitalization costs were estimated using Medicare reimbursement rates, and economic benefits from increased survival were estimated using a willingness to pay of $50,000 per quality-adjusted life year (QALY) gained. Results: At the optimal model threshold, the total cost savings per patient (including benefits from QALYs gained) for the AI-ECG algorithm vs. standard-of-care on the overall study population were projected to be $79.11 (95% CI: [$53.77, $104.44]) and $124.93 (95% CI: [$83.99, $165.88]) for the 2 and 4 year periods, respectively. For patients over 50 years old, the projected total cost savings were $85.56 (95% CI: [$56.35, $114.78]) and $135.28 (95% CI: [$88.18, $182.38]) per patient for the 2 and 4 year periods, respectively. See Figure. Conclusions: The AI-ECG algorithm to detect low ejection fraction is a cost-effective early screening tool for HFrEF.

Abstract Background Atherosclerotic cardiovascular disease (ASCVD), of which coronary artery disease (CAD) is the primary manifestation, is the leading cause of death worldwide. The AHA/ACC Pooled Cohort Equation (PCE) facilitates risk stratification and initiation of primary prevention of ASCVD, however their performance is suboptimal. Purpose We develop an artificial intelligence-based electrocardiogram (ECG) analysis (AI-ECG) algorithm to detect different stages of CAD and assessed the value of AI-ECG over PCE to predict incident myocardial infarction (MI) and survival over 3-years. Methods Using electronic health record data from approximately 7 million patients seen at over 70 hospitals and clinics across 5 states in the USA, we created independent AI-ECG models to identify elevated coronary artery calcium (CAC) by computer tomography, the presence of obstructive CAD by stress test or angiography, and left ventricular regional akinesis by echocardiogram, presumed to represent prior MI. Deep structured data, AI-enabled natural language processing of unstructured data were used to ensure cohort, and physician review of labeling were used to ensure cohort accuracy. These models were ensembled into a single model that was used to assess the utility of AI-ECG for predicting new occurrence MI in PCE- and age-stratified cohorts of patient with no prior history of ASCVD, over a 3-year timeframe using Cox’s proportional hazards analysis. Results A total of 6,080 patients (15,192 ECGs) were labeled positive for CAC ≥ 300, 12,561 patients (18,915 ECGs) positive for obstructive CAD, and 10,767 patients (23,870 ECGs) positive for regional akinesis. AI-ECG for identifying elevated CACS, obstructive CAD, and regional akinesis achieved AUROCs of 0.86, 0.80, and 0.93, respectively. All AI-ECG models identified increased 3-year risk for MI and death in AI-ECG positive versus AI-ECG negative patients and subgroups with 2 or 3 positive results had stepwise increased risk (Figure A). In a 3-year survival analysis of all-comer patients, the hazard ratio for those that tested positive on 1, 2, or 3 versus 0 positive disease-specific models were 1.61 (1.53-1.69). 2.40 (2.27-2.55), 3.42 (3.21-3.65), respectively. Risk stratification was also observed in cohorts without prior ASCVD stratified by low and high PCE-derived 10-year ASCVD risk (Figure B) and in cohorts stratified by age, including under the age of 40. Conclusions AI-ECG has potential to address unmet need for accessible risk stratification in patients in whom PCE under, over, or insufficiently estimate ASCVD risk, and in whom risk assessment over periods of time shorter than 10 years is desired.

INTRODUCTION: Screening for Barrett's esophagus (BE) is suggested in those with risk factors, but remains underutilized. BE/esophageal adenocarcinoma (EAC) risk prediction tools integrating multiple risk factors have been described. However, accuracy remains modest (area under the receiver-operating curve [AUROC] ≤0.7), and clinical implementation has been challenging. We aimed to develop machine learning (ML) BE/EAC risk prediction models from an electronic health record (EHR) database. METHODS: The Clinical Data Analytics Platform, a deidentified EHR database of 6 million Mayo Clinic patients, was used to predict BE and EAC risk. BE and EAC cases and controls were identified using International Classification of Diseases codes and augmented curation (natural language processing) techniques applied to clinical, endoscopy, laboratory, and pathology notes. Cases were propensity score matched to 5 independent randomly selected control groups. An ensemble transformer-based ML model architecture was used to develop predictive models. RESULTS: We identified 8,476 BE cases, 1,539 EAC cases, and 252,276 controls. The BE ML transformer model had an overall sensitivity, specificity, and AUROC of 76%, 76%, and 0.84, respectively. The EAC ML transformer model had an overall sensitivity, specificity, and AUROC of 84%, 70%, and 0.84, respectively. Predictors of BE and EAC included conventional risk factors and additional novel factors, such as coronary artery disease, serum triglycerides, and electrolytes. DISCUSSION: ML models developed on an EHR database can predict incident BE and EAC risk with improved accuracy compared with conventional risk factor-based risk scores. Such a model may enable effective implementation of a minimally invasive screening technology.

Abstract Introduction Narcolepsy is a chronic sleep disorder with multiple comorbid conditions. Sodium oxybate (SXB) is strongly recommended for treatment of narcolepsy. This study used aggregate electronic health record (EHR) data to characterize demographic characteristics and comorbidities of patients with narcolepsy treated with or without SXB. Methods An EHR-based search identified first-time Mayo Clinic patients between 1975–2020. Patients had ≥1 narcolepsy-specific ICD-9/-10 code and ≥1 diagnostic mention of narcolepsy in clinical notes (natural-language-processing algorithm). Patients with narcolepsy treated with SXB were age/sex matched with a cohort without SXB treatment. Common comorbidities were identified using ICD-9/-10 codes and compared between cohorts (odds ratio [OR]). Results Of 4387 patients identified, 8% received SXB treatment (n=351; mean [IQR] age at first diagnosis code observed at Mayo Clinic, 32 [23.2-46.1] y; 65.5% female; 92.3% white); 4036 patients had no SXB treatment (mean [IQR] age at first diagnosis code observed at Mayo Clinic, 44.8 [29.8-59.0] y; 58.0% female; 88.9% white). The 10 most commonly reported comorbidities (overall population) were insomnia (46.6%), fatigue (46.3%), depression (42.2%), hypertension (36.2%), hyperlipidemia (34.2%), obstructive sleep apnea (32.3%), diabetes mellitus (31.3%), arrhythmia (27.3%), idiopathic hypersomnia (IH; 26.8%), and coronary artery disease (17.6%). A cohort of 351 patients without SXB were age/sex matched to patients with SXB for comparison of comorbidities. In the unadjusted analysis, P values were significant for differences between cohorts (OR [95% CI] SXB vs no SXB) for fatigue (0.72 [0.54-0.97]; P&lt; 0.05; adjusted P&gt;0.9) and IH (0.60 [0.43-0.84]; P&lt; 0.01; adjusted P=0.29). After P value adjustment (Bonferroni correction), there were no significant differences between these cohorts in ORs for any comorbidity. The matched cohorts, which were younger than the overall population, had numerically lower rates of these diagnoses vs the overall population, except fatigue and IH in patients without SXB and depression in patients with SXB. Conclusion Among age-/gender-matched cohorts of patients with/ without SXB, there were no significant differences in comorbidities. Prevalence of comorbid IH/narcolepsy diagnoses highlight the diagnostic challenge in differentiating IH from narcolepsy type 2. SXB is highly effective but used by &lt; 1:10 patients with narcolepsy in the Mayo Clinic health system. Support (if any) Avadel Pharmaceuticals

Abstract Introduction Sodium oxybate (SXB) is strongly recommended for treatment of narcolepsy. Patients treated with available immediate-release oxybates are required to awaken for a second dose 2.5–4 hours after the bedtime dose to cover a full night of sleep. This study characterized SXB treatment patterns and discontinuation in narcolepsy patients. Methods Mayo Clinic patients from 1975–2020 were identified using an electronic health record–based search. Patients with ≥1 narcolepsy-specific ICD-9/-10 code and ≥1 diagnostic mention of narcolepsy in clinical notes (identified using a natural-language-processing [NLP] algorithm) were included. NLP was used to identify SXB non-use events with manual chart reviews performed to characterize reasons for discontinuation, switching, or missing the second dose and effects of missing the second dose. Results Of the patients with narcolepsy prescribed SXB (n=351; mean age at first diagnosis code observed at Mayo Clinic, 32 y [IQR: 23.2–46.1]; 65.5% female; 92.3% white), 113 (32.2%) had clinical notes indicating discontinuation of SXB, with 71 (20.2%) including reasons for discontinuation. The most common reasons for discontinuation (n≥5) included lack of efficacy (n=11), side effects (n=10), gastrointestinal side effects (n=10), neurological side effects (n=9), psychiatric side effects (n=8), lack of access (insurance/cost, n=10), and resolution/absence of cataplexy (n=5). Mentions of switching from SXB to other drugs (n=12 patients) included switching to stimulants (mixed amphetamine salts, n=3; dextroamphetamine, n=2; methylphenidate, n=1), sedative hypnotics (zolpidem, n=2), antidepressants (mirtazapine, n=1), and wake-promoting agents (solriamfetol, n=1). Of the 38 patients (10.8%) with mentions of missing the second nightly SXB dose, reasons were recorded for 24 (63.2%), including inability to wake up (n=14), forgetting the second dose (n=3), and having to wake up early the following day (n=3). Consequences of missing the second dose were recorded for a small subset (increased cataplexy, n=4; lower daytime alertness, n=3; awakens earlier, n=1; tiredness, n=1). Conclusion Limited data are available regarding real-world use of SXB. This novel study used a combination of NLP algorithm and manual chart review to identify difficulties patients have with the second, middle-of-the night immediate-release oxybate dose and patient-experienced consequences of not adhering to the prescribed dosing regimen. Support (if any) Avadel Pharmaceuticals

Abstract Introduction Narcolepsy is a rare, chronic sleep disorder whose symptoms detrimentally impact quality of life. Narcolepsy has a complex phenotype associated with multiple comorbid conditions. This is the first study to use aggregate electronic health record (EHR) data to characterize the demographics and comorbidities of patients with narcolepsy. Methods An EHR-based search identified first-time Mayo Clinic patients between 2000–2020. Included patients had ≥1 narcolepsy-specific ICD-9/10 code and ≥1 disease-supportive statement in the clinical notes (identified using a natural language processing algorithm). A control cohort was propensity-matched for birth year, age at first institutional encounter, sex, race, ethnicity, number of diagnosis codes, and mortality. Common comorbidities were compared and ranked between cohorts by odds ratio (OR); P values were adjusted and calculated based on Bonferroni correction. Results In the EHR database (N=6,389,186 patients), 2057 patients with narcolepsy were identified (median age at first presentation to Mayo clinic, 32 y [range, 17–48]; 59.6% female; 92.6% white; 89.2% non-Hispanic) and propensity-matched with a control cohort of 2057 patients (median age at first presentation to Mayo clinic, 35 y [range, 12–52]; 58.9% female; 94.6% white; 84.5% non-Hispanic). Among the top comorbidities that occurred more frequently in the narcolepsy cohort compared to the control cohort (OR [95% CI]; P&lt; 0.001]) were sleep disorders (restless leg syndrome, 3.94 [3.09–5.02]; obstructive sleep apnea, 3.27 [2.83–3.79]; insomnia, 1.84 [1.57–2.17]); mood disorders (depression, 2.11 [1.86-2.40]; dysthymia, 1.86 [1.54–2.25]; anxiety, 1.67 [1.46–1.89]); and pain disorders (chronic pain syndrome, 2.20 [1.76–2.76]; migraine, 1.96 [1.66–2.31]; fibromyalgia, 1.90 [1.61–2.25]; carpal tunnel syndrome, 1.80 [1.46–2.22]; myalgia, 1.69 [1.45–1.97]). Other comorbidities statistically significantly associated with narcolepsy were (OR range, 1.33–1.95) irritable bowel syndrome (P&lt; 0.001), asthma (P&lt; 0.001), cervical spondylosis (P&lt; 0.01), syncope (P&lt; 0.01), and hypothyroidism (P&lt; 0.05). Conclusion This propensity-matched cohort study affirmed prior studies of increased psychiatric and sleep disorders in patients with narcolepsy. Narcolepsy patients were twice as likely to experience chronic pain syndrome compared to the matched control group. Understanding common narcolepsy comorbidities may help optimize treatment efficacy and increase understanding of the medical/psychiatric challenges of narcolepsy patients. Support (if any) Avadel Pharmaceuticals