Stratifying heart failure patients with graph neural network and transformer using Electronic Health Records to optimize drug response prediction.

OBJECTIVES: Heart failure (HF) impacts millions of patients worldwide, yet the variability in treatment responses remains a major challenge for healthcare professionals. The current treatment strategies, largely derived from population based evidence, often fail to consider the unique characteristics of individual patients, resulting in suboptimal outcomes. This study aims to develop computational models that are patient-specific in predicting treatment outcomes, by utilizing a large Electronic Health Records (EHR) database. The goal is to improve drug response predictions by identifying specific HF patient subgroups that are likely to benefit from existing HF medications. MATERIALS AND METHODS: A novel, graph-based model capable of predicting treatment responses, combining Graph Neural Network and Transformer was developed. This method differs from conventional approaches by transforming a patient's EHR data into a graph structure. By defining patient subgroups based on this representation via K-Means Clustering, we were able to enhance the performance of drug response predictions. RESULTS: Leveraging EHR data from 11 627 Mayo Clinic HF patients, our model significantly outperformed traditional models in predicting drug response using NT-proBNP as a HF biomarker across five medication categories (best RMSE of 0.0043). Four distinct patient subgroups were identified with differential characteristics and outcomes, demonstrating superior predictive capabilities over existing HF subtypes (best mean RMSE of 0.0032). DISCUSSION: These results highlight the power of graph-based modeling of EHR in improving HF treatment strategies. The stratification of patients sheds light on particular patient segments that could benefit more significantly from tailored response predictions. CONCLUSIONS: Longitudinal EHR data have the potential to enhance personalized prognostic predictions through the application of graph-based AI techniques.

Takotsubo syndrome and vaccines: a systematic review.

AIMS: Takotsubo syndrome (TTS) is a rare complication of vaccination. In this study, we sought to provide insight into the characteristics of reported TTS induced by vaccination. METHODS AND RESULTS: We did a systematic review, searching PubMed, Embase, Web of Science, Ovid MEDLINE, Journals@Ovid, and Scopus databases up to 26 April 2023 to identify case reports or case series of vaccine-induced TTS. We then extracted and summarized the data from these reports. Eighteen reports were identified, with a total of 19 patients with TTS associated with vaccinations. Of the 19 included patients, the majority were female (n = 13, 68.4%) with a mean age of 56.6 ± 21.9 years. Seventeen patients developed TTS after coronavirus disease 2019 vaccination, 14 of whom received an mRNA vaccination. Two cases of TTS occurred after influenza vaccination. Among the 19 patients, 17 (89.5%) completed transthoracic echocardiography and 16 (84.2%) underwent angiography procedures. Seven patients (36.8%) completed cardiac magnetic resonance imaging. The median time to symptom onset was 2 (inter-quartile range, 1-4) days. The most common symptoms were chest pain (68.4%), dyspnoea (57.9%), and digestive symptoms (31.6%). A total of 57.9% of patients developed nonspecific symptoms such as fatigue, myalgia, diaphoresis, and fever. Among the 16 reported cases of TTS, 15 patients (93.8%) exhibited elevated cardiac troponin levels, while among the nine reported cases, eight patients (88.9%) had elevated natriuretic peptide levels. All patients had electrocardiographic changes: ST-segment change (47.1%), T-wave inversion (58.8%), and prolonged corrected QT interval (35.3%). The most common TTS type was apical ballooning (88.2%). Treatment during hospitalization typically included beta-blockers (44.4%), angiotensin-converting enzyme inhibitors/angiotensin receptor blockers (33.3%), and diuretics (22.2%). After treatment, 81.3% of patients were discharged with improved symptoms. Among this group, nine patients (56.3%) were reported to have recovered ventricular wall motion during follow-up. Two patients (12.5%) died following vaccination without resuscitation attempts. CONCLUSIONS: TTS is a rare but potentially life-threatening complication of vaccination. Typical TTS symptoms such as chest pain and dyspnoea should be considered alarming symptoms, though nonspecific symptoms are common. The risks of such rare adverse events should be balanced against the risks of infection.

Technology engagement is associated with higher perceived physical well-being in stroke patients prescribed smartwatches for atrial fibrillation detection.

Background: Increasing ownership of smartphones among Americans provides an opportunity to use these technologies to manage medical conditions. We examine the influence of baseline smartwatch ownership on changes in self-reported anxiety, patient engagement, and health-related quality of life when prescribed smartwatch for AF detection. Method: We performed a post-hoc secondary analysis of the Pulsewatch study (NCT03761394), a clinical trial in which 120 participants were randomized to receive a smartwatch-smartphone app dyad and ECG patch monitor compared to an ECG patch monitor alone to establish the accuracy of the smartwatch-smartphone app dyad for detection of AF. At baseline, 14 days, and 44 days, participants completed the Generalized Anxiety Disorder-7 survey, the Health Survey SF-12, and the Consumer Health Activation Index. Mixed-effects linear regression models using repeated measures with anxiety, patient activation, physical and mental health status as outcomes were used to examine their association with smartwatch ownership at baseline. Results: Ninety-six participants, primarily White with high income and tertiary education, were randomized to receive a study smartwatch-smartphone dyad. Twenty-four (25%) participants previously owned a smartwatch. Compared to those who did not previously own a smartwatch, smartwatch owners reported significant greater increase in their self-reported physical health (ß = 5.07, P < 0.05), no differences in anxiety (ß = 0.92, P = 0.33), mental health (ß = -2.42, P = 0.16), or patient activation (ß = 1.86, P = 0.54). Conclusions: Participants who own a smartwatch at baseline reported a greater positive change in self-reported physical health, but not in anxiety, patient activation, or self-reported mental health over the study period.

Left bundle branch area pacing for heart failure patients requiring cardiac resynchronization therapy: A meta-analysis.

INTRODUCTION: Left bundle branch area pacing (LBBP) is a novel conduction system pacing method to achieve effective physiological pacing and an alternative to cardiac resynchronization therapy (CRT) with biventricular pacing (BVP) for patients with heart failure with reduced ejection fraction (HFrEF). We conduted this meta-analysis and systemic review to review current data comparing BVP and LBBP in patients with HFrEF and indications for CRT. METHODS: We searched PubMed/Medline, Web of Science, and Cochrane Library from the inception of the database to November 2022. All studies that compared LBBP with BVP in patients with HFrEF and indications for CRT were included. Two reviewers performed study selection, data abstraction, and risk of bias assessment. We calculated risk ratios (RRs) with the Mantel-Haenszel method and mean difference (MD) with inverse variance using random effect models. We assessed heterogeneity using the I2 index, with I2 > 50% indicating significant heterogeneity. RESULTS: Ten studies (9 observational studies and 1 randomized controlled trial; 616 patients; 15 centers) published between 2020 and 2022 were included. We observed a shorter fluoroscopy time (MD: 9.68, 95% confidence interval [CI]: 4.49-14.87, I2 = 95%, p < .01, minutes) as well as a shorter procedural time (MD 33.68, 95% CI: 17.80-49.55, I2 = 73%, p < .01, minutes) during the implantation of LBBP CRT compared to conventional BVP CRT. LBBP was shown to have a greater reduction in QRS duration (MD 25.13, 95% CI: 20.06-30.20, I2 = 51%, p < .01, milliseconds), a greater left ventricular ejection fraction improvement (MD: 5.80, 95% CI: 4.81-6.78, I2 = 0%, p < .01, percentage), and a greater left ventricular end-diastolic diameter reduction (MD: 2.11, 95% CI: 0.12-4.10, I2 = 18%, p = .04, millimeter). There was a greater improvement in New York Heart Association function class with LBBP (MD: 0.37, 95% CI: 0.05-0.68, I2 = 61%, p = .02). LBBP was also associated with a lower risk of a composite of heart failure hospitalizations (HFH) and all-cause mortality (RR: 0.48, 95% CI: 0.25-0.90, I2 = 0%, p = .02) driven by reduced HFH (RR: 0.39, 95% CI: 0.19-0.82, I2 = 0%, p = .01). However, all-cause mortality rates were low in both groups (1.52% vs. 1.13%) and similar (RR: 0.98, 95% CI: 0.21-4.68, I2 = 0%, p = .87). CONCLUSION: This meta-analysis of primarily nonrandomized studies suggests that LBBP is associated with a greater improvement in left ventricular systolic function and a lower rate of HFH compared to BVP. There was uniformity of these findings in all of the included studies. However, it would be premature to conclude based solely on the current meta-analysis alone, given the limitations stated. Dedicated, well-designed, randomized controlled trials and observational studies are needed to elucidate better the comparative long-term efficacy and safety of LBBP CRT versus BIV CRT.

Artificial Intelligence-based Efficacy Prediction of Phase 3 Clinical Trial for Repurposing Heart Failure Therapies.

Introduction: Drug repurposing involves finding new therapeutic uses for already approved drugs, which can save costs as their pharmacokinetics and pharmacodynamics are already known. Predicting efficacy based on clinical endpoints is valuable for designing phase 3 trials and making Go/No-Go decisions, given the potential for confounding effects in phase 2. Objectives: This study aims to predict the efficacy of the repurposed Heart Failure (HF) drugs for the Phase 3 Clinical Trial. Methods: Our study presents a comprehensive framework for predicting drug efficacy in phase 3 trials, which combines drug-target prediction using biomedical knowledgebases with statistical analysis of real-world data. We developed a novel drug-target prediction model that uses low-dimensional representations of drug chemical structures and gene sequences, and biomedical knowledgebase. Furthermore, we conducted statistical analyses of electronic health records to assess the effectiveness of repurposed drugs in relation to clinical measurements (e.g., NT-proBNP). Results: We identified 24 repurposed drugs (9 with a positive effect and 15 with a non-positive) for heart failure from 266 phase 3 clinical trials. We used 25 genes related to heart failure for drug-target prediction, as well as electronic health records (EHR) from the Mayo Clinic for screening, which contained over 58,000 heart failure patients treated with various drugs and categorized by heart failure subtypes. Our proposed drug-target predictive model performed exceptionally well in all seven tests in the BETA benchmark compared to the six cutting-edge baseline methods (i.e., best performed in 266 out of 404 tasks). For the overall prediction of the 24 drugs, our model achieved an AUCROC of 82.59% and PRAUC (average precision) of 73.39%. Conclusion: The study demonstrated exceptional results in predicting the efficacy of repurposed drugs for phase 3 clinical trials, highlighting the potential of this method to facilitate computational drug repurposing.

False Atrial Fibrillation Alerts from Smartwatches are Associated with Decreased Perceived Physical Well-being and Confidence in Chronic Symptoms Management.

Wrist-based wearables have been FDA approved for AF detection. However, the health behavior impact of false AF alerts from wearables on older patients at high risk for AF are not known. In this work, we analyzed data from the Pulsewatch (NCT03761394) study, which randomized patients (≥50 years) with history of stroke or transient ischemic attack to wear a patch monitor and a smartwatch linked to a smartphone running the Pulsewatch application vs to only the cardiac patch monitor over 14 days. At baseline and 14 days, participants completed validated instruments to assess for anxiety, patient activation, perceived mental and physical health, chronic symptom management self-efficacy, and medicine adherence. We employed linear regression to examine associations between false AF alerts with change in patient-reported outcomes. Receipt of false AF alerts was related to a dose-dependent decline in self-perceived physical health and levels of disease self-management. We developed a novel convolutional denoising autoencoder (CDA) to remove motion and noise artifacts in photoplethysmography (PPG) segments to optimize AF detection, which substantially reduced the number of false alerts. A promising approach to avoid negative impact of false alerts is to employ artificial intelligence driven algorithms to improve accuracy.

Using artificial intelligence to learn optimal regimen plan for Alzheimer's disease.

BACKGROUND: Alzheimer's disease (AD) is a progressive neurological disorder with no specific curative medications. Sophisticated clinical skills are crucial to optimize treatment regimens given the multiple coexisting comorbidities in the patient population. OBJECTIVE: Here, we propose a study to leverage reinforcement learning (RL) to learn the clinicians' decisions for AD patients based on the longitude data from electronic health records. METHODS: In this study, we selected 1736 patients from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. We focused on the two most frequent concomitant diseases-depression, and hypertension, thus creating 5 data cohorts (ie, Whole Data, AD, AD-Hypertension, AD-Depression, and AD-Depression-Hypertension). We modeled the treatment learning into an RL problem by defining states, actions, and rewards. We built a regression model and decision tree to generate multiple states, used six combinations of medications (ie, cholinesterase inhibitors, memantine, memantine-cholinesterase inhibitors, hypertension drugs, supplements, or no drugs) as actions, and Mini-Mental State Exam (MMSE) scores as rewards. RESULTS: Given the proper dataset, the RL model can generate an optimal policy (regimen plan) that outperforms the clinician's treatment regimen. Optimal policies (ie, policy iteration and Q-learning) had lower rewards than the clinician's policy (mean -3.03 and -2.93 vs. -2.93, respectively) for smaller datasets but had higher rewards for larger datasets (mean -4.68 and -2.82 vs. -4.57, respectively). CONCLUSIONS: Our results highlight the potential of using RL to generate the optimal treatment based on the patients' longitude records. Our work can lead the path towards developing RL-based decision support systems that could help manage AD with comorbidities.

Self-reported efficacy in patient-physician interaction in relation to anxiety, patient activation, and health-related quality of life among stroke survivors.

BACKGROUND: Early detection of AF is critical for stroke prevention. Several commercially available smartwatches are FDA cleared for AF detection. However, little is known about how patient-physician relationships affect patients' anxiety, activation, and health-related quality of life when prescribed smartwatch for AF detection. METHODS: Data were used from the Pulsewatch study (NCT03761394), which randomized adults (>50 years) with no contraindication to anticoagulation and a CHA2DS2-VASc risk score ≥2 to receive a smartwatch-smartphone app dyad for AF monitoring vs. conventional monitoring with an ECG patch (Cardea SoloTM) and monitored participants for up to 45 days. The Perceived Efficacy in Patient-Physician Interactions survey was used to assess patient confidence in physician interaction at baseline with scores ≥45 indicating high perceived efficacy in patient-provider interactions. Generalized Anxiety Disorder-7 Scale, Consumer Health Activation Index, and Short-Form Health Survey were utilized to examine anxiety, patient activation, and physical and mental health status, at baseline, 14, and 44 days, respectively. We used mixed-effects repeated measures linear regression models to assess changes in psychosocial outcomes among smartwatch users in relation to self-reported efficacy in physician interaction over the study period. RESULTS: A total of 93 participants (average age 64.1 ± 8.9 years; 43.0% female; 88.2% non-Hispanic white) were included in this analysis. At baseline, fifty-six (60%) participants reported high perceived efficacy in patient-physician interaction. In the fully adjusted models, high perceived efficacy (vs. low) at baseline was associated with greater patient activation and perceived mental health (ß 12.0, p-value <0.001; ß 3.39, p-value <0.05, respectively). High perceived self-efficacy was not associated with anxiety or physical health status (ß - 0.61, p-value 0.46; ß 0.64, p-value 0.77) among study participants. CONCLUSIONS: Higher self-efficacy in patient-physician interaction was associated with higher patient activation and mental health status among stroke survivors using smartwatches. Furthermore, we found no association between anxiety and smartwatch prescription for AF in participants with high self-efficacy in patient-physician interaction. Efforts to improve self-efficacy in patient-physician interaction may improve patient activation and self-rated health and subsequently may lead to better clinical outcomes.KEY MESSAGESHigher self-efficacy in patient-physician interaction was associated with higher patient activation and mental health status among stroke survivors using smartwatches.No association between anxiety and smartwatch prescription for AF in participants with high self-efficacy in patient-physician interaction.Efforts to improve self-efficacy in patient-physician interaction may improve patient activation and self-rated health and subsequently may lead to better clinical outcomes.

Using Artificial Intelligence to Learn Optimal Regimen Plan for Alzheimer's Disease.

Background: Alzheimer's Disease (AD) is a progressive neurological disorder with no specific curative medications. While only a few medications are approved by FDA (i.e., donepezil, galantamine, rivastigmine, and memantine) to relieve symptoms (e.g., cognitive decline), sophisticated clinical skills are crucial to optimize the appropriate regimens given the multiple coexisting comorbidities in this patient population. Objective: Here, we propose a study to leverage reinforcement learning (RL) to learn the clinicians' decisions for AD patients based on the longitude records from Electronic Health Records (EHR). Methods: In this study, we withdraw 1,736 patients fulfilling our criteria, from the Alzheimer's Disease Neuroimaging Initiative(ADNI) database. We focused on the two most frequent concomitant diseases, depression, and hypertension, thus resulting in five main cohorts, 1) whole data, 2) AD-only, 3) AD-hypertension, 4) AD-depression, and 5) AD-hypertension-depression. We modeled the treatment learning into an RL problem by defining the three factors (i.e., states, action, and reward) in RL in multiple strategies, where a regression model and a decision tree are developed to generate states, six main medications extracted (i.e., no drugs, cholinesterase inhibitors, memantine, hypertension drugs, a combination of cholinesterase inhibitors and memantine, and supplements or other drugs) are for action, and Mini-Mental State Exam (MMSE) scores are for reward. Results: Given the proper dataset, the RL model can generate an optimal policy (regimen plan) that outperforms the clinician's treatment regimen. With the smallest data samples, the optimal-policy (i.e., policy iteration and Q-learning) gained a lesser reward than the clinician's policy (mean -2.68 and -2.76 vs . -2.66, respectively), but it gained more reward once the data size increased (mean -3.56 and -2.48 vs . -3.57, respectively). Conclusions: Our results highlight the potential of using RL to generate the optimal treatment based on the patients' longitude records. Our work can lead the path toward the development of RL-based decision support systems which could facilitate the daily practice to manage Alzheimer's disease with comorbidities.

Predicting Physiological Response in Heart Failure Management: A Graph Representation Learning Approach using Electronic Health Records.

Heart failure management is challenging due to the complex and heterogenous nature of its pathophysiology which makes the conventional treatments based on the "one size fits all" ideology not suitable. Coupling the longitudinal medical data with novel deep learning and network-based analytics will enable identifying the distinct patient phenotypic characteristics to help individualize the treatment regimen through the accurate prediction of the physiological response. In this study, we develop a graph representation learning framework that integrates the heterogeneous clinical events in the electronic health records (EHR) as graph format data, in which the patient-specific patterns and features are naturally infused for personalized predictions of lab test response. The framework includes a novel Graph Transformer Network that is equipped with a self-attention mechanism to model the underlying spatial interdependencies among the clinical events characterizing the cardiac physiological interactions in the heart failure treatment and a graph neural network (GNN) layer to incorporate the explicit temporality of each clinical event, that would help summarize the therapeutic effects induced on the physiological variables, and subsequently on the patient's health status as the heart failure condition progresses over time. We introduce a global attention mask that is computed based on event co-occurrences and is aggregated across all patient records to enhance the guidance of neighbor selection in graph representation learning. We test the feasibility of our model through detailed quantitative and qualitative evaluations on observational EHR data.

Association of atrial fibrillation burden with in-hospital outcomes in patients with Takotsubo cardiomyopathy.

BACKGROUND: The effects of atrial fibrillation (AF) and its burden on in-hospital mortality in patients with Takotsubo cardiomyopathy (TCM) are unclear. Here, we examined the effect of AF and paroxysmal AF on in-hospital outcomes in patients with TCM. METHODS: We used ICD-10 codes to retrospectively identify patients with a primary diagnosis of TCM in the National Inpatient Sample database 2016-2018. We compared in-hospital outcomes in TCM patients with and without AF before and after propensity score matching. The effect of AF burden on outcomes was assessed in patients with paroxysmal AF and no AF. RESULTS: Of the 4,733 patients with a primary diagnosis of TCM, 650 (13.7%) had AF, and 4,083 (86.3%) did not. Of TCM patients with AF, 368 (56.6%) had paroxysmal AF. In-hospital mortality was higher in patients with AF before (3.4% vs 1.2%, P <  0.001) and after propensity matching (3.4% vs 1.7%, P = 0.021) but did not differ between the paroxysmal AF and the no AF groups (P = 0.205). In the matched cohorts, both AF and paroxysmal AF groups were associated with a higher rate of cardiogenic shock (AF, P < 0.001; paroxysmal AF, P < 0.001), ventricular arrhythmia (AF, P = 0.002; paroxysmal AF, P = 0.02), acute kidney injury (AF, P = 0.007; paroxysmal AF, P = 0.008), and acute respiratory failure (AF, P < 0.001; paroxysmal AF, P < 0.001) compared with the no AF group. CONCLUSIONS: Although AF was associated with increased in-hospital mortality, paroxysmal AF did not affect in-hospital mortality, suggesting a higher AF burden is associated with worse clinical outcome in patients with TCM.

Usefulness of Self-Reported Physical Activity and Clinical Outcomes in Older Patients With Atrial Fibrillation.

Current guidelines encourage regular physical activity (PA) to gain cardiovascular health benefit. However, little is known about whether older adults with atrial fibrillation (AF) who engage in the guideline-recommended level of PA are less likely to experience clinically relevant outcomes. We did a retrospective study based on the data from Systemic Assessment of Geriatric Elements in AF (SAGE-AF) prospective cohort study. The study population consisted of older participants with AF (≥65 years) and a congestive heart failure, hypertension, age, diabetes, stroke vascular disease, age 65 to 75 and sex(CHA2DS2-VASc) score ≥2. PA was quantified by self-reported Minnesota Leisure Time PA questionnaire. Competing risk models were used to examine the association between PA level and clinical outcomes over 2 years while controlling for several potentially confounding variables. A total of 1,244 participants (average age 76 years; 51% men; 85% non-Hispanic White) were studied. A total of 50.5% of participants engaged in regular PA. Meeting the recommended level of PA was associated with lower mortality over 2 years (adjusted hazard ratio 0.60, 95% confidence interval 0.38 to 0.95) but was not associated with rates of stroke or major bleeding. In conclusion, older adults with AF who engaged in guideline-recommended PA are more likely to survive in the long term. Healthcare providers should promote and encourage engagement in PA and tailor interventions to address barriers of engagement.

Takotsubo syndrome and respiratory diseases: a systematic review.

Takotsubo syndrome (TTS) is a rare cardiovascular condition characterized by reversible ventricular dysfunction and a presentation resembling that of acute myocardial infarction. An increasing number of studies has shown the association of respiratory diseases with TTS. Here, we comprehensively reviewed the literature and examined the available evidence for this association. After searching PubMed, EMBASE, and Cochrane Library databases, two investigators independently reviewed 3117 studies published through May 2021. Of these studies, 99 met the inclusion criteria (n = 108 patients). In patients with coexisting respiratory disease and TTS, the most common TTS symptom was dyspnoea (70.48%), followed by chest pain (24.76%) and syncope (2.86%). The most common type of TTS was apical, accounting for 81.13% of cases, followed by the midventricular (8.49%), basal (8.49%), and biventricular (1.89%) types. Among the TTS cases, 39.82% were associated with obstructive lung disease and 38.89% were associated with pneumonia. Coronavirus disease 2019 (COVID-19), which has been increasingly reported in patients with TTS, was identified in 29 of 42 (69.05%) patients with pneumonia. The overall mortality rate for patients admitted for respiratory disease complicated by TTS was 12.50%. Obstructive lung disease and pneumonia are the most frequently identified respiratory triggers of TTS. Medications and invasive procedures utilized in managing respiratory diseases may also contribute to the development of TTS. Furthermore, the diagnosis of TTS triggered by these conditions can be challenging due to its atypical presentation. Future prospective studies are needed to establish appropriate guidelines for managing respiratory disease with concurrent TTS.

Mortality in Stable Coronary Disease in Patients With Intermediate- or High-Risk Myocardial Perfusion Imaging.

The management of patients with stable coronary disease and intermediate- or high-risk features on single photon emission computed tomography myocardial perfusion imaging (SPECT MPI) continues to be controversial as to whether they should be treated with an initial invasive strategy (catheterization and revascularization when feasible) or medical therapy alone to improve mortality. We performed a retrospective observational study of 1,946 patients with intermediate- or high-risk SPECT MPI scans performed over a 6-year period (from 2014 to 2019). Each patient was followed from the time of SPECT MPI to 16 months after the last patient was enrolled. The primary end point was all-cause mortality and the secondary end point cardiovascular mortality. Of the eligible 1,697 patients, 1,144 had an intermediate-risk scan, 553 a high-risk scan, 915 had medical therapy alone, and 782 went on an initial invasive strategy. All patients were divided into the following three groups: combined SPECT MPI (both intermediate- and high-risk), high-risk SPECT MPI, and intermediate-risk SPECT MPI groups. After propensity score matching, there was a statistically significant difference in cardiovascular death (5.9% vs 2.7%; p = 0.038) in the medical therapy cohort compared with initial invasive cohort in the combined SPECT MPI group, but no difference in all-cause death (15.7% vs 13%; p = 0.318). On subgroup analysis, in intermediate-risk SPECT MPI group, there was no significant difference in either all-cause death (13.8 vs 11.7%; p = 0.583) or cardiac death (5.4% vs 2.5%; p = 0.16) in conservative cohort compared with invasive strategy cohort. In high-risk SPECT MPI group, conservative therapy cohort had higher cardiac death (11.7% vs 2.5%; p = 0.002) compared with initial invasive strategy cohort, but there was no significant difference in all-cause death (24.5% vs 15.3%; p = 0.052). In conclusion, this study supports that patients with intermediate- or high-risk SPECT MPI scans when considered together or only with high-risk features, derive a cardiovascular mortality benefit with an initial invasive strategy. Patients who had undergone intermediate-risk SPECT MPI had similar outcomes with either medical therapy alone or initial invasive evaluation.

Meta-Analysis of Randomized Clinical Trials Comparing the Impact of Implantable Loop Recorder Versus Usual Care After Ischemic Stroke for Detection of Atrial Fibrillation and Stroke Risk.

Implantable loop recorder (ILR) is recommended to detect subclinical atrial fibrillation (AF) after cryptogenic stroke; however, the clinical outcomes of this practice is unclear. We conducted a systematic review and meta-analysis of randomized controlled trials to evaluate 12-month AF detection, change in oral anticoagulation (OAC), and recurrent stroke in ILR versus usual care after ischemic stroke. We searched Medline, Embase, Web of Science, Cochrane Library for randomized controlled trials comparing ILR with usual care after any ischemic stroke. Primary outcomes were cumulative AF detection and recurrent stroke (ischemic or hemorrhagic) or transient ischemic attack over 12 months. Secondary outcome was OAC initiation. Meta-analysis was performed with Mantel-Haenszel pooled odds ratios (ORs) and random effects models. Of 200 identified articles, 3 trials were included (1,233 participants). Cryptogenic stroke and underlying AF included cryptogenic stroke only, stroke of known cause and underlying-AF included small or large vessel stroke only, and post embolic rhythm detection with implantable vs external monitoring included all ischemic strokes. The 12-month AF detection was 13% in the ILR group and 2.4% in controls. ILR was more likely to detect AF compared with usual care (OR 5.8, 95% confidence interval 3.2 to 10.2). Stroke or transient ischemic attack occurred in 7% with ILR and 9% with usual care (OR 0.8, 95% confidence interval 0.5 to 1.2). In patients with detected AF, 97% and 100% were started on OAC in cryptogenic stroke and underlying AF and post embolic rhythm detection with implantable vs external monitoring, respectively, compared with 68% in stroke of known cause and underlying-AF. In conclusion, ILR was superior to usual care in AF detection, but the relative low incidence of AF and the nondifferential risk of stroke between the ILR and usual care arms may suggest that most patients do not benefit from ILR implantation. Further studies are warranted to understand if patient selection can be improved to increase the diagnostic yield of ILR.

Machine learning predicting mortality in sarcoidosis patients admitted for acute heart failure.

Background: Sarcoidosis with cardiac involvement, although rare, has a worse prognosis than sarcoidosis involving other organ systems. Objective: We used a large dataset to train machine learning models to predict in-hospital mortality among sarcoidosis patients admitted with heart failure (HF). Method: Utilizing the National Inpatient Sample, we identified 4659 patients hospitalized with a primary diagnosis of HF. In this cohort, we identified patients with a secondary diagnosis of sarcoidosis using International Statistical Classification of Disease, Tenth Revision (ICD-10) codes. Patients were separated into a training group and a testing group in a 7:3 ratio. Least absolute shrinkage and selection operator regression was used to select variables to prevent model overfitting or underfitting. For machine learning models, logistic regression, random forest, and XGBoosting were applied in the training group. Parameters in each of the models were tuned using the GridSearchCV function. After training, all models were further validated in the testing group. Models were then evaluated using the area under curve (AUC) score, sensitivity, and specificity. Results: A total of 2.3% of sarcoidosis patients died in HF admission. Our machine learning model analysis found the RF model to have the highest AUC score and sensitivity. Feature analysis found that comorbid arrhythmias and fluid electrolyte disorders were the strongest factors in predicting in-hospital mortality. Conclusion: Machine learning methods can be useful in identifying predictors of in-hospital mortality in a given dataset.

The Association Between Hyperlipidemia and In-Hospital Outcomes in Takotsubo Cardiomyopathy.

PURPOSE: Hyperlipidemia (HLD) is one of the most common cardiovascular risk factors and is prevalent in patients with takotsubo cardiomyopathy (TCM), but the association between HLD and TCM patients' outcomes is unclear. We investigated the impact of HLD on the in-hospital outcomes of TCM patients. PATIENTS AND METHODS: Our retrospective cohort study used the latest available data from the National Inpatient Sample (2016-2017). Using the ICD-10 code, we identified 3139 patients with a primary diagnosis of TCM, 1530 of whom had HLD. We compared in-hospital outcomes between HLD and non-HLD groups before and after propensity score matching. RESULTS: In the unmatched cohort, the HLD group had lower incidences of cardiac arrest, cardiogenic shock, and acute respiratory failure (ARF); shorter length of stay (LOS); and lower total charges (All p<0.05). In-hospital mortality (p=0.102) and ventricular arrhythmia (p=0.235) rates did not differ. After propensity score matching, the HLD group had lower rates of in-hospital mortality (1.1% vs 2.4%, p=0.027), ARF (9.1% vs 12.1%, p = 0.022) and cardiogenic shock (3.4% vs 5.6%, p=0.012), shorter LOS (3.20 ± 3.27 days vs 3.57 ± 3.14 days, p=0.005), and lower total charges (p=0.013). The matched groups did not differ significantly regarding cardiac arrest (p=0.141), ventricular arrhythmia (p=0.662) or acute kidney injury (AKI) (p = 0.167). CONCLUSION: Counterintuitively, HLD was associated with better in-hospital outcomes in both the unmatched and propensity-matched cohorts of hospitalized TCM patients. Further studies are needed to investigate the mechanisms that may contribute to the association in TCM patients with HLD.

Identifying different phenotypes in takotsubo cardiomyopathy by latent class analysis.

AIMS: This study sought to determine whether clinical clusters exist in takotsubo cardiomyopathy. Takotsubo cardiomyopathy (TCM) is a heterogeneous disorder with a complex, poorly understood pathogenesis. To better understand the heterogeneity of TCM, we identified different clinical phenotypes in a large sample of TCM patients by using latent class analysis (LCA). METHODS AND RESULTS: Using the National Inpatient Sample (NIS) database, we identified 3139 patients admitted to hospitals in 2016-2017 with a primary diagnosis of TCM. We performed LCA based on several patient demographics and comorbidities: age, sex, hypertension, hyperlipidaemia, diabetes mellitus, obesity, current smoking, asthma, chronic obstructive pulmonary disease (COPD), and anxiety and depressive disorders. We then repeated LCA separately with the NIS 2016 and 2017 data sets and performed a robust test to validate our results. We also compared in-hospital outcomes among the different clusters identified by LCA. Four patient clusters were identified. C1 (n = 1228, 39.4%) had the highest prevalence of hyperlipidaemia (93.4%), hypertension (61.6%), and diabetes (34.3%). In C2 (n = 440, 14.0%), all patients had COPD, and many were smokers (45.8%). C3 (n = 376, 11.8%) largely comprised patients with anxiety disorders (98.4%) and depressive disorders (80.1%). C4 (n = 1097, 34.8%) comprised patients with isolated TCM and few comorbidities. Among all clusters, C1 had the lowest in-hospital mortality (1.0%) and the shortest length of stay (3.2 ± 3.1 days), whereas C2 had the highest in-hospital mortality (3.4%). CONCLUSIONS: Using LCA, we identified four clinical phenotypes of TCM. These may reflect different pathophysiological processes in TCM. Our findings may help identify treatment targets and select patients for future clinical trials.