Background Patients with heart failure and preserved ejection fraction (HFpEF) have a poor prognosis, and no therapies have been proven to improve outcomes. It has been proposed that heart failure, including HFpEF, represents overlapping syndromes that may have different prognoses. We present an exploratory study of patients enrolled in the Irbesartan in Heart Failure with Preserved Ejection Fraction Study (I-PRESERVE) using latent class analysis (LCA) with validation using the Candesartan in Heart failure: Assessment of Reduction in Mortality and morbidity (CHARM)-Preserved study to identify HFpEF subgroups. Methods and results In total, 4113 HFpEF patients randomized to irbesartan or placebo were characterized according to 11 clinical features. HFpEF subgroups were identified using LCA. Event-free survival and effect of irbesartan on the composite of all-cause mortality and cardiovascular hospitalization were determined for each subgroup. Subgroup definitions were applied to 3203 patients enrolled in CHARM-Preserved to validate observations regarding prognosis and treatment response. Six subgroups were identified with significant differences in event-free survival (p<0.001). Clinical profiles and prognoses of the 6 subgroups were similar in CHARM-Preserved. The two subgroups with the worst event-free survival in both studies were characterized by a high prevalence of obesity, hyperlipidemia, diabetes mellitus, anemia, and renal insufficiency (Subgroup C) and by female predominance, advanced age, lower body mass index, and high rates of atrial fibrillation, valvular disease, renal insufficiency, and anemia (Subgroup F). Conclusion Using a data-driven approach, we identified HFpEF subgroups with significantly different prognoses. Further development of this approach for characterizing HFpEF subgroups is warranted.

Use of machine learning (ML) in clinical research is growing steadily given the increasing availability of complex clinical data sets. ML presents important advantages in terms of predictive performance and identifying undiscovered subpopulations of patients with specific physiology and prognoses. Despite this popularity, many clinicians and researchers are not yet familiar with evaluating and interpreting ML analyses. Consequently, readers and peer-reviewers alike may either overestimate or underestimate the validity and credibility of an ML-based model. Conversely, ML experts without clinical experience may present details of the analysis that are too granular for a clinical readership to assess. Overwhelming evidence has shown poor reproducibility and reporting of ML models in clinical research suggesting the need for ML analyses to be presented in a clear, concise, and comprehensible manner to facilitate understanding and critical evaluation. We present a recommendation for transparent and structured reporting of ML analysis results specifically directed at clinical researchers. Furthermore, we provide a list of key reporting elements with examples that can be used as a template when preparing and submitting ML-based manuscripts for the same audience.

ABSTRACT:The learning healthcare system uses health information technology and the health data infrastructure to apply scientific evidence at the point of clinical care while simultaneously collecting insights from that care to promote innovation in optimal healthcare delivery and to fuel new scientific discovery. To achieve these goals, the learning healthcare system requires systematic redesign of the current healthcare system, focusing on 4 major domains: science and informatics, patient-clinician partnerships, incentives, and development of a continuous learning culture. This scientific statement provides an overview of how these learning healthcare system domains can be realized in cardiovascular disease care. Current cardiovascular disease care innovations in informatics, data uses, patient engagement, continuous learning culture, and incentives are profiled. In addition, recommendations for next steps for the development of a learning healthcare system in cardiovascular care are presented.H ealth care has never been more complex. Clinicians and patients must make decisions that integrate the continually evolving scientific evidence base with hundreds of individual data points such as patients' vital signs, symptoms, comorbidities, medications, test results, and preferences. Furthermore, once these decisions are made, little information is available about their impact, limiting the ability to learn from and ultimately improve care delivery. This inability of the healthcare system to learn from its operation results in significant inefficiencies, substantial costs, and suboptimal health outcomes.The creation of a learning healthcare system (LHS) can potentially address these issues. The LHS uses health information technology and the health data infrastructure to apply scientific evidence at the point of clinical care while simultaneously collecting insights from that care to promote innovation in optimal healthcare delivery and to fuel new scientific discovery 1,2 (Figure 1). Thus, the LHS enables rapid, iterative learning in which "evidence informs practice, and practice informs evidence." 2 The authors argued that recent advances in information processing and connectivity, healthcare organizational design, and reimbursement policies centered on quality rather than quantity of care provided the necessary tools for the creation of the LHS. Accordingly, they called for systematic redesign of the healthcare system, focusing on 4 major domains: science and informatics, patient-clinician partnerships, incentives, and development of a continuous learning culture.

No abstract

Objectives To identify clinical features associated with peripartum cardiomyopathy (PPCM) and possible racial differences, and to quantify in-hospital outcomes in delivering mothers with PPCM. Background Investigation of patient characteristics and outcomes in PPCM has been limited to small cohorts. Hospital discharge data allows assembly of the largest number of PPCM cases to date. Methods Hospital records from six states were screened for PPCM. Clinical profiles, maternal, and fetal outcomes in delivering mothers with and without PPCM were compared and stratified by race. A maternal major adverse event (MAE) was defined as death, cardiac arrest, heart transplantation, or mechanical circulatory support. Logistic regression was used to identify variables associated with PPCM. Results In total, 535/4,003,914 records of delivering mothers specified a diagnosis of PPCM. Prevalence of PPCM was highest among African-Americans and similar in Caucasians and Hispanics. Established risk factors including age ≥ 30 years, African-American race, hypertension, preeclampsia/eclampsia, and multigestational status were associated with PPCM, and novel associations such as anemia and asthma were identified. Autoimmune disease and substance abuse, which can cause cardiomyopathy independently, were also associated with PPCM. Maternal MAE (odds ratio=436, p<0.0001) and stillbirth (odds ratio=3.8, p<0.0001) occurred more frequently among women with PPCM. Conclusions The prevalence of PPCM at the time of delivery in Hispanics was similar to Caucasians and lower than African-Americans. Autoimmune disease, substance abuse, anemia and asthma were conditions associated with PPCM not consistently identified in smaller cohorts. PPCM was also associated with increased risk of stillbirth and maternal MAEs at delivery.