This paper presents a principled and scalable framework for systematically generating complex Question Answering (QA) data. In the core of this framework is a graphlet-anchored generation process, where small subgraphs from a Knowledge Graph (KG) are used in a structured prompt to control the complexity and ensure the factual grounding of questions generated by Large Language Models. The first instantiation of this framework is BioGraphletQA, a new biomedical KGQA dataset of 119,856 QA pairs. Each entry is grounded in a graphlet of up to five nodes from the OREGANO KG, with most of the pairs being enriched with relevant document snippets from PubMed. We start by demonstrating the framework’s value and the dataset’s quality through evaluation by a domain expert on 106 QA pairs, confirming the high scientific validity and complexity of the generated data. Secondly, we establish its practical utility by showing that augmenting downstream benchmarks with our data improves accuracy on PubMedQA from 49.2 https://zenodo.org/records/17381119 ) and framework code ( https://github.com/ieeta-pt/BioGraphletQA ), are publicly available to facilitate use, reproducibility and extension.
Abstract Background Telemonitoring (TM) improves outcomes in chronic heart failure (HF), but real-world evidence on whether TM also supports patient activation and treatment adherence beyond congestion management remains limited. Aim To assess whether a HF TM pathway is associated with improved adherence/patient activation, using changes in cardiometabolic risk-factor control (LDL-cholesterol, HbA1c) and smoking status as pragmatic surrogates, while also describing concomitant changes in healthcare utilization and functional status. Methods Retrospective before–after analysis of 31 consecutive HF patients enrolled in a TM program. Outcomes were compared within patients over the 12 months before versus after TM initiation. Endpoints included hospitalizations, days hospitalized, emergency department (ED) visits, NYHA class, diuretic dose, biomarkers, lipid profile, glycaemic control (diabetic subgroup), smoking status, and LVEF. Multivariable models explored baseline predictors of improvement. Results Patients were predominantly male (77%), mean age 67 years, with high cardiovascular risk burden (hypertension 61%, dyslipidaemia 74%, diabetes 36%); mean LVEF 34% and ischaemic etiology 81%. The most frequent TM alert was weight gain (59%); 67% of alerts were resolved without in-person assessment. After TM initiation, there were mean reductions in hospitalizations (−0.7; SD 0.95) and admission days (−5.8; SD 11.0), ED visits (−0.3; SD 1.4), and NYHA class (−0.4; SD 0.7), alongside an increase in LVEF (+6.4%; SD 8.4). Improvements were statistically significant for hospitalizations (p<0.01), days hospitalized (p<0.01), NYHA class (p<0.01), and LVEF (p<0.01). LDL levels decreased by −30.6 mg/dL (SD 51.9; p<0.01). HbA1c levels decreased by −0.6 %-points (SD 0.8; p<0.05) among patients with diabetes (n=11). Current smoking decreased from 5/31 (16.1%) pre-TM to 2/31 (6.5%) at 12 months, with 3/5 (60%) baseline smokers achieving cessation. In multivariable analysis, greater benefit clustered in patients with higher pre-TM event burden (more prior hospitalizations/longer admissions/more advanced symptoms and lower LVEF). Conclusions In a real-world HF TM cohort, TM was associated with fewer admissions and improved NYHA class and LVEF, with the largest gains in clinically advanced/high-utilization patients. Beyond clinical stabilization, TM coincided with improved cardiometabolic risk-factor control (LDL, HbA1c) and smoking cessation, supporting a potential additional value through enhanced adherence/patient engagement. Prospective studies should confirm whether TM-driven patient activation translates into sustained adherence improvements and long-term outcomes.
Heart failure (HF) guidelines primarily rely on left ventricular ejection fraction (LVEF) to classify patients and guide therapy implementation. However, LVEF alone may be insufficient to adequately characterize patients with HF, particularly those with HF with preserved ejection fraction (HFpEF). There is a growing need to advance cardiac imaging, especially echocardiography, to improve the phenotypic and prognostic characterization of HF, ultimately guiding more precise therapy and follow-up strategies. LV global longitudinal strain (GLS), which directly reflects the motion of the myocardium, is believed to provide a better assessment of myocardial function. We aim to explore whether LV GLS predicts adverse clinical cardiovascular outcomes. In this retrospective cohort study, patients diagnosed with HFmrEF/HFpEF (defined as LVEF ≥40%), according to the HFA–PEFF diagnostic algorithm, were included. Exclusion criteria inlcuded primary valvular disease, any prior LVEF <40%, infiltrative and hypertrophic cardiomyopathy, congenital heart disease, isolated right HF, and inadequate image quality for strain analysis. The primary outcome was a composite of cardiovascular mortality, HF hospitalization and HF-related emergency visits. The secondary outcomes included the evolution of NYHA functional class, diuretic dosage, and NT-proBNP levels. A time-to-event analysis was conducted, incorporating Kaplan-Meier analysis and Cox proportional hazard models. During a median follow-up of 11 months [IQR 7.25, 11], 15 patients developed the primary outcome, comprising 9 HF hospitalisations and one cardiovascular death. In univariable Cox regression analysis, neither LV GLS (HR=1.09, [95% CI, 0.92-1.30], p=0.3) nor LVEF (HR=0.95, [95% CI, 0.88-1.03], p=0.2) was associated with a higher risk of the primary outcome. When considering HF hospitalisations alone, LV GLS was not associated with an increased risk of HF hospitalisation (HR=1.26, [95% CI, 0.99-1.60], p=0.064). When combining LVEF and GLS, while examining the Kaplan-Meier curves, neither subgroup had a higher risk of the composite endpoint (log-rank test p=0.31, Figure 1). In comparison to stable patients, those whose NYHA functional class worsened during follow-up exhibited significantly worse baseline GLS (p=0.025), but no association was found with diuretic dose (p=0.14). In this cohort of HFmrEF/HFpEF patients, although a large proportion exhibited impaired myocardial function by GLS, LV global longitudinal strain was not associated with cardiovascular mortality, HF hospitalizations, or HF-related emergency visits.Univariable Cox regression Kaplan-Meier curves
Risk stratification in nonischemic dilated cardiomyopathy (NIDCM) remains challenging. Cardiac magnetic resonance (CMR) not only aids in identifying underlying etiology but also offers a noninvasive means of myocardial tissue characterisation, which might be a valuable tool for prognostic assessment - through late gadolinium enhancement (LGE), parametric mapping, and extracellular volume (ECV). While recent studies have focused on the presence and extent of LGE, parametric mapping remains less explored but may provide complementary insights into diffuse interstitial fibrosis. To explore the predictive value of cardiovascular magnetic resonance (CMR) findings - T1, T2 mapping and ECV - for heart failure (HF) and arrhythmia-related events in NIDCM patients. In this retrospective cohort study, patients diagnosed with NIDCM who underwent CMR at our center were included. All CMR images were acquired using a 1.5-T scanner. T1 mapping was quantified within the septal myocardium in areas without LGE enhancement (T1 native) and ECV was calculated using pre, post-contrast T1 and synthetic haematocrit. The primary outcomes were HF hospitalisation and arrhythmia-related events (defined as appropriate ICD therapy or sustained VT/VF or non-sustained VT). Among the 53 patients with NIDCM, 40% were women, with a median age of 64 years [IQR 52-69], the median LV ejection fraction was 40% [IQR 30-47] and 15% had an implantable device. During a median follow-up of 17 months [IQR 9-27], 6 patients (11%) had a HF hospitalisation and 11 patients (35%) had an arrhythmia-related event. ECV was similar across NYHA class (p=0.33), ECV was correlated with both NT-proBNP value (p=0.001), LV ejection fraction (p=0.001), as well as the presence of LGE (p=0.019, Figure 1). In univariable Cox regression analysis, although T1 mapping (HR=1.01, [95% CI, 0.99-1.02], p=0.3) and T2 mapping (HR=1.01, [95% CI, 0.83-1.23], p=>0.9) were not associated with higher risk of HF hospitalisation, patients with higher ECV had higher risk of HF hospitalisation (HR=1.11, [95% CI, 1.01-1.22], p=0.03, Table 1). Furthermore, neither CMR parameter was associated with increased risk of an arrhythmia-related event (Table 1). In multivariable Cox regression analysis, including age, gender and LV ejection fraction, ECV remain an independent predictor of HF hospitalisation (HR=1.13, [95% CI, 1.00-1.28], p=0.046, Table 1). In this cohort of patients diagnosed with NIDCM, only extracellular volume was an independent predictor of HF hospitalisation. Arrhythmia-related risk was not associated with any of the CMR parameters.Uni- and Multivariable Cox regression Correlation between ECV and LGE presence
Abstract Background Telemonitoring (TM) in chronic heart failure (HF) has been associated with fewer decompensation events, yet the real-world budget impact of TM remains a key barrier to wider implementation, particularly in cohorts with high-severity admissions. Aim To estimate the annual net direct-cost impact of a HF TM pathway by integrating changes in healthcare utilization with local tariffs and the annual TM cost. Methods We performed a retrospective before–after analysis of 31 consecutive HF patients enrolled in a TM program. Healthcare utilization was assessed over the 12 months before versus after TM initiation, including hospitalizations, inpatient days, and emergency department (ED) visits. A micro-costing model applied local unit costs: hospitalization €1,074 for medium/high severity and €2,230 for high severity; ED attendance €167. The annual TM cost was €1,562 per patient. Avoided admission costs were calculated using the cohort’s admission severity distribution (10 medium/high; 21 high). The primary economic endpoint was net budget impact per patient-year, defined as TM cost minus avoided hospitalization and ED costs. Results Patients were predominantly male (77%), with mean age 67 years; mean LVEF was 34% and ischemic etiology was present in 81%. Weight gain was the most frequent TM alert (59%), and 67% of alerts were managed without an in-person encounter. Compared with the year before enrolment, the year after TM initiation showed reductions in hospitalizations (−0.7 ± 0.95 per patient-year; p<0.01), inpatient days (−5.8 ± 11.0; p<0.01), and ED visits (−0.3 ± 1.4). Using the observed severity mix, the weighted mean hospitalization tariff was €1,857. The estimated direct cost offset from reduced admissions and ED visits was €1,350 per patient-year. After accounting for TM cost (€1,562 per patient-year), the net budget impact was +€212 per patient-year, equivalent to +€6,570 annually for the 31-patient cohort; TM therefore offset approximately 86% of its annual per-patient cost under these assumptions. Conclusions In a real-world HF cohort with predominantly medium/high-to-high severity admissions, TM was associated with lower acute-care utilization and substantial direct cost offsets, approaching cost neutrality at a program cost of €1,562 per patient-year. These findings support prioritizing TM for higher-risk, high-cost HF populations and justify prospective evaluation including full operational costing and broader healthcare resource capture.
