Predicting in-hospital all-cause mortality in heart failure using machine learning

Background

The age of onset and causes of heart failure differ between high-income and low-and-middle-income countries (LMIC). Heart failure patients in LMIC also experience a higher mortality rate. Innovative ways that can risk stratify heart failure patients in this region are needed. The aim of this study was to demonstrate the utility of machine learning in predicting all-cause mortality in heart failure patients hospitalised in a tertiary academic centre.

Methods

Six supervised machine learning algorithms were trained to predict in-hospital all-cause mortality using data from 500 consecutive heart failure patients with a left ventricular ejection fraction (LVEF) less than 50%.

Results

The mean age was 55.2 ± 16.8 years. There were 271 (54.2%) males, and the mean LVEF was 29 ± 9.2%. The median duration of hospitalisation was 7 days (interquartile range: 4-11), and it did not differ between patients discharged alive and those who died. After a prediction window of 4 years (interquartile range: 2-6), 84 (16.8%) patients died before discharge from the hospital. The area under the receiver operating characteristic curve was 0.82, 0.78, 0.77, 0.76, 0.75, and 0.62 for random forest, logistic regression, support vector machines (SVM), extreme gradient boosting, multilayer perceptron (MLP), and decision trees, and the accuracy during the test phase was 88, 87, 86, 82, 78, and 76% for random forest, MLP, SVM, extreme gradient boosting, decision trees, and logistic regression. The support vector machines were the best performing algorithm, and furosemide, beta-blockers, spironolactone, early diastolic murmur, and a parasternal heave had a positive coefficient with the target feature, whereas coronary artery disease, potassium, oedema grade, ischaemic cardiomyopathy, and right bundle branch block on electrocardiogram had negative coefficients.

Conclusion

Despite a small sample size, supervised machine learning algorithms successfully predicted all-cause mortality with modest accuracy. The SVM model will be externally validated using data from multiple cardiology centres in South Africa before developing a uniquely African risk prediction tool that can potentially transform heart failure management through precision medicine.

Machine learning and statistical methods for predicting mortality in heart failure

Heart failure is a debilitating clinical syndrome associated with increased morbidity, mortality, and frequent hospitalization, leading to increased healthcare budget utilization. Despite the exponential growth in the introduction of pharmacological agents and medical devices that improve survival, many heart failure patients, particularly those with a left ventricular ejection fraction less than 40%, still experience persistent clinical symptoms that lead to an overall decreased quality of life. Clinical risk prediction is one of the strategies that has been implemented for the selection of high-risk patients and for guiding therapy. However, most risk predictive models have not been well-integrated into the clinical setting. This is partly due to inherent limitations, such as creating risk predicting models using static clinical data that does not consider the dynamic nature of heart failure. Another limiting factor preventing clinicians from utilizing risk prediction models is the lack of insight into how predictive models are built. This review article focuses on describing how predictive models for risk-stratification of patients with heart failure are built.

Patient-specific simulations for planning treatment in congenital heart disease

Patient-specific computational models have been extensively developed over the last decades and applied to investigate a wide range of cardiovascular problems. However, translation of these technologies into clinical applications, such as planning of medical procedures, has been limited to a few single case reports. Hence, the use of patient-specific models is still far from becoming a standard of care in clinical practice. The aim of this study is to describe our experience with a modelling framework that allows patient-specific simulations to be used for prediction of clinical outcomes. A cohort of 12 patients with congenital heart disease who were referred for percutaneous pulmonary valve implantation, stenting of aortic coarctation and surgical repair of double-outlet right ventricle was included in this study. Image data routinely acquired for clinical assessment were post-processed to set up patient-specific models and test device implantation and surgery. Finite-element and computational fluid dynamics analyses were run to assess feasibility of each intervention and provide some guidance. Results showed good agreement between simulations and clinical decision including feasibility, device choice and fluid-dynamic parameters. The promising results of this pilot study support translation of computer simulations as tools for personalization of cardiovascular treatments.

Cardiac Changes and Their Association with Fetuin-A and Fibroblast Growth Factor-23 in Children with Chronic Kidney Disease

AIMS

In children with chronic kidney disease (CKD), fetuin-A and fibroblast growth factor-23 (FGF-23) have been implicated in the mechanism and progression of several cardiac changes. This study aimed to determine the types and rates of cardiac changes in children with CKD and their association with fetuin-A, FGF-23, and other cardiovascular risk factors (CVRFs).

METHODS

This comparative cross-sectional study recruited 88 children (5-18 years): 27 CKD I with a glomerular filtration rate (GFR) >90 mL/min/1.73 m2 and 61 with a GFR of <90 mL/min/1.73 m2 (29 CKD II-IV, 32 CKD V-dialysis). Each patient had a short demographic and clinical history taken along with a physical examination. Blood was taken and sent for routine tests and for fetuin-A and FGF-23 assay. All patients had an echocardiogram to evaluate cardiac structure and function.

RESULTS

The distribution of left atrial diameter (LAD) and left ventricular (LV) mass differed significantly (p < 0.05) across the different CKD groups. Abnormal LAD was seen in 10% of patients; LV hypertrophy (LVH) in 27%; LV systolic dysfunction in 6% and diastolic dysfunction in 1 patient. Fetuin-A was the only independent predictor for abnormal LAD; mean arterial pressure was independently associated with concentric LVH, and age and hypoalbuminemia with eccentric LVH. Overall, the dialysis group had the highest rate of cardiac changes and associated risk factors.

CONCLUSION

Though not common, the importance of left atrial changes in children with CKD is highlighted along with the need to address modifiable CVRFs such as hypertension and hypoalbuminemia.

Using 4D Cardiovascular Magnetic Resonance Imaging to Validate Computational Fluid Dynamics: A Case Study

Computational fluid dynamics (CFD) can have a complementary predictive role alongside the exquisite visualization capabilities of 4D cardiovascular magnetic resonance (CMR) imaging. In order to exploit these capabilities (e.g., for decision-making), it is necessary to validate computational models against real world data. In this study, we sought to acquire 4D CMR flow data in a controllable, experimental setup and use these data to validate a corresponding computational model. We applied this paradigm to a case of congenital heart disease, namely, transposition of the great arteries (TGA) repaired with arterial switch operation. For this purpose, a mock circulatory loop compatible with the CMR environment was constructed and two detailed aortic 3D models (i.e., one TGA case and one normal aortic anatomy) were tested under realistic hemodynamic conditions, acquiring 4D CMR flow. The same 3D domains were used for multi-scale CFD simulations, whereby the remainder of the mock circulatory system was appropriately summarized with a lumped parameter network. Boundary conditions of the simulations mirrored those measured in vitro. Results showed a very good quantitative agreement between experimental and computational models in terms of pressure (overall maximum % error = 4.4% aortic pressure in the control anatomy) and flow distribution data (overall maximum % error = 3.6% at the subclavian artery outlet of the TGA model). Very good qualitative agreement could also be appreciated in terms of streamlines, throughout the cardiac cycle. Additionally, velocity vectors in the ascending aorta revealed less symmetrical flow in the TGA model, which also exhibited higher wall shear stress in the anterior ascending aorta.

Reduced ascending aorta distensibility relates to adverse ventricular mechanics in patients with hypoplastic left heart syndrome: noninvasive study using wave intensity analysis

OBJECTIVE

To evaluate the aortic arch elastic properties and ventriculoarterial coupling efficiency in patients with single ventricle physiology, with and without a surgically reconstructed arch.

METHODS

We studied 21 children with single ventricle physiology after bidirectional superior cavopulmonary surgery: 10 with hypoplastic left heart syndrome, who underwent surgical arch reconstruction, and 11 with other types of single ventricle physiology but without arch reconstruction. All children underwent pre-Fontan magnetic resonance imaging. No patient exhibited aortic recoarctation. Data on aortic wave speed, aortic distensibility and wave intensity profiles were all extracted from the magnetic resonance imaging studies using an in-house-written plug-in for the Digital Imaging and Communications in Medicine viewer OsiriX.

RESULTS

Children with hypoplastic left heart syndrome had significantly greater wave speed (P = .002), and both stiffer (P = .004) and larger (P < .0001) ascending aortas than the patients with a nonreconstructed arch. Aortic distensibility was not influenced by ventricular stroke volume but depended on a combination of increased aortic diameter and abnormal wall mechanical properties. Those with hypoplastic left heart syndrome had a lower peak wave intensity and reduced energy carried by the forward compression and the forward expansion waves, even after correction for stroke volume, suggesting an abnormal systolic and diastolic function. Lower wave energy was associated with an increased aortic diameter.

CONCLUSIONS

Using a novel, noninvasive technique based on image analysis, we have demonstrated that aortic arch reconstruction in children with hypoplastic left heart syndrome is associated with reduced aortic distensibility and unfavorable ventricular-vascular coupling compared with those with single ventricle physiology without aortic arch reconstruction.

Cardiovascular magnetic resonance findings in repaired anomalous left coronary artery to pulmonary artery connection (ALCAPA)

BACKGROUND

Anomalous origin of the left coronary artery from the pulmonary artery (ALCAPA) is a rare coronary artery anomaly. This study shows the role of cardiovascular magnetic resonance (CMR) in assessing young patients following surgical repair of ALCAPA.

METHODS

6 patients, aged 9-21 years, with repaired ALCAPA (2 Tackeuchi method, 4 direct re-implantation) underwent CMR because of clinical suspicion of myocardial ischemia. Imaging used short and long axis cine images (assess ventricular function), late-gadolinium enhancement (LGE) (detect segmental myocardial fibrosis), adenosine stress perfusion (detect reversible ischaemia) and 3D whole-heart imaging (visualize proximal coronary arteries).

RESULTS

The left ventricular (LV) global systolic function was preserved in all patients (mean LV ejection fraction = 62.7% ± 4.23%). The LV volumes were within the normal ranges, (mean indexed LVEDV = 75.4 ± 3.5 ml/m², LVESV = 31.6 ± 9.4 ml/m²). In 1 patient, hypokinesia of the anterior segments was visualized. Five patients showed sub-endocardial LGE involving the basal, antero-lateral wall and the anterior papillary muscle. Three patients had areas of reversible ischemia. In these 3, 3D whole-heart MRA showed that the proximal course of the left coronary artery was occluded (confirmed with cardiac catheterisation).

CONCLUSIONS

CMR is a good, non-invasive, radiation-free investigation in the post-surgical evaluation of ALCAPA. In referred patients we show that basal, antero-lateral sub-endocardial myocardial fibrosis is a characteristic finding. Furthermore, stress adenosine CMR perfusion, can identify reversible ischemia in this group, and was indicative of left coronary artery occlusion.

Validation of an animal model of right ventricular dysfunction and right bundle branch block to create close physiology to postoperative tetralogy of Fallot

BACKGROUND

In the past 5 years a few number of studies and case reports have come out focusing on biventricular (BiV) stimulation for treatment of congenital heart disease related ventricular dysfunction. The few available studies include a diverse group of pathophysiological entities ranging from a previously repaired tetralogy of Fallot (TOF) to a functional single ventricle anatomy. Patient's status is too heterogeneous to build important prospective study. To well understand the implication of prolonged electromechanical dyssynchrony we performed a chronic animal model that mimics essential parameters of postoperative TOF.

METHODS

Significant pulmonary regurgitation, mild stenosis, as well as right ventricular outflow tract (RVOT) scars were induced in 15 piglets to mimic repaired TOF. 4 months after hemodynamics and dyssynchrony parameters were compared with a control group and with a population of symptomatic adult with repaired TOF.

RESULTS

Comparing the animal model with the animal control group on echocardiography, RV dilatation, RV and LV dysfunction, broad QRS complex and dyssynchrony were observed on the animal model piglets. Moreover, epicardial electrical mapping showed activation consistent with a right bundle branch block. The animal models displayed the same pathophysiological parameters as the post TOF repair patients in terms of QRS duration, pulmonary regurgitation biventricular dysfunction and dyssynchrony.

CONCLUSION

This chronic swine model mimics electromechanical ventricular activation delay, RV and LV dysfunction, as in adult population of repair TOF. It does appear to be a very useful and interesting model to study the implication of dyssynchrony and the interest of resynchronization therapy in TOF failing ventricle.