Dataset Information

Multi-modality machine learning approach for risk stratification in heart failure with left ventricular ejection fraction ≤ 45.

ABSTRACT: Heart failure (HF) involves complex remodelling leading to electrical and mechanical dysfunction. We hypothesized that machine learning approaches incorporating data obtained from different investigative modalities including atrial and ventricular measurements from electrocardiography and echocardiography, blood inflammatory marker [neutrophil-to-lymphocyte ratio (NLR)], and prognostic nutritional index (PNI) will improve risk stratification for adverse outcomes in HF compared to logistic regression. Consecutive Chinese patients referred to our centre for transthoracic echocardiography and subsequently diagnosed with HF, between 1 January 2010 and 31 December 2016, were included in this study. Two machine learning techniques, multilayer perceptron and multi-task learning, were compared with logistic regression for their ability to predict incident atrial fibrillation (AF), transient ischaemic attack (TIA)/stroke, and all-cause mortality. This study included 312 HF patients [mean age: 64 (55-73) years, 75% male]. There were 76 cases of new-onset AF, 62 cases of incident TIA/stroke, and 117 deaths during follow-up. Univariate analysis revealed that age, left atrial reservoir strain (LARS) and contractile strain (LACS) were significant predictors of new-onset AF. Age and smoking predicted incident stroke. Age, hypertension, type 2 diabetes mellitus, chronic kidney disease, mitral or aortic regurgitation, P-wave terminal force in V1, the presence of partial inter-atrial block, left atrial diameter, ejection fraction, global longitudinal strain, serum creatinine and albumin, high NLR, low PNI, and LARS and LACS predicted all-cause mortality. Machine learning techniques achieved better prediction performance than logistic regression. Multi-modality assessment is important for risk stratification in HF. A machine learning approach provides additional value for improving outcome prediction.

SUBMITTER: Tse G

PROVIDER: S-EPMC7754744 | biostudies-literature | 2020 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Multi-modality machine learning approach for risk stratification in heart failure with left ventricular ejection fraction ≤ 45.

Tse Gary G Zhou Jiandong J Woo Samuel Won Dong SWD Ko Ching Ho CH Lai Rachel Wing Chuen RWC Liu Tong T Liu Yingzhi Y Leung Keith Sai Kit KSK Li Andrew A Lee Sharen S Li Ka Hou Christien KHC Lakhani Ishan I Zhang Qingpeng Q

ESC heart failure 20201023 6

<h4>Aims</h4>Heart failure (HF) involves complex remodelling leading to electrical and mechanical dysfunction. We hypothesized that machine learning approaches incorporating data obtained from different investigative modalities including atrial and ventricular measurements from electrocardiography and echocardiography, blood inflammatory marker [neutrophil-to-lymphocyte ratio (NLR)], and prognostic nutritional index (PNI) will improve risk stratification for adverse outcomes in HF compared to lo ...[more]

PMID: 33094925

Similar Datasets

Project description:Aims: Evidence-based guidelines for heart failure management depend mainly on current left ventricular ejection fraction (LVEF). However, fewer studies have examined the impact of prior LVEF. Patients may enter the heart failure with midrange ejection fraction (HFmrEF) category when heart failure with preserved ejection fraction (HFpEF) deteriorates or heart failure with reduced ejection fraction (HFrEF) improves. In this study, we examined the association between change in LVEF and adverse outcomes. Methods: HFmrEF patients with at least two or more echocardiograms 3 months apart at the First Affiliated Hospital of Dalian Medical University between September 1, 2015 and November 30, 2019 were identified. According to the prior LVEF, the subjects were divided into improved group (prior LVEF < 40%), stable group (prior LVEF between 40 and 50%), and deteriorated group (prior LVEF ≥ 50%). The primary outcomes were cardiovascular death, all-cause mortality, hospitalization for worsening heart failure, and composite event of all-cause mortality or all-cause hospitalization. Results: A total of 1,168 HFmrEF patients (67.04% male, mean age 63.60 ± 12.18 years) were included. The percentages of improved, stable, and deteriorated group were 310 (26.54%), 334 (28.60%), and 524 (44.86%), respectively. After a period of follow-up, 208 patients (17.81%) died and 500 patients met the composite endpoint. The rates of all-cause mortality were 35 (11.29%), 55 (16.47%), and 118 (22.52%), and the composite outcome was 102 (32.90%), 145 (43.41%), and 253 (48.28%) for the improved, stable, and deteriorated groups, respectively. Cox regression analysis showed that the deterioration group had higher risk of cardiovascular death (HR: 1.707, 95% CI: 1.064-2.739, P = 0.027), all-cause death (HR 1.948, 95% CI 1.335-2.840, P = 0.001), and composite outcome (HR 1.379, 95% CI 1.096-1.736, P = 0.006) compared to the improvement group. The association still remained significant after fully adjusted for both all-cause mortality (HR = 1.899, 95% CI 1.247-2.893, P = 0.003) and composite outcome (HR: 1.324, 95% CI: 1.020-1.718, P = 0.035). Conclusion: HFmrEF patients are heterogeneous with three different subsets identified, each with different outcomes. Strategies for managing HFmrEF should include previously measured LVEF to allow stratification based on direction changes in LVEF to better optimize treatment.

Project description:ObjectivesTo determine clinical and prognostic differences between preserved and deteriorated systolic function (defined as left ventricular (LV) ejection fractions > or = 50% and < 50%, respectively) in patients with heart failure satisfying modified Framingham criteria.Patients and methodsRecords were studied of 1252 patients with congestive heart failure (CHF) (mean (SD) age 69.4 (11.7) years; 485 women, 767 men) who had been admitted to a cardiology service for CHF in the period 1991-2002 and whose LV systolic function had been echocardiographically evaluated within two weeks of admission. Data were collected on the main clinical findings, supplementary examinations, treatment, and duration of hospitalisation. Whether the patient was alive in the spring of 2003 was evaluated by searching the general archives of the hospital and by telephone survey.ResultsLV systolic function was preserved in 39.8% of patients. Age, female to male sex ratio, and prevalence of atrial fibrillation, valve disease, and other non-ischaemic, non-dilated cardiopathies were all significantly greater in the group with preserved systolic function. New York Heart Association functional class IV, third heart sound, jugular vein congestion, cardiomegaly, radiological signs of lung oedema, pathological Q waves, left bundle branch block, sinus rhythm, ischaemic cardiopathy, and dilated cardiomyopathy were all significantly more prevalent in the group with deteriorated systolic function, as was treatment with angiotensin converting enzyme inhibitors and most other antihypertensive drugs on discharge from hospital. There was no significant difference in survival between the groups with preserved and deteriorated systolic function (either survival regardless of age at admission or in subgroups aged < 75 and > or = 75 years at admission). In the whole group, survival rates after one, three, and five years were 84.0%, 66.7%, and 50.9%, respectively.ConclusionIn view of the poor prognosis of patients with CHF with preserved LV systolic function, who are currently treated empirically, it is to be hoped that relevant controlled clinical trials under way will afford information allowing optimisation of their treatment.

Dataset Information

Multi-modality machine learning approach for risk stratification in heart failure with left ventricular ejection fraction ≤ 45.

Publications

Multi-modality machine learning approach for risk stratification in heart failure with left ventricular ejection fraction ≤ 45.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets