Heart Failure Mortality Prediction Research Articles

A deep learning system for heart failure mortality prediction.

Heart failure (HF) is the final stage of the various heart diseases developing. The mortality rates of prognosis HF patients are highly variable, ranging from 5% to 75%. Evaluating the all-cause mortality of HF patients is an important means to avoid death and positively affect the health of patients. But in fact, machine learning models are difficult to gain good results on missing values, high dimensions, and imbalances HF data. Therefore, a deep learning system is proposed. In this system, we propose an indicator vector to indicate whether the value is true or be padded, which fast solves the missing values and helps expand data dimensions. Then, we use a convolutional neural network with different kernel sizes to obtain the features information. And a multi-head self-attention mechanism is applied to gain whole channel information, which is essential for the system to improve performance. Besides, the focal loss function is introduced to deal with the imbalanced problem better. The experimental data of the system are from the public database MIMIC-III, containing valid data for 10311 patients. The proposed system effectively and fast predicts four death types: death within 30 days, death within 180 days, death within 365 days and death after 365 days. Our study uses Deep SHAP to interpret the deep learning model and obtains the top 15 characteristics. These characteristics further confirm the effectiveness and rationality of the system and help provide a better medical service.

Abstract 12569: Analysis of Deep Learning Models for Prediction of Heart Failure Mortality

Introduction: Heart failure (HF) is a leading cause of hospitalization, morbidity and mortality. Deep learning (DL) techniques appear to show promising results in risk stratification and prognosis in several conditions in medicine. However, few methods using DL exist to help quantitatively estimate prognosis of HF. We hypothesized that deep learning (DL) techniques could prognosis of HF using simple variables. We propose application of a custom-built deep-neural-network model to identify mortality in HF patients. Methods: Custom-built deep-neural-networks were assessed using survey data from 42,147 participants from the National Health and Nutrition Examination Survey 1999-2016 (NHANES). Variables were selected using clinical judgment and stepwise backward regressions to develop prediction models. We partitioned the data into training and testing sets and repetitive experiments. We then evaluated model performance based on discrimination and calibration including the area under the receiver-operator characteristics curve (C-statistics), balanced accuracy, probability calibration with sigmoid, and the Brier score, respectively. As sensitivity analyses, we examined results limited to cases with complete clinical information available. We validated models’ performance using Mount Sinai database. Results: Of 42,147 participants with 4,060 variables, 1,491 (3.5%) had HF and HF mortality was 51.8%. In validation cohort, of 26,333 HF patients, the mortality in HF patients was 405 (1.5%). Final model using only 20 variables (age, race, gender, BMI, smoking, alcohol consumption, HTN, COPD, SBP, DBP, HR, HDL, LDL, CRP, A1C, BUN, creatinine, hemoglobin, sodium level, on statin) was tested. A state-of-the-art deep learning models achieved high accuracy for predicting mortality in HF patients with an AUC of 0.96 (95% CI: 0.95-0.99) in the first cohort and AUC of 0.93 (95% CI: 0.91-0.96) in validation cohort. Conclusions: A deep neural network model has shown to have high predictive accuracy and discriminative and calibrative power for prediction of HF mortality. Further research can delineate the clinical implications of DL in predicting HF mortality.

Quantitative determination of circulating L-cartinine and its derivates contributes to Heart failure diagnosis, etiology discrimination and clinical prognosis prediction

Abstract Background Despite the latest progress in heart failure therapy, early diagnosis and clinical prognosis prediction are still critical issues nowadays. It has been proved that carnitines play an essential role in fatty acid metabolism. However, it is unclear about the changes and clinical effects of circulating carnitines in heart failure. Objectives This study was designed to clarify the alteration of serum carnitine and its derivates in heart failure patients, and to verify the impact of carnitines on heart failure etiology discrimination and mortality prediction. Methods A total of 161 heart failure patients (Dilated cardiomyopathy: DCM, n=98; ischemia cardiomyopathy: ICM, n=63) and control patients (n=48) were enrolled from Feb to Sep in 2017. Serum L-carnitines were quantitatively measured by liquid chromatography/ mass spectrometry. All patients underwent follow-up (mean 30.8 months). Multi-variable Cox survival was performed to verify the impact of carnitines on heart failure mortality prediction. Results A total of 27 different carnitine derivates were detected. Compared with control group, 26 types of carnitines were increased significantly in heart failure patients. Several circulating carnitines were independent biomarkers for heart failure even adjusted by multi-variable logistic analysis. We also found 7 carnitines were obviously increased in DCM group than those in ICM group. Isobutyryl-L-carnitine and stearoyl-L-carnitine were independently associated with higher probability of DCM than ICM. DCM prediction model established by adding carnitines (isobutyryl-L-carnitine and stearoyl-L-carnitine) to age, serum creatinine and left ventricular ejection fraction,had favorable discrimination (C-index = 0.832, P<0.01, Figure 1A and B) and calibration efficiency (Hosmer-Lemeshow χ2=7.376, P=0.497>0.05). Meanwhile, a total of 43 mortality event occurred, 18 death (31.6%) in ICM group and 25 (27.2%) in DCM group. Independent clinical risk factors for the occurrence of mortality were serum creatinine >2mg/dl, left atrial diameter 0.55mm and N-terminal pro-B-type natriuretic peptide >4000 pg/ml. Using multi-variable COX survival analysis simultaneously adjusted by serum creatinine, left atrial diameter, NT-pro-BNP and age, oleoyl L-carnitine >300nmo/L (HR=2.364, 95% CI: 1.122–4.976, P=0.024) and isovaleryl-L-carnitine <100nmol/L (HR=2.108, 95% CI: 1.091–4.074, P=0.026) were also independently associated with higher mortality. Conclusions As one of critical participants in fatty acid metabolism, L-carnitines alteration not only differentiates DCM patients from ICM ones, but also independently predicts the risk of long-term mortality in heart failure patients. Figure 1 Funding Acknowledgement Type of funding source: Public grant(s) – National budget only. Main funding source(s): National Natural Science Foundation of China, Grant of Shanghai Municipal Commission of Health and Family Planning

Feature rearrangement based deep learning system for predicting heart failure mortality

Background and objective: Heart Failure is a clinical syndrome commonly caused by any structural or functional impairment. Fast and accurate mortality prediction for Heart Failure is essential to improve the health care of patients and prevent them from death. However, due to the imbalance problem and poor feature representation in Heart Failure data, mortality prediction of Heart Failure is difficult with some simple models. To handle these problems, this study is focused on proposing a fast and accurate Heart Failure mortality prediction framework.Methods: This paper proposes a feature rearrangement based deep learning system for heart failure mortality prediction. The proposed framework improves the performance of predicting heart failure mortality by handling imbalance problem and achieving better feature representation. This paper also proposes a method named Feature rearrangement based convolutional layer, which demonstrates that the order of the input features is essential for the convolutional network.Results: The proposed system is experimentally evaluated on real-world Heart Failure data collected from the EHR system of Shanghai Shuguang Hospital, where 10,198 in-patients records are extracted between March 2009 and April 2016. Internal comparison results illustrate that the proposed framework achieves the best performance for Heart Failure mortality prediction. Extensive experimental results compared with other machine learning methods demonstrate that the proposed method has the highest average accuracy and area under the curve while predicting the three goals of in-hospital mortality, 30-day mortality, and 1-year mortality. Finally, top 12 essential clinical features are mined with their chi-square scores, which can help to assist clinicians in the treatment and research of heart failure.Conclusions: The proposed method successfully predict different target in three observation windows. Feature rearrangement based convolutional layer and Focal loss are employed into the proposed framework, which helps promote the prediction accuracy of Heart Failure death. The proposed method is fast and accurate for predicting heart failure mortality, especially for imbalance situation. This paper also provide a reasonable pipeline to model EHRs data and handle imbalance problem in medical data.

Heart failure mortality prediction using PRISM score and development of a classification and regression tree model to refer patients for palliative care consultation

IntroductionWe sought to assess one-year mortality in heart failure (HF) patients by using (Placement Resource Indicator for Systems Management) PRISM, a disease nonspecific risk stratification score, and use it along with modified Seattle Heart Failure Model (SHFM) to guide patient selection for palliative care consultation. MethodsA retrospective study design was used to examine 1-year mortality in 689 HF patients admitted from 2012 to 2014. One-year mortality was calculated using Pmort30/PRISM and modified SHFM scores, and the predicted scores were validated using the area under the ROC curve. CART was used to develop an algorithm to classify patients based on their mortality risk. ResultsThe discriminatory ability of PRISM categorical score (AUC = 0.701) was not significantly different than the discriminatory ability of modified SHFM (AUC = 0.686) (DeLong's test p = 0.56) but improved significantly with the combination of PRISM (categorical) score + modified SHFM (AUC = 0.740) (p = 0.002). The predictive capability of the CART tree model after cross-validation was 72.2% (AUC 0.631). ConclusionOur study suggests PRISM score performed as well as modified SHFM for one-year mortality prediction. Moreover, the addition of modified SHFM to PRISM score increases discriminatory ability in predicting 1-year mortality in heart failure patients compared to either of the two models alone. Together, when combined in a CART model, they can be used to identify the population subset with the highest mortality risk and hence guide goals of care discussion.

Weight-based multiple empirical kernel learning with neighbor discriminant constraint for heart failure mortality prediction.

Heart Failure (HF) is one of the most common causes of hospitalization and is burdened by short-term (in-hospital) and long-term (6-12month) mortality. Accurate prediction of HF mortality plays a critical role in evaluating early treatment effects. However, due to the lack of a simple and effective prediction model, mortality prediction of HF is difficult, resulting in a low rate of control. To handle this issue, we propose a Weight-based Multiple Empirical Kernel Learning with Neighbor Discriminant Constraint (WMEKL-NDC) method for HF mortality prediction. In our method, feature selection by calculating the F-value of each feature is first performed to identify the crucial clinical features. Then, different weights are assigned to each empirical kernel space according to the centered kernel alignment criterion. To make use of the discriminant information of samples, neighbor discriminant constraint is finally integrated into multiple empirical kernel learning framework. Extensive experiments were performed on a real clinical dataset containing 10, 198 in-patients records collected from Shanghai Shuguang Hospital in March 2009 and April 2016. Experimental results demonstrate that our proposed WMEKL-NDC method achieves a highly competitive performance for HF mortality prediction of in-hospital, 30-day and 1-year. Compared with the state-of-the-art multiple kernel learning and baseline algorithms, our proposed WMEKL-NDC is more accurate on mortality prediction Moreover, top 10 crucial clinical features are identified together with their meanings, which are very useful to assist clinicians in the treatment of HF disease.

Visual Analytics for Congestive Heart Failure Mortality Prediction.

Several indices exist to classify Congestive Heart Failure (CHF) patients' propensity for early mortality; however, they are primarily based on limited data and are not intuitive to use at the point of care. We investigate a novel, data-driven, risk assessment and visualization approach to investigate mortality prediction of CHF patients using data retrieved from an intensively digitized hospital's data repository. Combining well-known, computationally efficient, dimensionality reduction (DR) methods with 2-d information visualization, the method classifies and visualizes CHF patients into high and low risk groups, contextualized by the factors driving their classification. The DR method performed similar to logistic regression (LR), but visualized the classification and its significant factors at the population level, individual level and the potential impact of interventions for an individual patient. These are encouraging results in favor of the proposed visualization approach, and contributes to the current focus on advancing patient care via large-scale visual analytics.

Predictive Modeling of Hospital Mortality for Patients With Heart Failure by Using an Improved Random Survival Forest

Identification of different risk factors and early prediction of mortality for patients with heart failure are crucial for guiding clinical decision-making in Intensive care unit cohorts. In this paper, we developed a comprehensive risk model for predicting heart failure mortality with a high level of accuracy using an improved random survival forest (iRSF). Utilizing a novel split rule and stopping criterion, the proposed iRSF was able to identify more accurate predictors to separate survivors and nonsurvivors and thus improve discrimination ability. Based on the public MIMIC II clinical database with 8059 patients, 32 risk factors, including demographics, clinical, laboratory information, and medications, were analyzed and used to develop the risk model for patients with heart failure. Compared with previous studies, more critical laboratory predictors were identified that could reveal difficult-to-manage comorbidities, including aspartate aminotransferase, alanine aminotransferase, total bilirubin, serum creatine, blood urea nitrogen, and their inherent effects on events; these were determined to be critical indicators for predicting heart failure mortality with the proposed iRSF. The experimental results showed that the developed risk model was superior to those used in previous studies and the conventional random survival forest-based model with an out-of-bag C-statistic value of 0.821. Therefore, the developed iRSF-based risk model could serve as a valuable tool for clinicians in heart failure mortality prediction.

Abstract 12899: A New Prognostic Marker for Functional Deterioration in Heart Failure: Coronary Vascular Resistance

Introduction: Heart failure (HF) is a challenging clinical syndrome with complex pathophysiology. Previous studies demonstrated that coronary flow reserve assessed by PET technique was impaired in HF. Transthoracic echocardiography (TTE) can provide reliable data for myocardial blood flow (MBF) which can be measured from coronary sinus (CS). In relation with MBF, coronary vascular resistance (CVR) can be evaluated. In this study, we aimed to evaluate the association of MBF and CVR with clinical data in HF. Methods: The MBF was measured in 100 patients (80 patients with HF (LVEF<30%) and 20 healthy control subjects. MBF was measured from anterograde CS flow TVI at PLAX RV inflow view and CS diameter at A2C view (2Πr2*TVI*heart rate) and indexed to left ventricular mass index. Patient group was categorized based on NYHA functional classes (NYHA FCs). Results: There was an incremental trend for MBF with advancing NYHA FCs (β=0.253, p=0.015). MBF was similar among HF patients due to NYHA FCs. However, MBF was higher in NYHA FC 4 group compared to healthy subjects (p=0.029). Also, CVR was lower in HF patients than the control group (11.15±6.54 vs.24.84±7.29, p<0.001). In the subgroup analysis, CVR was not statistically different between non-ischemic and ischemic etiology of HF (6.91±1.38 vs. 6.31±0.89, p=0.249). Additionally, CVR was significantly lower in NYHA FC 4 compared to other study groups. CVR was positively correlated with peak VO2 (β=0.596, p<0.001), and Seattle Heart Failure Model derived life expectancy (β=0.476, p<0.001) and negatively correlated with serum BNP level (β=-0.450, p=0.001), troponin I (β=-0.366, p=0.001) and serum uric acid level (β=-0.503, p<0.001). In multivariate analysis, CVR is an independent predictor of mortality (HR:0.763 95%CI:0.624-0.933 P=0.008). In ROC analysis, a cut of value 6.73 for CVR has a 74% sensitivity and 66.7 % specifity for prediction of mortality in HF(AUC 0.825 P<0.001). Conclusion: Our results showed that CVR was reduced with advancing FC in HF patients, despite MBF was similar among different FCs. In addition, CVR was significantly correlated with various prognostic markers in HF patients. Therefore, evaluation of MBF hemodynamics, particularly CVR, with TTE may utilize prognostic assessment of HF patients.

Prediction of Heart Failure Mortality in Emergent Care

Heart failure contributes to millions of emergency department (ED) visits, but hospitalization-versus-discharge decisions are often not accompanied by prognostic risk quantification. To derive and validate a model for acute heart failure mortality applicable in the ED. Clinical data abstraction with development of a broadly applicable multivariate risk index for 7-day death using initial vital signs, clinical and presentation features, and readily available laboratory tests. Multicenter study of 86 hospitals in Ontario, Canada. Population-based random sample of 12 591 patients presenting to the ED from 2004 to 2007. Death within 7 days of presentation. In the derivation cohort (n = 7433; mean age, 75.4 years [SD, 11.4]; 51.5% men), mortality risk increased with higher triage heart rate (adjusted odds ratio [OR], 1.15 [95% CI, 1.03 to 1.30] per 10 beats/min) and creatinine concentration (OR, 1.35 [CI, 1.14 to 1.60] per 1 mg/dL [88.4 µmol/L]), and lower triage systolic blood pressure (OR, 1.52 [CI, 1.31 to 1.77] per 20 mm Hg) and initial oxygen saturation (OR, 1.16 [CI, 1.01 to 1.33] per 5%). Nonnormal serum troponin levels (OR, 2.75 [CI, 1.86 to 4.07]) were associated with increased mortality risk. Areas under the receiver-operating characteristic curves of the multivariate model were 0.805 for the derivation data set (bootstrap-corrected, 0.811) and 0.826 for validation data set (n = 5158; mean age, 75.7 years [SD, 11.4]; 51.6% men). In the derivation cohort, a multivariate index score stratified 7-day mortality with rates of 0.3%, 0.3%, 0.7%, and 1.9% in quintiles 1 to 4, respectively. Mortality rates in the 2 highest risk groups were 3.5% and 8.2% in deciles 9 and 10, respectively. Left ventricular ejection fraction was not included in the model. A multivariate index comprising routinely collected variables stratified mortality risk with high discrimination in a broad group of patients with acute heart failure presenting to the ED. Canadian Institutes of Health Research.

88: Acute Heart Failure Mortality Prediction Using Copeptin: Results of the Biomarkers in ACute Heart Failure Trial

Emergency physicians have few tools providing objective accurate prediction of short term mortality risk in patients presenting to the emergency department (ED) with acute heart failure (AHF). Our purpose was to describe whether plasma C-terminal pro-vasopressin fragment (Copeptin), a surrogate for circulating arginine vasopressin (AVP), is associated with short term death rates in ED patients presenting with AHF.

Abstract 1252: A Multimarker Strategy for Risk Prediction in Heart Failure: Application of Novel Methods in a Community Cohort

Introduction: Mortality among patients with heart failure (HF) is high. Though individual biomarkers have been studied to determine their value in risk prediction, the role of a multimarker strategy in predicting mortality above established risk indicators requires evaluation in a relevant cohort using robust analytical methods. Objective: To determine whether adding C-reactive protein (CRP), B-type natriuretic peptide (BNP), and troponin T (TnT) to established risk indicators improved HF mortality risk prediction. Methods: Olmsted County residents presenting with HF from July 2004 to September 2007 were prospectively recruited to undergo biomarker measurement. We investigated whether addition of CRP, BNP, and TnT to a model including established risk indicators improved 6-month mortality risk prediction using multiple methods ( c statistic, net reclassification index, integrated discrimination improvement). Results: Of 566 participants (mean age 77±13 years, 48% male), 87(15%) patients died after 6 months follow-up. Patients with CRP (<12.1mg/L), BNP (<341pg/mL), and TnT (≥0.01ng/mL) below the median (n=112) had the lowest mortality (3%), those with one or two biomarkers above the median had intermediate mortality, and those with all biomarkers above the median (n=105) had markedly increased 6-month mortality (33%). After adjustment for other predictors, higher CRP, BNP, and TnT were each associated with increased mortality, with odds ratios (95% confidence intervals) of 1.66(1.24 –2.24), 2.06(1.42–3.07), and 1.45(1.12–1.87), respectively. While adding CRP, BNP, and TnT individually to the model improved mortality risk prediction, the combination of two or more biomarkers offered greater incremental benefit. Using all biomarkers resulted in an increase in the c statistic from 0.747 to 0.832 (p<0.001), a net reclassification index of 0.33 (p<0.001), and a 100% relative improvement in risk discrimination (integrated discrimination improvement 0.097, p<0.001). Conclusions: The combined use of CRP, BNP, and TnT markedly improved the prediction of 6-month mortality in HF beyond established risk indicators. Implementation of a multimarker strategy including CRP, BNP, and TnT for risk prediction in HF should be considered. This research has received full or partial funding support from the American Heart Association, AHA Midwest Affiliate (Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, South Dakota & Wisconsin).