scholarly journals Prediction of all-cause mortality in coronary artery disease patients with atrial fibrillation based on machine learning models

2021 ◽  
Vol 21 (1) ◽  
Author(s):  
Xinyun Liu ◽  
Jicheng Jiang ◽  
Lili Wei ◽  
Wenlu Xing ◽  
Hailong Shang ◽  
...  
Keyword(s):  
Machine Learning ◽  
Atrial Fibrillation ◽  
Coronary Artery Disease ◽  
Logistic Regression ◽  
Regression Model ◽  
Logistic Regression Model ◽  
Support Vector ◽  
Vector Machines ◽  
All Cause Mortality ◽  
Artery Disease

Abstract Background Machine learning (ML) can include more diverse and more complex variables to construct models. This study aimed to develop models based on ML methods to predict the all-cause mortality in coronary artery disease (CAD) patients with atrial fibrillation (AF). Methods A total of 2037 CAD patients with AF were included in this study. Three ML methods were used, including the regularization logistic regression, random forest, and support vector machines. The fivefold cross-validation was used to evaluate model performance. The performance was quantified by calculating the area under the curve (AUC) with 95% confidence intervals (CI), sensitivity, specificity, and accuracy. Results After univariate analysis, 24 variables with statistical differences were included into the models. The AUC of regularization logistic regression model, random forest model, and support vector machines model was 0.732 (95% CI 0.649–0.816), 0.728 (95% CI 0.642–0.813), and 0.712 (95% CI 0.630–0.794), respectively. The regularization logistic regression model presented the highest AUC value (0.732 vs 0.728 vs 0.712), specificity (0.699 vs 0.663 vs 0.668), and accuracy (0.936 vs 0.935 vs 0.935) among the three models. However, no statistical differences were observed in the receiver operating characteristic (ROC) curve of the three models (all P > 0.05). Conclusion Combining the performance of all aspects of the models, the regularization logistic regression model was recommended to be used in clinical practice.

The impact of atrial fibrillation on hospitalization outcomes for patients with chronic lymphocytic leukemia: A large database analysis.

2021 ◽  
Vol 39 (15_suppl) ◽  
pp. e19508-e19508
Author(s):  
Mohammad Ammad Ud Din ◽  
Samarthkumar Thakkar ◽  
Harsh P. Patel ◽  
Syed Ather Hussain ◽  
Aneeqa Zafar ◽  
...  
Keyword(s):  
Heart Failure ◽  
Atrial Fibrillation ◽  
Logistic Regression ◽  
Chronic Lymphocytic Leukemia ◽  
Regression Model ◽  
Logistic Regression Model ◽  
Lymphocytic Leukemia ◽  
Total Cost ◽  
Patient Admissions ◽  
All Cause Mortality

e19508 Background: With the increased use of novel agents like Bruton tyrosine kinase inhibitors (BTKi) for the treatment of chronic lymphocytic leukemia (CLL), the incidence of atrial fibrillation (AF) is on the rise in these patients. However, the excess burden added by AF to the morbidity and mortality of CLL patients is unclear. Methods: Using the appropriate ICD-9 and ICD-10 codes, the National Inpatient Sample (NIS) database was accessed to gather data of hospitalized CLL patients with AF from 2008 to 2019. Propensity-score matching (PSM) and logistic regression model were performed to control for baseline patient factors like age, sex, income, and the relevant co-morbidities to match 7265 CLL patient admissions with AF and 7265 CLL patient admissions without AF. The primary outcome was all-cause mortality (ACM), while secondary outcomes included stroke, acute heart failure (AHF), and total cost of hospital stay. Results: The mean age of the cohorts was 82 years. Females made up 44% of both groups. The AF group had similar prevalence of systemic hypertension (62.38% vs 62.10%; p= 0.73), diabetes mellitus (5.09% vs 5.43%; p= 0.35), congestive heart failure (5.57% vs 5.36%; p= 0.58), valvular heart disease (1.17% vs 1.44%; p= 0.14), and pulmonary hypertension (0.21% vs 0.14%; p= 0.31) compared to the group without AF. PSM revealed CLL patients with AF had a higher rate of ACM (6.06% vs 4.47%; p= <0.0001), AHF (7.50% vs 3.85%; p= <0.001), and stroke (3.09% vs 1.65%; p= <0.0001). Admission in the AF group also had a higher median total cost of hospital stay ($9097 vs $7646). A logistic regression model was done to adjust for confounders which revealed similar results for the AF group with increased adjusted odd’s ratio (aOR) of ACM (aOR:1.39, 95% confidence interval (CI): 1.19-1.61; p= <0.001), AHF (aOR: 2.16, 95% CI: 1.85-2.52; p= <0.001), and stroke (aOR:1.94, 95% CI: 1.54-2.44; P= <0.001) (Table). Conclusions: Our data suggest that hospitalized CLL patients with AF are at a significantly increased risk of all-cause mortality, AHF, and stroke. Several limitations like the inability to establish the temporal relationship between CLL and AF and the lack of data regarding medications of individual patients are important to keep in mind while interpreting the results.[Table: see text]

P5710Clinical applications of machine learning for prediction of incident atrial fibrillation from the general population: a nationwide cohort study

2019 ◽  
Vol 40 (Supplement_1) ◽  
Author(s):  
I.-S Kim ◽  
P S Yang ◽  
H T Yu ◽  
T H Kim ◽  
J S Uhm ◽  
...  
Keyword(s):  
Machine Learning ◽  
Atrial Fibrillation ◽  
Logistic Regression ◽  
General Population ◽  
Regression Model ◽  
National Health ◽  
Logistic Regression Model ◽  
Learning System ◽  
Health Examination ◽  
Clinical Variables

Abstract Background To evaluate the ability of machine learning algorithms to predict incident atrial fibrillation (AF) from the general population using health examination items. Methods We included 483,343 subjects who received national health examinations from the Korean National Health Insurance Service-based National Sample Cohort (NHIS-NSC). We trained deep neural network model (DNN) of a deep learning system and decision tree model (DT) of a machine learning system using clinical variables and health examination items (including age, sex, body mass index, history of heart failure, hypertension or diabetes, baseline creatinine, and smoking and alcohol intake habits) to predict incident AF using a training dataset of 341,771 subjects constructed from the NHIS-NSC database. The DNN and DT were validated using an independent test dataset of 141,572 remaining subjects. C-indices of DNN and DT for prediction of incident AF were compared with that of conventional logistic regression model. Results During 1,874,789 person·years (mean±standard-deviation age 47.7±14.4 years, 49.6% male), 3,282 subjects with incident AF were observed. In the validation dataset, 1,139 subjects with incident AF were observed. The c-indices of the DNN and DT for incident AF prediction were 0.828 [0.819–0.836] and 0.835 [0.825–0.844], and were significantly higher (p<0.01) than conventional logistic regression model (c-index=0.789 [0.784–0.794]). Conclusions Application of machine learning using simple clinical variables and health examination items was helpful to predict incident AF in the general population. Prospective study is warranted to construct an individualized precision medicine.

Abstract 16195: Inverse Relationship Between Inflammation and Cardiorespiratory Fitness: The Henry Ford Hospital ExercIse Testing Project (FIT Project)

Circulation ◽  
2014 ◽  
Vol 130 (suppl_2) ◽  
Author(s):  
Waqas Qureshi ◽  
Michael J Blaha ◽  
Clinton Brawner ◽  
Steven Keteyian ◽  
Mouaz Al-Mallah
Keyword(s):  
Coronary Artery Disease ◽  
Logistic Regression ◽  
Coronary Artery ◽  
Regression Model ◽  
Exercise Capacity ◽  
Cardiorespiratory Fitness ◽  
Logistic Regression Model ◽  
Henry Ford ◽  
Hs Crp ◽  
Artery Disease

Introduction: High sensitivity-C reactive protein (hs-CRP) is a measure of the degree of inflammation and has been associated with risk of coronary artery disease. Hypothesis: Hs-CRP is independently inversely associated with cardiorespiratory fitness (CRF). Methods: This is a cross sectional study of patients without heart failure or known coronary artery disease underwent graded exercise stress test in Henry Ford Health System (Detroit) according to Bruce protocol and CRF was estimated as peak metabolic equivalents (METs). Hs-CRP was dichotomized by 75th percentile and the relationship between CRF and hs-CRP was assessed by unadjusted and adjusted logistic regression model for age, sex, race, hypertension, diabetes, hyperlipidemia, smoking, total cholesterol, HDL, aspirin use, statin use and obesity. Results: A total of 4381 patients (mean age 54±11 years, 58% male, 79% whites) were analyzed. The 75th percentile of hs-CRP was 4.2 mg/L. Mean hs-CRP was higher in women, blacks, hypertensives, diabetics, smokers, and obese patients. In unadjusted analysis, CRF was inversely associated with highest quartile of hs-CRP (OR per 1 MET increase in CRF 0.82; 95% CI 0.81 - 0.84, p <0.0001). In multiple logistic regression model, CRF remained inversely associated with hs-CRP (OR per 1 MET increase in CRF 0.83; 95% CI 0.80 - 0.86, p 12 METS) compared to individuals with poor exercise capacity (<6 METS) had a 47% lower risk of having highest quartile of hs-CRP (OR 0.53; 95% CI 0.46 - 0.62, p <0.0001). Mean hs-CRP level decreased with increasing category of CRF (figure). There was a negative linear relationship between log transformed hs-CRP and peak METs achieved (r = -0.32, p<0.0001). Conclusions: In conclusion, exercise capacity, as measured by estimated Mets, was associated with reduced inflammation as measured by hs-CRP. The underlying pathophysiology between hs-CRP and CRF warrants further investigation.

scholarly journals Machine Learning Models for Breast Lesions Based on Ultrasound Imaging Features: A Observational Study

2020 ◽  
Author(s):  
Yao Tan ◽  
Ling Huo ◽  
Shu Wang ◽  
Cuizhi Geng ◽  
Yi Li ◽  
...  
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Regression Model ◽  
Ultrasound Imaging ◽  
Logistic Model ◽  
Logistic Regression Model ◽  
Support Vector ◽  
Imaging Features ◽  
Data Set ◽  
Class A

Abstract Background: The accuracy of breast cancer (BC) screening based on conventional ultrasound imaging examination largely depends on the experience of clinicians. Further, the effectiveness of BC screening and diagnosis in primary hospitals need to be improved. This study aimed to establish and evaluate the usefulness of a simple, practical, and easy-to-promote machine learning model based on ultrasound imaging features for diagnosing BC.Methods: Logistic regression, random forest, extra trees, support vector, multilayer perceptron, and XG boost models were developed. The modeling data set was divided into a training set and test set in a 75%:25% ratio, and these were used to establish the models and test their performance, respectively. The validation data set of primary hospitals was used for external validation of the model. The area under the receiver operating characteristic curve (AUC) was used as the main evaluation index, and pathological biopsy was used as the gold standard for evaluating each model. Diagnostic capability was also compared with those of clinicians. Results: Among the six models, the logistic model showed superior capability, with an AUC of 0.771 and 0.906 in the test and validation sets, respectively, and Brier scores of 0.18 and 0.165. The AUC of the logistic model in tertiary class A hospitals and primary hospitals was 0.875 and 0.921, respectively. The AUCs of the clinician diagnosis and the logistic model were 0.913 and 0.906. Their AUCs in the tertiary class A hospitals were 0.890 and 0.875, respectively, and were 0.924 and 0.921 in primary hospitals, respectively. Conclusions: The logistic regression model has better overall performance in primary hospitals, and the logistic regression model can be further extended to the basic level. A more balanced clinical prediction model can be further established on the premise of improving accuracy to assist clinicians in decision making and improve diagnosis.Trial Registration: http://www.clinicaltrials.gov. ClinicalTrials.gov ID: NCT03080623.

scholarly journals Machine-Learning vs. Expert-Opinion Driven Logistic Regression Modelling for Predicting 30-Day Unplanned Rehospitalisation in Preterm Babies: A Prospective, Population-Based Study (EPIPAGE 2)

2021 ◽  
Vol 8 ◽  
Author(s):  
Robert A. Reed ◽  
Andrei S. Morgan ◽  
Jennifer Zeitlin ◽  
Pierre-Henri Jarreau ◽  
Héloïse Torchin ◽  
...  
Keyword(s):  
Machine Learning ◽  
Logistic Regression ◽  
Random Forest ◽  
Regression Model ◽  
Expert Opinion ◽  
Logistic Regression Model ◽  
Population Based ◽  
Regression Modelling ◽  
Preterm Babies ◽  
Logistic Regression Modelling

Introduction: Preterm babies are a vulnerable population that experience significant short and long-term morbidity. Rehospitalisations constitute an important, potentially modifiable adverse event in this population. Improving the ability of clinicians to identify those patients at the greatest risk of rehospitalisation has the potential to improve outcomes and reduce costs. Machine-learning algorithms can provide potentially advantageous methods of prediction compared to conventional approaches like logistic regression.Objective: To compare two machine-learning methods (least absolute shrinkage and selection operator (LASSO) and random forest) to expert-opinion driven logistic regression modelling for predicting unplanned rehospitalisation within 30 days in a large French cohort of preterm babies.Design, Setting and Participants: This study used data derived exclusively from the population-based prospective cohort study of French preterm babies, EPIPAGE 2. Only those babies discharged home alive and whose parents completed the 1-year survey were eligible for inclusion in our study. All predictive models used a binary outcome, denoting a baby's status for an unplanned rehospitalisation within 30 days of discharge. Predictors included those quantifying clinical, treatment, maternal and socio-demographic factors. The predictive abilities of models constructed using LASSO and random forest algorithms were compared with a traditional logistic regression model. The logistic regression model comprised 10 predictors, selected by expert clinicians, while the LASSO and random forest included 75 predictors. Performance measures were derived using 10-fold cross-validation. Performance was quantified using area under the receiver operator characteristic curve, sensitivity, specificity, Tjur's coefficient of determination and calibration measures.Results: The rate of 30-day unplanned rehospitalisation in the eligible population used to construct the models was 9.1% (95% CI 8.2–10.1) (350/3,841). The random forest model demonstrated both an improved AUROC (0.65; 95% CI 0.59–0.7; p = 0.03) and specificity vs. logistic regression (AUROC 0.57; 95% CI 0.51–0.62, p = 0.04). The LASSO performed similarly (AUROC 0.59; 95% CI 0.53–0.65; p = 0.68) to logistic regression.Conclusions: Compared to an expert-specified logistic regression model, random forest offered improved prediction of 30-day unplanned rehospitalisation in preterm babies. However, all models offered relatively low levels of predictive ability, regardless of modelling method.

Abstract 211: Association of Patient Characteristics With the Initiation of Dabigatran Versus Warfarin Among Anticoagulant Naive Patients With Non-valvular Atrial Fibrillation

2013 ◽  
Vol 6 (suppl_1) ◽  
Author(s):  
David R Walker ◽  
Jasmina Ivanova ◽  
Keith A Betts ◽  
Sapna Rao ◽  
Eric Q Wu
Keyword(s):  
Atrial Fibrillation ◽  
Logistic Regression ◽  
Regression Model ◽  
Primary Care Physicians ◽  
Online Survey ◽  
Logistic Regression Model ◽  
Stroke Risk ◽  
Dabigatran Etexilate ◽  
Patient Characteristics ◽  
Categorical Variables

Background and Objective: Dabigatran etexilate (DE) and warfarin, both oral anticoagulants used for stroke risk reduction in patients with non-valvular atrial fibrillation (NVAF), have been or are being compared in several comparative effectiveness studies. Understanding patient characteristics of those prescribed DE vs. warfarin are important for interpreting such studies. The objective of this study is to identify the characteristics that differentiate NVAF patients prescribed DE versus warfarin as first-line anticoagulation. Methods: An online survey was administered in October 2012 to an established panel of cardiologists and primary care physicians (PCPs) in the US. Physicians were asked to identify medical charts of their patients diagnosed with NVAF and who had at least one prescription for DE or warfarin between 1/1/2011 and 6/30/2012. Patients were further required to be anticoagulant naïve prior to the first prescription of DE or warfarin. A computer generated random dice was applied to direct the random selection of the patients. Patient characteristics, comorbidities and clinical risk measures were compared between DE and warfarin patients using Chi-square tests for categorical variables and t-tests for continuous variables. A logistic regression model was utilized to evaluate patient characteristics associated with DE vs. warfarin use among anticoagulant naïve NVAF patients. Results: A total of 288 physicians (144 cardiologists and 144 PCPs) completed the survey. 262 medical records for DE patients and 247 for warfarin patients were randomly selected. The mean age of the DE and warfarin patients, respectively were 61.6 and 65.8 years (p < 0.01). The proportion of females was 20.6% and 41.7% in the DE and warfarin patients respectively (p<0.01). 86.3% of DE patients vs. 68.4% of warfarin patients were Caucasian (p<0.01). Other differences between DE and warfarin patients respectively included: previous myocardial infarction (3.8%, 9.3%; p<0.05), previous transient ischemic attack (8.4%, 16.2%; p <0.01), and CHA 2 DS 2 -VASc stroke risk score (2.21, 2.98; p<0.01). The logistic regression model found age (OR = 0.96; p=0.001), female gender (OR=0.46; p = 0.002), Hispanic/Latino (OR = 0.33; p=.007), Black (OR= 0.37; p = 0.006), and > 6 months and < 1 year for time from first NVAF diagnosis to first prescription date (OR = 0.38; p = 0.02) were associated with initiation of DE vs. warfarin. However, CHA 2 DS 2 -VASc was not found to be a significant predictor of anticoagulant prescription. Conclusions: Patients who are younger, male, Caucasian, and recently diagnosed with NVAF were significantly more likely to be initiated by their physician on DE vs. warfarin. These findings should be considered when doing comparative analyses of outcomes between patients on DE vs. warfarin.

Sign in / Sign up

Export Citation Format

Share Document