Risk Factors for Post-Transplant Outcomes in Patients with LVAD Support: A Machine Learning and Logistic Regression of the UNOS Database

Abstract Background Despite a wide range of proposed risk factors and theoretical models, prediction of eating disorder (ED) onset remains poor. This study undertook the first comparison of two machine learning (ML) approaches [penalised logistic regression (LASSO), and prediction rule ensembles (PREs)] to conventional logistic regression (LR) models to enhance prediction of ED onset and differential ED diagnoses from a range of putative risk factors. Method Data were part of a European Project and comprised 1402 participants, 642 ED patients [52% with anorexia nervosa (AN) and 40% with bulimia nervosa (BN)] and 760 controls. The Cross-Cultural Risk Factor Questionnaire, which assesses retrospectively a range of sociocultural and psychological ED risk factors occurring before the age of 12 years (46 predictors in total), was used. Results All three statistical approaches had satisfactory model accuracy, with an average area under the curve (AUC) of 86% for predicting ED onset and 70% for predicting AN v. BN. Predictive performance was greatest for the two regression methods (LR and LASSO), although the PRE technique relied on fewer predictors with comparable accuracy. The individual risk factors differed depending on the outcome classification (EDs v. non-EDs and AN v. BN). Conclusions Even though the conventional LR performed comparably to the ML approaches in terms of predictive accuracy, the ML methods produced more parsimonious predictive models. ML approaches offer a viable way to modify screening practices for ED risk that balance accuracy against participant burden.

Download Full-text

Predicting the Risk of Hypertension Based on Several Easy-to-Collect Risk Factors: A Machine Learning Method

Frontiers in Public Health ◽

10.3389/fpubh.2021.619429 ◽

2021 ◽

Vol 9 ◽

Author(s):

Huanhuan Zhao ◽

Xiaoyu Zhang ◽

Yang Xu ◽

Lisheng Gao ◽

Zuchang Ma ◽

...

Keyword(s):

Machine Learning ◽

Risk Factors ◽

Logistic Regression ◽

Risk Prediction ◽

Disease Risk ◽

Learning Algorithms ◽

Large Population ◽

Machine Learning Algorithms ◽

Hypertension Risk ◽

Model Training

Hypertension is a widespread chronic disease. Risk prediction of hypertension is an intervention that contributes to the early prevention and management of hypertension. The implementation of such intervention requires an effective and easy-to-implement hypertension risk prediction model. This study evaluated and compared the performance of four machine learning algorithms on predicting the risk of hypertension based on easy-to-collect risk factors. A dataset of 29,700 samples collected through a physical examination was used for model training and testing. Firstly, we identified easy-to-collect risk factors of hypertension, through univariate logistic regression analysis. Then, based on the selected features, 10-fold cross-validation was utilized to optimize four models, random forest (RF), CatBoost, MLP neural network and logistic regression (LR), to find the best hyper-parameters on the training set. Finally, the performance of models was evaluated by AUC, accuracy, sensitivity and specificity on the test set. The experimental results showed that the RF model outperformed the other three models, and achieved an AUC of 0.92, an accuracy of 0.82, a sensitivity of 0.83 and a specificity of 0.81. In addition, Body Mass Index (BMI), age, family history and waist circumference (WC) are the four primary risk factors of hypertension. These findings reveal that it is feasible to use machine learning algorithms, especially RF, to predict hypertension risk without clinical or genetic data. The technique can provide a non-invasive and economical way for the prevention and management of hypertension in a large population.

Download Full-text

Prediction of incident myocardial infarction using machine learning applied to harmonized electronic health record data

BMC Medical Informatics and Decision Making ◽

10.1186/s12911-020-01268-x ◽

2020 ◽

Vol 20 (1) ◽

Author(s):

Divneet Mandair ◽

Premanand Tiwari ◽

Steven Simon ◽

Kathryn L. Colborn ◽

Michael A. Rosenberg

Keyword(s):

Neural Network ◽

Machine Learning ◽

Risk Factors ◽

Myocardial Infarction ◽

Logistic Regression ◽

Large Scale ◽

Deep Neural Network ◽

Low Frequency ◽

Common Data Model ◽

Electronic Health Record Data

Abstract Background With cardiovascular disease increasing, substantial research has focused on the development of prediction tools. We compare deep learning and machine learning models to a baseline logistic regression using only ‘known’ risk factors in predicting incident myocardial infarction (MI) from harmonized EHR data. Methods Large-scale case-control study with outcome of 6-month incident MI, conducted using the top 800, from an initial 52 k procedures, diagnoses, and medications within the UCHealth system, harmonized to the Observational Medical Outcomes Partnership common data model, performed on 2.27 million patients. We compared several over- and under- sampling techniques to address the imbalance in the dataset. We compared regularized logistics regression, random forest, boosted gradient machines, and shallow and deep neural networks. A baseline model for comparison was a logistic regression using a limited set of ‘known’ risk factors for MI. Hyper-parameters were identified using 10-fold cross-validation. Results Twenty thousand Five hundred and ninety-one patients were diagnosed with MI compared with 2.25 million who did not. A deep neural network with random undersampling provided superior classification compared with other methods. However, the benefit of the deep neural network was only moderate, showing an F1 Score of 0.092 and AUC of 0.835, compared to a logistic regression model using only ‘known’ risk factors. Calibration for all models was poor despite adequate discrimination, due to overfitting from low frequency of the event of interest. Conclusions Our study suggests that DNN may not offer substantial benefit when trained on harmonized data, compared to traditional methods using established risk factors for MI.

Download Full-text

Risk Factors of Changes in Hepatitis B Antibody Status after Allogeneic Hematopoietic Stem Cell Transplantation in Chinese Children with Thalassemia Major

Blood ◽

10.1182/blood.v124.21.5809.5809 ◽

2014 ◽

Vol 124 (21) ◽

pp. 5809-5809

Author(s):

Xiaoqin Feng ◽

Lina Long ◽

Chunfu Li

Keyword(s):

Risk Factors ◽

Logistic Regression ◽

Odds Ratio ◽

Thalassemia Major ◽

Enzyme Linked Immunosorbent Assay ◽

Hbv Reactivation ◽

Type I ◽

Multivariate Logistic Regression ◽

Post Transplant

Abstract Objective: This retrospective study evaluated the risk factors involved in the changes in HBsAb status in patients with thalassemia major at a single center in China. Methods: A total of 104 children who underwent allo-HSCT, using NF-08-TM transplant protocol in our center, between January 2010 and June 2012 were recruited.Hepatitis B markers, including HBsAg, anti-HBs, HBeAg, anti-HBe and anti-HBc were examined by TRFIA (time-resolved fluoroimmunoassay) or ELISA (Enzyme-Linked Immunosorbent Assay) for recipients before and after allo-HSCT (at least up to 6 months) and for donors prior to transplantation. HBsAg positive recipients and donors received lamivudine antiviral therapy before allo-HSCT and the treatment was continued in recipients up to 6 months post transplantation. The demographic and clinical characteristics of the patients and their donors were summarized by descriptive statistics. For identification of risk factors that influenced the post-transplant anti-HBs loss and HBV reactivation, both univariate and multivariate logistic regression was used, and odds ratio (OR) and 95% confidence interval (CI) were determined for the covariates that were shown to be statistically significant. All tests were 2-sided, with the type I error rate fixed at 0.05. Statistical analyses were performed using IBM SPSS 20 (SPSS Statistics V20, IBM Corporation, Somers, New York). Results: Of the 104 patients, 2（1.9%） recipients were positive for HBsAg and 102（98.1%） recipients were negative for HBsAg. Of the 102 patients negative for HBsAg before transplantation, the proportion of positive anti-HBs was 69.6% (71 of 102 patients). Of the 104 donors, 99 (95.2%)were negative for HBsAg and 5 (4.8%)were positive for HBsAg. Of the 99 donors negative for HBsAg before transplantation, 72 donors (72.7%) had anti-HBs. After transplantation, of the 69 patients, 27 (39.1%) patients lost their HBV immunity in a median follow-up period of 30 months (range: 21–45); the remaining 42 (60.9 %) patients maintained the immunity against HBV after a median follow-up period of 28.5 months (range: 19–46). 33 patients were anti-HBs negative before the allo-HSCT. The 33 patients included 11 patients with donors who had no anti-HBs and 22 patients with donors who had anti-HBs. After the allo-HSCT, 15 of the 33 patients were found to have newly gained HBV immunity, as represented by the presence of anti-HBs. While 14 of them who developed adoptive immunity had immunized donors (63.6%; 14 out of 22), 1 of them (9.1%; 1 out of 11) with a non-immunized donor (donors without anti-HBs) also had developed HBV immunity. Multivariate logistic regression analysis of 104 patients who underwent allo-HSCT revealed that, patients with pre-HSCT titer of HBsAb < 257.47mIU/mL (adjusted odds ratio, 10.5, 95% CI, 2.1–53.3) and HBsAb-immunized donors (51.3, 2.8–938.6) were significant risk factors for post allo-HSCT HBV loss and acquisition, respectively. In addition, the post-transplant HBV reactivation rate was 11.1%. Conclusions: Current results indicate that pre-transplant HBsAb titer is a key determinant in the loss of HBV immunity after allo-HSCT and HBsAb negative patients with immunized donors are more likely to gain HBV immunity after allo-HSCT than those with non-immunized donors. Further, preemptive antiviral treatment with lamivudine significantly reduces HBV reactivation. This is the first study to have indicated the significant predictors of changes in HBsAg status in children with thalassemia major. Disclosures No relevant conflicts of interest to declare.

Download Full-text

Machine learning methods for diagnosis of chronic obstructive pulmonary disease in healthy subjects: Analysis of Risk factors To DEtect COPD (ARTDECO) (Preprint)

10.2196/preprints.24796 ◽

2020 ◽

Author(s):

Shigeo Muro ◽

Masato Ishida ◽

Yoshiharu Horie ◽

Wataru Takeuchi ◽

Shunki Nakagawa ◽

...

Keyword(s):

Machine Learning ◽

Risk Factors ◽

Chronic Obstructive Pulmonary Disease ◽

Logistic Regression ◽

Pulmonary Disease ◽

Vital Capacity ◽

Chronic Obstructive ◽

Obstructive Pulmonary Disease ◽

Copd Diagnosis

BACKGROUND Airflow limitation is a critical physiological feature in chronic obstructive pulmonary disease (COPD), for which long-term exposure to noxious substances including tobacco smoke is an established risk. However, not all long-term smokers develop COPD, meaning that other risk factors exist. OBJECTIVE To predict risk factors for COPD diagnosis using machine learning in an annual medical check-up database. METHODS In this retrospective, observational cohort study (Analysis of Risk factors To DEtect COPD [ARTDECO]), annual medical check-up records for all Hitachi Ltd. employees in Japan collected from April 1998 to March 2019 were analyzed. Employees who provided informed consent via an opt-out model were screened and those aged 30–75 years, without prior diagnosis of COPD, asthma, or history of cancer were included. The database included clinical measurements (e.g., pulmonary function tests) and questionnaire responses. To predict risk factors for COPD diagnosis within a 3-year period, the Gradient Boosting Decision Tree machine learning method (XGBoost) was applied as a primary approach, with logistic regression as a secondary method. A diagnosis of COPD was made when the ratio of the pre-bronchodilator forced expiratory volume in 1 second (FEV1) to pre-bronchodilator forced vital capacity (FVC) was <0.7 during two consecutive examinations. RESULTS Of the 26,101 individuals screened, 1,213 met the exclusion criteria and thus 24,815 individuals were included in the analysis. The top 10 predictors for COPD diagnosis were FEV1/FVC, smoking status, allergic symptoms, cough, pack years, hemoglobin A1c, serum albumin, mean corpuscular volume, percent predicted vital capacity value, and percent predicted value of FEV1. The area under the receiver operating characteristic curves of the XGBoost model and the logistic regression model were 0.956 and 0.943, respectively. CONCLUSIONS Using a machine learning model in this longitudinal database, we identified a set multiple of parameters as risk factors other than smoking exposure or lung function to support general practitioners and occupational health physicians to predict the development of COPD. Further research to confirm our results is warranted, as our analysis involved a database used only in Japan. CLINICALTRIAL Not applicable.

Download Full-text

Predicting Obesity in Adults Using Machine Learning Techniques: An Analysis of Indonesian Basic Health Research 2018

Frontiers in Nutrition ◽

10.3389/fnut.2021.669155 ◽

2021 ◽

Vol 8 ◽

Author(s):

Sri Astuti Thamrin ◽

Dian Sidik Arsyad ◽

Hedi Kuswanto ◽

Armin Lawi ◽

Sudirman Nasir

Keyword(s):

Machine Learning ◽

Risk Factors ◽

Logistic Regression ◽

Health Research ◽

Health Data ◽

Basic Health ◽

Multiple Risk Factors ◽

Obesity Status ◽

Address Data ◽

Multiple Risk

Obesity is strongly associated with multiple risk factors. It is significantly contributing to an increased risk of chronic disease morbidity and mortality worldwide. There are various challenges to better understand the association between risk factors and the occurrence of obesity. The traditional regression approach limits analysis to a small number of predictors and imposes assumptions of independence and linearity. Machine Learning (ML) methods are an alternative that provide information with a unique approach to the application stage of data analysis on obesity. This study aims to assess the ability of ML methods, namely Logistic Regression, Classification and Regression Trees (CART), and Naïve Bayes to identify the presence of obesity using publicly available health data, using a novel approach with sophisticated ML methods to predict obesity as an attempt to go beyond traditional prediction models, and to compare the performance of three different methods. Meanwhile, the main objective of this study is to establish a set of risk factors for obesity in adults among the available study variables. Furthermore, we address data imbalance using Synthetic Minority Oversampling Technique (SMOTE) to predict obesity status based on risk factors available in the dataset. This study indicates that the Logistic Regression method shows the highest performance. Nevertheless, kappa coefficients show only moderate concordance between predicted and measured obesity. Location, marital status, age groups, education, sweet drinks, fatty/oily foods, grilled foods, preserved foods, seasoning powders, soft/carbonated drinks, alcoholic drinks, mental emotional disorders, diagnosed hypertension, physical activity, smoking, and fruit and vegetables consumptions are significant in predicting obesity status in adults. Identifying these risk factors could inform health authorities in designing or modifying existing policies for better controlling chronic diseases especially in relation to risk factors associated with obesity. Moreover, applying ML methods on publicly available health data, such as Indonesian Basic Health Research (RISKESDAS) is a promising strategy to fill the gap for a more robust understanding of the associations of multiple risk factors in predicting health outcomes.

Download Full-text

A Machine Learning Model to Predict Intravenous Immunoglobulin-Resistant Kawasaki Disease Patients: A Retrospective Study Based on the Chongqing Population

Frontiers in Pediatrics ◽

10.3389/fped.2021.756095 ◽

2021 ◽

Vol 9 ◽

Author(s):

Jie Liu ◽

Jian Zhang ◽

Haodong Huang ◽

Yunting Wang ◽

Zuyue Zhang ◽

...

Keyword(s):

Machine Learning ◽

Risk Factors ◽

Logistic Regression ◽

Regression Analysis ◽

Retrospective Study ◽

Kawasaki Disease ◽

Prediction Model ◽

Prediction Models ◽

Ivig Resistance ◽

Independent Risk Factors

Objective: We explored the risk factors for intravenous immunoglobulin (IVIG) resistance in children with Kawasaki disease (KD) and constructed a prediction model based on machine learning algorithms.Methods: A retrospective study including 1,398 KD patients hospitalized in 7 affiliated hospitals of Chongqing Medical University from January 2015 to August 2020 was conducted. All patients were divided into IVIG-responsive and IVIG-resistant groups, which were randomly divided into training and validation sets. The independent risk factors were determined using logistic regression analysis. Logistic regression nomograms, support vector machine (SVM), XGBoost and LightGBM prediction models were constructed and compared with the previous models.Results: In total, 1,240 out of 1,398 patients were IVIG responders, while 158 were resistant to IVIG. According to the results of logistic regression analysis of the training set, four independent risk factors were identified, including total bilirubin (TBIL) (OR = 1.115, 95% CI 1.067–1.165), procalcitonin (PCT) (OR = 1.511, 95% CI 1.270–1.798), alanine aminotransferase (ALT) (OR = 1.013, 95% CI 1.008–1.018) and platelet count (PLT) (OR = 0.998, 95% CI 0.996–1). Logistic regression nomogram, SVM, XGBoost, and LightGBM prediction models were constructed based on the above independent risk factors. The sensitivity was 0.617, 0.681, 0.638, and 0.702, the specificity was 0.712, 0.841, 0.967, and 0.903, and the area under curve (AUC) was 0.731, 0.814, 0.804, and 0.874, respectively. Among the prediction models, the LightGBM model displayed the best ability for comprehensive prediction, with an AUC of 0.874, which surpassed the previous classic models of Egami (AUC = 0.581), Kobayashi (AUC = 0.524), Sano (AUC = 0.519), Fu (AUC = 0.578), and Formosa (AUC = 0.575).Conclusion: The machine learning LightGBM prediction model for IVIG-resistant KD patients was superior to previous models. Our findings may help to accomplish early identification of the risk of IVIG resistance and improve their outcomes.

Download Full-text

Impact of mechanical circulatory support on post-transplant stroke risk

The International Journal of Artificial Organs ◽

10.1177/03913988211035143 ◽

2021 ◽

Vol 44 (10) ◽

pp. 675-680

Author(s):

Nandini Nair ◽

Shengping Yang ◽

Enrique Gongora

Keyword(s):

Risk Factors ◽

Logistic Regression ◽

Mechanical Circulatory Support ◽

Stroke Risk ◽

Total Artificial Heart ◽

Circulatory Support ◽

Study Cohort ◽

Mechanical Support ◽

Ischemic Time ◽

Post Transplant

The effect of type of mechanical circulatory support on stroke risk during the early post-transplant period remains undefined in patients bridged to transplant. This study assesses if the type of circulatory support device affects stroke risk in this population. The study cohort of 4257 adult patients bridged with mechanical support to cardiac transplant were derived from the UNOS transplant registry data. Risk factors assessed were age, gender, ischemic time, diabetes (recipient), durable mechanical support at listing and mechanical ventilation pre-transplant. Descriptive statistics were used to describe characteristics of the study cohort. Univariate logistic regression was used to test if there is a significant association between stroke event and all the potential risk factors. Multivariate logistic regression was used to test such associations while adjusting for all other risk factors. Odds ratios (ORs) and their 95% confidence intervals (CIs) in parenthesis, were calculated. p < 0.05 was considered significant. Patients on Extracorporeal membrane oxygenation (ECMO) had the highest risk of stroke immediately post-transplant prior to discharge (OR 3.03, {1.16, 7.95}) followed by Total Artificial Heart (TAH) (OR 2.03, {1.01, 4.07) as compared to those only on a Left Ventricular Assist Device (LVAD). Ischemic time (OR 1.3 {1.09, 1.45}) and diabetes (OR 1.8 {1.29, 2.51}) were significant risk factors. Patients on ECMO and TAH had a 203% and 103% increase respectively in the odds of having a stroke prior to discharge as compared to those only on LVADS.

Download Full-text

PredictMed: A Machine Learning Model for Identifying Risk Factors of Neuromuscular Hip Dysplasia: A Multicenter Descriptive Study

Neuropediatrics ◽

10.1055/s-0040-1721703 ◽

2020 ◽

Author(s):

Carlo M. Bertoncelli ◽

Paola Altamura ◽

Domenico Bertoncelli ◽

Virginie Rampal ◽

Edgar Ramos Vieira ◽

...

Keyword(s):

Machine Learning ◽

Risk Factors ◽

Logistic Regression ◽

Motor Function ◽

Hip Dysplasia ◽

Descriptive Study ◽

Learning Model ◽

Trunk Muscles ◽

Lateral Migration ◽

Machine Learning Model

AbstractNeuromuscular hip dysplasia (NHD) is a common and severe problem in patients with cerebral palsy (CP). Previous studies have so far identified only spasticity (SP) and high levels of Gross Motor Function Classification System as factors associated with NHD. The aim of this study is to develop a machine learning model to identify additional risk factors of NHD. This was a cross-sectional multicenter descriptive study of 102 teenagers with CP (60 males, 42 females; 60 inpatients, 42 outpatients; mean age 16.5 ± 1.2 years, range 12–18 years). Data on etiology, diagnosis, SP, epilepsy (E), clinical history, and functional assessments were collected between 2007 and 2017. Hip dysplasia was defined as femoral head lateral migration percentage > 33% on pelvic radiogram. A logistic regression-prediction model named PredictMed was developed to identify risk factors of NHD. Twenty-eight (27%) teenagers with CP had NHD, of which 18 (67%) had dislocated hips. Logistic regression model identified poor walking abilities (p < 0.001; odds ratio [OR] infinity; 95% confidence interval [CI] infinity), scoliosis (p = 0.01; OR 3.22; 95% CI 1.30–7.92), trunk muscles' tone disorder (p = 0.002; OR 4.81; 95% CI 1.75–13.25), SP (p = 0.006; OR 6.6; 95% CI 1.46–30.23), poor motor function (p = 0.02; OR 5.5; 95% CI 1.2–25.2), and E (p = 0.03; OR 2.6; standard error 0.44) as risk factors of NHD. The accuracy of the model was 77%. PredictMed identified trunk muscles' tone disorder, severe scoliosis, E, and SP as risk factors of NHD in teenagers with CP.

Download Full-text