Comparison of machine learning techniques to predict all-cause mortality using fitness data: the Henry ford exercIse testing (FIT) project

Predicting all-cause mortality and premature death using interpretable machine learning among a middle-aged and elderly Chinese population

Objective

To develop machine learning-based prediction models for all-cause and premature mortality among the middle-aged and elderly population in China.

Method

Adults aged 45 years or older at baseline of 2011 from the China Health and Retirement Longitudinal Study (CHARLS) were included. The stacked ensemble model was built utilizing five selected machine learning algorithms. These models underwent training and testing using the CHARLS 2011-2015 cohort (derivation cohort) and subsequently underwent external validation using the CHARLS 2015-2018 cohort (validation cohort). SHapley Additive exPlanations (SHAP) was introduced to quantify the importance of risk factors and explain machine learning algorithms.

Result

In derivation cohort, a total of 10,677 subjects were included, 478 died during the follow-up. The stacked ensemble model demonstrated the highest efficacy in terms of its discrimination capability for predicting all-cause mortality and premature death, with an AUC[95 % CI] of 0.826[0.792-0.859] and 0.773[0.725-0.821], respectively. In validation cohort, the corresponding AUC[95 % CI] were 0.803[0.743-0.864] and 0.791[0.719-0.863], respectively. Risk factors including age, sex, self-reported health, activities of daily living, cognitive function, ever smoker, levels of systolic blood pressure, Cystatin C and low density lipoprotein were strong predictors for both all-cause mortality and premature death.

Conclusion

Stacked ensemble models performed well in predicting all-cause and premature death in this Chinese cohort. Interpretable techniques can aid in identifying significant risk factors and non-linear relationships between predictors and mortality.

Genome-wide association study of exercise-induced skeletal muscle hypertrophy and the construction of predictive model

The aim of the current study was to investigate interindividual differences in muscle thickness of the rectus femoris (MTRF) following 12 wk of resistance training (RT) or high-intensity interval training (HIIT) to explore the genetic architecture underlying skeletal muscle hypertrophy and to construct predictive models. We conducted musculoskeletal ultrasound assessments of the MTRF response in 440 physically inactive adults after the 12-wk exercise period. A genome-wide association study was used to identify variants associated with the MTRF response, separately for RT and HIIT. Using the polygenic predictor score (PPS), we estimated the genetic contribution to exercise-induced hypertrophy. Predictive models for the MTRF response were constructed using random forest (RF), support vector mac (SVM), and generalized linear model (GLM) in 10 cross-validated approaches. MTRF increased significantly after both RT (8.8%, P < 0.05) and HIIT (5.3%, P < 0.05), but with considerable interindividual differences (RT: -13.5 to 38.4%, HIIT: -14.2 to 30.7%). Eleven lead single-nucleotide polymorphisms in RT and eight lead single-nucleotide polymorphisms in HIIT were identified at a significance level of P < 1 × 10-5. The PPS was associated with the MTRF response, explaining 47.2% of the variation in response to RT and 38.3% of the variation in response to HIIT. Notably, the GLM and SVM predictive models exhibited superior performance compared with RF models (P < 0.05), and the GLM demonstrated optimal performance with an area under curve of 0.809 (95% confidence interval: 0.669-0.949). Factors such as PPS, baseline MTRF, and exercise protocol exerted influence on the MTRF response to exercise, with PPS being the primary contributor. The GLM and SVM predictive model, incorporating both genetic and phenotypic factors, emerged as promising tools for predicting exercise-induced skeletal muscle hypertrophy.NEW & NOTEWORTHY The interindividual variability induced muscle hypertrophy by resistance training (RT) or high-intensity interval training (HIIT) and the associated genetic architecture remain uncertain. We identified genetic variants that underlie RT- or HIIT-induced muscle hypertrophy and established them as pivotal factors influencing the response regardless of the training type. The genetic-phenotype predictive model developed has the potential to identify nonresponders or individuals with low responsiveness before engaging in exercise training.

Evaluating current assessment techniques of cardiorespiratory fitness

INTRODUCTION

Considerable and convincing global data from cohorts across the health spectrum (i.e. apparently healthy to known disease) indicate that cardiorespiratory fitness (CRF) is a major predictor of overall and cardiovascular disease (CVD)-survival, seemingly with greater prognostic resolution compared to other traditional CVD risk factors. Therefore, the assessment of CRF in research and clinical settings is of major importance.

AREAS COVERED

In this manuscript, we review the technology of measuring CRF assessed by the 'gold standard,' cardiopulmonary exercise testing (CPET), as well as with various other methods (e.g. estimated metabolic equivalents, 6-minute walk tests, shuttle tests, and non-exercise equations that estimate CRF), all of which provide significant prognostic information for CVD- and all-cause survival. The literature through May 2024 has been cited.

EXPERT OPINION

The promotion of physical activity in efforts to improve levels of CRF is needed throughout the world to improve lifespan and, more importantly, healthspan. The routine assessment of CRF should be considered a vital sign that is routinely assessed in clinical practice.

Stone decision engine accurately predicts stone removal and treatment complications for shock wave lithotripsy and laser ureterorenoscopy patients

Kidney stones form when mineral salts crystallize in the urinary tract. While most stones exit the body in the urine stream, some can block the ureteropelvic junction or ureters, leading to severe lower back pain, blood in the urine, vomiting, and painful urination. Imaging technologies, such as X-rays or ureterorenoscopy (URS), are typically used to detect kidney stones. Subsequently, these stones are fragmented into smaller pieces using shock wave lithotripsy (SWL) or laser URS. Both treatments yield subtly different patient outcomes. To predict successful stone removal and complication outcomes, Artificial Neural Network models were trained on 15,126 SWL and 2,116 URS patient records. These records include patient metrics like Body Mass Index and age, as well as treatment outcomes obtained using various medical instruments and healthcare professionals. Due to the low number of outcome failures in the data (e.g., treatment complications), Nearest Neighbor and Synthetic Minority Oversampling Technique (SMOTE) models were implemented to improve prediction accuracies. To reduce noise in the predictions, ensemble modeling was employed. The average prediction accuracies based on Confusion Matrices for SWL stone removal and treatment complications were 84.8% and 95.0%, respectively, while those for URS were 89.0% and 92.2%, respectively. The average prediction accuracies for SWL based on Area-Under-the-Curve were 74.7% and 62.9%, respectively, while those for URS were 77.2% and 78.9%, respectively. Taken together, the approach yielded moderate to high accurate predictions, regardless of treatment or outcome. These models were incorporated into a Stone Decision Engine web application (http://peteranoble.com/webapps.html) that suggests the best interventions to healthcare providers based on individual patient metrics.

FIT calculator: a multi-risk prediction framework for medical outcomes using cardiorespiratory fitness data

Accurately predicting patients' risk for specific medical outcomes is paramount for effective healthcare management and personalized medicine. While a substantial body of literature addresses the prediction of diverse medical conditions, existing models predominantly focus on singular outcomes, limiting their scope to one disease at a time. However, clinical reality often entails patients concurrently facing multiple health risks across various medical domains. In response to this gap, our study proposes a novel multi-risk framework adept at simultaneous risk prediction for multiple clinical outcomes, including diabetes, mortality, and hypertension. Leveraging a concise set of features extracted from patients' cardiorespiratory fitness data, our framework minimizes computational complexity while maximizing predictive accuracy. Moreover, we integrate a state-of-the-art instance-based interpretability technique into our framework, providing users with comprehensive explanations for each prediction. These explanations afford medical practitioners invaluable insights into the primary health factors influencing individual predictions, fostering greater trust and utility in the underlying prediction models. Our approach thus stands to significantly enhance healthcare decision-making processes, facilitating more targeted interventions and improving patient outcomes in clinical practice. Our prediction framework utilizes an automated machine learning framework, Auto-Weka, to optimize machine learning models and hyper-parameter configurations for the simultaneous prediction of three medical outcomes: diabetes, mortality, and hypertension. Additionally, we employ a local interpretability technique to elucidate predictions generated by our framework. These explanations manifest visually, highlighting key attributes contributing to each instance's prediction for enhanced interpretability. Using automated machine learning techniques, the models simultaneously predict hypertension, mortality, and diabetes risks, utilizing only nine patient features. They achieved an average AUC of 0.90 ± 0.001 on the hypertension dataset, 0.90 ± 0.002 on the mortality dataset, and 0.89 ± 0.001 on the diabetes dataset through tenfold cross-validation. Additionally, the models demonstrated strong performance with an average AUC of 0.89 ± 0.001 on the hypertension dataset, 0.90 ± 0.001 on the mortality dataset, and 0.89 ± 0.001 on the diabetes dataset using bootstrap evaluation with 1000 resamples.

Integrated transcriptomic meta-analysis and comparative artificial intelligence models in maize under biotic stress

Biotic stress imposed by pathogens, including fungal, bacterial, and viral, can cause heavy damage leading to yield reduction in maize. Therefore, the identification of resistant genes paves the way to the development of disease-resistant cultivars and is essential for reliable production in maize. Identifying different gene expression patterns can deepen our perception of maize resistance to disease. This study includes machine learning and deep learning-based application for classifying genes expressed under normal and biotic stress in maize. Machine learning algorithms used are Naive Bayes (NB), K-Nearest Neighbor (KNN), Ensemble, Support Vector Machine (SVM), and Decision Tree (DT). A Bidirectional Long Short Term Memory (BiLSTM) based network with Recurrent Neural Network (RNN) architecture is proposed for gene classification with deep learning. To increase the performance of these algorithms, feature selection is made from the raw gene features through the Relief feature selection algorithm. The obtained finding indicated the efficacy of BiLSTM over other machine learning algorithms. Some top genes ((S)-beta-macrocarpene synthase, zealexin A1 synthase, polyphenol oxidase I, chloroplastic, pathogenesis-related protein 10, CHY1, chitinase chem 5, barwin, and uncharacterized LOC100273479 were proved to be differentially upregulated under biotic stress condition.

Integrative Interpretation of Cardiopulmonary Exercise Tests for Cardiovascular Outcome Prediction: A Machine Learning Approach

Integrative interpretation of cardiopulmonary exercise tests (CPETs) may improve assessment of cardiovascular (CV) risk. Here, we identified patient phenogroups based on CPET summary metrics and evaluated their predictive value for CV events. We included 2280 patients with diverse CV risk who underwent maximal CPET by cycle ergometry. Key CPET indices and information on incident CV events (median follow-up time: 5.3 years) were derived. Next, we applied unsupervised clustering by Gaussian Mixture modeling to subdivide the cohort into four male and four female phenogroups solely based on differences in CPET metrics. Ten of 18 CPET metrics were used for clustering as eight were removed due to high collinearity. In males and females, the phenogroups differed significantly in age, BMI, blood pressure, disease prevalence, medication intake and spirometry. In males, phenogroups 3 and 4 presented a significantly higher risk for incident CV events than phenogroup 1 (multivariable-adjusted hazard ratio: 1.51 and 2.19; p ≤ 0.048). In females, differences in the risk for future CV events between the phenogroups were not significant after adjustment for clinical covariables. Integrative CPET-based phenogrouping, thus, adequately stratified male patients according to CV risk. CPET phenomapping may facilitate comprehensive evaluation of CPET results and steer CV risk stratification and management.

Systematic review and meta-analysis of prediction models used in cervical cancer

BACKGROUND

Cervical cancer is one of the most common cancers in women with an incidence of around 6.5 % of all the cancer in women worldwide. Early detection and adequate treatment according to staging improve the patient's life expectancy. Outcome prediction models might aid treatment decisions, but a systematic review on prediction models for cervical cancer patients is not available.

DESIGN

We performed a systematic review for prediction models in cervical cancer following PRISMA guidelines. Key features that were used for model training and validation, the endpoints were extracted from the article and data were analyzed. Selected articles were grouped based on prediction endpoints i.e. Group1: Overall survival, Group2: progression-free survival; Group3: recurrence or distant metastasis; Group4: treatment response; Group5: toxicity or quality of life. We developed a scoring system to evaluate the manuscript. As per our criteria, studies were divided into four groups based on scores obtained in our scoring system, the Most significant study (Score > 60 %); Significant study (60 % > Score > 50 %); Moderately Significant study (50 % > Score > 40 %); least significant study (score < 40 %). A meta-analysis was performed for all the groups separately.

RESULTS

The first line of search selected 1358 articles and finally 39 articles were selected as eligible for inclusion in the review. As per our assessment criteria, 16, 13 and 10 studies were found to be the most significant, significant and moderately significant respectively. The intra-group pooled correlation coefficient for Group1, Group2, Group3, Group4, and Group5 were 0.76 [0.72, 0.79], 0.80 [0.73, 0.86], 0.87 [0.83, 0.90], 0.85 [0.77, 0.90], 0.88 [0.85, 0.90] respectively. All the models were found to be good (prediction accuracy [c-index/AUC/R²] >0.7) in endpoint prediction.

CONCLUSIONS

Prediction models of cervical cancer toxicity, local or distant recurrence and survival prediction show promising results with reasonable prediction accuracy [c-index/AUC/R² > 0.7]. These models should also be validated on external data and evaluated in prospective clinical studies.

Machine-learning predicts time-series prognosis factors in metastatic prostate cancer patients treated with androgen deprivation therapy

Machine learning technology is expected to support diagnosis and prognosis prediction in medicine. We used machine learning to construct a new prognostic prediction model for prostate cancer patients based on longitudinal data obtained from age at diagnosis, peripheral blood and urine tests of 340 prostate cancer patients. Random survival forest (RSF) and survival tree were used for machine learning. In the time-series prognostic prediction model for metastatic prostate cancer patients, the RSF model showed better prediction accuracy than the conventional Cox proportional hazards model for almost all time periods of progression-free survival (PFS), overall survival (OS) and cancer-specific survival (CSS). Based on the RSF model, we created a clinically applicable prognostic prediction model using survival trees for OS and CSS by combining the values of lactate dehydrogenase (LDH) before starting treatment and alkaline phosphatase (ALP) at 120 days after treatment. Machine learning provides useful information for predicting the prognosis of metastatic prostate cancer prior to treatment intervention by considering the nonlinear and combined impacts of multiple features. The addition of data after the start of treatment would allow for more precise prognostic risk assessment of patients and would be beneficial for subsequent treatment selection.

Prediction of Prednisolone Dose Correction Using Machine Learning

Wrong dose, a common prescription error, can cause serious patient harm, especially in the case of high-risk drugs like oral corticosteroids. This study aims to build a machine learning model to predict dose-related prescription modifications for oral prednisolone tablets (i.e., highly imbalanced data with very few positive cases). Prescription data were obtained from the electronic medical records at a single institute. Cluster analysis classified the clinical departments into six clusters with similar patterns of prednisolone prescription. Two patterns of training datasets were created with/without preprocessing by the SMOTE method. Five ML models (SVM, KNN, GB, RF, and BRF) and logistic regression (LR) models were constructed by Python. The model was internally validated by five-fold stratified cross-validation and was validated with a 30% holdout test dataset. Eighty-two thousand five hundred fifty-three prescribing data for prednisolone tablets containing 135 dose-corrected positive cases were obtained. In the original dataset (without SMOTE), only the BRF model showed a good performance (in test dataset, ROC-AUC:0.917, recall: 0.951). In the training dataset preprocessed by SMOTE, performance was improved on all models. The highest performance models with SMOTE were SVM (in test dataset, ROC-AUC: 0.820, recall: 0.659) and BRF (ROC-AUC: 0.814, recall: 0.634). Although the prescribing data for dose-related collection are highly imbalanced, various techniques such as the following have allowed us to build high-performance prediction models: data preprocessing by SMOTE, stratified cross-validation, and BRF classifier corresponding to imbalanced data. ML is useful in complicated dose audits such as oral prednisolone.

Supplementary Information

The online version contains supplementary material available at 10.1007/s41666-023-00128-3.

Development and validation of questionnaire-based machine learning models for predicting all-cause mortality in a representative population of China

Background

Considering that the previously developed mortality prediction models have limited applications to the Chinese population, a questionnaire-based prediction model is of great importance for its accuracy and convenience in clinical practice.

Methods

Two national cohort, namely, the China Health and Nutrition Survey (8,355 individual older than 18) and the China Health and Retirement Longitudinal Study (12,711 individuals older than 45) were used for model development and validation. One hundred and fifty-nine variables were compiled to generate predictions. The Cox regression model and six machine learning (ML) models were used to predict all-cause mortality. Finally, a simple questionnaire-based ML prediction model was developed using the best algorithm and validated.

Results

In the internal validation set, all the ML models performed better than the traditional Cox model in predicting 6-year mortality and the random survival forest (RSF) model performed best. The questionnaire-based ML model, which only included 20 variables, achieved a C-index of 0.86 (95%CI: 0.80-0.92). On external validation, the simple questionnaire-based model achieved a C-index of 0.82 (95%CI: 0.77-0.87), 0.77 (95%CI: 0.75-0.79), and 0.79 (95%CI: 0.77-0.81), respectively, in predicting 2-, 9-, and 11-year mortality.

Conclusions

In this prospective population-based study, a model based on the RSF analysis performed best among all models. Furthermore, there was no significant difference between the prediction performance of the questionnaire-based ML model, which only included 20 variables, and that of the model with all variables (including laboratory variables). The simple questionnaire-based ML prediction model, which needs to be further explored, is of great importance for its accuracy and suitability to the Chinese general population.

Identification of clinical factors related to prediction of alcohol use disorder from electronic health records using feature selection methods

BACKGROUND

High dimensionality in electronic health records (EHR) causes a significant computational problem for any systematic search for predictive, diagnostic, or prognostic patterns. Feature selection (FS) methods have been indicated to be effective in feature reduction as well as in identifying risk factors related to prediction of clinical disorders. This paper examines the prediction of patients with alcohol use disorder (AUD) using machine learning (ML) and attempts to identify risk factors related to the diagnosis of AUD.

METHODS

A FS framework consisting of two operational levels, base selectors and ensemble selectors. The first level consists of five FS methods: three filter methods, one wrapper method, and one embedded method. Base selector outputs are aggregated to develop four ensemble FS methods. The outputs of FS method were then fed into three ML algorithms: support vector machine (SVM), K-nearest neighbor (KNN), and random forest (RF) to compare and identify the best feature subset for the prediction of AUD from EHRs.

RESULTS

In terms of feature reduction, the embedded FS method could significantly reduce the number of features from 361 to 131. In terms of classification performance, RF based on 272 features selected by our proposed ensemble method (Union FS) with the highest accuracy in predicting patients with AUD, 96%, outperformed all other models in terms of AUROC, AUPRC, Precision, Recall, and F1-Score. Considering the limitations of embedded and wrapper methods, the best overall performance was achieved by our proposed Union Filter FS, which reduced the number of features to 223 and improved Precision, Recall, and F1-Score in RF from 0.77, 0.65, and 0.71 to 0.87, 0.81, and 0.84, respectively. Our findings indicate that, besides gender, age, and length of stay at the hospital, diagnosis related to digestive organs, bones, muscles and connective tissue, and the nervous systems are important clinical factors related to the prediction of patients with AUD.

CONCLUSION

Our proposed FS method could improve the classification performance significantly. It could identify clinical factors related to prediction of AUD from EHRs, thereby effectively helping clinical staff to identify and treat AUD patients and improving medical knowledge of the AUD condition. Moreover, the diversity of features among female and male patients as well as gender disparity were investigated using FS methods and ML techniques.

Interpretable Machine Learning-Based Prediction of Intraoperative Cerebrospinal Fluid Leakage in Endoscopic Transsphenoidal Pituitary Surgery: A Pilot Study

Purpose  Transsphenoidal surgery (TSS) for pituitary adenomas can be complicated by the occurrence of intraoperative cerebrospinal fluid (CSF) leakage (IOL). IOL significantly affects the course of surgery predisposing to the development of postoperative CSF leakage, a major source of morbidity and mortality in the postoperative period. The authors trained and internally validated the Random Forest (RF) prediction model to preoperatively identify patients at high risk for IOL. A locally interpretable model-agnostic explanations (LIME) algorithm is employed to elucidate the main drivers behind each machine learning (ML) model prediction. Methods  The data of 210 patients who underwent TSS were collected; first, risk factors for IOL were identified via conventional statistical methods (multivariable logistic regression). Then, the authors trained, optimized, and audited a RF prediction model. Results  IOL reported in 45 patients (21.5%). The recursive feature selection algorithm identified the following variables as the most significant determinants of IOL: Knosp's grade, sellar Hardy's grade, suprasellar Hardy's grade, tumor diameter (on X, Y, and Z axes), intercarotid distance, and secreting status (nonfunctioning and growth hormone [GH] secreting). Leveraging the predictive values of these variables, the RF prediction model achieved an area under the curve (AUC) of 0.83 (95% confidence interval [CI]: 0.78; 0.86), significantly outperforming the multivariable logistic regression model (AUC = 0.63). Conclusion  A RF model that reliably identifies patients at risk for IOL was successfully trained and internally validated. ML-based prediction models can predict events that were previously judged nearly unpredictable; their deployment in clinical practice may result in improved patient care and reduced postoperative morbidity and healthcare costs.

Prediction of Mortality in Coronary Artery Disease: Role of Machine Learning and Maximal Exercise Capacity

OBJECTIVE

To develop a prediction model for survival of patients with coronary artery disease (CAD) using health conditions beyond cardiovascular risk factors, including maximal exercise capacity, through the application of machine learning (ML) techniques.

METHODS

Analysis of data from a retrospective cohort linking clinical, administrative, and vital status databases from 1995 to 2016 was performed. Inclusion criteria were age 18 years or older, diagnosis of CAD, referral to a cardiac rehabilitation program, and available baseline exercise test results. Primary outcome was death from any cause. Feature selection was performed using supervised and unsupervised ML techniques. The final prognostic model used the survival tree (ST) algorithm.

RESULTS

From the cohort of 13,362 patients (60±11 years; 2400 [18%] women), 1577 died during a median follow-up of 8 years (interquartile range, 4 to 13 years), with an estimated survival of 67% up to 21 years. Feature selection revealed age and peak metabolic equivalents (METs) as the features with the greatest importance for mortality prediction. Using these 2 features, the ST generated a long-term prediction with a C-index of 0.729 by splitting patients in 8 clusters with different survival probabilities (P<.001). The ST root node was split by peak METs of 6.15 or less or more than 6.15, and each patient's subgroup was further split by age or other peak METs cut points.

CONCLUSION

Applying ML techniques, age and maximal exercise capacity accurately predict mortality in patients with CAD and outperform variables commonly used for decision-making in clinical practice. A novel and simple prognostic model was established, and maximal exercise capacity was further suggested to be one of the most powerful predictors of mortality in CAD.

Application of machine learning methods for predicting infant mortality in Rwanda: analysis of Rwanda demographic health survey 2014-15 dataset

BACKGROUND

Extensive research on infant mortality (IM) exists in developing countries; however, most of the methods applied thus far relied on conventional regression analyses with limited prediction capability. Advanced of Machine Learning (AML) methods provide accurate prediction of IM; however, there is no study conducted using ML methods in Rwanda. This study, therefore, applied Machine Learning Methods for predicting infant mortality in Rwanda.  METHODS: A cross-sectional study design was conducted using the 2014-15 Rwanda Demographic and Health Survey. Python software version 3.8 was employed to test and apply ML methods through Random Forest (RF), Decision Tree, Support Vector Machine and Logistic regression. STATA version 13 was used for analysing conventional methods. Evaluation metrics methods specifically confusion matrix, accuracy, precision, recall, F1 score, and Area under the Receiver Operating Characteristics (AUROC) were used to evaluate the performance of predictive models.

RESULTS

Ability of prediction was between 68.6% and 61.5% for AML. We preferred with the RF model (61.5%) presenting the best performance. The RF model was the best predictive model of IM with accuracy (84.3%), recall (91.3%), precision (80.3%), F1 score (85.5%), and AUROC (84.2%); followed by decision tree model with model accuracy (83%), recall (91%), precision (79%), F1 score (84.67%) and AUROC(82.9%), followed by support vector machine with model accuracy (68.6%), recall (74.9%), precision(67%), F1 score (70.73%) and AUROC (68.6%) and last was a logistic regression with the low accuracy of prediction (61.5%), recall (61.1%), precision (62.2%), F1 score (61.6%) and AUROC (61.5%) compared to other predictive models. Our predictive models showed that marital status, children ever born, birth order and wealth index are the 4 top predictors of IM.

CONCLUSIONS

In developing a predictive model, ML methods are used to classify certain hidden information that could not be detected by traditional statistical methods. Random Forest was classified as the best classifier to be used for the predictive models of IM.

A Sustainable Model for Emergency Medical Services in Developing Countries: A Novel Approach Using Partial Outsourcing and Machine Learning

INTRODUCTION

Unlike Western countries, many low- and middle-income countries (LMIC), like India, have a de-centralized emergency medical services (EMS) involving both semi-government and non-government organizations. It is alarming that due to the absence of a common ecosystem, the utilization of resources is inefficient, which leads to shortage of available vehicles and larger response time. Fragmentation of emergency supply chain resources motivates us to propose a new vehicle routing and scheduling model equipped with novel features to ensure minimal response time using existing resources.

MATERIALS AND METHODS

The data set of medical and fire-related emergencies from January 2018 to May 2018 of Uttarakhand State in India was provided by GVK Emergency Management and Research Institute (GVK EMRI) also known as 108 EMSs was used in the study. The proposed model integrates all the available EMS vehicles including partial outsourcing to non-ambulatory vehicles like police vans, taxis, etc., using a novel two-echelon heuristic approach. In the first stage, an offline learning model is developed to yield the deployment strategy for EMS vehicles. Seven well researched machine learning (ML) algorithms were analyzed for parameter prediction namely random forest (RF), convolutional neural network (CNN), k-nearest neighbor (KNN), classification and regression tree (CART), support vector machine (SVM), logistic regression (LR), and linear discriminant analysis (LDA). In the second stage, a real-time routing model is proposed for EMS vehicle routing at the time of emergency, considering partial outsourcing.

RESULTS AND DISCUSSION

The results indicate that the RF classifier outperforms the LR, LDA, SVM, CNN, CART and NB classifier in terms of both accuracy as well as F-1 score. The proposed vehicle routing and scheduling model for automated decision-making shows an improvement of 42.1%, 54%, 27.9% and 62% in vehicle assignment time, vehicle travel time from base to scene, travel time from scene to hospital, and total response time, respectively, in urban areas.

Artificial intelligence model validation before its application in clinical diagnosis assistance

The process of selecting an artificial intelligence (AI) model to assist clinical diagnosis of a particular pathology and its validation tests is relevant since the values of accuracy, sensitivity and specificity may not reflect the behavior of the method in a real environment. Here, we provide helpful considerations to increase the success of using an AI model in clinical practice.

Virtual reality: a powerful technology to provide novel insight into treatment mechanisms of addiction

Due to its high ecological validity, virtual reality (VR) technology has emerged as a powerful tool for mental health research. Despite the wide use of VR simulations in research on mental illnesses, the study of addictive processes through the use of VR environments is still at its dawn. In a systematic literature search, we identified 38 reports of research projects using highly immersive head-mounted displays, goggles, or CAVE technologies to provide insight into treatment mechanisms of addictive behaviors. So far, VR research has mainly addressed the roles of craving, psychophysiology, affective states, cognition, and brain activity in addiction. The computer-generated VR environments offer very realistic, dynamic, interactive, and complex real-life simulations requesting active participation. They create a high sense of immersion in users by combining stereoscopic three-dimensional visual, auditory, olfactory, and tactile perceptions, tracking systems responding to user movements, and social interactions. VR is an emerging tool to study how proximal multi-sensorial cues, contextual environmental cues, as well as their interaction (complex cues) modulate addictive behaviors. VR allows for experimental designs under highly standardized, strictly controlled, predictable, and repeatable conditions. Moreover, VR simulations can be personalized. They are currently refined for psychotherapeutic interventions. Embodiment, eye-tracking, and neurobiological factors represent novel future directions. The progress of VR applications has bred auspicious ways to advance the understanding of treatment mechanisms underlying addictions, which researchers have only recently begun to exploit. VR methods promise to yield significant achievements to the addiction field. These are necessary to develop more efficacious and efficient preventive and therapeutic strategies.

Machine Learning-Based Prediction of Early Recurrence in Glioblastoma Patients: A Glance Towards Precision Medicine

BACKGROUND

Ability to thrive and time-to-recurrence following treatment are important parameters to assess in patients with glioblastoma multiforme (GBM), given its dismal prognosis. Though there is an ongoing debate whether it can be considered an appropriate surrogate endpoint for overall survival in clinical trials, progression-free survival (PFS) is routinely used for clinical decision-making.

OBJECTIVE

To investigate whether machine learning (ML)-based models can reliably stratify newly diagnosed GBM patients into prognostic subclasses on PFS basis, identifying those at higher risk for an early recurrence (≤6 mo).

METHODS

Data were extracted from a multicentric database, according to the following eligibility criteria: histopathologically verified GBM and follow-up >12 mo: 474 patients met our inclusion criteria and were included in the analysis. Relevant demographic, clinical, molecular, and radiological variables were selected by a feature selection algorithm (Boruta) and used to build a ML-based model.

RESULTS

Random forest prediction model, evaluated on an 80:20 split ratio, achieved an AUC of 0.81 (95% CI: 0.77; 0.83) demonstrating high discriminative ability. Optimizing the predictive value derived from the linear and nonlinear combinations of the selected input features, our model outperformed across all performance metrics multivariable logistic regression.

CONCLUSION

A robust ML-based prediction model that identifies patients at high risk for early recurrence was successfully trained and internally validated. Considerable effort remains to integrate these predictions in a patient-centered care context.

Estimated Artificial Neural Network Modeling of Maximal Oxygen Uptake Based on Multistage 10-m Shuttle Run Test in Healthy Adults

We aimed to develop an artificial neural network (ANN) model to estimate the maximal oxygen uptake (VO₂max) based on a multistage 10 m shuttle run test (SRT) in healthy adults. For ANN-based VO₂max estimation, 118 healthy Korean adults (59 men and 59 women) in their twenties and fifties (38.3 ± 11.8 years, men aged 37.8 ± 12.1 years, and women aged 38.8 ± 11.6 years) participated in this study; data included age, sex, blood pressure (systolic blood pressure (SBP), diastolic blood pressure (DBP)), waist circumference, hip circumference, waist-to-hip ratio (WHR), body composition (weight, height, body mass index (BMI), percent skeletal muscle, and percent body), 10 m SRT parameters (number of round trips and final speed), and VO₂max by graded exercise test (GXT) using a treadmill. The best estimation results (R² = 0.8206, adjusted R² = 0.7010, root mean square error; RMSE = 3.1301) were obtained in case 3 (using age, sex, height, weight, BMI, waist circumference, hip circumference, WHR, SBP, DBP, number of round trips in 10 m SRT, and final speed in 10 m SRT), while the worst results (R² = 0.7765, adjusted R² = 0.7206, RMSE = 3.494) were obtained for case 1 (using age, sex, height, weight, BMI, number of round trips in 10 m SRT, and final speed in 10 m SRT). The estimation results of case 2 (using age, sex, height, weight, BMI, waist circumference, hip circumference, WHR, number of round trips in 10 m SRT, and final speed in 10 m SRT) were lower (R² = 0.7909, adjusted R² = 0.7072, RMSE = 3.3798) than those of case 3 and higher than those of case 1. However, all cases showed high performance (R²) in the estimation results. This brief report developed an ANN-based estimation model to predict the VO₂max of healthy adults, and the model’s performance was confirmed to be excellent.

Using data mining technology to predict medication-taking behaviour in women with breast cancer: A retrospective study

Aims

Medication‐taking behaviours of breast cancer survivors undergoing adjuvant hormone therapy have received considerable attention. This study aimed to determine factors affecting medication‐taking behaviours in people with breast cancer using data mining.

Design

A longitudinal observational retrospective cohort study with a hospital‐based survey.

Methods

A total of 385 subjects were surveyed, analysing existing data from January 2010 to December 2017 in Taiwan. Three data mining approaches—multiple logistic regression, decision tree and artificial neural network—were used to build the prediction models and rank the importance of influencing factors. Accuracy, specificity and sensitivity were used as assessment indicators for the prediction models.

Results

Multiple logistic regression was the most effective approach, achieving an accuracy of 96.37%, specificity of 96.75% and sensitivity of 96.12%. The duration of adjuvant hormone therapy discontinuation, duration of adjuvant hormone therapy use and age at diagnosis by data mining were the three most critical factors influencing the medication‐taking behaviours of people with breast cancer.

Health informatics publication trends in Saudi Arabia: a bibliometric analysis over the last twenty-four years

OBJECTIVE

Understanding health informatics (HI) publication trends in Saudi Arabia may serve as a framework for future research efforts and contribute toward meeting national "e-Health" goals. The authors' intention was to understand the state of the HI field in Saudi Arabia by exploring publication trends and their alignment with national goals.

METHODS

A scoping review was performed to identify HI publications from Saudi Arabia in PubMed, Embase, and Web of Science. We analyzed publication trends based on topics, keywords, and how they align with the Ministry of Health's (MOH's) "digital health journey" framework.

RESULTS

The total number of publications included was 242. We found 1 (0.4%) publication in 1995-1999, 11 (4.5%) publications in 2000-2009, and 230 (95.0%) publications in 2010-2019. We categorized publications into 3 main HI fields and 4 subfields: 73.1% (n=177) of publications were in clinical informatics (85.1%, n=151 medical informatics; 5.6%, n=10 pharmacy informatics; 6.8%, n=12 nursing informatics; 2.3%, n=4 dental informatics); 22.3% (n=54) were in consumer health informatics; and 4.5% (n=11) were in public health informatics. The most common keyword was "medical informatics" (21.5%, n=52). MOH framework-based analysis showed that most publications were categorized as "digitally enabled care" and "digital health foundations."

CONCLUSIONS

The years of 2000-2009 may be seen as an infancy stage of the HI field in Saudi Arabia. Exploring how the Saudi Arabian MOH's e-Health initiatives may influence research is valuable for advancing the field. Data exchange and interoperability, artificial intelligence, and intelligent health enterprises might be future research directions in Saudi Arabia.

Bayesian Network as a Decision Tool for Predicting ALS Disease

Clinical diagnosis of amyotrophic lateral sclerosis (ALS) is difficult in the early period. But blood tests are less time consuming and low cost methods compared to other methods for the diagnosis. The ALS researchers have been used machine learning methods to predict the genetic architecture of disease. In this study we take advantages of Bayesian networks and machine learning methods to predict the ALS patients with blood plasma protein level and independent personal features. According to the comparison results, Bayesian Networks produced best results with accuracy (0.887), area under the curve (AUC) (0.970) and other comparison metrics. We confirmed that sex and age are effective variables on the ALS. In addition, we found that the probability of onset involvement in the ALS patients is very high. Also, a person’s other chronic or neurological diseases are associated with the ALS disease. Finally, we confirmed that the Parkin level may also have an effect on the ALS disease. While this protein is at very low levels in Parkinson’s patients, it is higher in the ALS patients than all control groups.

Machine learning method for predicting pacemaker implantation following transcatheter aortic valve replacement

BACKGROUND

An accurate assessment of permanent pacemaker implantation (PPI) risk following transcatheter aortic valve replacement (TAVR) is important for clinical decision making. The aims of this study were to investigate the significance and utility of pre- and post-TAVR ECG data and compare machine learning approaches with traditional logistic regression in predicting pacemaker risk following TAVR.

METHODS

Five hundred fifity seven patients in sinus rhythm undergoing TAVR for severe aortic stenosis (AS) were included in the analysis. Baseline demographics, clinical, pre-TAVR ECG, post-TAVR data, post-TAVR ECGs (24 h following TAVR and before PPI), and echocardiographic data were recorded. A Random Forest (RF) algorithm and logistic regression were used to train models for assessing the likelihood of PPI following TAVR.

RESULTS

Average age was 80 ± 9 years, with 52% male. PPI after TAVR occurred in 95 patients (17.1%). The optimal cutoff of delta PR (difference between post and pre TAVR PR intervals) to predict PPI was 20 ms with a sensitivity of 0.82, a specificity of 0.66. With regard to delta QRS, the optimal cutoff was 13 ms with a sensitivity of 0.68 and a specificity of 0.59. The RF model that incorporated post-TAVR ECG data (AUC 0.81) more accurately predicted PPI risk compared to the RF model without post-TAVR ECG data (AUC 0.72). Moreover, the RF model performed better than logistic regression model in predicting PPI risk (AUC: 0.81 vs. 0.69).

CONCLUSIONS

Machine learning using RF methodology is significantly more powerful than traditional logistic regression in predicting PPI risk following TAVR.

Time series prediction of under-five mortality rates for Nigeria: comparative analysis of artificial neural networks, Holt-Winters exponential smoothing and autoregressive integrated moving average models

BACKGROUND

Accurate forecasting model for under-five mortality rate (U5MR) is essential for policy actions and planning. While studies have used traditional time series modeling techniques (e.g., autoregressive integrated moving average (ARIMA) and Holt-Winters smoothing exponential methods), their appropriateness to predict noisy and non-linear data (such as childhood mortality) has been debated. The objective of this study was to model long-term U5MR with group method of data handling (GMDH)-type artificial neural network (ANN), and compare the forecasts with the commonly used conventional statistical methods-ARIMA regression and Holt-Winters exponential smoothing models.

METHODS

The historical dataset of annual U5MR in Nigeria from 1964 to 2017 was obtained from the official website of World Bank. The optimal models for each forecasting methods were used for forecasting mortality rates to 2030 (ending of Sustainable Development Goal era). The predictive performances of the three methods were evaluated, based on root mean squared errors (RMSE), root mean absolute error (RMAE) and modified Nash-Sutcliffe efficiency (NSE) coefficient. Statistically significant differences in loss function between forecasts of GMDH-type ANN model compared to each of the ARIMA and Holt-Winters models were assessed with Diebold-Mariano (DM) test and Deming regression.

RESULTS

The modified NSE coefficient was slightly lower for Holt-Winters methods (96.7%), compared to GMDH-type ANN (99.8%) and ARIMA (99.6%). The RMSE of GMDH-type ANN (0.09) was lower than ARIMA (0.23) and Holt-Winters (2.87). Similarly, RMAE was lowest for GMDH-type ANN (0.25), compared with ARIMA (0.41) and Holt-Winters (1.20). From the DM test, the mean absolute error (MAE) was significantly lower for GMDH-type ANN, compared with ARIMA (difference = 0.11, p-value = 0.0003), and Holt-Winters model (difference = 0.62, p-value< 0.001). Based on the intercepts from Deming regression, the predictions from GMDH-type ANN were more accurate (β₀ = 0.004 ± standard error: 0.06; 95% confidence interval: - 0.113 to 0.122).

CONCLUSIONS

GMDH-type neural network performed better in predicting and forecasting of under-five mortality rates for Nigeria, compared to the ARIMA and Holt-Winters models. Therefore, GMDH-type ANN might be more suitable for data with non-linear or unknown distribution, such as childhood mortality. GMDH-type ANN increases forecasting accuracy of childhood mortalities in order to inform policy actions in Nigeria.

Identification of early liver toxicity gene biomarkers using comparative supervised machine learning

Screening agrochemicals and pharmaceuticals for potential liver toxicity is required for regulatory approval and is an expensive and time-consuming process. The identification and utilization of early exposure gene signatures and robust predictive models in regulatory toxicity testing has the potential to reduce time and costs substantially. In this study, comparative supervised machine learning approaches were applied to the rat liver TG-GATEs dataset to develop feature selection and predictive testing. We identified ten gene biomarkers using three different feature selection methods that predicted liver necrosis with high specificity and selectivity in an independent validation dataset from the Microarray Quality Control (MAQC)-II study. Nine of the ten genes that were selected with the supervised methods are involved in metabolism and detoxification (Car3, Crat, Cyp39a1, Dcd, Lbp, Scly, Slc23a1, and Tkfc) and transcriptional regulation (Ablim3). Several of these genes are also implicated in liver carcinogenesis, including Crat, Car3 and Slc23a1. Our biomarker gene signature provides high statistical accuracy and a manageable number of genes to study as indicators to potentially accelerate toxicity testing based on their ability to induce liver necrosis and, eventually, liver cancer.

Using machine-learning approach to distinguish patients with methamphetamine dependence from healthy subjects in a virtual reality environment

BACKGROUND

The aim of this study was to evaluate whether machine learning (ML) can be used to distinguish patients with methamphetamine dependence from healthy controls by using their surface electroencephalography (EEG) and galvanic skin response (GSR) in a drug-simulated virtual reality (VR) environment.

METHODS

A total of 333 participants with methamphetamine (METH) dependence and 332 healthy control subjects were recruited between January 2018 and January 2019. EEG (five electrodes) and GSR signals were collected under four VR environments: one neutral scenario and three METH-simulated scenarios. Three ML classification techniques were evaluated: random forest (RF), support vector machine (SVM), and logistic regression (LR).

RESULTS

The MANOVA showed no interaction effects among the two subject groups and the 4 VR scenarios. Taking patient groups as the main effect, the METH user group had significantly lower GSR, lower EEG power in delta (p < .001), and alpha bands (p < .001) than healthy subjects. The EEG power of beta band (p < .001) and gamma band (p < .001) was significantly higher in METH group than the control group. Taking the VR scenarios (Neutral versus METH-VR) as the main effects, the GSR, EEG power in delta, theta, and alpha bands in neutral scenario were significantly higher than in the METH-VR scenario (p < .001). The LR algorithm showed the highest specificity and sensitivity in distinguishing methamphetamine-dependent patients from healthy controls.

CONCLUSION

The study shows the potential of using machine learning to distinguish methamphetamine-dependent patients from healthy subjects by using EEG and GSR data. The LR algorithm shows the best performance comparing with SVM and RF algorithm.

Forecasting mortality rates using hybrid Lee–Carter model, artificial neural network and random forest

Inaccurate prediction would cause the insurance company encounter catastrophic losses and may lead to overpriced premiums where low-earning consumers cannot afford to insure themselves. The ability to forecast mortality rates accurately can allow the insurance company to take preventive measures to introduce new policies with reasonable prices. In this paper, several Lee–Carter (LC) based models are used to forecast the mortality rates in a case study of the Malaysian population. The LC-ARIMA model and also a combination of the LC model with two machine learning (ML) methods, namely the random forest (RF) and artificial neural network (ANN) methods are utilized on the prediction of mortality rates for males and females in Malaysia, whereby the LC-Random Forest (LC-RF) hybrid model is a new model that is introduced in this paper. Seventeen years of mortality data in Malaysia are selected as the dataset for this research. To analyze how the forecasting models perform for other countries, we have determined the model that has the best fit and produced the best forecasted mortality rates for all the other countries that are studied. This research has showed that LC-ANN and LC-ARIMA are the best model in predicting the mortality rates of males and females in Malaysia, respectively. This study has also found that the LC-ARIMA model is the best performing model in forecasting the mortality rates in countries that have longer life expectancy and a good healthcare system such as Sweden, Ireland, Japan, Hong Kong, Norway, Switzerland and Czechia. In contrast, the LC-ANN model is the best performing model in forecasting the mortality rates in countries that have a less efficiency, less accessibility healthcare system, and bad personal behavior such as Malaysia, Canada and Latvia.

Predicting long-term multicategory cause of death in patients with prostate cancer: random forest versus multinomial model

The majority of patients with prostate cancer die of non-cancer causes of death (COD). It is thus important to accurately predict multi-category COD in these patients. Random forest (RF), a popular machine learning model, has been shown useful for predicting binary cancer-specific deaths. However, its accuracy for predicting multi-category COD in cancer patients is unclear. We included patients in Surveillance, Epidemiology, and End Results-18 cancer registry-program with prostate cancer diagnosed in 2004 (followed-up through 2016). They were randomly divided into training and testing sets with equal sizes. We evaluated prediction accuracies of RF and conventional statistical/multinomial models for 6-category COD by data-encoding types using the 2-fold cross-validation approach. Among 49,864 prostate cancer patients, 29,611 (59.4%) were alive at the end of follow-up, and 5,448 (10.9%) died of cardiovascular disease, 4,607 (9.2%) of prostate cancer, 3,681 (7.4%) of non-prostate cancer, 717 (1.4%) of infection, and 5,800 (11.6%) of other causes. We predicted 6-category COD among these patients with a mean accuracy of 59.1% (n=240, 95% CI, 58.7%-59.4%) in RF models with one-hot encoding, and 50.4% (95% CI, 49.7%-51.0%) in multinomial models. Tumor characteristics, prostate-specific antigen level, and diagnosis confirmation-method were important in RF and multinomial models. In RF models, no statistical differences were found between the accuracies of training versus cross-validation phases, and those of categorical versus one-hot encoding. We here report that RF models can outperform multinomial logistic models (absolute accuracy-difference, 8.7%) in predicting long-term 6-category COD among prostate cancer patients, while pathology diagnosis itself and tumor pathology remain important factors.

Machine learning algorithms performed no better than regression models for prognostication in traumatic brain injury

OBJECTIVE

We aimed to explore the added value of common machine learning (ML) algorithms for prediction of outcome for moderate and severe traumatic brain injury.

STUDY DESIGN AND SETTING

We performed logistic regression (LR), lasso regression, and ridge regression with key baseline predictors in the IMPACT-II database (15 studies, n = 11,022). ML algorithms included support vector machines, random forests, gradient boosting machines, and artificial neural networks and were trained using the same predictors. To assess generalizability of predictions, we performed internal, internal-external, and external validation on the recent CENTER-TBI study (patients with Glasgow Coma Scale <13, n = 1,554). Both calibration (calibration slope/intercept) and discrimination (area under the curve) was quantified.

RESULTS

In the IMPACT-II database, 3,332/11,022 (30%) died and 5,233(48%) had unfavorable outcome (Glasgow Outcome Scale less than 4). In the CENTER-TBI study, 348/1,554(29%) died and 651(54%) had unfavorable outcome. Discrimination and calibration varied widely between the studies and less so between the studied algorithms. The mean area under the curve was 0.82 for mortality and 0.77 for unfavorable outcomes in the CENTER-TBI study.

CONCLUSION

ML algorithms may not outperform traditional regression approaches in a low-dimensional setting for outcome prediction after moderate or severe traumatic brain injury. Similar to regression-based prediction models, ML algorithms should be rigorously validated to ensure applicability to new populations.

An evaluation of time series summary statistics as features for clinical prediction tasks

Background

Clinical prediction tasks such as patient mortality, length of hospital stay, and disease diagnosis are highly important in critical care research. The existing studies for clinical prediction mainly used simple summary statistics to summarize information from physiological time series. However, this lack of statistics leads to a lack of information. In addition, using only maximum and minimum statistics to indicate patient features fails to provide an adequate explanation. Few studies have evaluated which summary statistics best represent physiological time series.

Methods

In this paper, we summarize 14 statistics describing the characteristics of physiological time series, including the central tendency, dispersion tendency, and distribution shape. Then, we evaluate the use of summary statistics of physiological time series as features for three clinical prediction tasks. To find the combinations of statistics that yield the best performances under different tasks, we use a cross-validation-based genetic algorithm to approximate the optimal statistical combination.

Results

By experiments using the EHRs of 6,927 patients, we obtained prediction results based on both single statistics and commonly used combinations of statistics under three clinical prediction tasks. Based on the results of an embedded cross-validation genetic algorithm, we obtained 25 optimal sets of statistical combinations and then tested their prediction results. By comparing the performances of prediction with single statistics and commonly used combinations of statistics with quantitative analyses of the optimal statistical combinations, we found that some statistics play central roles in patient representation and different prediction tasks have certain commonalities.

Conclusion

Through an in-depth analysis of the results, we found many practical reference points that can provide guidance for subsequent related research. Statistics that indicate dispersion tendency, such as min, max, and range, are more suitable for length of stay prediction tasks, and they also provide information for short-term mortality prediction. Mean and quantiles that reflect the central tendency of physiological time series are more suitable for mortality and disease prediction. Skewness and kurtosis perform poorly when used separately for prediction but can be used as supplementary statistics to improve the overall prediction effect.

A comparative study of machine learning classifiers for risk prediction of asthma disease

Asthma is a chronic disease characterized by wheezing, chest tightening and difficulty in breathing due to inflammation of lung airways. Early risk prediction of asthma is crucial for proper and effective management. This study presents the use of machine learning approach for risk prediction of asthma by evaluating Raman spectral variations between asthmatic as well as healthy sera samples. Specifically, Raman spectra from 150 asthma and 52 healthy control blood sera samples were acquired. Spectral analyses illustrated significant spectral variations (p < 0.0001) in the asthmatic samples when compared with healthy sera. The existing spectral differences were further exploited by using artificial neural network (ANN) along with support vector machine (SVM) and random forest (RF) algorithms towards machine-assisted classification of the two groups. Quantitative comparison of the evaluation metrics of the classification algorithms showed superior performance of SVM model. Our results indicate that Raman spectroscopy in tandem with SVM can be used in the diagnosis and machine-assisted classification of asthma patients with promising accuracy.

Exploration of Machine Learning for Hyperuricemia Prediction Models Based on Basic Health Checkup Tests

BACKGROUND

Machine learning (ML) is a promising methodology for classification and prediction applications in healthcare. However, this method has not been practically established for clinical data. Hyperuricemia is a biomarker of various chronic diseases. We aimed to predict uric acid status from basic healthcare checkup test results using several ML algorithms and to evaluate the performance.

METHODS

We designed a prediction model for hyperuricemia using a comprehensive health checkup database designed by the classification of ML algorithms, such as discrimination analysis, K-nearest neighbor, naïve Bayes (NBC), support vector machine, decision tree, and random forest classification (RFC). The performance of each algorithm was evaluated and compared with the performance of a conventional logistic regression (CLR) algorithm by receiver operating characteristic curve analysis.

RESULTS

Of the 38,001 participants, 7705 were hyperuricemic. For the maximum sensitivity criterion, NBC showed the highest sensitivity (0.73), and RFC showed the second highest (0.66); for the maximum balanced classification rate (BCR) criterion, RFC showed the highest BCR (0.68), and NBC showed the second highest (0.66) among the various ML algorithms for predicting uric acid status. In a comparison to the performance of NBC (area under the curve (AUC) = 0.669, 95% confidence intervals (CI) = 0.669⁻0.675) and RFC (AUC = 0.775, 95% CI 0.770⁻0.780) with a CLR algorithm (AUC = 0.568, 95% CI = 0.563⁻0.571), NBC and RFC showed significantly better performance (p < 0.001).

CONCLUSIONS

The ML model was superior to the CLR model for the prediction of hyperuricemia. Future studies are needed to determine the best-performing ML algorithms based on data set characteristics. We believe that this study will be informative for studies using ML tools in clinical research.

Using machine learning on cardiorespiratory fitness data for predicting hypertension: The Henry Ford ExercIse Testing (FIT) Project

This study evaluates and compares the performance of different machine learning techniques on predicting the individuals at risk of developing hypertension, and who are likely to benefit most from interventions, using the cardiorespiratory fitness data. The dataset of this study contains information of 23,095 patients who underwent clinician- referred exercise treadmill stress testing at Henry Ford Health Systems between 1991 and 2009 and had a complete 10-year follow-up. The variables of the dataset include information on vital signs, diagnosis and clinical laboratory measurements. Six machine learning techniques were investigated: LogitBoost (LB), Bayesian Network classifier (BN), Locally Weighted Naive Bayes (LWB), Artificial Neural Network (ANN), Support Vector Machine (SVM) and Random Tree Forest (RTF). Using different validation methods, the RTF model has shown the best performance (AUC = 0.93) and outperformed all other machine learning techniques examined in this study. The results have also shown that it is critical to carefully explore and evaluate the performance of the machine learning models using various model evaluation methods as the prediction accuracy can significantly differ.