Prediction of adverse cardiac events in emergency department patients with chest pain using machine learning for variable selection

Abstract

BackgroundThe key aim of triage in chest pain patients is to identify those with high risk of adverse cardiac events as they require intensive monitoring and early intervention. In this study, we aim to discover the most relevant variables for risk prediction of major adverse cardiac events (MACE) using clinical signs and heart rate variability.MethodsA total of 702 chest pain patients at the Emergency Department (ED) of a tertiary hospital in Singapore were included in this study. The recruited patients were at least 30 years of age and who presented to the ED with a primary complaint of non-traumatic chest pain. The primary outcome was a composite of MACE such as death and cardiac arrest within 72 h of arrival at the ED. For each patient, eight clinical signs such as blood pressure and temperature were measured, and a 5-min ECG was recorded to derive heart rate variability parameters. A random forest-based novel method was developed to select the most relevant variables. A geometric distance-based machine learning scoring system was then implemented to derive a risk score from 0 to 100.ResultsOut of 702 patients, 29 (4.1%) met the primary outcome. We selected the 3 most relevant variables for predicting MACE, which were systolic blood pressure, the mean RR interval and the mean instantaneous heart rate. The scoring system with these 3 variables produced an area under the curve (AUC) of 0.812, and a cutoff score of 43 gave a sensitivity of 82.8% and specificity of 63.4%, while the scoring system with all the 23 variables had an AUC of 0.736, and a cutoff score of 49 gave a sensitivity of 72.4% and specificity of 63.0%. Conventional thrombolysis in myocardial infarction score and the modified early warning score achieved AUC values of 0.637 and 0.622, respectively.ConclusionsIt is observed that a few predictors outperformed the whole set of variables in predicting MACE within 72 h. We conclude that more predictors do not necessarily guarantee better prediction results. Furthermore, machine learning-based variable selection seems promising in discovering a few relevant and significant measures as predictors.

Phenotyping: Using Machine Learning for Improved Pairwise Genotype Classification Based on Root Traits

“…SVMs models using all available traits or including five randomly selected root traits (R_5) were not able to increase the overall accuracy, which confirmed the necessity of root traits selection through RF in cultivar differentiation. This finding is in accordance with previous ML approaches in other scientific fields (Wang et al, 2010; Löw et al, 2012; Liu et al, 2014). The improved accuracy probably benefits from alleviating the ‘curse of dimensionality’ through root traits selection, removing non-informative signals (Chu et al, 2012).…”

“…The validation accuracy was treated as final prediction accuracy of SVMs/RF classifications. Classifications with an average prediction accuracy ≥80% were regarded as a high accuracy classifications (HACCs); the 80% level was determined acceptable by previous ML studies (Wang et al, 2010; Liu et al, 2014; Shang and Chisholm, 2014; Zheng et al, 2014; Sacchet et al, 2015). The whole process – RF ranking of root traits in each cultivar pair, SVMs and RF classification of pairs using different mtrys and Timp s – was repeated three times; the average accuracy with standard error was calculated.…”

“…RF has been applied in several biological fields, like gene (Díaz-Uriarte and De Andres, 2006) and protein sequence (Pan and Shen, 2009) selection, and disease prediction (Yang et al, 2014). However, most previous studies mainly focused on improving classification accuracy with variable selection rather than variable interpretation (Liu et al, 2014; da Costa et al, 2015; Gowin et al, 2015) because the variable importance measure is biased in the standard RF algorithm – overestimating the importance of correlated predictor variables (Strobl and Zeileis, 2008). However, unbiased variable selection is essential to stable classification and meaningful interpretation of plant traits and other data and can be achieved by using an improved RF algorithm – based on a conditional permutation scheme as a computational means to determine variable importance (Strobl and Zeileis, 2008; Strobl et al, 2008).…”

Root traits of European Vicia faba cultivars-Using machine learning to explore adaptations to agroclimatic conditions

“…The current neglect of SML techniques in plant phenotyping is partially based on earlier studies who failed to show how plant traits reflect environmental differences (Bari et al, , ) or even produced (partially) misleading results due to a biased trait selection method (Khazaei, Street, Bari, et al, ). Furthermore, widely different classification accuracies have been deemed acceptable in previous studies (Bari et al, ; Liu et al, ; Wang, Huang, & Yang, ; Zheng, Yoon, & Lam, )—restraining “trust” in the resilience of SML‐based data analysis within the scientific community. Because the classification accuracy is a result of both data and analysis method, high generalization accuracies cannot be expected per se and are also only a prerequisite for discovering an important phenotype.…”

“…Furthermore, widely different classification accuracies have been deemed acceptable in previous studies (Bari et al, 2016;Liu et al, 2014;Wang, Huang, & Yang, 2010;Zheng, Yoon, & Lam, 2014)restraining "trust" in the resilience of SML-based data analysis within the scientific community. Because the classification accuracy is a result of both data and analysis method, high generalization accuracies cannot be expected per se and are also only a prerequisite for discovering an important phenotype.…”