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

Abstract

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.

AUD-DSS: a decision support system for early detection of patients with alcohol use disorder

“…In terms of ADs, DK70 (alcohol-induced liver disease), DS82 (fracture of the lower leg, including the ankle), DI63 (cerebral infarction), DK86 (other diseases of the pancreas), and DM19 (another arthrosis) are the top five AD factors. Clinical factors and comorbidities associated with the prediction of AUD have been identified in our previous studies [ 34 , 75 ], which provide a more in-depth analysis of ADs. However, no study has examined clinical and risk factors such as systolic and diastolic blood pressure, BMI, weight, saturation, temperature, and AD for the prediction of patients with AUD from EHRs in a single study.…”

“…Clinical factors related to the prediction of patients with AUD were also presented. In the literature, risk factors such as gender and age have been discovered in many studies [ 27 , 30 , 34 ]. In comparison to the work in [ 34 ], we can see that age was still the most important factor related to the early detection of patients with AUD.…”

“…One of the main disadvantages of their study was that they used a filter feature selection method. In such a method, classifier biases cannot be included in the classification model [ 34 , 79 ]. Another disadvantage of their study is that noise removal and balancing techniques were not considered.…”

“…Previous studies have shown that ML algorithms such as artificial neural networks [ 24 – 26 ], logistic regression (LR) [ 25 , 27 – 29 ], support vector machines (SVM) [ 24 – 26 , 29 – 32 ], random forests (RF) [ 24 , 25 , 28 , 29 , 31 ], elastic nets [ 24 , 31 ], k-nearest neighbour (KNN) [ 25 ], decision trees (DT) [ 25 , 28 ], naive bayes [ 28 ], etc., have been successfully utilized to develop predictive models for the early detection of patients with AUD based on EHRs. However, in the ML field, missing values, feature redundancy, noisy datasets, and imbalanced data may arise and impact the performance of such prediction models [ 33 , 34 ]. Previous studies have reported that by addressing the feature redundancy and imbalanced class distribution problems and handling missing values, the predictive performance of ML models can be enhanced significantly [ 35 , 36 ].…”