Development, validation, and visualization of a novel depression screening tool for hypertensive patients based on NHANES data: A cross-sectional study.

Figures

• 

  Figure 1.. Flowchart for participant screening and study design. ACC = accuracy, DT = decision tree, KNN = K-nearest neighbor, LGBM = light gradient boosting machine, LR = logistic regression, MCC = Matthews correlation coefficient, NHANES = National Health and Nutrition Examination Survey, NPV = negative predictive value, PPV = positive predictive value, RF = random forest, SEN = sensitivity, SPE = specificity, SVM = support vector machine, XGBoost = extreme gradient boosting.

• 

  Figure 2.. Clinical feature selection via the LASSO regression model. (A) LASSO coefficients of a total of 23 clinical features. (B) The partial likelihood deviance (binomial deviance) curve was plotted against log (lambda). The dotted vertical lines represent the optimal predictors using the minimum criteria (min. criteria) and the 1 SE of the minimum criteria (1 − SE criteria). LASSO = least absolute shrinkage and selection operator, SE = standard error.

• 

  Figure 3.. Summary of ROC curves for the 7 machine learning models. DT = decision tree, KNN = K-nearest neighbor, LGBM = light gradient boosting machine, LR = logistic regression, RF = random forest, ROC = receiver operating characteristic, SVM = support vector machine, XGBoost = extreme gradient boosting.

• 

  Figure 4.. Confusion matrix comparison of 7 machine learning models. DT = decision tree, KNN = K-nearest neighbor, LGBM = light gradient boosting machine, LR = logistic regression, RF = random forest, SVM = support vector machine, XGBoost = extreme gradient boosting.

• 

  Figure 5.. Decision curve analysis for predicting depression in hypertensive patients. The x-axis represents threshold probability, and the y-axis represents net benefit. The RF-based model exhibits superior net benefit compared with other models and strategies when the threshold probability ranges from 0.07 to 0.39. DT = decision tree, KNN = K-nearest neighbor, LightGBM = light gradient boosting machine, RF = random forest, rSVM = regularized support vector machine, XGBoost = extreme gradient boosting.

• 

  Figure 6.. Calibration plots of depression prediction models in hypertensive patients. The x-axis denotes the midpoint of the predicted probability intervals (25th, 50th, and 75th percentiles), and the y-axis represents the actual percentage of patients with depression (Perct). Models include logistic regression, DT, RF, XGBoost, rSVM, LightGBM, and KNN. The RF model shows the closest alignment between predicted and actual probabilities, indicating optimal calibration performance. DT = decision tree, RF = random forest, XGBoost = extreme gradient boosting, rSVM = regularized support vector machine, LightGBM = light gradient boosting machine, KNN = K-nearest neighbor.

• 

  Figure 7.. Summary plot of SHAP values for the model built by the RF algorithm. The horizontal position “SHAP value” indicates whether the effect of the value is associated with a higher or lower prediction, and the color of each SHAP value point indicates whether the observed value is higher (blue) or lower (red). The vertical coordinates show the importance of the features, sorted in descending order of importance of the variables, with the variables above being more important to the model. COPD = chronic obstructive pulmonary disease, PIR = poverty-to-income ratio, RF = random forest, SHAP = SHapley Additive exPlanations, SLQ = sleep questionnaire.

• 

  Figure 8.. Screenshot of an example for the results presentation. COPD = chronic obstructive pulmonary disease, PIR = poverty-to-income ratio, SLQ = sleep questionnaire.