Answering Clinical Questions Using Machine Learning: Should We Look at Diastolic Blood Pressure When Tailoring Blood Pressure Control?

Figures

• 

  Figure 1. Flowchart presenting ML experiment phases. DBP—diastolic blood pressure, SMOTE—synthetic minority oversampling technique.

• 

  Figure 2. Comparison of the performance of the classifiers when DBP was included (DBP (+)) and excluded (DBP (−)) as the predictor presented as boxplots for all outcome variables examined in the unselected population. AUC—area under curve, mF1—mean F-measure, wF1—mean weighted F-measure, no differences detected for all of the comparisons.

• 

  Figure 3. Comparison of the performance of the classifiers when DBP was included (DBP (+)) and excluded (DBP (−)) as the predictor presented as boxplots for all outcome variables examined in the selected population of subjects with DBP < 70 mmHg. AUC—area under curve, mF1—mean F-measure, wF1—mean weighted F-measure, no differences detected for all of the comparisons.

• 

  Figure 4. Comparison of the performance of the classifiers in the selected (selection) and unselected (no selection) population presented as boxplots for all of outcome variables when diastolic blood pressure was included and excluded as the predictor. AUC—area under curve, mF1—mean F-measure, wF1—mean weighted F-measure, no differences detected for all of the comparisons.

• 

  Figure 5. ROC curves of the classifiers detecting of primary outcome when DBP was included as the risk factor. DT—decision tree , KNN—k-nearest neighbor classifier , LR—logistic regression , MLP—multi-layer perceptron , NB—naive Bayesian classifier , RF—random forest .