Prediction and Analysis of Heart Failure Decompensation Events Based on Telemonitored Data and Artificial Intelligence Methods

Abstract

Cardiovascular diseases are the leading cause of death globally, taking an estimated 17.9 million lives each year. Heart failure (HF) occurs when the heart is not able to pump enough blood to satisfy metabolic needs. People diagnosed with chronic HF may suffer from cardiac decompensation events (CDEs), which cause patients' worsening. Being able to intervene before decompensation occurs is the major challenge addressed in this study. The aim of this study is to exploit available patient data to develop an artificial intelligence (AI) model capable of predicting the risk of CDEs timely and accurately. Materials and Methods: The vital variables of patients (n = 488) diagnosed with chronic heart failure were monitored between 2014 and 2022. Several supervised classification models were trained with these monitoring data to predict CDEs, using clinicians' annotations as the gold standard. Feature extraction methods were applied to identify significant variables. Results: The XGBoost classifier achieved an AUC of 0.72 in the cross-validation process and 0.69 in the testing set. The most predictive physiological variables for CAE decompensations are weight gain, oxygen saturation in the final days, and heart rate. Additionally, the answers to questionnaires on wellbeing, orthopnoea, and ankles are strongly significant predictors.

Keywords: XGBoost; decompensation; heart failure; logistic regression; machine learning; monitoring; supervised classification.

Figure 1

The boxplots in this figure represent the different AUC values obtained in each folder of the CV (10 AUC values per classifier; in blue). The green dot represents the AUC value obtained applying the model to the training set.

Figure 2

The ROC curves of the classifiers in the test partition. Next to them is the value of the area under this curve. Dotted line represents AUC value of 0.5, which means random prediction.

Figure 3

Confusion matrix of the XGBoost model results, divided into low-, medium-, and high-risk categories.

Figure 4

The importance given to each variable by the XGBoost model. The variable name is the acronym for the variable or the tag for the questionnaire. The number next to the name stands for the day of the variable value report (e.g., if it is zero, it is the variable value for the same day of risk prediction). Note: Variables with zero importance are not shown in the figure.

Figure 5

Confusion matrix of the logistic regression model results, divided into low-, medium-, and high-risk categories.

Figure 6

Feature importance calculated from the logistic regression model.