IDHPre: Intradialytic Hypotension Prediction Model Based on Fully Observed Features

In the course of dialysis, hypotension is a common complication, known as Intradialytic Hypotension (IDH). To improve the predictive accuracy and practicality of IDH, we conducted research on machine learning techniques. To effectively impute missing values while maintaining the interpretability and confidence intervals of the data, we utilized the covariance structure of observed data to estimate missing values for imputation, preserving the correlation between data features and variables. We also introduced a feature dimensionality reduction module that supports nonlinear features, employing a tree-based method capable of capturing nonlinear relationships in the data. This method prioritizes basic functions with a small number of features to effectively describe the complex structure of the data. We used a weighted Lasso optimization criterion to select a sparse subset of features for the basic functions to ensure the retention of the most important features while reducing the complexity of the model. Experimental results demonstrate that our proposed model achieves high accuracy and stability in predicting IDH. Compared to traditional methods, it can more accurately identify patient risks, providing strong support for clinical decision-making. Therefore, our research provides a new approach and perspective for predicting and preventing IDH, with promising clinical application prospects.