Background: The occurrence of perioperative heart failure will affect the quality of medical services and threaten the safety of patients. It is of great significance to improve the quality of medical service and ensure the safety of patients with perioperative heart failure by scientific method.
Methods: Firstly, the data of preoperative patients are preprocessed, and the patient data is divided into the structural data of numerical data and the unstructured data of textual data. The numerical data of the patient is constructed by the gradient boosting tree model, and the textual features of the patients are extracted by the topic model of the text-based data. Fusion of textual features and numerical features, and finally through a simple logistic regression to predict the patient critical illness. Results: To evaluate the performance of the proposed method, we made preoperative prediction of critical events of heart failure in perioperative period based on the real operative data of patients from a hospital, and the results showed that the sensitivity and specificity of the model proposed could reach 94% and 97%, thus verifying the feasibility and effectiveness of the model.
Conclusions: a novel method of predicting preoperative critical disease is proposed for the properties of medical heterogeneity data in perioperative patients. This work is the first study that integrates the numerical laboratory data and textual diagnostic data of patient for extracting the predictive feature by XGBOOST and LDA model respectively, and builds a low cost, scalable and effective model of predicting preoperative critical disease by using logistic regression. The experimental demonstrate show that it is better to predict whether the patients have heart failure with Numerical and Textual Attributes of patients