Postoperative Sleep Disturbance in Patients Undergoing Arthroplasty: Risk Prediction Modeling Based on Logistic Regression

Abstract

To construct a risk predictive model for postoperative sleep disturbance (PSD) in patients undergoing arthroplasty by using logistic regression. We retrospectively collected the data of 4286 patients who underwent joint replacement surgeries at a tertiary-care hospital in Chengdu, China between January 1, 2017 and September 30, 2021. With 3001 cases in the training set and 1285 cases in the test set, we constructed the model by using a logistic regression algorithm to screen for predictors in Matlab, displaying the predicted risks of postoperative sleep disturbance with nomographs. The performance of the model was assessed by the area under the curve ( AUC) of the receiver operating characteristic curve, accuracy, precision, recall, F1 value, and calibration curve. A total of 9 predictors, including post-admission preoperative sleep disturbance, ward type, body mass index, smoking status, range of diseases, joint mobility (flexion), joint mobility (extension), preoperative last hemoglobin, and type of surgery, were eventually included in the study for predictive modeling . The performance assessment findings of the predictive model were as follows, AUC value, 0.708 (95% confidence interval: 0.677-0.740), accuracy, 75.20%, precision, 65.80%, recall, 43.70%, and F1 value, 0.525. The calibration curve showed good agreement between the predicted probabilities and the actual data. The model constructed in the study has good predictive efficacy and the nomographs are simple and easy to use. With this model, health workers can make preoperative prediction of the risk of PSD in arthroplasty patients based on the predictors, which facilitates early prevention and reduces the risk of postoperative sleep disturbance in patients.