Novel insight into prediction model for sleep quality among college students: A LASSO-derived sleep evaluation

Ling Yao ^1,2 Qingquan Chen ³ Yang Kang ² Zhihua Zheng ² Zhihan Chen ² Danna Wang ² Yining Xia ² Dingquan Chen ² Lufeng Chen ^1*

• ¹ The Second Affiliated Hospital of Fujian Medical University, Quanzhou, China
• ² School of Clinical Medicine, Fujian Medical University, Fuzhou, Fujian Province, China
• ³ School of Public Health, Fujian Medical University, Fuzhou, Fujian Province, China

Background: Sleep disturbance has become a significant concern among college students, as it can lead to various mental and physical disorders. This study aims to provide a fresh perspective by developing and validating a predictive model for sleep quality among college students.Methods: Data from 20,645 college students in Fujian Province, China, collected between 5th April and 16th April 2022, were analyzed. Participants completed the Pittsburgh Sleep Quality Index (PSQI) scale, a self-designed general data questionnaire, and a sleep quality influencing factor questionnaire. Multinomial logistic regression, LASSO regression, and Boruta feature selection methods were utilized to select relevant variables. The data were then divided into a training–testing set (70%) and an independent validation set (30%) using stratified sampling. Six machine learning techniques, including artificial neural network (ANN), decision tree, gradient-boosting tree, k-nearest neighbor, naïve Bayes, and random forest, were developed and validated. Finally, an online sleep evaluation website was established based on the best-fitting prediction model.Results: The mean global PSQI score was 6.02 ± 3.112, with a sleep disturbance prevalence of 28.9% (defined as a global PSQI score > 7). The LASSO regression model identified eight predictors: age, specialty, respiratory history, coffee consumption, staying up late, prolonged online activity, sudden changes, and impatient closed-loop management. Among the evaluated models, the ANN demonstrated superior performance with an area under the receiver operating characteristic curve (AUC) of 0.713 (95% CI: 0.696–0.730), accuracy of 0.669 (95% CI: 0.669–0.669), sensitivity of 0.682 (95% CI: 0.699–0.665), specificity of 0.637 (95% CI: 0.665–0.610). Decision curve analysis and clinical impact analysis further confirmed the model’s clinical utility.Conclusions: This study developed a prediction model for sleep disturbance among college students using a LASSO regression and ANN, incorporating eight predictors. The model can serve as an intuitive and practical tool for predicting sleep quality and supporting effective management and healthcare on college campuses.