Predicting sleep quality among college students during COVID-19 lockdown using a LASSO-based neural network model

Abstract

Background

In March 2022, a new outbreak of COVID-19 emerged in Quanzhou, leading to the implementation of strict lockdown management measures in colleges. While existing research has indicated that the pandemic has had a significant impact on sleep quality, the specific effects of containment measures on college students’ sleep patterns remain understudied.

Objective

This study aimed to understand the sleep quality of college students in Fujian Province during the epidemic and determine sensitive variables, in order to develop an efficient prediction model for the early screening of sleep problems in college students.

Methods

A cross-sectional survey was conducted April 5–16, 2022 to survey college students in Quanzhou. A total of 4959 college students in Quanzhou were enrolled in this study. Descriptive analysis, univariate analysis, correlation analysis, and multiple regression analysis were used to explore the influencing factors regarding sleep quality. In addition, we constructed eight sleep quality risk prediction models to predict sleep quality.

Results

A mean PSQI total score of 6.03 ± 3.21 and a sleep disorder rate of 29.4% (PSQI > 7) were obtained. Sleep quality, sleep latency, sleep efficiency, diurnal dysfunction, and PSQI score were all higher than the national norm (P < 0.05). A total of eight significant predictors finally identified by the LASSO algorithm was incorporated into prediction models. Through a series of assessments, we identified the artificial neural network model as the best model, achieving an area under curve of 73.8% an accuracy of 67.3%, a precision of 84.0%, a recall of 66.3%, and an F1 score of 69.3%. These performance indices suggest that the ANN model outperforms other models. It is noteworthy that the threshold probabilities for net benefit were found to be between 0.81 and 0.92 and the clinical impact curve confirmed that the models’ predictions were particularly effective in identifying individuals with poor sleep quality when the threshold probability was set above 70%. These findings underscore the potential clinical utility of our models for the early detection of sleep disorders.

Conclusions

In Quanzhou, under COVID-19 quarantine management, the sleep quality of college students was affected to a certain extent, and their PSQI scores were higher than the national average in China. The artificial neural network model had the best performance, and it is expected to be used to provide early interventions to prevent sleep disorders.