Development and validation of prediction models for poor sleep quality among older adults in the post-COVID-19 pandemic era

Abstract

Background

Corona Virus Disease 2019 (COVID-19) has a significant impact on sleep quality. However, the effects on sleep quality in the post-COVID-19 pandemic era remain unclear, and there is a lack of a screening tool for Chinese older adults. This study aimed to understand the prevalence of poor sleep quality and determine sensitive variables to develop an effective prediction model for screening sleep problems during infectious diseases outbreaks.

Materials and Methods

The Peking University Health Cohort included 10,156 participants enrolled from April to May 2023. The Pittsburgh Sleep Quality Index (PSQI) scale was used to assess sleep quality. The data were randomly divided into a training-testing cohort (n = 7109, 70%) and an independent validation cohort (n = 3027, 30%). Five prediction models with 10-fold cross validation including the Least Absolute Shrinkage and Selection Operator (LASSO), Stochastic Volatility Model (SVM), Random Forest (RF), Artificial Neural Network (ANN), and XGBoost model based on the area under curve (AUC) were used to develop and validate predictors.

Results

The prevalence of poor sleep quality (PSQI >7) was 30.69% (3117/10,156). Among the generated models, the LASSO model outperformed SVM (AUC 0.579), RF (AUC 0.626), ANN (AUC 0.615) and XGBoost (AUC 0.606), with an AUC of 0.7. Finally, a total of 12 variables related to sleep quality were used as parameters in the prediction models. These variables included age, gender, ethnicity, educational level, residence, marital status, history of chronic diseases, SARS-CoV-2 infection, COVID-19 vaccination, social support, depressive symptoms, and cognitive impairment among older adults during the post-COVID-19 pandemic. The nomogram illustrated that depressive symptoms contributed the most to the prediction of poor sleep quality, followed by age and residence.

Conclusions

This nomogram, based on twelve-variable, could potentially serve as a practical and reliable tool for early identification of poor sleep quality among older adults during the post-pandemic period.

KEY MESSAGE

• The poor sleep quality (PSQI >7) was still prevalent among older adults during the post-COVID-19 pandemic.

• The LASSO model was the best model to predict poor sleep quality among older adults, compared with SVM, RF, ANN and XGBoost.

• This prediction model, based on twelve variables, may potentially serve as a practical and reliable tool for the early identification of poor sleep quality among older adults during the post-pandemic period.

Keywords:

• Sleep quality
• COVID
• older adults
• machine learning
• regression model

Acknowledgments

We are also grateful to all family physicians and health workers on collection of data.

Authors’ contribution statement

JL conceptualized and designed the study, MD, YW, WY, QL, ML, XY and SW did data acquisition, MD did data curation, formal analysis, and visualization, MD did writing - original draft, ML, XY, SW, ML and JL did writing- reviewing and editing.

Ethical approval and consent to participate

This study was approved by the institutional review boards at Peking University (IRB00001052-21126). All participants had oral informed consent at the time of participation. The research has been performed in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Disclosure statement

The authors declare that they have no competing interests.

Data availability statement

Data are obtained according to corresponding author permission.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [grant numbers 72122001, 72211540398].