Risk factors for depression in China based on machine learning algorithms: A cross-sectional survey of 264,557 non-manual workers

Hui Li,Ying Li,Yinglong Duan,Sha Wang,Min Liu,Yating Luo,Jiangang Wang,Zhiheng Chen,Pinting Yang,Jianfei Xie
DOI: https://doi.org/10.1016/j.jad.2024.09.016
2024-09-05
Abstract:Background: Factors related to depression differ depending on the population studied, and studies focusing on the population of non-manual workers are lacking. Thus, we aimed to identify the risk factors related to depression in non-manual workers in China. Method: A large-scale cross-sectional survey was conducted between January 1, 2015 and December 31, 2020, which included 264,557 non-manual workers from one large physical examination institution in China. The Patient Health Questionnaire (PHQ-2) was used to measure depression. A total of 73 variables covering aspects of sociodemographic characteristics, general examination data, health history, symptoms, eating habits, work situation, general sleep conditions and laboratory findings were included in the collection and analysis. Machine learning algorithms of neural networks and logistic regressions were used to assess the risk of depression and explore its factors. Results: Age, feeling fatigue, sleep quality, overeating, waist-to-hip ratio (WHR), and high-density lipoprotein cholesterol (HDLC) were found to be factors of depression. Two prediction models for depression among Chinese non-manual workers were developed with good AUC (0.820), accuracy (0.943), sensitivity (0.743-0.773), and specificity (0.700-0.729). Limitations: Data were collected at one time point only, meaning that this study cannot explain the causality of the factor on depression. Conclusions: Our finding that age, feeling fatigue, sleep quality, overeating, WHR, and HDL-C were risk factors for depression in non-manual workers may provide strong evidence for health care facilities to develop preventive measures or government policies for non-manual workers.
What problem does this paper attempt to address?