Zobrazit v EDS

Interpretable machine learning for classification and risk factor identification of anxiety, depression, and insomnia symptoms after the full opening of China’s COVID-19 lockdown

Uloženo v:
Podrobná bibliografie
Název: Interpretable machine learning for classification and risk factor identification of anxiety, depression, and insomnia symptoms after the full opening of China’s COVID-19 lockdown
Autoři: Zhihao Li, Jinfeng Li, Anqi Zhang, Pengpeng Liu, Guanli Su, Zidan Wang, Yujing Zhang, Yujie Wang, Hao Wu, Yuxia Ma, Jun Ge, Mengyang Liu
Zdroj: BMC Psychiatry, Vol 25, Iss 1, Pp 1-14 (2025)
Informace o vydavateli: BMC, 2025.
Rok vydání: 2025
Sbírka: LCC:Psychiatry
Témata: Mental health, Machine learning, Influencing factor, Categorical Boost, SHapley Additive exPlanations, Psychiatry, RC435-571
Popis: Abstract Background Mental health disorders remain a critical global public health issue, exacerbated by the coronavirus disease 2019 (COVID-19) pandemic. Post-lockdown mental health assessments are scarce. This study aims to develop interpretable machine learning (ML) models to classify and identify key risk factors for anxiety, depression, and insomnia symptoms in mainland China after the opening of COVID-19 lockdown. Methods Cross-sectional data collected in 2023 and 2024 were used as the training set (with test set divided in a 7:3 ratio) and internal validation set, respectively. The Generalized Anxiety Disorder-7 (GAD-7) scale, Patient Health Questionnaire-9 (PHQ-9) scale, and Insomnia Severity Index (ISI) scale were employed to assess symptoms of anxiety, depression, and insomnia, respectively. Seven interpretable machine learning models incorporating 27 relevant influencing factors were applied for the classification and risk factor identification of three mental health symptoms. Results The survey included 65,292 respondents with 36,996 cases in the training set (with test set) and 28,296 cases in the external validation set. The Categorical Boosting (CatBoost) model achieved the highest Area Under the Curve (AUC) value among seven ML models, with 0.817, 0.829, and 0.819 for anxiety, depression, and insomnia symptoms in the test set, and with 0.815, 0.820, and 0.822 for anxiety, depression, and insomnia symptoms in the external validation set. Several factors associated with a lower prevalence of anxiety symptom, including satisfactory neighborhood relationship, extensive knowledge about COVID-19, maintaining a regular sleep-wake cycle, daily vegetable consumption, and sufficient time spent sunbathing. Conversely, factors associated with a higher prevalence comprised history of anxiety, history of insomnia, history of depression, external noise at home, and fear of COVID-19 infection. Similar results were observed for depression and insomnia symptoms. Conclusions The explainable ML model found that healthy habits and positive social ties were linked to lower anxiety, depression, and insomnia symptoms. In contrast, past mental health issues and stressors were tied to higher levels. Public health efforts should focus on encouraging these modifiable habits and strengthening community support, to help improve residents’ mental health, especially in the aftermath of a pandemic.
Druh dokumentu: article
Popis souboru: electronic resource
Jazyk: English
ISSN: 1471-244X
Relation: https://doaj.org/toc/1471-244X
DOI: 10.1186/s12888-025-07569-7
Přístupová URL adresa: https://doaj.org/article/f974ad13e61443b9afa17b609d639ef2
Přístupové číslo: edsdoj.f974ad13e61443b9afa17b609d639ef2
Databáze: Directory of Open Access Journals
Popis
Abstrakt:Abstract Background Mental health disorders remain a critical global public health issue, exacerbated by the coronavirus disease 2019 (COVID-19) pandemic. Post-lockdown mental health assessments are scarce. This study aims to develop interpretable machine learning (ML) models to classify and identify key risk factors for anxiety, depression, and insomnia symptoms in mainland China after the opening of COVID-19 lockdown. Methods Cross-sectional data collected in 2023 and 2024 were used as the training set (with test set divided in a 7:3 ratio) and internal validation set, respectively. The Generalized Anxiety Disorder-7 (GAD-7) scale, Patient Health Questionnaire-9 (PHQ-9) scale, and Insomnia Severity Index (ISI) scale were employed to assess symptoms of anxiety, depression, and insomnia, respectively. Seven interpretable machine learning models incorporating 27 relevant influencing factors were applied for the classification and risk factor identification of three mental health symptoms. Results The survey included 65,292 respondents with 36,996 cases in the training set (with test set) and 28,296 cases in the external validation set. The Categorical Boosting (CatBoost) model achieved the highest Area Under the Curve (AUC) value among seven ML models, with 0.817, 0.829, and 0.819 for anxiety, depression, and insomnia symptoms in the test set, and with 0.815, 0.820, and 0.822 for anxiety, depression, and insomnia symptoms in the external validation set. Several factors associated with a lower prevalence of anxiety symptom, including satisfactory neighborhood relationship, extensive knowledge about COVID-19, maintaining a regular sleep-wake cycle, daily vegetable consumption, and sufficient time spent sunbathing. Conversely, factors associated with a higher prevalence comprised history of anxiety, history of insomnia, history of depression, external noise at home, and fear of COVID-19 infection. Similar results were observed for depression and insomnia symptoms. Conclusions The explainable ML model found that healthy habits and positive social ties were linked to lower anxiety, depression, and insomnia symptoms. In contrast, past mental health issues and stressors were tied to higher levels. Public health efforts should focus on encouraging these modifiable habits and strengthening community support, to help improve residents’ mental health, especially in the aftermath of a pandemic.
ISSN:1471244X
DOI:10.1186/s12888-025-07569-7