Investigating the role of depression in obstructive sleep apnea and predicting risk factors for OSA in depressed patients: machine learning-assisted evidence from NHANES

Comprehensive Summary

The study being conducted by Cheng et al. aims to investigate the relationship between obstructive sleep apnea (OSA) and depression, which has been a topic of debate for years. The study utilized data from four cycles of the U.S. National Health and Nutrition Examination Survey (NHANES), in which 14,492 adults total participants including 1,212 individuals with depression were analyzed. For each patient, they were given a PHQ-9 depression questionnaire and OSA was defined with sleep-related questionnaire material. Researchers found that depression was consistently associated with higher odds of OSA, even after adjustment for confounding variables. Additionally, subgroup analyses (by sex, BMI, smoking, alcohol use, education, income, hypertension, and diabetes) showed no significant interaction effects, suggesting that the association is broadly consistent across populations. Among participants with depression, researchers used machine learning (ML) to predict OSA risk. Model interpretability (through SHAP analysis) showed that BMI was the most important predictor of OSA, followed by age and marital status, with additional contributions from hypertension and caffeine intake. The findings support a positive association between depression and OSA, with potential biological explanations including inflammation, oxidative stress, circadian rhythm disruption, and obesity-related neuroendocrine changes. Particularly, the use of ML is important in highlighting the importance of integrated screening and management strategies for mental health and sleep disorders.

Outcomes and Implications

Obstructive sleep apnea (OSA) affects nearly 1 billion people worldwide and is linked to serious health outcomes such as hypertension, cardiovascular disease, and stroke. Depression is also increasing globally and is a leading contributor to disability and mortality. While many studies suggest a connection between OSA and depression, findings have been largely inconsistent. This study reinforces that depression is not just a comorbidity, but a meaningful risk marker for OSA, even after adjusting for other lifestyle factors. Therefore, mental health clinics, primary care, and psychiatry settings could become frontline screening points for sleep apnea, especially in patients who do not fit the classic stereotype. This work is particularly relevant for primary care physicians, psychiatrists and psychologists, behavioral health clinics, sleep medicine referral triage, and population health programs. In the future, this could manifest itself as a risk calculator embedded in electronic health records (EHRs) or ML-based screening becoming a part of routine mental health assessment. Treating OSA may improve depressive symptoms, and vice versa. Overall, ML models offer a promising tool for risk stratification and early detection of OSA in depressed individuals, especially by capturing non-linear relationships.