Development and temporal validation of a probable depression risk prediction model for US adults with questionnaire-ascertained probable obstructive sleep apnea

Using NHANES, we developed and temporally validated a nomogram for screen-positive probable depression (PHQ-9 ⩾ 10) in U.S. adults with questionnaire-ascertained probable obstructive sleep apnea (OSA). Probable OSA used an OR-based symptom rule: any one of habitual snoring ⩾ 3 nights/week, snorting/gasping/breathing pauses ⩾ 3 nights/week, or excessive daytime sleepiness 16–30 times/month despite ⩾ 7 hours sleep was sufficient. Development/internal validation pooled NHANES 2005–2008 and 2015–2016 (n = 2802; training = 1961; internal = 841); temporal validation used 2017–2018 (n = 508). Participants were mainly middle-aged, 59.5%–61.4% male, and 44.1%–47.1% non-Hispanic White. LASSO retained 11 candidates; multivariable logistic regression selected seven predictors: gender, marital status, poverty-income ratio, self-rated health, cardiovascular disease, smoking, and sleep hours. AUCs were 0.822, 0.817, and 0.785 in training, internal, and temporal sets, with good calibration and decision-curve benefit. The nomogram may help prioritize depression assessment and follow-up triage in large survey and real-world primary-care settings, while objective sleep testing remains necessary for diagnostic confirmation of OSA.