Empty-nest operationalisation and depressive symptoms among rural older Chinese adults: a CHARLS study

Background Empty-nest status describes household circumstances and is used to identify older adults for administrative attention. We examined how its operationalisation affects classification coverage and the information it provides about concurrent depressive symptoms. Methods This cross-sectional China Health and Retirement Longitudinal Study (CHARLS) 2018 analysis included rural adults aged ≥60 years with living children and observed 10-item Center for Epidemiologic Studies Depression Scale (CES-D-10) scores. Proximity-defined empty-nest status required no child living with or nearby the participant; co-residence-defined status required no co-resident child. We compared person-level, household-expanded, and Harmonized CHARLS constructions. Primary association analyses used multiple imputation with community-clustered inference (N = 5,847), with continuous CES-D-10 as the primary outcome and CES-D-10 ≥ 10 as the binary outcome. We assessed fixed-label classification and incremental information beyond a demographic and health benchmark. Complete-case associations were sensitivity analyses; household-composition and 2020 analyses were exploratory. Results Among 5,917 eligible participants, proximity status was classifiable for 41.5%, 67.5%, and 98.8% using person-level, household-expanded, and Harmonized constructions, respectively. The primary adjusted mean difference was 0.198 CES-D-10 points (95% confidence interval (CI) − 0.153 to 0.549); the binary-outcome odds ratio was 1.019 (0.897 to 1.158). Direct-label areas under the receiver operating characteristic curve (AUCs) were 0.493 for proximity and 0.495 for co-residence. Adding either label changed apparent AUC by less than 0.001; community-fold cross-validation similarly showed little incremental discrimination. In exploratory analyses, empty-nest participants living alone had higher scores than non-empty-nest participants (mean difference 1.208; 95% CI 0.429 to 1.987), but their direct comparison with empty-nest participants living with other non-spouse members was uncertain (0.858; −0.505 to 2.222). Conclusions Among rural older adults with living children in this sample, classification coverage depended on how the empty-nest label was constructed, and the aggregate label conveyed little concurrent symptom information. Studies should report the label definition, construction route, and classification coverage; household status should not substitute for direct symptom information. Exploratory household-group differences warrant longitudinal study; these findings do not validate a screening or referral pathway.