Mental Health World Conference 2026

Abstract

Background :
Subjective cognitive complaints are frequently encountered in psychiatry, yet they often differ from objectively assessed cognitive performance. We investigated whether actigraphy relates these complaints more closely to affective distress or to objective cognition.
Methods :
Baseline data from a prospective transdiagnostic cohort in Korea were analysed. Participants underwent wrist actigraphy for one week and completed self-report assessments covering depression, anxiety, insomnia, stress, loneliness, and subjective cognition. A subset also underwent CANTAB assessments of memory, working memory, and attention. We compared capacity-matched single task neural networks with shared–private multi-task neural networks using repeated five-fold cross-validation. Task-representation similarity was assessed with CKA, and missing CANTAB outcomes were handled through label masking.
Results :
The final analysis included 2,797 participants. Under the full configuration, similarity among affective psychosocial private representations exceeded similarity between affective and cognitive domains at both the mild threshold (0.49 vs 0.08 ratio, 6.1) and the moderate threshold (0.37 vs 0.14 ratio, 2.7). PDQD exhibited greater similarity to affective psychosocial targets than to objective cognitive targets at the mild threshold (0.39 vs 0.08;0.31 95 CI, 0.28 0.33) and the moderate threshold (0.39 vs 0.13 difference,  0.26; 95CI, 0.24 0.28). Across all clustering runs (100, PDQD grouped with affective psychosocial targets. Multi task learning did not yield consistent predictive gains relative to single task models; in the full configuration, a significant improvement was observed only for GAD 7 at the moderate threshold (ΔAUROC, 0.012; q0.011).

Conclusion :
Subjective cognitive complaints were embedded in affective psychosocial representations rather than objective cognitive representations. This finding supports an account of subjective objective cognitive dissociation at the representation level.