In 8,464 SARS-CoV-2-infected participants from the German National Cohort (NAKO), pre-infection mental health — measured via PHQ-9 and GAD-7 depression and anxiety questionnaires collected 3–8 years before infection — emerged as the dominant predictor of Post-COVID Condition (PCC), contributing 38% of feature importance in a stacked machine-learning ensemble. The odds ratio was 2.06 (95% CI 1.83–2.32), and critically, the association persisted after adjusting for subsequent symptom trajectories (OR 1.38, CI 1.20–1.58). By contrast, structural brain MRI across six parcellations performed at or near chance (ROC-AUC 0.496–0.516), suggesting pre-existing brain volume differences do not meaningfully predict PCC.
This preprint, not yet peer-reviewed, carries several important caveats: infection status and PCC were self-reported without clinical confirmation, introducing potential recall and classification bias. The finding that baseline mental health predicted self-reported smell loss but not objectively measured hyposmia (OR 0.89 vs. 1.92) is a striking internal signal suggesting symptom perception or reporting bias partly drives the association — a limitation that meaningfully tempers causal interpretation. Nevertheless, the large cohort size and prospective pre-infection measurement are genuine strengths rarely achieved in long-COVID research. For adults, this reinforces the practical value of managing mental health proactively, not solely for wellbeing but potentially as a resilience factor against post-viral syndromes. Whether treating anxiety or depression before infection would reduce PCC risk remains an open and clinically important question.