Across four aging cohorts (CHARLS, HRS, ELSA, SHARE; 91,043 adults aged 50+), depression's link to incident cognitive impairment depended heavily on how impairment was measured. With self-reported or non-adjudicated outcomes, worsening depressive symptoms predicted impairment in only two cohorts (HR 1.31 and 1.99), and persistent high symptoms never reached significance (HR 1.07-1.15). Under clinical consensus diagnosis (n=3,196), every non-low trajectory was associated with mild cognitive impairment or dementia (OR 1.66-2.29; dementia-only persistent high OR 3.32). Effect sizes rose 0.314 log HR per rigor level.
This offers a plausible explanation for why published estimates of depression as a dementia risk factor scatter so widely: crude outcomes dilute real associations through misclassification. The pattern fits the broader literature, where depression is among the 14 modifiable risk factors in the Lancet Commission framework, yet effect sizes vary between cohorts. The practical takeaway for adults is that persistent or worsening depressive symptoms in midlife and later deserve attention as a possible cognitive health signal, not merely a mood problem. Important caveats apply. This is a preprint and has not been peer reviewed. The design is observational, so reverse causation (depression as an early dementia symptom) cannot be excluded, though lagged analyses were reportedly robust. The clinical-diagnosis subsample is small relative to the full cohort, and the meta-regression with only four cohorts, reported as exploratory with R2 of 100%, should be read cautiously. The claim that treating depression yields cognitive benefit remains untested here. Overall, a useful methodological clarification rather than a paradigm shift.