A 14-minute digital assessment recognizes mild cognitive impairment—but what happens after the score?

A future cognitive-health clinic may measure how someone speaks, moves and solves problems, rather than rely only on a short paper questionnaire. The NeuroMarker study offers a promising look at that possibility. Across an adjudicated research cohort, a combination of tasks distinguished mild cognitive impairment from unimpaired cognition reasonably well.

AUC measures discrimination across thresholds, not whether a positive result means a particular person has Alzheimer’s disease. Clinical utility requires realistic disease prevalence, prespecified thresholds, confirmatory evaluation, false-positive burden, accessibility and longitudinal outcomes. The modest classifier agreement on repeated measurements also deserves careful examination.

The valuable question is whether such tools help patients receive better evaluation earlier and improve subsequent decisions—not whether a dashboard can draw a persuasive curve.

The evidence: 638 adults aged 50 or older in three countries, 441 cognitively unimpaired and 197 with adjudicated mild cognitive impairment; self-administered multimodal tablet tasks involving motor, speech and augmented reality; independent cohort n=75. Median completion time 13.8 minutes. ROC AUC was 0.90 (95% CI 0.88–0.92), and 0.87 (0.80–0.93) in the 75-person independent sample. Test-retest classifier agreement κ=0.60; composite/domain ICCs ranged 0.55–0.80. Funded in part by Altoida Inc., a platform developer.

dementia-prevention claim.