Digital Cognitive Phenotyping for Differential Diagnosis and Monitoring in Neurological Conditions

To assess the utility, accessibility, and equivalence to supervised scales of online cognitive assessment in older individuals with cognitive impairment.

Patients with Alzheimer's disease (AD, n = 31), idiopathic normal pressure hydrocephalus (iNPH, n = 26), and traumatic brain injury (TBI, n = 23) completed online cognitive tasks (Cognitron). We evaluated cognition relative to a large normative dataset (N ≈ 400,000), adjusting for device and demographics which can affect performance. Principal Component Analysis (PCA) was used to derive domain-specific and total composite scores. We compared clinical groups and correlated performance with standard assessments.

Uptake was ~70%. PCA identified components across memory, processing speed, language, and executive functions. AD showed memory and language impairments compared with the norms and other groups. iNPH had greater executive and processing speed deficits, consistent with a subcortical impairment profile. TBI showed milder deficits in memory, working memory, and language. Cognitron total composite was associated with standard supervised tests (ADAS-Cog: β = -0.76, p < 0.001 and ACE-III: β = 0.69, p < 0.001). In iNPH, Cognitron composite predicted walking speed (estimate = 1.10, p < 0.001), a core clinical feature of the disease which is difficult to evaluate remotely. We selected five tasks with high completion rates, discriminability between conditions, and broad cognitive coverage. The derived short composite showed very high accuracy in separating AD (AUC = 0.94) and iNPH (AUC = 0.90) from age-matched norms; performance was weaker for TBI (AUC = 0.66).

Online assessment in older clinical populations is feasible and sensitive to subtle disease-specific cognitive deficits. A demographically adjusted, 15-min battery offers a scalable adjunct to standard testing, with potential to reduce burden on patients and healthcare systems.