B-201: Predicting Conversion to MCI in ADNI: Traditional Neuropsychological Vs. Digital Cognitive Tests

Abstract

Objective: Compare the predictive capacity of traditional neuropsychological (NP) and self-administered digital cognitive tests for identifying future conversion from cognitively unimpaired (CU) to mild cognitive impairment (MCI). Method/Data Selection: Participants were 427 CU older adults (mean age: 72.6; 59.7% female; 82.2% White; 76.8% college graduates) from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Fifty-five converted to MCI (mean 3.8 years from baseline to diagnosis), while 372 remained cognitively stable. Baseline scores were selected as the earliest available timepoint with NP and digital testing within 30 days. Traditional NP included five measures: Rey Auditory Verbal Learning Test (RAVLT) total learning and delayed recall, Trail Making Test Parts A and B, and category fluency (animals). Digital included two Cogstate Brief Battery (CBB) composites (Psychomotor/Attention, Learning/Working Memory). Logistic regression controlling for age, sex, education, and race with leave-one-out cross-validation classified converters versus cognitively stable participants. Area under the curve (AUC) with bootstrap 95% confidence intervals (CIs; 10,000 iterations) quantified accuracy. Results/Data Synthesis: Predictive accuracy for MCI conversion was comparable: NP achieved AUC=.70 (95% CI: .62–.78), CBB achieved .68 (.61–.76), and the difference was not significant (p = .54). For NP, the strongest predictor was RAVLT total learning (odds ratio [OR] = 0.56 per standard deviation), followed by RAVLT delayed recall (OR=0.71). For CBB, Psychomotor/Attention was strongest (OR=0.69), followed by Learning/Working Memory (OR=0.70). Conclusions: Brief, self-administered digital measures predicted conversion to MCI with similar accuracy to traditional NP. As digital cognitive technologies evolve, determining whether they are best suited for screening, monitoring, or standalone assessment across varied clinical contexts warrants further investigation.