Public Health

Background: Cross-national comparisons of dementia prevalence are essential for identifying unique determinants and cultural-specific risk factors, but methodological differences in dementia ascertainment across countries hinder global comparisons. This study maps the 10/66 Dementia Research Group algorithm for dementia ascertainment, widely used and validated in lower- and middle-income countries, to the U.S.-based Aging, Demographics, and Memory Study (ADAMS), and validates it for use in the U.S.

Methods: We conducted pre-statistical harmonization across the ADAMS and the 10/66 studies in Latin America to identify common measures, which were then utilized to develop a modified 10/66 algorithm. We computed dementia probabilities using the modified 10/66 algorithm and the optimal threshold for dementia ascertainment was established by maximizing sensitivity and specificity against ADAMS' clinical gold standard diagnosis. We then compared this algorithm to a TICS-based algorithm developed in ADAMS and commonly used in the U.S. Health and Retirement Survey. Further analysis included an examination of the age-adjusted educational gradients in predicted dementia prevalence, allowing for a comprehensive assessment of the modified algorithm's validity and its comparative performance.

Results: The sample comprised 5,649 adults ages 70 and over (539 from ADAMS and 5,109 from 10/66 studies).10/66 participants had a lower education level (60% had primary school or less) relative to the ADAMS cohort (69.8% possessed at least a high school education). The modified 10/66 diagnostic algorithm performed similarly in the 10/66 data as compared to the original 10/66 algorithm. In the ADAMS data, the modified 10/66 algorithm outperformed the TICS algorithm across several metrics: it demonstrated higher sensitivity (90% vs 77%), specificity (90% vs 74%), accuracy (90% vs 75%), and Area Under the Curve (AUC) (90% vs 76%). The modest ADAMS sample size precluded statistically precise comparisons by education, but education gradients in predicted dementia were broadly similar across algorithms.

Conclusions: The modified 10/66 algorithm effectively classifies dementia in US studies and is suitable for cross-national comparisons, supporting its use for international studies. This facilitates future harmonization efforts of dementia diagnostic tools and enhances cross-country dementia research.