Assessing the 10/66 dementia classification algorithm for international comparative analyses with the United States

Area of Science:

• Neuroscience
• Epidemiology
• Gerontology

Background:

• Cross-national dementia prevalence studies are vital but hindered by varied classification methods.
• The 10/66 algorithm is a validated dementia classification tool for low- and middle-income countries (LMICs).

Purpose of the Study:

• To adapt and validate the 10/66 algorithm for dementia classification within the U.S. Aging, Demographics, and Memory Study (ADAMS).
• To assess the modified 10/66 algorithm's performance against existing U.S. dementia classification methods.

Main Methods:

• Mapped comparable 10/66 algorithm items to ADAMS data.
• Re-trained the 10/66 algorithm using ADAMS clinical diagnoses and k-fold cross-validation.
• Compared the modified 10/66 algorithm's accuracy and education gradient estimation against four other ADAMS-validated algorithms.

Main Results:

• The modified 10/66 algorithm demonstrated high sensitivity (87%) and specificity (93%) in ADAMS.
• It outperformed the four comparison algorithms in dementia classification accuracy.
• All algorithms, including the modified 10/66, tended to over-estimate the education gradient in dementia.

Conclusions:

• The modified 10/66 algorithm is a reliable tool for dementia classification in the U.S.
• This validation facilitates more accurate cross-national comparisons of dementia risk factors between the U.S. and LMICs.