Related Experiment Video
Updated: Dec 10, 2025
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Approximating dementia prevalence in population-based surveys of aging worldwide: An unsupervised machine learning
Laurent Cleret de Langavant1,2,3,4,5, Eléonore Bayen5,6, Anne-Catherine Bachoud-Lévi1,2,3,4
1Département d'Etudes Cognitives École normale supérieure PSL University Paris France.
Introduction:
Ability to determine dementia prevalence in low- and middle-income countries (LMIC) remains challenging because of frequent lack of data and large discrepancies in dementia case ascertainment.
Methods:
High likelihood of dementia was determined with hierarchical clustering after principal component analysis applied in 10 population surveys of aging: HRS (USA, 2014), SHARE (Europe and Israel, 2015), MHAS (Mexico, 2015), ELSI (Brazil, 2016), CHARLS (China, 2015), IFLS (Indonesia, 2014-2015), LASI (India, 2016), SAGE-Ghana (2007), SAGE-South Africa (2007), SAGE-Russia (2007-2010). We approximated dementia prevalence using weighting methods.
Results:
Estimated numbers of dementia cases were: China, 40.2 million; India, 18.0 million; Russia, 5.2 million; Europe and Israel, 5.0 million; United States, 4.4 million; Brazil, 2.2 million; Mexico, 1.6 million; Indonesia, 1.3 million; South Africa, 1.0 million; Ghana, 319,000.
Discussion:
Our estimations were similar to prior ones in high-income countries but much higher in LMIC. Extrapolating these results globally, we suggest that almost 130 million people worldwide were living with dementia in 2015.
