Development of a brief assessment and algorithm for ascertaining dementia in low-income and middle-income countries: the 10/66 short dementia diagnostic schedule.

摘要

Objectives: To develop and evaluate a shorter form of the current 10/66 dementia diagnostic algorithm for use in low- and middle-income countries, replacing the Geriatric Mental State (GMS) interview with the Euro-D depression screening scale. Design: Split-half analysis for algorithm development and testing; cross-evaluation of short-form and standard algorithms in data from a series of community surveys. Settings: 1) The 10/66 pilot sample dataset of people aged 60 years and over in 25 international centres each recruiting the following samples: a) people with dementia; b) people with depression but no dementia; c) people without dementia, with high education; d) people without dementia, with low education. 2) 10/66 cross-sectional surveys of people aged 65 years and over from 12 urban and rural sites in 8 countries (Cuba, Dominican Republic, Peru, Mexico, Venezuela, India, China and Puerto Rico). Participants: The short-form diagnostic algorithm was developed in 1218 randomly selected participants in the 10/66 pilot samples, and were tested in a separate 1211 randomly selected participants from the same samples with sufficient data. The short-form algorithm was evaluated against the standard algorithm in 16,536 survey participants. Outcome measures: The diagnostic algorithm was derived from output from the Community Screening Instrument for Dementia, the CERAD 10-word list-recall task, and the Euro-D depression screening scale. The algorithm was evaluated against clinically assigned groups in the pilot data and against the standard 10/66 algorithm in the survey data. Results: In the test half of the pilot sample, the short-form algorithm ascertained dementia with 94.2% sensitivity. Specificities were 80.2% in depression, 96.6% in high education and 92.7% in low education groups. In survey samples it coincided with standard algorithm dementia classifications with over 95% accuracy in most sites. Estimated dementia prevalences in the survey samples were not consistently higher or lower using the short-form compared to standard algorithm. Conclusions: For epidemiological studies of dementia in low- and middle-income settings where the GMS interview (and/or interviewer training for this) is not feasible, the short-form 10/66 algorithm provides an alternative with acceptable levels of performance.