Aim: The aim of this project was to develop a dementia screening instrument for use in the hospital or community in populations with low levels of formal education.

Methods: A screening instrument was developed from retrospective data collected in a rural area of Tanzania in 2010. The community screening instrument for dementia was administered to over 95% of the population aged 70 years and older of 6 villages (n = 1198) in Hai district, Tanzania. Factor analysis, regression modeling, and Mokken scale analysis (MSA) were used to develop screening instruments from these data, which were then tested and refined during prospective fieldwork.

Results: A 5-item screening instrument with an area under the receiver–operating characteristic (AUROC) curve of 0.871, sensitivity of 91.7%, and specificity of 61.7% was developed using a combination of factor analysis and logistic regression modeling and had a higher AUROC (0.786) than a 7-item screening instrument developed using MSA. During prospective testing and refinement (n = 60), the 5-item instrument performed well (AUROC 0.867) and took an average of less than 10 minutes to administer. Its performance was improved by including a matchstick design item added to measure praxis, AUROC 0.888.

Conclusions: The 6-item brief dementia screening instrument has acceptable properties and will be further tested and validated during future fieldwork. Although developed for use in sub-Saharan Africa, it may be of use in other world regions where the use of other cognitive screening instruments may result in bias due to low levels of formal education.