Differing Methodologies Are Required to Estimate Prevalence of Dementia: Single Study Types Are No Longer Reliable
Leon Flicker, Kaarin J. Anstey, Osvaldo P. Almeida, Michael Waller, Patrick Fitzgerald, Fleur de Crespigny, Thao Vu, Annette J. Dobson
Abstract
Population-based surveys were used to estimate community prevalence of dementia, but have low response fractions due, among other things, to difficulties in obtaining informed consent from people with diminished capacity. Cohort studies of younger people are subject to recruitment bias and non-random drop-outs. Dementia registries can delineate sub-types of dementia but have limited population coverage and are costly to maintain. Administrative datasets have low costs but may be subject to selection bias and uncertain sensitivity. We propose that astute combination of methodologies, including assessment of coverage and validity of administrative datasets, is the most cost-effective process to estimate and monitor community prevalence.