Ecological evidence for later dementia diagnosis in areas with higher rates of bilingualism: Evidence fromY Fro Gymraegin Wales

Background Bilingualism has been proposed as a protective factor against dementia, putatively delaying onset by several years. However, evidence is equivocal. One possibility is that the pattern of confounding varies across contexts, with confounders such as socioeconomic status, rurality, and migrant status being positively associated with bilingualism in some settings and negatively in others. The present study examines this in Welsh-speaking Wales, where the confounding structure differs from much of the literature. Methods Routine clinical data from the Memory Clinic service covering the counties of Gwynedd and Ynys Môn were used to examine the ecological correlation between Welsh-speaking rates and dementia diagnosis rates by lower super output area. Analyses were conducted using Poisson generalised linear mixed effects models with spatial auto-correlation terms. Results A total of 404 cases were identified. No overall association was observed between dementia diagnosis rates and Welsh-speaking rates, but the association was moderated by age band, such that Welsh-speaking rates were protective in 65–74 year-olds but a risk factor in 75+ year olds (rate ratio=.71, .53-.94). This was robust to adjustment for area-level poverty. Conclusion Although causation cannot be inferred, the findings are consistent with bilingualism delaying dementia onset. Welsh-English bilingualism provides a useful complement to research on this topic elsewhere in the world, as the confounding structure is unusual.