Identification of dementia in epidemiological research: a study on the usefulness of various data sources

Abstract

Background and aims: Prevalence and incidence ratios of dementia in epidemiological studies vary according to the data source used. Medical records, cognitive tests, and registry information are sources frequently used to differentiate dementia from normal aging. The aim of the present study was to compare the identification of dementia from these different sources with that from consensus diagnosis.

Methods: 498 elderly people (age range 70-81 at baseline) enrolled in a Swedish population-based longitudinal twin study (Gender) were evaluated on physical and mental health and interviewed for their socio-demographic background three times during an eight-year period. Reviews of medical records and the Swedish Discharge Registry (DR) were conducted. The 10th percentile was used to differentiate between dementia and non-dementia in all cognitive tests. Scores of 24 or below on the Mini-Mental State Examination (MMSE) (range 1-30) indicated dementia. A consensus conference diagnosed dementia on the basis of total information. The consensus diagnosis was used as the gold standard.

Results: MMSE scores (sensitivity 64%, specificity 96%, kappa 0.65) and the review of medical records (sensitivity 57%, specificity 99%, kappa 0.65) were good sources for dementia identification. The precision of medical records increased when recordings of cognitive impairment were included (sensitivity 83%, specificity 98%, kappa 0.84). The discharge registry had low sensitivity (26%) and kappa coefficient (0.31).

Conclusions: The present study shows that both review of medical records and MMSE scores are good although not perfect identifiers of dementia. The discharge registry is an uncertain source of dementia identification.