Validity of dementia and Alzheimer's disease diagnoses in Finnish national registers

BACKGROUND: We investigated dementia and Alzheimer disease (AD) diagnoses in three national registers in Finland: the Hospital Discharge Register (HDR), the Drug Reimbursement Register, and the Causes of Death Register (CDR). METHODS: The Cardiovascular Risk Factors, Aging and Dementia (CAIDE) study was used as the gold standard. Participants were first evaluated in 1972 to 1987, and were reexamined in 1998 and in 2005 to 2008. RESULTS: Two approaches were used for the HDR: with a time restriction (considering "positive" only those cases recorded in the HDR before CAIDE study evaluations) and without a time restriction. Sensitivity of the HDR was 13.7% with time restriction and 51% without time restriction (dementia), and 15.6% with time restriction 55.6% without time restriction (AD). The positive predictive value (PPV) was 87.5% with time restriction and 96.3% without time restriction (dementia), and 100% for AD. Sensitivity and PPV of the HDR were greater after 1998. For AD in the Drug Reimbursement Register alone, sensitivity was 63.5% and PPV was 97.1%; together with the HDR, sensitivity became 65.4% with time restriction and 71.1% without time restriction, and PPV was 100%. For dementia in the CDR, sensitivity was 62.2% and PPV was 100%. CONCLUSIONS: Diagnoses in registers have very good accuracy, but underestimation of dementia/AD occurrence may cause an underestimation of associations with risk/protective factors.