Objective: to determine the documentation rate of dementia in primary health care, the clinical characteristics of patients with documented and undocumented dementia, and the diagnostic evaluations made in cognitive impairment. Design: cross-sectional population-based study with a retrospective review of medical history. Setting: primary health care in the municipality of Lieto, Southwestern Finland. Subjects: all the inhabitants aged 64 and over in Lieto. Participation rate 82%, numbers = 1260. Measurements: assessment of dementia according to DSM-IV criteria, and severity according to Clinical Dementia Rating. Possible documentation of dementia and evaluations done were reviewed from primary health care medical records. Results: 112 patients with dementia were found. The sensitivity of the general practitioners’ judgment of dementia was 48.2% and the specificity 99.6%. The documentation rate of dementia was 73% in severe, 46% in moderate and 33% in mild dementia. A greater proportion of the patients with undocumented dementia were male (P = 0.003), lived at home (P = 0.003), coped better with the instrumental activities of daily living (P = 0.006), had more depression (P = 0.029) and milder dementia (P = 0.005) than patients with documented dementia. Thyroid stimulating hormone was measured in 51% of the patients with suspected memory impairment or dementia, B12 vitamin in 20%, and serum calcium in 18%. Twenty-eight per cent of the patients had been tested for cognitive function, 68% for depressive symptoms, and 88% for social abilities. Forty-two per cent of patients were referred to a specialist, 32% of patients who were over 75 years. Conclusions: less than half of the patients with dementia had their diagnosis documented in primary care medical records. Documentation increased in more advanced dementia. The diagnostic evaluations for reversible causes of dementia were insufficient in primary care, and they were done at a late phase of cognitive impairment.