A method of defining blood pressure (BP) status from a review of primary care patient records was developed and then validated using the case notes of a general practitioner with an interest in hypertension. Data were drawn from the records of the previous 6 years of all 65 to 80-year-old patients in the practice (n = 263). Patients were then categorised as hypertensive, normotensive or 'undetermined' by using a flowchart based on the mean of the three most recent BP measurements, antihypertensive medication and comorbidities of ischaemic heart disease, myocardial infarction, angina, oedema or cardiac failure. Mean systolic BP of > or = 160 mm Hg and/or diastolic BP of > or = 90 mm Hg were used as a threshold definition of hypertension and of BP control. Disagreement between general practitioner and the notes based definition occurred in 11% of patients (5% hypertensive, 6% normotensive). Reasons for disagreement were: controlled hypertensives with comorbidities such as angina or heart failure (4%), isolated elevated readings (3%), use of antihypertensive medication for separate indications (2%), other reasons (2%). The resulting sensitivity and specificity was 86% and 88% respectively. Including the recording of a diagnosis of hypertension in the definition increased the sensitivity to 98% with specificity unchanged at 88%. Actual sensitivity of the instrument when used in other practices is likely to lie between 88% and 98% depending on the quality of the doctor's recording of the diagnosis of hypertension. These findings suggest that data from primary care case notes can provide a ready and valid means of defining cases of hypertension for studying the management of hypertension in primary care and for research purposes.