Objective To study stroke incidence among women over 32 years of age with a focus on subdividing by stroke type, to consolidate end points and associations with risk factors.

Design Prospective population study initiated in 1968–1969 with follow-ups in 1974, 1980, 1992 and 2001.

Setting Gothenburg, Sweden.

Participants A sample of 1462 women from five age strata examined in 1968–1969, representative of women in the general population.

Main outcome measures Main types of first-ever stroke and fatal stroke during 1968–2001 identified and validated. Stroke incidence rates in different age strata. Association with baseline smoking, body mass index (BMI), waist–hip ratio, hypertension, serum lipids, physical inactivity, perceived mental stress and education. Associations with atrial fibrillation (AF), diabetes, baseline hypertension and myocardial infarction (MI). Blood pressure (BP) levels 1–3, corresponding to modern guidelines, in relation to stroke risk.

Results 184 (12.6%) cases of first-ever stroke, 33 (18%) of them fatal. Validation reduced unspecified stroke diagnoses from 37% to 11%. Age-standardised incidence rate per 100 000 person-years was 448. A multivariate model showed a significant association between ischaemic stroke and high BMI: HR 1.07 (95% CI 1.02 to 1.12), smoking 1.78 (1.23 to 2.57) and low education 1.17 (1.01 to 1.35). Significant association was seen between haemorrhagic stroke and, besides age, physical inactivity 2.18 (1.04 to 4.58) and for total stroke also hypertension 1.45 (1.02 to 2.08). Survival analysis showed a significantly increased risk of stroke in participants with diabetes (p<0.001), AF (p<0.001) and hypertension (p=0.001), but not MI. Stroke risk increased with increasing BP levels but was already seen for diastolic pressure grade 1 and particularly when combined with systolic BP grade 1; 1.62 (1.17 to 2.25).

Conclusions Hypertension, smoking, AF, diabetes and high BMI were associated with increased stroke risk. Low education was associated with stroke. Validation of National Patient Registry diagnoses to increase specified diagnoses improved data quality.