Risk assessment for osteoporotic fracture within a primary care context, in old age, has received little attention. We aimed to develop such a risk score and assess its feasibility and validity. This was a 100% population-based, prospective cohort study, with a minimum 5 1/2 year follow-up among women aged 70 years and over, set in a large single general practice in Melton Mowbray, Leicestershire, UK. The main outcome measures were hip fracture, death and migration. Baseline measures included calcaneal broadband ultrasound attenuation (BUA), reported falls, balance, previous fracture history, medical problems, visual acuity, foot problems, body size, lifestyle factors and cognitive impairment. Seventy percent of the sample (1289) participated, including those in residential accommodation. Independent predictors of hip fracture over 3 years were low weight, kyphosis, poor circulation in the foot, epilepsy, short-term use of steroids and poor trunk maneuver. Using the highest tertile, a risk score based on these variables identified 84% (95% CI: 70% to 98%) of the hip fractures with a specificity of 68% (95% CI: 65% to 71%). BUA did not independently predict hip fracture in women of this age group. This study shows that a combination of readily obtained risk factors can identify elderly women who will sustain a hip fracture in the next 3 years more accurately than bone measurements alone in younger women. It also suggests that a risk score approach to universal assessment in the elderly is a feasible proposition in the primary care setting.