Refusal rates in a longitudinal study of older people: implications for field methods.