Does observer bias contribute to variations in the rate of retinopathy of prematurity between centres?