Design of cohort studies in chronic diseases using routinely collected databases when a prescription is used as surrogate outcome