The use of standardised patients to assess clinical competence: does practice make perfect?