A critique of revision rate as an outcome measure: re-interpretation of knee joint registry data.