Underestimation of the reliability of codes for rheumatoid arthritis within administrative data: comment on the article by Ng et al.