Methodological and statistical approaches to criteria development in rheumatic diseases.

We have discussed methodological and statistical considerations in developing disease classification and assessment criteria. In choosing cases with disease and nondisease controls, classification criteria should be developed with an eye toward face, content and construct validity, and toward their ultimate applicability. The validity of disease criteria should be tested in a patient sample different from the one used to develop the criteria. Several analytic approaches are available to reduce the candidate diagnostic elements to those that will define the presence of disease. Methodological issues in the assessment of disease severity, activity or damage are similar to those faced in criteria development studies, although the analytic concerns are different and the options more varied.