Measuring clinical status in cystic fibrosis: internal validity and reliability of a modified NIH score.

Abstract

We examined measurement properties of the NIH Clinical Score for Cystic Fibrosis (CF) as an index of disease status. This score is being employed as a research tool for defining study populations and as an outcome measure, yet there are no published data on its reliability or how its items contribute to the overall measure of disease status. Criteria for scoring some items in the original index lack specificity. In this study, we used a modified score to have more clearly specified criteria, while retaining the original weightings and structure. For 200 patients with CF in two centers, we analyzed the total NIH Score and its subscores for internal consistency, interrater reliability, and factor analysis. Internal consistency indicates how inter-related the items are. The pulmonary subscore and overall score had fairly high internal consistency. However, the general subscore had low internal consistency, suggesting that the items are not measuring a single element of disease status and should not be added. Factor analysis provides additional information on the underlying structure and relationships among items. Five factors (groups of items) were identified accounting for 85% of the consistent variance of 14 items. These factors were designated by items accounting for most of their variance: general pulmonary, weight, disability, psychosocial, and acute infiltrate. While inter-rater reliability for the overall index was high, individual items showed less agreement. The results indicate that most of the variability in the NIH Score is attributable to pulmonary items in the first factor. The analyses suggest a new scoring structure for the NIH Score; the general subscore items do not contribute to the reliability or account for significant variance. Therefore, they will likely require further refinement or be eliminated.