BACKGROUND: We aimed to compare the judgmental and mathematical approaches in weighting the Pediatric Ulcerative Colitis Activity Index (PUCAI). METHODS: The PUCAI was previously weighted mathematically using multivariate regression modeling on 157 children with ulcerative colitis (UC). Independently, a Delphi group of 36 experts in pediatric UC judgmentally provided weights to the PUCAI's items. The agreement between the tools was evaluated using the 95% limits of agreement method. Validity was assessed on a prospective cohort of 48 UC children, using three constructs: colonoscopic appearance, physician's global assessment, and the Mayo score. Responsiveness was compared on a longitudinal cohort of 75 children. RESULTS: The weights of the resulting PUCAI tools were quite similar, but the Delphi group retained the laboratory items, excluded by the mathematical modeling. This difference was reflected by the Bland and Altman method. Both tools performed equally well in the validation and responsiveness cohorts. CONCLUSION: The judgmentally weighted PUCAI had good validity and responsiveness. The mathematical weighting, however, performed just as well without the laboratory items, resulting in a more feasible index. Therefore, the mathematical modeling has proven to be superior in weighting the PUCAI.
BACKGROUND: We aimed to compare the judgmental and mathematical approaches in weighting the Pediatric Ulcerative Colitis Activity Index (PUCAI). METHODS: The PUCAI was previously weighted mathematically using multivariate regression modeling on 157 children with ulcerative colitis (UC). Independently, a Delphi group of 36 experts in pediatric UC judgmentally provided weights to the PUCAI's items. The agreement between the tools was evaluated using the 95% limits of agreement method. Validity was assessed on a prospective cohort of 48 UC children, using three constructs: colonoscopic appearance, physician's global assessment, and the Mayo score. Responsiveness was compared on a longitudinal cohort of 75 children. RESULTS: The weights of the resulting PUCAI tools were quite similar, but the Delphi group retained the laboratory items, excluded by the mathematical modeling. This difference was reflected by the Bland and Altman method. Both tools performed equally well in the validation and responsiveness cohorts. CONCLUSION: The judgmentally weighted PUCAI had good validity and responsiveness. The mathematical weighting, however, performed just as well without the laboratory items, resulting in a more feasible index. Therefore, the mathematical modeling has proven to be superior in weighting the PUCAI.
Authors: Boyko Kabakchiev; Dan Turner; Jeffrey Hyams; David Mack; Neal Leleiko; Wallace Crandall; James Markowitz; Anthony R Otley; Wei Xu; Pingzhao Hu; Anne M Griffiths; Mark S Silverberg Journal: PLoS One Date: 2010-09-30 Impact factor: 3.240