Premium
Development and Testing of New Candidate Psoriatic Arthritis Screening Questionnaires Combining Optimal Questions From Existing Tools
Author(s) -
Coates Laura C.,
Walsh Jessica,
Haroon Muhammad,
FitzGerald Oliver,
Aslam Tariq,
Al Balushi Farida,
Burden A. D.,
BurdenTeh Esther,
Caperon Anna R.,
Cerio Rino,
Chattopadhyay Chandrabhusan,
Chinoy Hector,
Goodfield Mark J. D.,
Kay Lesley,
Kelly Stephen,
Kirkham Bruce W.,
Lovell Christopher R.,
MarzoOrtega Helena,
McHugh Neil,
Murphy Ruth,
Reynolds Nick J.,
Smith Catherine H.,
Stewart Elizabeth J. C.,
Warren Richard B.,
Waxman Robin,
Wilson Hilary E.,
Helliwell Philip S.
Publication year - 2014
Publication title -
arthritis care and research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.032
H-Index - 163
eISSN - 2151-4658
pISSN - 2151-464X
DOI - 10.1002/acr.22284
Subject(s) - medicine , discriminative model , receiver operating characteristic , youden's j statistic , psoriatic arthritis , logistic regression , weighting , artificial intelligence , arthritis , computer science , radiology
Objective Several questionnaires have been developed to screen for psoriatic arthritis (PsA), but head‐to‐head studies have found limitations. This study aimed to develop new questionnaires encompassing the most discriminative questions from existing instruments. Methods Data from the CONTEST study, a head‐to‐head comparison of 3 existing questionnaires, were used to identify items with a Youden index score of ≥0.1. These were combined using 4 approaches: CONTEST (simple additions of questions), CONTESTw (weighting using logistic regression), CONTESTjt (addition of a joint manikin), and CONTESTtree (additional questions identified by classification and regression tree [CART] analysis). These candidate questionnaires were tested in independent data sets. Results Twelve individual questions with a Youden index score of ≥0.1 were identified, but 4 of these were excluded due to duplication and redundancy. Weighting for 2 of these questions was included in CONTESTw. Receiver operating characteristic (ROC) curve analysis showed that involvement in 6 joint areas on the manikin was predictive of PsA for inclusion in CONTESTjt. CART analysis identified a further 5 questions for inclusion in CONTESTtree. CONTESTtree was not significant on ROC curve analysis and discarded. The other 3 questionnaires were significant in all data sets, although CONTESTw was slightly inferior to the others in the validation data sets. Potential cut points for referral were also discussed. Conclusion Of 4 candidate questionnaires combining existing discriminatory items to identify PsA in people with psoriasis, 3 were found to be significant on ROC curve analysis. Testing in independent data sets identified 2 questionnaires (CONTEST and CONTESTjt) that should be pursued for further prospective testing.