AB1255 A review of case-mix and centre effect adjustment in early rheumatoid arthritis cohorts. (12th June 2018)

• Record Type: Journal Article
• Authors: Yates, M.; Bechman, K.; Norton, S.; Galloway, J.
• Is Part Of: Annals of the rheumatic diseases. Volume 77(2018)Supplement 2
• Page Start: 1723
• Page End: 1723
• Publication Date: 2018-06-12
• Subjects: Rheumatism -- Periodicals; 616.723005
• DOI: 10.1136/annrheumdis-2018-eular.1913
• Languages: English
• ISSNs: 0003-4967

Abstract

Abstract : Background: Observational cohort studies have been utilised extensively in early Rheumatoid Arthritis (RA), regularly conducted across multiple centres spanning regional and national boundaries. Case-mix and centre effect are considerations essential for determining comparability of results, and likely prevalence of bias. There is currently no standardised approach for case-mix and centre effect adjustment in early RA observational cohorts. Objectives: Describe the spectrum of methodologies used to address case-mix and centre level effects on outcomes in multi-centre early RA observational cohort studies. Methods: Inclusion criteria were cohorts recruiting from 2 or more centres with 100 or more subjects, with a Rheumatologist diagnosis of RA or EIA within the last 24 months. A systematic electronic search of publications was undertaken. Papers were reviewed by two researchers independently. Reference lists of included papers were reviewed for further relevant publications. A search of all included papers' authors was also conducted. Detail on cohort characteristics, case-mix data collection and adjustment, and consideration of centre-level effect in analyses were collected. Results: 1047 papers were identified from the initial search. A total of 20 unique cohorts were identified. Reference review and author search produced 14 more, to make a total of 34 unique observational cohorts drawn from 205 papers. The cohorts were mainly conducted in Europe (24/34, 71%), With 2 (6%) from less economically developed regions. The period of data collection was between 1955 and 2017. Case-mix: All cohorts considered case-mix in some form (e.g. age and gender), but with heterogenous approaches. The figure displays the relative frequencies of sociodemographic variable consideration across all included papers. Centre effect: 18/205 (9%) of the included papers accounted for centre in their results, utilising a range methodologies. Where reported, centre had a significant impact. Conclusions: The degree of case-mix reporting varied widely, and few studies addressed centre effect. Where analysed, a centre level impact was clearly apparent. A failure to incorporate centre into analyses can lead to unrecognised bias as a result of confounding by centre. It must be acknowledged that including case-mix variables and adjusting for centre substantially reduces power, and it is likely that many of the reported observations may have lost statistical significance had case-mix and centre effect been addressed more completely. This is the first systematic review of centre effect and case-mix in early RA, and highlights a challenging field deserving further research. Disclosure of Interest: None declared