Meta-analysis identifies Back Pain Questionnaire reliability influenced more by instrument than study design or population

Objective
To assess if predefined variables in study design, instrument type, and patient characteristics account for variance in reported retest reliability for the Oswestry Disability Index (ODI) and Roland–Morris Questionnaire (RMQ). A second aim was a more precise estimate of instrument reliability.

Study Design and Setting
A systematic literature search of MEDLINE, EMBASE, AMED, PsycINFO, and CINAHL was performed from inception to January 2011. Hand search, gray literature, and reference retrieval completed the search. Two blinded reviewers extracted the data. Original authors were contacted for the missing data. A meta-analysis was performed with the intraclass correlation coefficient as the outcome measure.

Results
Fifty studies on 31 ODI and 28 RMQ cohorts were retrieved that met the inclusion criteria. Meta-analysis found the ODI more reliable than the RMQ (coefficient, -0.2840; P = 0.009) and lower reliability on increased days to retest (coefficient, -0.0089; P = 0.005) and in low back pain (LBP) with leg pain than LBP only cohorts (coefficient, -0.2194; P = 0.046). The use of a transition item to identify stable patients and percentage of cohort not included at retest were significant only on single variable analysis.

Conclusion
Study design and population influence the reliability of a given instrument, however, a greater difference in reliability exists between instruments.