Arthritis & Rheumatism, Volume 62,
November 2010 Abstract Supplement
Abstracts of the American College of
Rheumatology/Association of Rheumatology Health Professionals
Annual Scientific Meeting
Atlanta, Georgia November 6-11, 2010.
Application of the Generalizability Theory for Determining Inter-Rater and Inter-Trial Reliability of Spine Curvature Measures in Postmenopausal Women with Osteoporosis of the Spine Using the Flexicurve Ruler and the Digital Inclinometer.
MacIntyre1, Norma J., Bennett2, Lisa, Bonnyman2, Alison, Stratford2, Paul W.
Spine curvatures are often measured in physical therapy practice using a flexicurve ruler to calculate kyphotic index (KI) and lordotic index (LI) or a digital inclinometer to quantify the angle at the cervicothoracic (CT), thoracolumbar (TL) and lumbosacral (LS) junctions. Generalizability theory (G-Theory) provides tools that characterize the sources of variation, or facets, in a measurement procedure and permit identification of a protocol that provides optimal reliability. The purpose of this study was to use the tools of G-Theory to investigate the inter-rater and inter-trial reliability of these spine curvature measures and to establish an optimal measurement protocol.
Nine postmenopausal women over the age of 60 years with established osteoporosis of the spine were recruited for this cross-sectional observational study. Two raters completed triplicate measures of spine curvatures using both the flexicurve ruler and the digital inclinometer according to a standardized protocol. G-Theory was applied through a Generalizability Study to estimate G-coefficients (analogous to the reliability coefficient in the Classical Test Theory). The facets included were participants (P), raters (R) and trials (T). Three-way ANOVAs were used to generate variance components for the following terms: P, R, T, PxR, PxT, RxT and error. G-coefficients were calculated as the proportion of the overall variance that can be attributed to the facets and interactions and the proportions of variance attributed to rater and trial. Follow-up Decision Studies were performed to identify a measurement protocol that minimized error and optimized reliability. G-coefficients >= 0.8 were considered desirable and 0.70 to 0.79 were considered acceptable.
Inter-rater reliability is excellent for KI, LI, TL and LS (G-coefficient = 0.97, 0.91, 0.85 and 0.80, respectively) and acceptable for CT (G-coefficient = 0.74) using our protocol for measuring spine curvatures. Three raters are required to achieve excellent inter-rater reliability for CT (0.81). A single trial would result in excellent reliability for all measures (G-coefficient >= 0.94 for all except LI where G-coefficient = 0.84). Triplicate trials improve inter-trial reliability (G-coefficient >= 0.98 for all except LI where G-coefficient = 0.94). Little gains are achieved by increasing trials beyond three; eight trials are needed to raise the inter-trial reliability of LI measures to 0.98.
This study demonstrates the application of G-Theory to identify a protocol for measuring spine curvatures in osteoporotic women which optimizes reliability by testing the gains achieved by increasing the number of raters or the number of trials. If more than one rater is involved in assessing spine curvatures, triplicate measures acquired using the flexicurve ruler will be most reliable. For a given rater, inter-trial reliability is optimized by measuring in triplicate whether using the flexicurve ruler or the digital inclinometer.
To cite this abstract, please use the following information:
MacIntyre, Norma J., Bennett, Lisa, Bonnyman, Alison, Stratford, Paul W.; Application of the Generalizability Theory for Determining Inter-Rater and Inter-Trial Reliability of Spine Curvature Measures in Postmenopausal Women with Osteoporosis of the Spine Using the Flexicurve Ruler and the Digital Inclinometer. [abstract]. Arthritis Rheum 2010;62 Suppl 10 :1316