FRI0587-HPR Development of the Instrumented Bath Ankylosing Spondylitis Functional Index (IBASFI) in Axial Spondyloarthritis: Reliability and Validity of Automated Accelerometry-Based Algorithms. (10th June 2014)

• Record Type: Journal Article
• Authors: Swinnen, T.; Milosevic, M.M.; Van Huffel, S.; Dankaerts, W.; Westhovens, R.; de Vlam, K.
• Is Part Of: Annals of the rheumatic diseases. Volume 73:Supplement 2(2014)
• Page Start: 1207
• Page End: 1207
• Publication Date: 2014-06-10
• Subjects: Rheumatism -- Periodicals; 616.723005
• DOI: 10.1136/annrheumdis-2014-eular.6021
• Languages: English
• ISSNs: 0003-4967

Abstract

Abstract : Background: Performance-based in addition to self-reported assessment of physical function has become increasingly popular in the field of axial spondyloarthritis 1 . Unfortunately, the observer capturing movement time with a chronometer requires extensive training in standardized assessment. A body-worn sensor may enable automated detection of movement time during performance-based tests. Objectives: To establish the within-session test-retest reliability of novel automated movement time detection algorithms using a body-worn accelerometer. To compare automated versus manual segmentation of acceleration signals. To compare movement time captured with an accelerometer versus a chronometer (i.e. concurrent validity). Methods: Twenty-eight consecutively evaluated patients with axial spondyloarthritis (Age: 43.69 (10.45); BASMI: 3.11 (1.60); BASFI: 3.41 (2.19); Sex: 16M, 12F) completed ten performance-based tests derived from the BASFI questionnaire. All patients wore a two-axial accelerometer fixed on the upper arm or the trunk. An observer captured movement time with a chronometer. All tests were repeated twice and all clinical data were collected up front. Movement time was extracted from filtered accelerometer signals using custom-written automated algorithms in MatLab and by calculating the mean of manual signal segmentations by two blinded evaluators. Trials or methods were compared with the intraclass correlation coefficient and the standard error of measurement (SEM) clinically expressed in % of total movement time. Results: Good to excellent test-retest reliability of automated movement time was found with the lowest ICC values for sock test (ICC 0.765, 0.482-0.893), maximal reach test (ICC 0.528, 0.185-0.753), shoulder speed test (ICC 0.785, 0.441-0.911) and sit-to-stand test (ICC 0.753, 0.537-0.877), but very high ICC values for pen test (ICC 0.812, 0.637-0.908), 5 pens speed test (ICC 0.974, 0.887-0.991), sit-to-stand speed test (ICC 0.959, 0.803-0.986), lying down (ICC 0.979, 0.954-0.991), getting up (ICC 0.993, 0.985-0.997) and stair climbing (ICC 0.863, 0.714-0.935). Overall, single/self-selected pace movements were less reproducible than repeated/fast pace movements. SEM values ranged from 13 to 46%, excluding the pen (46%) and maximal reach test (64%) . Automated segmentation of acceleration excellently mimicked the mean value of manual segmentation (ICC range 0.900-0.998), except for the maximal reach test (ICC 0.727, 0.063-0.906) and sit-to-stand test (ICC 0.599, 0.189-0.881). In comparison to to the chronometer, automated movement time was validly assessed (ICC range 0.878-0.998), apart from the maximal reach (ICC 0.532, 0.133-0.768) and sit-to-stand test (ICC 0.770, 0.565-0.886). SEM values ranged from 8 to 44%, ignoring the maximal reach test (66%). Conclusions: We developed automated movement time detection algorithms that are psychometrically sound and viable for clinical application. Automated analysis for repeated and fast-paced performance tests shows excellent test-retest reliability and criterion-based validity and can replace the chronometer. The maximal reach test is nor reproducible nor valid. References: van Weely SF, et al. Rheumatology (Oxford). 2009 Oct;48(10):1254-60. Disclosure of Interest: None declared DOI: 10.1136/annrheumdis-2014-eular.6021