Prediction bands and intervals for the scapulo-humeral coordination based on the Bootstrap and two Gaussian methods

• Published in: Journal of biomechanics 2014-03, Vol.47 (5), p.1035-1044
• Main Authors: Cutti, A.G, Parel, I, Raggi, M, Petracci, E, Pellegrini, A, Accardo, A.P, Sacchetti, R, Porcellini, G
• Format: Article
• Language: English
• Subjects: Adolescent; Adult; Age; Aged; Arm - physiology; Bands; Bootstrap; Classification; Computation; Confidence intervals; Female; Gaussian; Humans; Humerus; Humerus - physiology; Intervals; Kinematics; Male; Methods; Middle Aged; Motion analysis; Normal Distribution; Optimization; Patients; Physical Medicine and Rehabilitation; Physiology; Prediction; Reference Values; Rotation; Scapula; Scapula - physiology; Scapular kinematics; Shoulder; Shoulder - physiology; Shoulder Joint - physiology; Young Adult
• ISSN: 0021-9290; 1873-2380
• DOI: 10.1016/j.jbiomech.2013.12.028

Abstract Quantitative motion analysis protocols have been developed to assess the coordination between scapula and humerus. However, the application of these protocols to test whether a subject's scapula resting position or pattern of coordination is “normal”, is precluded by the unavailability of reference prediction intervals and bands, respectively. The aim of this study was to present such references for the “ISEO” protocol, by using the non-parametric Bootstrap approach and two parametric Gaussian methods (based on Student's T and Normal distributions). One hundred and eleven asymptomatic subjects were divided into three groups based on their age (18–30, 31–50, and 51–70). For each group, “monolateral” prediction bands and intervals were computed for the scapulo-humeral patterns and the scapula resting orientation, respectively. A fourth group included the 36 subjects (42±13 year-old) for whom the scapulo-humeral coordination was measured bilaterally, and “differential” prediction bands and intervals were computed, which describe right-to-left side differences. Bootstrap and Gaussian methods were compared using cross-validation analyses, by evaluating the coverage probability in comparison to a 90% target. Results showed a mean coverage for Bootstrap from 86% to 90%, compared to 67–70% for parametric bands and 87–88% for parametric intervals. Bootstrap prediction bands showed a distinctive change in amplitude and mean pattern related to age, with an increase toward scapula retraction, lateral rotation and posterior tilt. In conclusion, Bootstrap ensures an optimal coverage and should be preferred over parametric methods. Moreover, the stratification of “monolateral” prediction bands and intervals by age appears relevant for the correct classification of patients.