File
Gómez-Conesa A.1, Carles-Hernandez R.1, Meseguer-Henarejos A.B.1, López-Pina J.A.2
1University of Murcia, Physical Therapy, Murcia, Spain, 2University of Murcia, Basic Psychology and Methodology, Murcia, Spain
Background: The Berg Balance Scale (BBS) is the instrument most used to assess the balance in different environments and populations, and adaptations to different languages have been developed. The BBS is considered the gold standard for evaluating balance.
Purpose: The aim of this study was perform to a metaanalytic reliability generalization (RG) study de la BBS, to check whether reliability estimates can be generalized in the studies where BBS has been applied and, if variability in the estimates of this parameter is found, to examine the variables that explain its heterogeneity.
Methods: A metaanalytic RG study of the BBS has been made. To locate the studies, the following electronic databases were consulted: PubMed, Scopus, PEDro y Cochrane Library Plus.
To be included, studies were required to meet five criteria:
a) studies which had applied one of version to BBS to a sample of participants,
b) they had to report a reliability coefficient computed from their own sample data, either inter-rater reliability, intra-rater reliability, or coefficients alpha,
c) they had to be published from 1989 to 2014,
d) they had to be written in English, French or Spanish, and
e) they must be clinical trial, comparative study, controlled clinical trial, evaluation studies, multicenter study, randomized controlled trial, validation studies, meta-analysis, review, or systematic reviews.
Of the 597 studies BBS applied to a sample of subjects, 45 studies reported some estimates of reliability were selected.
Results: The reliability coefficient was a source of heterogeneity. As the studies describe one or more coefficients of internal consistency (alpha coefficient) and inter- or intra-rater reliability, based on different suppositions, three meta-analyses of the RG were made one for each measurement of reliability. To test the influence of different characteristics of the studies (as the age or disease of the subjects) in the variability of the different reliability coefficients obtained with the BBS, we assume a fixed effect model as the number of articles included in each study of reliability generalization is relatively low: 19 for internal consistency, in which an average alpha coefficient of .897 was obtained; 29 for inter- rater reliability, with an average CCIb of .939; and 30 for intra- rater reliability, with an average CCIw of .937.
Conclusion(s): The reliability of BBS scores cannot be generalized to all its applications, since the different characteristics of the studies may influence estimations of reliability.
The results are only generalizable to other studies designed in such conditions.
Implications: The generalizability of our results will be only for those studies that have characteristics similar to those used in the present study, future research needs to evaluate on the psychometric properties of BBS for each particular sample.
Funding acknowledgements: We have no funding research. No acknowledgement is required.
Topic: Outcome measurement
Ethics approval: It is not required.
All authors, affiliations and abstracts have been published as submitted.
