Interrater and test-retest reliability and validity of the Norwegian version of the BESTest and mini-BESTest in people with increased risk of falling

Translation

The BESTest and the Mini-BESTest were translated into Norwegian following international guidelines [16]. Both tests were translated into Norwegian by three experienced physiotherapists, fluent in both English and Norwegian. These three versions were compared and discussed until agreement was reached. In addition, a senior researcher commented on the translation. A professional translator then back-translated both tests to English. Throughout the translation process we had communication with the original author (Fay Horak) who gave us permission to conduct the translation and who also approved the back-translated version. In the Norwegian versions, meters and centimetres rounded up to the closest centimetre are used instead of inches and feet.

Design and subjects

This was an observational study with a test-retest design. Three groups of participants, elderly over 65 years of age, people with a history of stroke or Multiple Sclerosis (MS), representing different clinical conditions and fall-risk profiles were recruited to ensure heterogeneity of balance impairments. In order to be included the participants had to be able to walk 6 m without a walking aid, and to be able to meet for testing on two occasions with a two-day interval. Exclusion criteria were to be unable to understand or follow oral instructions. Eligible people were asked to participate by their treating physiotherapist from three inclusion sites: The Geriatric rehabilitation ward at Oslo University Hospital (OUS), the Department of physiotherapy at Oslo and Akershus University College of Applied Sciences (HiOA) and from the Multiple Sclerosis Centre Hakadal in the period of 01.09.11–01.06.12. Forty-two participants were included in the study.

Qualification of raters

Two raters with extensive experience conducted the assessments, rater A had 20 years of experience both as a clinical physiotherapist and as a teacher in physiotherapy education at the university, rater B had 16 years of experience as a clinical physiotherapist.

Both raters had attended a 3-day BESTest workshop led by the developer of the tests and also watched the training videos available at the BESTests web portal [19]. Before the study, the raters had three training sessions where they were allowed to discuss how to score the test items with each other. The raters were not allowed to discuss the scoring during the study period.

Procedures

The test sessions took place at the three inclusion sites, and the same test equipment was used at all three sites. The participants were tested with two-day interval; both test sessions were conducted in the same room and at the same time of the day. Instructions to the participants were to live their normal life, and take their medication according to their normal regime during the test period. Before the second test session, the participants were asked about any relevant changes in self-perceived balance state since the first test session. The participants performed the BESTest and the Mini-BESTest barefoot except for the tasks in section VI where they were allowed to wear flat-heeled shoes. All participants used the same shoes at both sessions.

Both the BESTest and the mini-BESTest were scored at both test sessions. Rater B administered all the tests at both sessions, while both raters scored the participants performance from the same test trials. All items in the mini-BESTest are included in the BESTest, so each task (item) was only performed once and scored according to the test criteria. The participants were allowed to rest as needed during the test sessions.

Demographic information (age, weight, height, diseases, number of medications and number of falls during the last year) was obtained by interviewing the subjects before the first test session. At the end of the first session, the FES-I was administered by rater B as a structured interview. Total time for first session was approximately 60 min. The second session was without the interviewing, and took approximately 40 min to administrate.

Assessment tools

Data analysis

All statistical analyses were performed using the IBM SPSS Statistics, version 22.0 (IBM Corp., Armonk, New York). For calculation of relative and absolute reliability the criteria for evaluation of measurements developed by the prevention of Falls Network Europa [21] and The COSMIN checklist was followed [22].

The sample size calculations were based on the formula n = 2×(SD/Δ)2xk [23], where the expected standard deviation (SD) was based on Horak’s study with SD = ±9.6% [2]. At the time, no study had established clinical relevance change for the BESTest; we therefore chose the least clinically difference, which is the difference in score which patients perceive as important, to be 7.0% based on other clinical outcome measures [24]. We used α = 0.05 with a power of 0.80. This gave a sample size of minimum 30 participants [23].

Intraclass Correlation Coefficient (ICC) with 95% confidence interval was used as measures of relative reliability [25, 26]. For the interrater reliability the ICC(2,1) and the ICC(3,1) were used. ICC(2,1) is based on a two-ways random absolute agreement, shows variability between raters, and the results can be generalized to other raters. ICC(3,1), using two-ways mixed, consistency, is a measure of the consistency of the scoring of each rater. Systematic errors are not included as measurement error. When ICC(2,1) and ICC(3,1) are identical or shows only minor differences, there is no systematic error e.g. learning effect, present [25].

For the test-retest reliability ICC(1,1) using a one-way random model, and ICC(3,1) was used. With ICC(1,1) all systematic and random intrasubject variability is seen as measurements error. Again, if ICC(1,1) and ICC(3,1) is identical or shows only minor differences, there is no systematic error present [25].

Measurement error is the systematic and random error of a subject’s score that is not attributed to the true changes in the construct to be measured [22]. For assessment of absolute reliability Standard Error of Measurement (SEM) and Smallest Detectable Change (SDC₉₅) were used. SEM represents the standard deviation of repeated measures in one participant (SEM = SD/√2). SDC₉₅ represents the smallest change that a participant must show to ensure that the observed change is real and not just a measurement error (SDC₉₅ = SEMx√2 × 1.96) [22]. Since the FES-I demonstrated a skewed distribution we used the Spearman Rank Correlation to examine concurrent validity between the FES-I and rater A’s total scores of the BESTest and the Mini-BESTest. Correlation coefficients of 0.00–0.25 were interpreted as little to no correlation, 0.25–0.49 as fair, 0.50–0.75 as a moderate to good, and above 0.75 as a strong correlation [18].

The presence of floor and ceiling effects was defined as 15% or more of the participants having the lowest or the respectively the highest possible score on the BESTest and the Mini-BESTest [27].

Descriptive statistics

Reliability

Validity

There were moderate correlations for the BESTest and the Mini-BESTest against the FES-I; r _{s =}-0.51 (p = 0.01) and r _{s =}-0.50 (p = 0.01), respectively.

Reliability

The absolute reliability, presented by SEM and SDC in actual scale units, is probably the most important reliability measures for clinical purposes. SDC values of 6.9 have previously been reported for the BESTest [28], and in the range of 2.0–4.4 points on the Mini-BESTest [12, 13, 28–30]. We found SDC to be 10.8–11.8 for the BESTest, and 4.9–5.2 for the Mini-BESTest, for the two raters in our study. The discrepancy between our results and the previous studies may be explained by the fact that we used two raters and did not score the tests from video as have been done in the other studies. Furthermore, we included patients with neurological disease who have earlier been found to have high variability in behaviour [31], which might also explain the larger SDC in our study. Biological variability is however a characteristic by the sample and should not be regarded a measurement error.

Validity

Ideally, the concurrent validity of newly developed assessments methods should be established by examining how well the new method reflects the existing gold standard method. However, because the BESTest is developed based on other balance and mobility tests such as the Berg Balance scale, the Dynamic Gait Index and the Timed up and Go, and incorporates modified items from these tests, this is not a suitable approach for evaluation of the validity of the BESTest [2]. Similar to previous studies we chose to examine concurrent validity of the BESTest with measures that addresses related, but not identical constructs. While previous studies have used the ABC Scale, we used the FES-I in our study since the Norwegian translation has previously been translated to Norwegian and tested for psychometric properties [20, 32]. The correlations between the FES-I and the BESTests in our study were moderate (BESTest r _s = −0.51/Mini-BESTest r _s = 0–.50). This is in accordance with our a priori hypothesis, as fear of falling and balance are related but not identical constructs. Fear of falling is naturally related to balance performance, however other factors than balance will also influence fear of falling. Previous studies have observed moderate to high correlations between the BESTest and the ABC Scale [2, 28, 30, 33]. Although the ABC Scale and the FES-I have similar items and are highly correlated (r = 0.68), there are also differences between the scales. The ABC Scale has more questions on gait, while the FES-I has more focus on social activities [34]. This may also explain the higher correlations between the BESTests and the ABC Scale compared to our findings.

The study sample in a reliability and validity study should reflect the population of interest [24]. To cover a heterogeneous population of people at risk of falling we included subgroups of older persons and persons with diagnoses of stroke or MS. All three groups have earlier been found to have increased risk of falling, and are thought to display a variety of balance impairments [1, 6, 35, 36]. We succeeded to recruit a heterogeneous sample, as the total scores for the BESTest ranged from 31 to 102 and there was also a wide range of scores for the different test sections. Correspondingly, the scores for the Mini-BESTest showed a considerable variability with scores ranging between 1 and 27 points (max score 28) (Table 2). Thus, our results can likely be generalised to a wide group of people with increased risk of falling.

A major strength of the study is the thoroughly conducted test procedures. Both versions of the BESTest were translated according to cross-cultural validity procedures [16]. Further, the two test sessions were close to identical as all the participants are tested by the same raters, with the same equipment in the same room at the same time of day. Another strength of the study is the evaluation of absolute reliability of the BESTest and the Mini-BESTest. By determining SEM and SDC₉₅ for all subscales and total scores in a sample of persons with fall risk, this increases the interpretability of BESTest results both in clinical practice and in research.

The study findings will be useful for directing interventions and fall-preventions aimed at reducing falls and improving balance in patients coming to the physiotherapist for treatment. Conclusively this study indicates that the Norwegian version of BESTest and its short form the Mini-BESTest are reliable and valid instruments for assessing balance in community-dwelling people with increased risk of falling, but that change in balance performance measured with the BESTest and the Mini-BESTest should be interpreted cautiously.