Cross-cultural adaption and validation of the German version of the Mini-BESTest in individuals after stroke: an observational study

According to the guideline suggested by Beaton et al. (2000), the original version of the Mini-BESTest was translated and cross-culturally adapted from May to August 2018 [17]. All six recommended stages of cross-cultural adaption were conducted; for instance, the translation of the original version into the German language and the back-translation by two independent persons, respectively. A committee of experts created a pre-final version. This pre-final version was pilot-tested with ten participants and linguistic expressions in the GVMBT were finalised. In correspondence with the developers of the original Mini-BESTest, an experts’ committee created a final German version.

Data acquisition took place at two neurorehabilitation centers in North Rhine-Westphalia, Germany, from October 2018 to January 2019. Two experienced physiotherapists assessed all participants. In order to guarantee standardized implementation, the physiotherapists watched the original training videos of the Mini-BESTest and practiced the handling of the MIs on healthy persons.

All participants provided written informed consent. Inclusion criteria were: individuals (1) after stroke, (2) with hemiplegia, (3) able to walk at least six meters (walking aids allowed), (4) with sufficient cognitive abilities regarding verbal communication and understanding instructions in the German language, (5) aged ≥18 years and (6) physically able to perform different tasks for 1.5 h. Exclusion criteria were: individuals (1) with ataxia, (2) with deafness or blindness, (3) in clinical isolation, (4) with contraindications for mobilization, (5) with acute pain, (6) who were pregnant, (7) with symptoms of vertigo and (8) with other conditions influencing the balance ability for example polyneuropathy. Since the “COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN)” guidelines do not recommend a specific sample size for validity analysis [19], the sample size was therefore oriented on similar validity studies [20‐22]. Due to the wide range of different sample sizes reported in similar studies, sample size calculation for validity analysis was done a priori with an effect size of 0.5, an alpha error of 0.05 (two-tailed) and a power of 0.9 [23].

Initially, the National Institute Health Stroke Scale (NIHSS) was performed to describe the study population. The testing sequence (Mini-BESTest, TUG and BBS) was randomly assigned to the participants, avoiding bias because of fatigue [18]. The administration time of the Mini-BESTest was recorded for each participant. Additionally, each participant was asked about the comprehensibility of the test instructions. Blinding of the physiotherapists was not possible.

The Mini-BESTest is a 14-item balance test with four subcategories: ‘anticipatory control’, ‘reactive postural control’, ‘sensory orientation’ and ‘dynamic gait’. Each item is scored from 0 to 2. A higher score indicates better postural control. Administration time of the Mini-BESTest is reported with 15 min [14].

The BBS consists of 14 items; each item is scored out of 4. Individuals need to maintain positions and perform specific balance tasks. Higher scores reflect better postural control. It takes 10–15 min for administration [8]. The BBS revealed good validity, reliability and internal consistency in patients after stroke except for floor and ceiling effects [9].

The TUG measures the time in seconds needed for standing up from a chair, walking three meters straight-ahead, turning around, returning to the chair, and sitting down again. Individuals perform the TUG with their regular walking aids and footwear, but without physical assistance. Healthy controls perform the TUG in 10 s or less [11]. The TUG demonstrated good convergent validity, apart from being a single-task-MI [10].

Statistical calculation was performed for 50 participants using the statistic software “R v3.5.2”. Descriptive statistics were calculated for gender, age, number of strokes, time and stage after stroke, the use of walking aids and the NIHSS. The MIs were assessed for normal distribution with the calculation of skewness, kurtosis and Q-Q-Plots. Floor and ceiling effects were analysed by calculating percentages for the lowest or highest possible score. The established threshold to identify a floor or ceiling effect is 15% [5]. For hypothesis testing and examining the relationship between the GVMBT and the BBS and the TUG, Spearman’s rank correlation coefficient (r_s) was used, since the Mini-BESTest and the BBS are ordinal-scaled. The correlation coefficient (r_s) is interpreted as follows: 0–0.25 little or no relationship; 0.25–0.50 fair relationship; 0.50–0.75 moderate to good relationship and above 0.75 good to excellent relationship [24]. A significant, moderate to high association between MIs would provide evidence of convergent validity [5]. Bland Altman plots were compiled to compare agreement between MIs. Data transformation was necessary, since all tests have different scales. It was performed with the following formulas:

In order to achieve comparable results, data transformation was calculated with the highest possible score of the BBS and the longest time required in the TUG and then based on the scoring of the Mini-BESTest. Good agreement between two MIs is given if the limits of agreement include the line of equality (0) in the Bland Altman plots [25]. Corresponding scatter plots were calculated with non-transformed data. The internal consistency for the GVMBT was assessed as Cronbach’s Alpha. The minimum value for clinical application is α = 0.90 [5].

A total of 62 individuals after stroke participated in the study. Two persons were excluded because of limited cognitive abilities or cardio respiratory capacity. Ten participants completed pilot-testing and 50 participants were included in statistical analysis. Descriptive characteristics of the participants and scores on all MIs are provided in Table 1. There were no missing data.

Translation and back translation of the Mini-BESTest were conducted and a committee of investigators discussed and resolved discrepancies. All units were converted to the international metric system. The GVMBT was performed without any problems. Participants reported the Mini-BESTest as comprehensible during performance and relevant concerning their activities of daily living. The average application time of the GVMBT was 16 min.

No significant floor or ceiling effects were observed for the GVMBT with 2% of the participants reaching the maximum score and none achieving the lowest possible score. In contrast to the GVMBT, the BBS and TUG demonstrated significant deviations of the normal distribution (Table 2 and Fig. 1a-c).

Significant relationships were found between the GVMBT and the BBS (r_s = 0.93, p < 0.001) and TUG (r_s = − 0.85, p < 0.001) (Table 3). These high associations confirm convergent validity of the GVMBT.

In addition, Bland Altman analysis revealed sufficient agreement between all MIs (GVMBT-BBS/TUG). The biases between the GVMBT and the BBS or the TUG were − 5.36 or 4.92, respectively (Table 4). The bias describes the difference between two methods. The comparison between the GVMBT and the BBS demonstrated a smaller variation compared to the GVMBT and the TUG. Both comparisons included the line of equality, which shows that there is no significant difference between the MIs. However, Fig. 2a+c visualized that the line of equality was more centred between the GVMBT and the TUG. The corresponding scatter plots revealed the association between the compared MIs (Fig. 2b+d).

Linguistic and content-related discrepancies were clarified in correspondence with the original developer [14] and an interdisciplinary committee of experts. The GVMBT was comprehensible and well accepted by participants and physiotherapists, which is in accordance with other translational studies of the Mini-BESTest [20‐22]. Consistently, the application of the German version was comparable to the original version.

The GVMBT was significantly less skewed and displayed the least ceiling effect compared to the BBS and TUG. In contrast, the BBS and the TUG demonstrated significant skewness. As expected, the BBS almost revealed a ceiling effect of 14% in our participants reaching the highest possible score, which is close to the established threshold of 15% [5]. A recent review of the psychometric properties of the Mini-BESTest reported a ceiling effect ranging from 0.9 to 4.3% in various populations [16], which is similar to our results, showing a ceiling effect of 2%. Tsang and colleagues [26] determined a skewness of 2.69 in the BBS and 1.69 in the TUG, which are similarly distributed to our findings. Furthermore, the authors reported a ceiling effect of 32.1% for the BBS in persons with chronic stroke [26]. Our results revealed no floor effects in any MI, since the inclusion criteria assume a walking ability. A previous review reported a floor effect for the Mini-BESTest in a small group of individuals with strong impairments [16]. Additionally, the Mini-BESTest covers, with its four subcategories, different aspects of postural control and includes more challenging tasks, in contrast to other MIs. The BBS and TUG are more limited and do not include the comprehensive examination of anticipatory postural control, reactive postural control, sensory orientation and dynamic gait aspects, including dual-task [14]. As a result, the subcategories may have improved the discrimination between participants in the GVMBT, too.

The excellent construct validity of the GVMBT is reflected by very high correlations between the GVMBT and the BBS or TUG. The correlations strengthen the assumption that the MIs assess the construct postural control and thus support the construct validity. Our results are in agreement with previous validity studies on the original [14, 16], Greek [22] and Swedish [20] version and in individuals after stroke [26], persons with balance disorders, Parkinson’s disease, spinal cord injury [21] and adults aged 50 years and older [16]. Other studies also investigated the construct validity of the Mini-BESTest with different subjective outcomes like fall efficacy and health-related quality of life. For instance, five papers assessed the concurrent validity between the Mini-BESTest and measures of fall efficacy in different populations [20, 27‐30]. A low correlation with the fall efficacy scale (r = 0.26) [20] was reported and moderate correlations with the activities-specific balance confidence scale (r between 0.53 and 0.66, p = 0.52) [27‐30]. The Mini-BESTest may not reflect the performance in everyday life in detail and may therefore not be transferable to the estimated fear of falling. However, the Mini-BESTest may indicate, if there is an increased fear of falling. To date, only one study investigated the relation of health-related quality of life (Quality of Life Questionnaire-C30) and postural control (Mini-BESTest) in patients with cancer. It showed positive correlations between the Mini-BESTest and all domains of the Quality of Life Questionnaire-C30 [31]. Further studies, especially with other populations are needed.

However, since high correlations do not automatically imply an agreement between MIs, Bland Altman analysis was applied [25]. Our results suggest that the comparisons of the GVMBT with the BBS and TUG expose no significant bias, since the lines of equality were included in the limits of agreement and the mean differences were relatively low. Additionally, the standard deviation was smaller for the comparison of the GVMBT and BBS. In this study the GVMBT showed no over- or underestimation of balance performance.

Internal consistency of the GVMBT with a Cronbach’s α of 0.90 was considered as excellent. Our values are in agreement with those previously reported in different populations with Cronbach’s α ranging from 0.89–0.96 [16]. Moreover, it is comparable to other translated versions as for instance the Greek version with Cronbach’s α of 0.95 [22].

Although the present study has shown that the GVMBT is valid and comprehensible, it has some limitations. Our findings are only generalizable to individuals in the subacute or chronic stages of stroke recovery and the sample size was too small for further analysis such as item-response theory or Rasch analysis. Although there are many different MIs to assess postural control such as the Dynamic Gait Index [6, 13], we decided to administer the most commonly used balance MIs in stroke rehabilitation and research to ensure feasibility and to avoid patient fatigue.