Empowering individuals to prevent falls better

Psychometric properties of the Balance Recovery Confidence scale

Content development, content validity and face validity

The content of the BRC scale was developed using a two-stage approach. The first stage of development was to construct the content with the target population. Twelve community-dwelling older adults aged 65 were invited to participate in focus group sessions. The nominal group technique was used. Ninety items were generated. Thirty-two items were selected to be important to measure the construct of interest and congruent with Bandura’s self-efficacy theory 11 and literature.

The second stage of development was to validate the content with the target population and clinical experts. Delphi Technique and RAND/UCLA appropriateness method were used. A new group of ten community-dwelling older adults aged 65 years and older and twenty-eight healthcare professions representing physiotherapy, occupational therapy, nursing, podiatry and medicine participated to evaluate the instrument’s name, instructions, recall period, response options, and the thirty-two items. The below aspects of the instruments achieved consensus of appropriate. The instrument achieved face validity.

On acceptability
The BRC scale was found to be well-accepted by the target population. All items were completed without missing scores. The distribution of responses was broad across the score categories. The distribution of the BRC scale scores was identified to have no significant departures from normality (W=.97, p = 0.05), with skewness of -0.32, kurtosis of -0.63 (SEM 0.22).

On internal structure
The BRC scale was identified to be measuring a single factor. Since the BRC scale was constructed to be unidimensional, confirmatory factor analysis (CFA) was applied. The standardized root mean square residual (SRMR) was 0.057, meeting the criteria of < 0.08 for a good fit. Each item had a single factor loading that ranged between 0.727 to 0.921. However, comparative fit index (CFI) and Tucker-Lewis Index (TFL) scores of 0.792 and 0.767, respectively, suggested that some improvement could be made to improve the structural validity of the BRC scale for the sample population. Nevertheless, the uniqueness of the items in the BRC scale was low (0.15 to 0.47), which indicated that the variation in each item could be explained by the latent construct of balance recovery confidence.

There was no flooring effect for all items (Table 3). However, nine items (Items 3, 8, 9, 10, 11, 14, 17, 18, 19) were identified as having a ceiling effect. This showed that these items were easy for the sample population. Based on the Wright map (Figure 1), the BRC scale was well-targeted to the sample. This reflected that the items were generally not extremely difficulty or extremely easy for the target population with a range of confidence levels from low to high certainty to recover balance.

On reliability
The BRC scale had excellent internal consistency (α = .975). The reliability of the person ability estimates and the item difficulty estimates were high, indicated by 0.93 (person separation =3.76) and 0.94 (item separation = 3.99), respectively. The test-retest reliability was excellent (ICC = 0.944, CI of 0.891 to 0.969).

On validity
The modelling fit statistics showed that the sample population had a person ability estimate means of 0.86 logits (SD 1.46). This implied that this sample of older adults found the items in the BRC scale to be sufficiently challenging. The standard deviation of 1.46 logits for the person estimate indicated an adequate spread of person measures. The mean of the infit and outfit mean squares at 1.03 and 1.01, respectively, were harmonious with the Rasch-modelled expectations of 1. This indicated the data fitted well to the probabilistic Rasch model specification. The standardized fit Z values were around zero (infit Z = -0.5; outfit Z = -0.5). The variation of modelled fit scores for persons (infit Z SD = 2.3 and outfit Z SD = 2.3) suggested that most person ability estimates were able to be transformed within the fit statistics. RMT had identified that some items were potentially redundant and that the instrument could be further refined for better utility.

On construct validity
The field test used the multiple outcome measures to evaluate the construct validity of the BRC scale. The patient-reported outcome measures used were the Activities-specific Balance Confidence (ABC) scale, the Falls Efficacy Scale-International (FES-I) scale, and the Late Life Function and Disability Instrument (LLFDI) scale. The performance-based measures were the hand-held dynamometer for grip strength, the 30-second sit-to-stand test for lower limb strength, and the Mini BESTest for balance performance. The sub-section of the Mini BESTest on reactive balance was used for the BRC scale to compare against reactive balance control ability.
There were moderate positive correlations between the BRC scale, ABC scale, LLFDI scale, and Mini BESTest (Figure 2). The BRC scale and the FES-I scale had a moderate negative correlation. There were weak correlations between the BRC scale, handgrip strength dynamometer, and 30-second chair stand test. The BRC scale and the reactive postural control section of Mini-BESTest (Section 4-6) were strongly correlated.

Further work evaluating the Balance Recovery Confidence scale is underway. Contact Dr Shawn Soh at [email protected] if you want to join the team advancing the understanding of balance recovery confidence.