Psychometric properties of Fullerton Advanced Balance Scale in patients with stroke

ABSTRACT Background Patients experience falls frequently after stroke. Preserved or acquired balance skills decrease fall risk and improve independence. Feasibility of Fullerton Advanced Balance Scale (FAB) has been shown in balance assessment in some neurological diseases except stroke. Objectives The purpose of this study was to investigate the reliability and validity of Turkish version of FAB (FAB-T) in patients with stroke (PwS). Methods This cross-sectional study included 51 PwS (60.64 ± 7.66 years). Reliability analyses were conducted with Cronbach’s alpha, intraclass correlation coefficient (ICC), and Spearman correlation analysis. Intra-rater and inter-rater reliability were assessed with three raters. FAB-T, Stroke Rehabilitation Assessment of Movement (STREAM), Brunnstrom Recovery Stages (BRS), Barthel Index (BI), and 36-Item Short Form Health Survey (SF-36) were used for convergent validity. Correlations of FAB-T with Berg Balance Scale (BBS) and Mini-Balance Evaluation Systems Test (Mini-BESTest) were measured for concurrent validity. Spearman correlation analysis was used for convergent and concurrent validity. For predictive validity patients’ self-reports of falling were analyzed with ROC. Results Intra-rater (ICC = 0.998) and inter-rater reliability (ICCs = 0.984; 0.984; 0.990), and internal consistency (Cronbach’s alpha = 0.930) were excellent. FAB-T had good correlations with STREAM (ρ = 0.677) and BI (ρ = 0.628), moderate correlations with BRS (ρ = 0.504 and ρ = 0.579) and physical function of SF-36 (ρ = 0.436). FAB-T excellently correlated with Mini-BESTest and BBS (ρ = 0.928 and ρ = 0.942). The cutoff score of FAB-T was determined to be 21.5 points, with sensitivity of 84% and specificity of 61% (AUC = 0.749). Conclusions FAB-T is a reliable and valid balance assessment tool with an acceptable accuracy of fall prediction in PwS.


Introduction
2][3] Abnormal postural sway, impaired weight distribution and weight shifting due to muscle paresis with or without spasticity were reported.PwS also have problems in sensory integration and reweighting, where they exhibit greater postural sway in changing somatosensory and visual information. 4,5n effective stroke rehabilitation program depends on eligible assessment of special requirements.][10] Currently, balance evaluation of PwS in clinical settings is mostly conducted with Mini-BESTest and BBS which both have functional tasks.Although BBS is the most frequently used test in this field, it has ceiling effect and deficiency of evaluating reactive postural control which is closely related to falling risk. 11As an advantage, Mini-BESTest examines different systems contributing to balance control such as verticality, anticipatory postural adjustments, automatic postural responses, sensory organization, and gait stability. 12Unlike BBS, any floor or ceiling effects do not exist for Mini-BESTest 13 evaluating balance in more dynamic conditions than BBS and dual-task performance.Nevertheless, the scale has still disadvantages such as long administration time and insensitive scoring system. 14ullerton Advanced Balance Scale (FAB) is a performance-based tool originally developed to assess balance and determine fall risk of highfunctioning older adults.The scale evaluates functions likely to be affected after stroke such as sensory and musculoskeletal systems, sensory strategies, proactive and reactive mechanisms.In addition to its easy application, it has potential to evaluate balance comprehensively with its multidimensional item composition and sensitive scoring system. 14,155][16] FAB has been translated into Turkish as FAB-T, and found to be reliable and valid just in older adults, 17 but its reliability and validity has not been studied in PwS yet.The purpose of this study is to investigate reliability and validity of FAB-T in PwS.

Study design
This cross-sectional study was completed from January 2020 to August 2020 in Hacettepe University Faculty of Physical Therapy and Rehabilitation, and Ankara Physical Medicine and Rehabilitation Training and Research Hospital with the patients with stroke diagnosed by neurologist undergoing their routine rehabilitation program.Study was approved by Hacettepe University Non-Interventional Clinical Researches Ethics Board.All participants signed the written informed consent after being elucidated by the researcher.Our study complies with (The STAndards for Reporting of Diagnostic Accuracy) STARD 2015 checklist.

Participants
Fifty-one PwS participated in this study.All participants were screened and assessed by the first rater for convenience before the study when the first rater existed in rehabilitation unit.Inclusion criteria were; age ranging between 18 and 80, having first time unilateral anterior circulation stroke at least 3 months ago, able to walk at least 10 m without human assistance, lower extremity level 3 to 6 for Brunnstrom Recovery Stages (BRS) and at least 24 points at Mini Mental State Exam (MMSE).Patients who have vestibular and visual dysfunction, other movement disorders and comorbid diseases preventing study process were excluded.Sample size was determined using a priori analysis.Based on sample sizes suggested by Bonett 18 for reliability analysis between two sets of data (i.e.two judges or two measurements in time), a sample of 15 patients is adequate at alpha level of 0.05 and correlation coefficient above 0.9.Our study was based on alpha level of 0.05 (2-tailed) and a power of 0.8.The studies of Iyigün et al, 17 Kim et al, 19 Schott et al 20 and Schenledt et al 14 have shown ICC 0.92 and above for rater reliability so we expected a value of ICC at least 0.9 in the present study.Schott et al, 20 Iyigün et al 17 and Kim et al 19 also found strong to excellent correlations of FAB with BBS (rho = 0.68, 0.70 and 0.89 respectively) in older adults and Schenledt et al 14 showed excellent correlation between FAB and Mini-BESTest (rho = 0,87).We just in case expected at least a moderate effect size (here correlation coefficient) for validity analyses between FAB-T with BBS and Mini-BESTest.Using the value of a moderate effect size of rho = 0.4 we calculated 46 participants as sample size for validity anlayses with the help of G*Power software. 21aking into account the probability of dropout of 10%, we determined our sample as at least 51 participants and started study with 53 participants.

Measures
Balance a. FAB.FAB measures dynamic and static balance skills.It has 10 items and 5-point ordinal scoring system (0-4).Total score ranges between 0 to 40 and higher scores indicate better balance capability.Average administration time of test ranges 10-12 minutes.Test materials used are stopwatch, pen, masking tape, tape measure, 15 cm x 35 cm x 45 cm bench, 2 Airex balance pads (Airex AG, Sins, Switzerland), metronome (100 bpm) and non-slip material. 15Score of 25 and below indicates high risk of falling in older adults. 22b.BBS.BBS evaluates balance in different tasks difficulty levels.It is an ordinal scale and contains a total of 14 items scored between 0-4 depending on time and total score is ranging 0-56.There is high risk of falling with a score of 44 and below. 23BS has been found reliable and valid to assess balance in stroke. 11

Motor ability a. Stroke Rehabilitation Assessment of Movement (STREAM).
STREAM is a motor performance scale evaluating extremities and basic mobility.It consists of three subscales, each of which has 10 items.Upper and lower extremities have 20 total points each and mobility dimension has 30 points.Each item in extremity scoring ranges between 0-2 while 0-3 in mobility and higher scores suggest better performance.It is reliable and valid in clinical practice and research in stroke. 24b.Brunnstrom Recovery Stages (BRS).BRS helps to describe motor recovery phases after stroke through considering spasticity and voluntary movement presence.BRS includes six stages for upper and lower extremities, and hand.Stage 1 is flaccid stage and there is no active movement.In stage 6, spasticity disappears and isolated joint movements can be performed.Recovery increases as the stage advances. 25

Activities of daily living (a) Barthel Index for Activities of Daily Living (BI).
BI has 100 points in total with 10 items evaluating patient's independence in ADL.Score of 0-20 indicates complete dependence while 100 indicates complete independence.It is a reliable and valid scale for rehabilitation outcomes in stroke. 26ality of Life (QoL) (a) 36-Item Short Form Health Survey (SF-36).SF-36 is a questionnaire evaluating eight dimensions of health-related QoL.Each item of 36 total has full score between 2-6 and each section is evaluated out of 100 points.Higher section score is directly proportional to QoL. 27

Procedure
Each patient was evaluated by the same physiotherapist and FAB-T assessments were videotaped with permission of patients during initial evaluations.The video recordings of 30% of participants picked by simple random numbers table were watched and rated by the first rater for intra-rater reliability 10 days after first assessment. 14Then, same video recordings were watched and rated by second and third rater for inter-rater reliability among three raters. 28,29Raters other than the first rater and statistician were blind to participant's clinical information and other test scores.Face to face evaluations of first rater were analyzed for internal consistency.Construct validity was assessed with exploratory factor analysis and convergent validity.Within convergent validity, the relationship of FAB-T with STREAM, SF-36, BRS and BI was examined.Concurrent validity was investigated by considering correlations of FAB-T with Mini-BESTest and BBS.For predictive validity, the relationship between FAB-T scores and self-reported falling frequencies of 39 patients in last 6 months was searched. 14

Statistical analysis
IBM SPSS Statistics 22 (Statistical Package for the Social Sciences) analysis program was used for statistical analysis.Normal distribution was tested via Shapiro-Wilk test, and descriptive statistics were given as mean (SD), median (minimum-maximum) or number (percentage), where appropriate.Cronbach's alpha coefficients (0.70 or higher is accepted as ''sufficient'') 30 were used for internal consistency.ICC values (excellent >0.9, good 0.9-0.75,moderate 0.5-0.75 or poor < 0.5) 31 were calculated for rater reliability analysis.Sampling adequacy for factor analysis was evaluated by Kaiser-Meyer-Olkin (KMO) (<0.8 indicates adequacy) and Bartlett's Test of Sphericity (BTS) (<.05 is acceptable) results. 32For predictive validity, AUC (Area Under Curve) value (acceptable when < 0.7 22 was calculated by Receiver Operating Characteristic (ROC) analysis.Spearman's rho (ρ) value (negligible to poor < 0.40, moderate 0.40-0.60,strong 0.60-0.80 or excellent > 0.80) 33 was used for all correlation analyses.The floor and ceiling effects were considered with frequency table of total scores.Statistical significance level was accepted as p < .05and2-tailed 95% confidence interval (95% CI) was utilized.Estimation of correlation CIs was based on Fisher's r-to-z transformation.

Subject characteristics
Fifty-one PwS of seventy-five interviewed were included in this study (Figure 1).Table 1 shows demographic and clinical characteristics of participants.In balance tests mean scores of FAB-T, Mini-BESTest and BBS were 19.94 (10.93), 16.90 (7.75) and 48.69 (8.56) respectively.Mean score of STREAM was 47.76 (13.8).Majority of the extremity scores were in 3 rd stage of BRS.BI and SF-36 scores showed that patients were moderately dependent and their QoL was mildly to moderately affected.

Discussion
This is the first study to examine whether FAB-T is reliable and valid to assess balance disorders in PwS.Our study concluded with satisfactory consequences presenting reliability and validity of FAB-T in PwS.Current study confirmed that FAB-T has excellent rater reliability and concurrent validity.Also we had acceptable to excellent convergent validity results and showed that FAB-T can predict falling status at an acceptable level.As a result of our study, it was determined that FAB-T did not show ceiling and floor effects.In this case, it can be said that it is more advantageous to use FAB-T instead of BBS in balance assessments in subacute and chronic stroke patients.FAB-T showed higher sensitivity, specificity and accuracy than Mini-BESTest in determining fall risk in subacute and chronic stroke patients.It has also been observed that FAB-T is a faster test than Mini-BESTest.Therefore, FAB-T seems to be more preferable than Mini-BESTest in the evaluation of balance in post-stroke clinical conditions.We found moderate to excellent internal consistency results supported by literature on older adults, 17,19,20,34 PD 14 and CP. 16These results show that FAB-T items are purposefully related to each other.Also Cronbach's alpha coefficient decreased when items were deleted except 10 th which has negligible effect.Items most contributed to internal consistency were 4th and 5 th items confirming Schott et al. who found item-total score  correlations moderate to very good, and items 8, 5 and 4 had the most positive effect on the scale consistency in older adults. 20Climbing up and down a step relates very closely to standing balance, 35 and item 4 tests postural preparation before activity, weight bearing on one leg dynamically and adaptive responses to stepping up and down.Also 5th item testing tandem walking of which performance worsens in neurological problems and advancing ages 36,37 was highly correlated with overall score in present study.Considering age of PwS, it is quite natural for them to experience loss of balance when asked to maintain an active gait on a narrow base of support with present sensory and motor problems.We had excellent ICC values and Spearman correlations for rater reliability in this study similarly to the studies with older adults, older adults with balance problems and PD. 14,15,17In our opinion, to use of most of items in FAB-T already in different scales and to be the instruction to raters clear provides convenience to raters, even if in different experience levels.In addition, comprehensibility of instructions to the participant and ease of application may have improved high rater reliability.
It has been shown that lower 24,38,39 and surprisingly upper extremity 24,40-42 recovery levels are associated with functional balance in PwS similar to our motor ability related balance results extending moderate to excellent correlations.It is crucial to have sufficient motor outputs for a good balance.STREAM has basic mobility, standing balance and functional activities components such as walking and going down stairs, again STREAM and also BRS include isolated extremity movements.Thus, FAB-T has potential to determine motor ability related balance status and extremity functionality-balance relationship.The BI is closely related to balance, such as transfer, stair climbing and mobility, and has close relationship with disease severity. 43,44Very good to excellent correlation between balance and BI in PwS has been reported 23,45 and our results were similar.Considering the balance related components of STREAM and BI, correlations of these scales with FAB-T strengthen FAB-T's con- vergence in PwS. 43,44Loss of balance is among the factors affecting QoL after stroke. 46,47Several researchers showed effect of balance skill on QoL and increase of QoL parallels to improve of balance in PwS. 48,49Current study showed moderate correlation between SF-36 physical component and FAB-T similar to Garland's study. 50ivisions questioned in physical function of SF-36 are doing sports, squatting, climbing stairs and limitations in functions such as in walking.To us, it is clear that each of these items have an equilibrium component and explains equilibriumrelated results.FAB-T was perfectly correlated with both Mini-BESTest and BBS in present study.Several studies found good to perfect correlations between FAB with BBS and Mini-BESTest in PD, older adults with or without balance disorder. 14,15,17,19,34BBS aims to measure standing and sitting balance in terms of static and dynamic aspects.Although it has ceiling effect, BBS is still used in chronic and high functional PwS.Also to us BBS is more convenient for acute and subacute of PwS as shown in literature. 51ini-BESTest measures balance in more dynamic conditions than BBS, but long application time may become clinically impractical causing fatigue for patients.FAB, on the other hand, is considered more advantageous in terms of examining gait balance with dual task, static and dynamic balance from proactive and reactive perspective in a short time.Each use of assistive devices decreases 1 point in each item of Mini-BESTest.Differently, FAB allows patients to exhibit their potentials without rigid restrictions but in a non-compensatory way.FAB seems to prioritize functionality mostly on outputs in ADL.While Mini-BESTest offers 3-point ordinal scoring, the 5-point scoring of FAB can better identify subtle differences in balance performance and offer more detailed classified functional levels.For clinical evaluation the cost effectiveness, time investment, (clinical) feasibility, and the close relationship with functional ADL are crucial for selecting a tool.Correspondingly FAB-T differs from other scales compared in terms of space saving, application time and clinical practicality.It can also be easily fulfilled with its 10-item structure in 10-12 minutes with already existing materials (balance pad, step, stopwatch, tape measure and 100 bpm tuned metronom on mobile phone) in clinical settings.
The prevalence of aphasia in PwS is 12%. 52atient is asked to count backwards 3 by 3 while performing the timed-up and go dual-task component of Mini-BESTest.So it is not possible to evaluate aphasic patients with Mini-BESTest.Fortunately, there is no item in FAB-T requiring patient's verbal expression.In this respect, we think FAB-T allows to evaluate PwS not having a cognitive problem for understanding the instructions but have speech problems.Evaluation of the ''rise to toes'' item in Mini-BESTest was difficult to score.Because patients who could not rise on toes on the hemiplegic side compensated it with healthy side and was scored with 3 seconds as full score.Similarly, during ''stand on one leg'' item in Mini-BESTest, standing on hemiplegic side is taken into account with healthy side.This results in the same score for patients who can stand on both feet for 5 seconds, and for patients who can stand for more than 20 seconds on one side and less than 5 seconds on the other, for example.Since FAB-T evaluates standing on preferred foot and prioritizes postural control, reliability of the evaluation for this item increases.
Postural control is possible with a dynamic balance built on a well-established static balance, which is both affected after stroke.Mini-BESTest includes standing on one leg and keeping eyes closed on soft foam as a static challenge.Then with a segmental transition to dynamic balance, walking parameters are evaluated during rest of test.While standing on one foot and on soft ground are in static characteristics in FAB similar to Mini-BESTest, differently a smooth transition is made to dynamic balance evaluation with 360degree rotation and functional reach tasks reflecting "quasi-dynamic" balance where center of mass is at the margins of supporting surface, in other words, at stability limits.Narrowing support surface is an effective method challenging and measuring balance ability of PwS. 37Therefore, FAB has an advantage to have tandem walking component.While PwS evaluated in reliability and validity study of Mini-BESTest were community dwelling, 13 we had both inpatient and community dwelling patients in both subacute and chronic periods.Therefore, we have proved the inclusivity of FAB-T by showing its usability in different PwS populations.
Given the current literature about fall prediction, our AUC value (AUC = 0.749, 95% CI = 0.59-0.91 with a sensitivity of 84% and a specificity of 61%) remains higher than the other frequently used balance scales (Mini-BESTest AUC = 0.64, 95% CI = 0.51-0.77;BBS AUC = 0.72, 95% CI = 0.61-0.83;TUG AUC = 0.66, 95% CI = 0.53-0.80;on the paretic leg stand AUC = 0.67, 95% CI = 0.54-0.80;on the nonparetic leg stand AUC = 0.64, 95% CI = 0.52-0.77and FRT AUC = 0.67, 95% CI = 0.55-0.79).As seen there are limited associations with fall history (AUC <0.7) among scales except BBS showing a reasonable AUC value.Because FAB-T AUC value was both at a reasonable level and superior to other scales, we assumed that FAB-T has the ability on predicting fall of stroke patients.The cutoff scores for fall risk of balance scales vary in different disease groups.In previous studies cutoff scores of FAB has been found as 25 for older adults 22 and as 27 for PD. 53Our cutoff score (21.5 points) was lower compared to older adults.Furthermore, Shumway-Cook et al. reported BBS cutoff score as 40 in older adults, 54 but Maeda et al. defined BBS score of 29 in PwS. 55imilarly, the cutoff point for Mini-BESTest was determined as 19.5 in older adults, 56 but as 17.5 in chronic PwS. 13 PwS lower their activity levels due to fear of falling and less involved in community. 57,58However, our patients had high ADL scores and this may be due to indoor activities repeated more often than those outside and learned automatically.In addition, limited environment and lack of free space in house may enable patients to use protective reactions effectively to prevent falls.Another finding that should be emphasized is that none of our participants had the lowest score (0 point) and only one had the highest score (40 points) on FAB-T, which both are under %15 of all patients.Consequently we did not find any floor or ceiling effect parallel to previous studies. 14,17,22ence, we assumed that FAB-T had a good score distribution in assessing balance of PwS.
This study has some limitations.Relatively small sample size may have limited the potential of generalization of results.Nevertheles we tried to hold our sample homogenus in terms of BRS.
As the patients could not be reevaluated at different time intervals, the test-retest method could not be applied and this can be considered as an other limitation of our study.Lastly, the study could benefit from including comparisons with other computerized balance assessment tools such as posturography or inertial sensors.This would provide a more comprehensive understanding of the performance of the FAB-T.
In our study, no relationship was found between FAB-T and the SF 36 QoL questionnaire, except for one subcomponent.In future balance studies, it may be recommended to use strokespecific QoL scales in order to examine the relationship of FAB-T with quality of life.In addition, considering the physical activity levels of the patients may contribute to the interpretation of the fall risk cutoff scores.In order to determine whether FAB-T can predict recovery, the scale can be applied before and after rehabilitation to investigate the ability of the scale to show changes in patients' performance over time.Finally, a detailed examination of the effect of muscle tone on balance in stroke patients may help fill the gap in the literature.

Conclusion
Our study showed that FAB-T is a reliable and valid tool which is applicable in a short time easily and addresses all aspects of balance assessment in subacute and chronic patients with anterior circulation stroke.Allowing to evaluate via video recording and assess independent of rater experience are noteworthy for FAB-T.In addition, the scale does not show any ceiling or floor effect and is a good fall predictor.

Figure 2 .
Figure 2. Scree plot and eigenvalues of FAB scale items.

Table 2 .
Item-total score correlation and Cronbach's alpha values if item deleted of FAB Scale.

Table 4 .
Correlations of FAB scale with other scales within convergent and concurrent validity.