Berg Balance Scale Score as a Predictor of Independent Walking at Discharge among Adult Stroke Survivors

Abstract

Purpose: The purpose of this study was to determine which admission clinical assessment or assessments best predict independent walking at discharge (IW-DC) among adults with unilateral impairments hospitalized for rehabilitation post-stroke. Method: On admission, we collected measures of balance (Berg Balance Scale [BBS]), physical function (Chedoke McMaster Stroke Assessment – Activity Inventory), postural and leg motor control (Chedoke McMaster Stroke Assessment – Impairment Inventory), functional independence (FIM), sensation and proprioception, and pushing behaviour (Four-Point Pusher Score). Logistic regression determined which measures influenced the odds of IW-DC. A receiver operating characteristic (ROC) curve determined the cut-points for variables retained in a multivariable model. Results: Data were available for 68 participants, aged a median of 57 (interquartile range [IQR] 16) years, who had received inpatient rehabilitation for a median of 8 (IQR 10) weeks. The odds of IW-DC were reduced with greater impairments in motor control, sensation, or proprioception and with pusher behaviour and increased with lesser impairments in balance, physical function, and functional independence. Only the BBS was retained in the multivariable model (OR 1.23; 95% CI: 1.02, 1.49). An admission BBS score of 14 or more points (sensitivity 0.73; specificity 0.89) predicted IW-DC (area under the ROC curve 0.81; 95% CI: 0.71, 0.92). Conclusions: Among adult stroke survivors, a BBS score of 14 or more provides information on the odds of achieving IW-DC.

Clinicians describe independent walking as the most common rehabilitation goal for people after stroke. With the global trend toward a shorter length of stay (LOS) at inpatient stroke rehabilitation facilities, clinicians face increased pressure to prognosticate, soon after admission, the walking level that a patient will likely reach by discharge. The ability to predict, at the point of admission, whether a person is likely to achieve independence in walking by the time of discharge (IW-DC) gives the stroke rehabilitation team members an initial basis for establishing treatment goals, setting target treatments, and performing discharge planning. Team members can then use this information to guide the expectations of patients and their families. In this way, costs may be reduced because accurate predictions would increase the efficiency of stroke services.¹

Most previous research on the prediction of walking independence post-stroke has tended to evaluate independence at 6 or 12 months because this is when recovery is thought to plateau and a person’s ultimate walking capacity is generally represented.¹ A few studies, however, have explored the predictors of independent ambulation at the time of discharge from an inpatient stroke facility.^2–4 Moreover, there has been no consensus on which physiotherapy clinical assessments offer the most predictive information regarding whether a patient will achieve IW-DC.

Therefore, the aim of our study was to determine, among adults admitted to an inpatient stroke rehabilitation facility, which combination of clinical assessments – assessing balance (Berg Balance Scale [BBS]), physical function (Chedoke McMaster Stroke Assessment – Activity Inventory [CMSA–AI]), postural and leg motor control (Chedoke McMaster Stroke Assessment – Impairment Inventory), functional independence (FIM), sensation and proprioception, and pushing behaviour (Four-Point Pusher Score [4PPS]) – completed on admission best predict being able to achieve IW-DC.

Methods

Protocol

We used data collected on adults transferred to Royal Perth Rehabilitation Hospital (Shenton Park, WA, Australia) between June 2012 and January 2014 after an acute stroke. To be included in these analyses, the participants needed to present with unilateral impairments and be unable to ambulate independently at the time of admission. The local Human Research Ethics Committee approved our study, and participants provided written consent to the use of their data.

Within 2 days of admission, physiotherapists recorded demographic data and completed assessments of (1) balance (BBS),⁵ (2) physical function (CMSA–AI),⁶ (3) postural control (Chedoke McMaster Stroke Assessment – Impairment Inventory, denoted CMSA–PC),⁶ (4) leg motor control (Chedoke McMaster Stroke Assessment – Impairment Inventory, denoted CMSA–LC),⁶ (5) FIM,⁷ (6) sensation and proprioception at the ankle and metatarsophalangeal joints on the affected side (each ranked on a 5-point ordinal scale), and (7) pushing behaviour (4PPS).⁸ These assessments are collected as part of routine clinical practice. A round-table discussion by senior physiotherapy staff at Western Australian hospitals selected these assessments as the most useful measures.

Participants received physiotherapy for an average of 1.5 hours per day, 5 days a week. Other members of the comprehensive stroke team, such as occupational therapists, speech therapists, psychologists, social workers, nurses, and doctors, also provided care. We coded participants’ capacity to walk independently (without supervision or physical assistance) for 10 metres or more. Participants who achieved this were coded as 1 (IW-DC). Participants who were unable to walk independently for 10 metres or more were coded as 2. Those who could walk independently with the use of walking aids or orthoses were also coded as 1 (IW-DC).

Data management and statistical analyses

All analyses were performed using IBM SPSS Statistics for Windows, Version 22.0 (IBM Corporation, Armonk, NY). Frequency histograms displayed the distribution of continuous data and reported means and standard deviations (parametric) or medians and interquartile ranges (IQR; non-parametric). We implemented univariate (unadjusted and adjusted for length of stay [LOS]) and multivariable logistic regression models to determine which independent variables influenced the odds of IW-DC. When necessary, for independent variables collected on ordinal scales, we collapsed categories so that the minimum sample size in a category was 10. For example, for the CMSA–PC, we collapsed scores ranging from 1 to 7 into categories labelled no or mild impairment (score of 5–7), moderate impairment (score of 3 or 4), or severe impairment (score of 1 or 2).

Variables that were significant at p < 0.1 (in any category) in the univariate models were considered candidate predictors and entered into the multivariable model. Covariates that remained significant in the initial multivariable model were retained in the final model. Thereafter, for the continuous variables retained in the final model, we constructed a receiver operating characteristic (ROC) curve and determined the optimal operating point using Youden’s index. For all analyses, a p-value < 0.05 was considered significant.

Because these analyses were opportunistic and used data entered into an existing database, we did not undertake a priori sample size calculations.

Results

Consent was obtained from 85 (73%) of the 116 patients who were approached to provide their data. Of these, 8 (9%) patients were excluded because of bilateral symptomatology, and 9 (11%) were excluded because they were able to walk independently at the time of admission. The median LOS in the inpatient stroke rehabilitation facility was 8 (IQR 10) weeks. Table 1 summarizes characteristics and admission data collected for the 68 participants who contributed to these analyses.

Table 1.

Participants Characteristics on Admission (N = 68)

Characteristic	No. (%) of participants*
Age, y, median (IQR)	57 (16)
Gender, male	49 (72)
Side of lesion
Right	44 (65)
Left	23 (34)
Bilateral	1(2)
Infarct	56 (82)
Haemorrhage	12 (18)
Days between stroke onset and admission, median (IQR)	14 (16)
Stroke syndrome
TACS	33 (49)
PACS	19(28)
POCS	9(13)
LACS	7(10)
Continent (n = 62)
Bladder	37 (60)
Bowel	37 (60)
BBS (out of 56), median (IQR)	7 (18)
FIM (out of 126), median (IQR)	60 (27)
CMSA, median (IQR)
PC (out of 7)	3 (2)
LC (out of 7)	3 (2)
Al (out of 100)	32 (24)
Proprioception, median (IQR)
At ankle (out of 5); n = 56	3 (2)
At MTPJ (out of 5); n = 58	3 (4)
Sensation, light touch LL (out of 5), median (IQR); n = 56	3 (2)
4PPS (out of 3), median (IQR)	2 (2)

^*Unless otherwise noted.

IQR = interquartile range; TACS = total anterior circulation stroke syndrome; PACS = partial anterior circulation stroke syndrome; POCS = posterior circulation stroke syndrome; LACS = lacunar stroke syndrome; BBS = Berg Balance Scale; CMSA = Chedoke-McMaster Stroke Assessment Impairment Inventory; PC = postural control; LC = leg control; Al = Activity Inventory; MTPJ = metatarsophalangeal joint; LL = lower limb; 4PPS = Four-Point Pusher Score.

At the time of discharge, 33 of 68 (49%) individuals could ambulate independently. The results of the regression analyses are presented in Table 2. Briefly, in the univariate models, the variables that influenced the odds of IW-DC were BBS, CMSA–AI, CMSA–PC (moderate and severe impairments), CMSA–LC (severe impairment only), FIM, impairment in proprioception at ankle or metatarsophalangeal joint, impairment in sensation in the lower limb, and pushing behaviour. In the multivariable model, the only variable retained was the BBS (OR 1.23; 95% CI: 1.02, 1.49). To interpret the 95% CI for each odds ratio, only those that do not traverse 1 are considered statistically significant. For example, for every 1-point increase in BBS score, there is an increase in the odds of IW-DC of 1.23. Because the 95% CI excluded 1, this is a significant result.

Table 2.

Results of Logistic Regression Using Variables Collected on Admission to Predict Independent Walking on Discharge

Admission variable	Univariate model,OR (95% Cl)	Adjusted model (LOS), OR (95% Cl)
BBS	1.14 (1.07, 1.22)	1.19 (1.07, 1.31)
CMSA
Al (per 1-unit increase in score)	1.09 (1.04, 1.14)	1.11 (1.04, 1.19)
LC, mild (ref.; score 5–7)	1(-)	1(-)
LC, moderate (score 3–4)	0.26 (0.05, 1.44)	0.26 (0.04, 1.47)
LC, severe (score 1–2)	0.05 (0.01, 0.29)	0.02 (0.00, 0.18)
PC, mild (ref.; score 5–7)	1(-)	1(-)
PC, moderate (score 3–4)	0.10 (0.01, 0.86)	0.10 (0.01, 0.95)
PC, severe (score 1–2)	0.03 (0.00, 0.29)	0.04 (0.00, 0.45)
FIM (per 1 -unit increase in score)	1.06 (1.02, 1.09)	1.05 (1.01, 1.10)
Proprioception
At ankle, no impairment (ref.)	1(-)	1(-)
At ankle, evidence of impairment	0.21 (0.07, 0.66)	0.27 (0.07, 0.96)
At MTPJ, no impairment (ref.)	1(-)	1(-)
At MTPJ, evidence of impairment	0.08 (0.02, 0.38)	0.09 (0.02, 0.49)
Sensation
Light touch LL, no impairment (ref.)	1(-)	1(-)
Light touch LL, evidence of impairment	0.20 (0.06, 0.65)	0.26 (0.08, 0.90)
4PPS
No pushing (ref. 0 out of 3)	1(-)	1(-)
Evidence of pushing (1–3 out of 3)	0.06 (0.01, 0.31)	0.08 (0.01, 0.45)

LOS = length of stay; BBS = Berg Balance Scale; CMSA = Chedoke-McMaster Stroke Assessment Impairment Inventory; Al = Activity Inventory; LC = leg control; ref. = reference; PC = postural control; MTPJ = metatarsophalangeal joint; LL = lower limb; 4PPS = Four-Point Pusher Score.

Figure 1 presents the ROC curve for the BBS as a predictor of IW-DC. The area under the curve was 0.81 (95% CI: 0.71, 0.92). The optimal operating point on the BBS that discriminated IW-DC from requiring physical assistance to walk on discharge was 14 (95% CI: 9, 21); sensitivity was 0.73 (95% CI: 0.63, 0.82), and specificity was 0.89 (95% CI: 0.81, 0.94).

Figure 1.

ROC curve showing BBS scores to predict IW-DC.

ROC = receiver operating characteristic; BBS = Berg Balance Scale; IW-DC = independent walking on discharge.

Discussion

Our study reports data for 68 stroke survivors, admitted for inpatient stroke rehabilitation, who were unable to walk independently at the time of admission. These analyses suggest that the BBS provides information on the odds of IW-DC. Although the BBS was the only variable retained in the multivariable model, this result should be seen as hypothesis generating. That is, our sample size limits our certainty around the result, and our analyses require repetition in a much larger sample. A BBS score of 14 points or more on admission had the best capacity to discriminate between those who achieved, versus those who did not achieve, IW-DC, irrespective of their LOS in the rehabilitation facility.

Systematic reviews have suggested the following physiotherapy clinical variables to be predictors of walking post-stroke: severity of sensory and motor dysfunction of the affected leg, leg power, initial disability in activities of daily living and ambulation, and sitting balance.^1,9–11 Evidence is emerging that standing and functional balance are also highly important and may even have superior capacity compared with other admission measures in distinguishing those who will achieve independence with walking.^3,12,13 Our data support this contention.

Two earlier studies have explored the BBS as a predictor of walking independence.^3,13 The first demonstrated that an admission BBS score of 13 points or more could predict independent walking when assessed after patients had spent 3 months on a rehabilitation ward. Nevertheless, that study did not enter the BBS into a multivariable model; therefore, it did not establish whether the BBS had contributed information over and above other measures. It also used the set time point of 3 months in its prediction model.¹¹ The second study found that when compared with measures of stroke type, stroke severity, cognition, and depression in a multivariable model, a BBS score of 12 points or more had superior capacity to predict IW-DC after patients had spent 4 weeks in an inpatient rehabilitation hospital. That study also used a set time point (4 wk).³

The data from our study extend these findings, but to our knowledge our study is the first to find that the BBS is both (1) the most useful measure of predicting IW-DC from a suite of physiotherapy measurements and (2) a predictor of IW-DC from an inpatient rehabilitation facility, independent of LOS.

We think it is important to point out that the BBS is relatively quick to administer (10–20 min), has acceptable reliability and validity in the stroke population, has excellent sensitivity to change,¹⁴ and is strongly recommended for use as a balance outcome measure in the American Physical Therapy Association’s Core Set of Outcome Measures for Adults with Neurologic Conditions undergoing rehabilitation.¹⁵ For adults admitted to a rehabilitation facility who present with unilateral impairments after a stroke and who are unable to ambulate independently, we recommend that physiotherapists prioritize using the BBS to gain information about their likelihood of achieving IW-DC.

Predicting IW-DC is clinically important because it will enable therapists to have enhanced communication with patients and their families (with goal setting and expectation management) and ensure more prepared and strategic discharge planning with the team. For example, if it seems unlikely that a person will achieve IW-DC, the team can commence home modifications early (delays in making home modifications are often lengthy discharge barriers), organize equipment and care providers, and train family members in transfer strategies or gait facilitation. If a person is likely to achieve IW-DC, the treatment goals can focus on gait retraining, and unnecessary home modifications or extensive equipment may be avoided.

Our study had several limitations. First, arguably the most important one was the somewhat modest sample size. This limited our statistical power (increased risk of type II error) and compromised the level of precision that we could provide in our estimate of the odds ratios. Also, our finding that the BBS was the only variable retained in the multivariable model may be a Type I error. Second, our results do not extend to those with bilateral symptomatology. Third, because we did not follow the participants beyond discharge, we can only comment on the capacity of the BBS to provide information on walking at the time of discharge from our rehabilitation facility. Finally, although it is possible that the BBS may predict other parameters related to walking, such as walking speed, investigating those relationships was beyond the scope of this study.

Conclusion

In conclusion, a number of measures collected on admission influenced the odds of IW-DC; however, only the BBS remained a significant predictor in a multivariable model. A BBS score of 14 points or more on admission had the best capacity to predict those who would achieve IW-DC.

Key Messages

What is already known on this topic

Early prediction of independent walking at discharge (IW-DC) after stroke in the rehabilitation setting is clinically important for communication, guiding rehabilitation, and discharge planning. Currently, there is a lack of consensus on which physiotherapy admission outcome measures are able to predict, or best predict, IW-DC.

What this study adds

Although multiple physiotherapy admission outcome measures predicted whether stroke survivors would achieve IW-DC, regardless of their LOS in rehabilitation, our data suggest that that none of the measures provided information over and above the Berg Balance Scale (BBS). These results suggest that clinicians should prioritize administering the BBS, and the cutoff score of 14 or more on admission, to predict whether a patient is likely to achieve IW-DC.