Development and Validation of the Balance Outcome Measure for Elder Rehabilitation

Phase 1 

Phase 1 included item selection and was designed to establish content validity and establish practical applicability. To achieve this, we used a cross-sectional survey and expert panel. Participants included rehabilitation physiotherapy (PT) service providers who were members of the Queensland Rehabilitation Physiotherapists Network (QRPN). The QRPN is a coalition of PT practitioners from publicly and privately funded PT services across Queensland, Australia. The expert panel was made up of 8 senior clinical physiotherapists with 5 to 20 years of clinical experience from QRPN member sites including metropolitan and regional and public and private facilities.

Information on balance measures currently used, their ease of applicability, and clinicians’ willingness to use them was sought through the cross-sectional survey. Twenty-two QRPN member sites were provided with a list of 12 measures. For each, facilities were asked if the measure was currently used at that facility. If the measure was not currently used, facilities were asked to indicate if they would be willing to use that measure as a component of their standard care if it was a part of this state-wide dataset. An option for facilities to nominate measures other than those listed that they used and would be willing to use in the state-wide dataset was also provided.

The survey was distributed in January 2005. Results were viewed by the expert panel, which then identified 4 single-item tests for inclusion in the outcome measure to be constructed. These measures were selected on the criteria that they were already used by responding facilities or that the responding facilities indicated that they would be happy to use the measure in future (practical applicability) and that the tests selected collectively examined a diverse range of underlying domains of the overall standing balance construct (content validity). These measures would then form the nucleus of the core dataset being collected by participating sites in phase 2 of this research program.

Phase 2 

Phase 2 included item scaling and attempted to establish the instrument’s sensitivity to change, to establish its internal consistency, and to identify minimum clinically significant change criterion. To achieve this, we used a multicenter prospective cohort, with split-sample validation.

Participants included people admitted consecutively for rehabilitation PT (inpatient, outpatient, domiciliary) at a participating site across 2 states in Australia. Patients with spinal cord injury who were wheelchair bound and patients with recent amputation (and had not yet received a prosthesis) were excluded from this trial because measures of standing balance were deemed to be irrelevant to their care at this stage.

The 4 measures selected from phase 1 of this trial included the step test,14 the Timed Up & Go (TUG) test,24 the FRT,19 and timed static standing with eyes closed and feet together (a component of the CTSIB).15, 16 These measures, selected by the expert clinician workgroup consulted in this research, were considered representative of the essential domains within the standing balance construct (static, dynamic, function) and that the measurement of these domains was essential to a “global” measure of standing balance.

The clinometric properties of each of these tests has previously been examined by using older adults with a range of clinical conditions and in a range of clinical settings.13, 14, 16, 18, 19, 24, 25, 26, 27, 28 After 2 practice trials, only 1 test trial of the FRT was recorded because (1) previous research indicates that although agreement varies, the correlation between the first trial and the average of 3 trials is excellent (intraclass correlation coefficient=.90),29 and (2) the expert workgroup noted that 1 test trial of the FRT was more reflective of current clinical practice. The static timed standing with eyes closed retained its 3 trials; however, if the first trial recorded the maximum score of 30 seconds, then the subsequent 2 trials were automatically scored as 30 seconds as per the original procedure.16, 30 The step test is a measure of a dynamic single-limb stance task in which the foot is placed onto the top of a 7.5-cm block and back to the floor as many times as possible in 15 seconds.14 Participants performed the step test with each foot, and the average of 2 step test trials (1 with each foot) was recorded. The TUG test is a dynamic and functional performance measure commonly used in rehabilitation practice as a measure of overall mobility, balance, and gait disorders.24, 31 In addition, the TUG samples the ability of an individual to turn 180° while maintaining the upright position and the ability to maintain the upright standing position immediately after transition from a seated posture.

Patient demographic data were collected along with measures of cognition (Mini-Mental State Examination32) and function (FIM instrument motor score33 or Modified Barthel Index [MBI]34) if these were collected as a part of routine care at each site. MBI scores were converted to FIM motor scores following a conversion system previously developed for the Australian setting.35 Physiotherapists also recorded if improving standing balance was the first-, second-, third-, fourth-, or lower-ranked goal on their prioritized treatment plan for that patient. Standard PT practice is to develop a prioritized problem list so that the problems of greatest importance (that can be addressed by a physiotherapist) can be focused on in treatment sessions. From this, prioritized treatment goals are formed.

The project investigative team contacted all member facilities of the QRPN along with other facilities with which professional contacts were held. Sites collected data for between 1 and 12 months from May 2005 to May 2006. Each site that participated used the 4 balance measures collected as a part of their routine admission and discharge assessments. Each site was asked to collate data from their routine assessments of consecutive patients meeting the inclusion and exclusion criteria for 6 months. A site representative was selected at each participating site. These representatives were responsible for disseminating the study protocol manual to all physiotherapists at their sites and to ensure consistent data collection in line with the instructions set out in the study protocol manual. Site representatives were also responsible for collating data at each site before forwarding data to the project manager (TH) for entry and analysis. All data were deidentified with subject- and site-specific coding before being forwarded.

Ethics approval for collation of these data was provided from the Medical Research Ethics Committee of the University of Queensland, the Queensland Health Human Research Ethics Committee, and the hospital human research ethics committees of each of the participating sites.

Phase 3 

Phase 3 established construct (criterion) validity. To do this, we used a prospective cohort (single site) design. Patients were selected from 1 site participating in phase 2 of this research program.

Criterion reference tests were the Modified Elderly Mobility Scale (MEMS),40 a measure of mobility among older adults, and the FIM motor score,33 a measure of independence in performing functional tasks. Because standing balance is a construct closely related to mobility and function,4, 23, 41, 42 these measures were considered appropriate criterion reference tests with which to assess the validity of the BOOMER. The MEMS,40 commonly used within Queensland, Australia, is a measure of global function comprising 8 items including lying to sitting, sitting to standing, walking, functional reach, and stairs. The MEMS has shown criterion-related validity with the FIM motor score and interrater and test-retest reliability.

The MEMS and FIM motor scores were collected at this site at patient admission and discharge along with demographic details and other test items that constituted the BOOMER. This was a component of routine care at this site. Data were not examined until after formation of the BOOMER so that knowledge of the data distribution would not influence the design of the BOOMER.

Phase 1 

Responses to the phase 1 survey were received from 16 of the 22 QRPN member sites contacted. Eight tests (step test, TUG test, FRT, static standing with eyes closed and feet together, static tandem stance, single-leg stance, standing with feet together and eyes open, ten-meter walk test) were currently used by at least 11 of the 16 sites. One site responded that they would not be willing to use 2 of the tests (static tandem stance, single-leg stance) as a part of their standard practice. From this list, the expert panel selected the step test, TUG test, FRT, and static standing with eyes closed. The expert panel deemed that these tests encompassed a range of static and dynamic tasks incorporating a range of functional activities including both motor and sensory challenges to balance and that these tests adequately sampled the range of domains that constitutes the overall standing balance construct while preserving the desire for the tool to be practically applicable.

Phase 2 

Data from 1769 participants across 18 sites (16 from Queensland, 2 from Victoria) were collated. Data from 1 site did not include 1 of the 4 balance measures and were not allocated to either the development or validation dataset. Patient demographics of these datasets are presented along with the demographics of patients contributing data to phase 3 (table 1).

Table 1. Baseline Demographics of Participants in Development and Validation Datasets and for the Single Site in Phase 3 of the Study

	Demographic	Development	Validation	Phase 3
	N	683	784	272
	Age (y)	74±14	74±13	75±14
	Men (n)	271(39.6)	315(40.2)	105(38.6)
	Diagnosis (n)
	Stroke	155(24)	176(23)	66(24)
	Other neurologic condition⁎	88(13)	52(7)	37(14)
	Elective orthopedic	34(5)	72(9)	2(1)
	Other orthopedic†	203(30)	200(26)	82(30)
	Other set diagnoses combined‡	125(18)	188(24)	53(19)
	Other geriatric impairment§	67(10)	94(12)	32(12)
	Aid before admission nothing or single point stick	527(77)	550(70)	194(71)
	Living arrangements before admission community dwelling	651(96)	721(92)	255(94)
	Mini-Mental State Examination (/30)	25±5	24±6	24±5
	Admission FIM motor (/91)	59±17	54±17	58±17
	Discharge FIM motor (/91)	75±15	73±16	75±14
	Length of stay (d)	40±32	32±23	50±32

NOTE. Values are mean±standard deviation or frequency (%).

⁎ Refers to neurologic conditions other than stroke (eg, Parkinson’s disease). † Refers to orthopedic diagnoses that were not elective, for example, fractured neck of femur. ‡ A merged category including the categories of amputation, spinal cord injury, arthritis, pain, musculoskeletal, major multiple trauma, cardiac, pulmonary, and burns. § Category for conditions associated with older age not adequately described by the above categories (eg, deconditioning and poor mobility).

To scale each item, 4 cutoff points were selected creating 5 ordinal categories (scores range, 0−4). An overall score was created by summing the scores for each item (scores range, 0−16). This was initially conducted by using the admission dataset for inpatients in the development dataset on the basis of generating a skewed data distribution toward poorer scores. This was then assessed graphically by using frequency histograms. The cutoff points were applied to the discharge scores of the outpatient and domiciliary patients of the development dataset combined to examine whether a skew toward higher scores had been attained. This was again checked graphically, and modifications to the selected cutoff points were made when deviations from the desired properties were observed (for both individual items and the overall scale) before repeating the process. Multiple iterations of this process were pursued before settling on the final set of cutoff points (table 2), which the authors determined best satisfied the 3 rationale for item scaling.

Table 2. Cutoff Points for BOOMER Items

	Test	0	1	2	3	4
	Step test (average no. of steps)	Unable	>0–5	>5–8	>8–12	>12
	TUG test (s)	Unable	≥30	<30–20	<20–10	<10
	FRT (cm)	0	>0–15	>15–20	>20–30	>30
	Static standing eyes closed (s)	Unable	>0–30	>30–60	>60–<90	90

The distribution of each item for inpatient admission and outpatient and domiciliary discharge scores arising from the development and validation datasets are presented (fig 1). These frequency histograms indicated that the primary rationale for item scaling had been met as skew toward poorer scores for admission inpatient data, and skew toward superior balance scores for discharge outpatient and domiciliary data were evident in both individual items and the overall BOOMER scale (fig 2). One possible exception (the static standing with eyes closed and feet together item) was evident because it showed a bimodal distribution for admission inpatient scores. A mild floor effect was also apparent when examining the admission inpatient datasets. Comparison between development and validation datasets indicated that the data distributions were consistent.

View Large Image Download to PowerPoint

Fig 1. Distribution of scores for each item from inpatient admission and outpatient and domiciliary discharge data.

View Large Image Download to PowerPoint

Fig 2. Distribution of BOOMER scores from inpatient admission and outpatient and domiciliary discharge data.

The internal consistency of the items (α) ranged from .87 to .89 for the admission and discharge assessments of the development and validation datasets. Interitem correlations (r) ranged from .77 to .51, with the lowest interitem correlations consistently being found in those involving the static standing with eyes closed and feet together test (table 3). Factor analysis resulted in single-factor solutions for admission and discharge data of development and validation datasets (analyzed separately). Uniqueness (1 – communality) varied for each of the following items: step test (.29−.35), TUG test (.27−.35), FRT (.32–.40), and static stance eyes closed (.48−.55).

Table 3. Interitem Correlation (Spearman ρ) for BOOMER Items at Admission and Discharge Across Development and Validation Datasets

				Step Test	TUG Test	FRT
	Development	Admission	TUG test	.71
			FRT	.72	.73
			Static standing eyes closed	.65	.64	.68
		Discharge	TUG test	.77
			FRT	.66	.74
			Static standing eyes closed	.52	.52	.53
	Validation	Admission	TUG test	.70
			FRT	.67	.65
			Static standing eyes closed	.59	.61	.66
		Discharge	TUG test	.77
			FRT	.65	.69
			Static standing eyes closed	.56	.51	.52

Point estimates and 95% CIs for each of the 5 approaches to calculating the minimum clinically significant difference are presented (fig 3). For the second approach (comparing those for whom standing balance was rated as a higher- or a lower-ranked goal), the cutoff point of “fourth or lower-ranked problem” was identified and used. The differences ranged between 2 and 6 points on the BOOMER scale, with 95% CIs overlapping between 3 to 4 points. The minimum clinically significant differences identified by using each approach were consistent between development and validation datasets.

View Large Image Download to PowerPoint

Fig 3. Point estimates and 95% CIs for minimum clinically significant differences for the BOOMER calculated from development and validation datasets. Data presented are point estimate and 95% CI of the mean difference between groups. *Discharge BOOMER – admission BOOMER (within subjects), for inpatients discharged to the community. †Admission BOOMER for subjects for whom improving standing balance was a fourth- or lower-ranked treatment goal – admission BOOMER for subjects who had improving standing balance as 1 of their top 3 treatment goals (between subjects). ‡Admission BOOMER for outpatient or domiciliary patients – admission BOOMER for inpatients (between subjects). §Discharge BOOMER for patients whose living status at discharge was the community – discharge BOOMER for patients whose living status at discharge was not the community (eg, nursing home, hospital) (between subjects). ∥Change in BOOMER (discharge – admission) for people whose improvement in FIM motor score was 10 points or more – change in BOOMER (discharge – admission) for people whose improvement in FIM motor score was less than 10 points (between subjects).

Phase 3 

The BOOMER showed good correlation with both the motor FIM (admission data, ρ=.73; discharge data, ρ=.72) and the MEMS (admission data, ρ=.88; discharge data, ρ=.83). Each of these associations was statistically significant (P<.001).

Study Limitations 

The analysis did raise a concern that the BOOMER may be subject to a floor effect and, thus, potentially be unable to discriminate between subjects with greatly diminished standing balance ability and be insensitive to change across the rehabilitation period. In considering this, it should first be noted that this study received data from facilities and services primarily providing inpatient rehabilitation. In Australia, geriatric inpatient rehabilitation is commonly provided to older adults who have become medically stable after a recent illness or surgery and require an extended rehabilitation period before returning to their place of residence. Consequently, intensive therapy focused on improving standing balance commonly commences after admission to the rehabilitation wards. It is not surprising then that many people do not have sufficient standing balance to register more than the minimum score on this standing balance scale at this time.

Another study limitation was that we considered only existing standing balance measures. It is plausible that as-yet-undeveloped measures of standing balance could better measure the standing balance construct. However, pursuing only existing assessment strategies, particularly those already in use in clinical practice, is likely to enhance the practical applicability and clinician acceptability of the BOOMER.