The Gait Disorientation Test: A New Method for Screening Adults With Dizziness and Imbalance

Abstract

Objective:

To develop and evaluate a new method for identifying gait disorientation due to vestibular dysfunction.

Design:

The gait disorientation test (GDT) involves a timed comparison of the ability to walk 6.096 m with eyes open versus eyes closed. In this prospective study, participants were grouped based on vestibular function. All participants completed a clinical examination, self-report- and performance-based measures relevant to vestibular rehabilitation, and the tasks for the GDT. Vestibular-impaired participants underwent the criterion standard, videonystagmography and/or rotational chair testing.

Setting:

Ambulatory clinic, tertiary referral center.

Participants:

Participants (N=40) (20 vestibular-impaired, 30 women, 49.9±16.1years old) were enrolled from a convenience/referral sample of 52 adults.

Main Outcome and Measure(s):

We determined test-retest reliability using the intraclass correlation coefficient model 3,1; calculated the minimal detectable change (MDC); examined concurrent validity through Spearman correlation coefficients; assessed criterion validity with the area under the curve (AUC) from receiver operator characteristic analysis; and computed the sensitivity, specificity, diagnostic odds ratio (DOR), likelihood ratios for positive (LR+) and negative (LR−) tests, and posttest probabilities of a diagnosis of vestibulopathy. The 95% confidence interval demonstrates measurement uncertainty.

Results:

Test-retest reliability was 0.887 (0.815, 0.932). The MDC was 3.7 seconds. Correlations with other measures ranged from 0.59 (0.34, 0.76) to −0.85 (−0.92, −0.74). The AUC was 0.910 (0.822, 0.998), using a threshold of 4.5 seconds. The sensitivity and specificity were 0.75 (0.51, 0.91) and 0.95 (0.75, 1), respectively. The DOR=57 (6, 541.47), LR+ =15 (2.18, 103.0), and LR− =0.26 (0.12, 0.9). Positive posttest probabilities were 89%–94%.

Conclusions and Relevance:

The GDT has good reliability, excellent discriminative ability, strong convergent validity, and promising clinical utility.

An estimated 33 million American adults annually experience imbalance or dizziness.^1,2 Vestibular dysfunction is a common cause of impaired balance.^3–5 Presently, improving the diagnosis, treatment, and prevention of imbalance and dizziness is a top priority in the United States.⁶ Screening persons for vestibular dysfunction is an important step in determining whether further diagnostic workup, referral to specialty providers, and/or therapies for vestibular dysfunction may be indicated.⁷

Vestibular disorders result in impairments in gaze stability,⁸ spatial orientation,⁹ postural control,¹⁰ and gait.¹¹ Tests of vestibular-ocular reflex function,^8,12,13 spatial orientation,^4,5 standing balance,^14–17 and walking^18–20 have been used to screen for the presence of vestibular dysfunction.^21–26 Despite the fact that many vestibular-impaired persons perform poorly on these tests, their screening utility is limited by suboptimal discriminative validity.^21–26

Persons with peripheral,^27–35 central,^36–38 and age-related³⁹ vestibular dysfunction also have impaired spatial navigation, visual dependence,⁴⁰ and gait disorientation, that is, difficulty walking under challenging sensory conditions. Path integration tasks, during which a person walks while blindfolded along a memorized path or toward a previously viewed target, are used to assess accuracy in spatial navigation.⁴¹ Prior studies of vestibular-impaired adults used linear^27,28,33,34 or geometrically shaped^42–44 paths. Impairment in spatial navigation is responsive to vestibular rehabilitation^27,28,31,45; however, deficits may persist.^46,47

To our knowledge, spatial navigation assessments have not been used to screen for vestibular dysfunction. The objectives of this study were to develop and validate a screening method for gait disorientation. We hypothesized that greater difficulty navigating with eyes closed compared to with eyes open would discriminate vestibular impaired from healthy adults.

Methods

The main dataset was collected prospectively at the University of Wisconsin (UW)-Madison (site 1) from adults who were either healthy or had documented vestibular loss (fig 1). The Gait Disorientation Test (GDT), which was developed from specific items from the Functional Gait Assessment (FGA),⁴⁸ was the index test. Vestibular-impaired adults underwent the criterion tests (videonystagmography [VNG] and/or rotational chair [RC] testing)⁴⁹ prior to enrollment (supplemental appendix S1, available online only at http://www.archives-pmr.org/). All participants were 20–79 years old; spoke English fluently; were independently functioning; did not have a history of neurologic, musculoskeletal, vision, or pain conditions; had normal, bilateral lower-extremity strength⁵⁰ and sensation⁵¹; could stand for 20 minutes without sitting and walk 6.096 m unaided; could abstain from alcohol and withhold any antivertigo, sedative, and narcotic or barbiturate medications for 48 hours prior to study visits; and were not pregnant during the study.

Fig 1.

Data collection flowchart. These data are reported for time point A in the main study. The criteria for abnormal or positive results on VNG, RC, and the GDT are provided in supplemental appendix S1. All healthy and vestibular-impaired participants underwent a thorough clinical examination to screen for previously undetected vestibular dysfunction in healthy participants or to confirm vestibular loss in impaired participants, as well as to rule out somatosensory, visual, musculoskeletal, and neuromuscular conditions that might interfere with testing (see supplemental appendix S1). Abbreviation: Hx/o, history of musculoskeletal or neurological exclusion criteria.

The GDT was externally validated using a second dataset from a study conducted at the University of Utah (site 2) in which the effects of vestibular loss on gait were investigated.⁵² Adults who were 4 to 8 weeks post–unilateral, vestibular schwannoma resection, and healthy adults with no history of vestibulopathy were recruited. Participants in this study were 18–70 years old, able to walk unaided, and had no history of lower extremity injuries within the previous 12 months. Persons with unstable medical conditions, for example, angina, seizures, were excluded.

Each study was approved by an institutional review board from the site at which the data were collected and was conducted in accordance with the Declaration of Helsinki.⁵³ Informed consent was obtained from all participants.

The assessor at site 1 (C.R.G.) was a physical therapist with 26 years of experience in vestibular rehabilitation and was aware of group assignment and VNG/RC results. The Activities-specific Balance Confidence Scale (ABCS),⁵⁴ Dizziness Handicap Inventory (DHI),⁵⁵ FGA,⁴⁸ GDT, and Sensory Organization Test (SOT)^56,57 were collected at 2 time points (A and B) that were 8 weeks (±3days) apart. The Five-times Sit-to-stand Test (FTSTST), horizontal head impulse test (hHIT), head-shaking nystagmus (HSN), and non-instrumented dynamic visual acuity test (DVAT) were administered at time point A only (see supplemental appendix S1).

ABCS⁵⁴:

This 16-question survey measures how confident the responder feels about their balance in specific situations. The average score ranges between 0% and 100% with higher scores indicating greater balance-related confidence.

DHI⁵⁵:

This 25-question survey is used to assess how often dizziness has affected the person’s function. The DHI total score, which ranges from 0 to 100, records perceived handicap with higher scores indicating greater perceived handicap.

FGA⁴⁸:

This standardized test includes 10 gait-related balance tasks. The total score for the FGA ranges from 0 to 30 and higher scores indicate better balance. The GDT was created using FGA items 1 (FGA1: walking, eyes open) and 8 (FGA8: walking, eyes closed). The GDT result was calculated by subtracting the time to complete FGA1 in seconds from the time to complete FGA8.

Participants at site 1 performed the FGA in their preferred, flat-soled shoes along a dedicated path. Markings designating the beginning, ending, and lateral boundaries of a 6.096-m-long by 1.067-m-wide scoring section are integrated into the design pattern of the flooring. We used visual observation and a stopwatch to record performance on timed tasks. Participants began walking at their preferred speed prior to entering the scoring section and continued walking until they were told to stop at a point beyond the end of the scoring section. Timing began when the leading foot crossed the marking indicating the beginning of the scoring section and stopped when the leading foot crossed the marking indicating the end of the scoring section.

Participants at site 2 performed the FGA while wearing their preferred, flat-soled shoes and a suite of triplanar, inertial measurement units.^a The times to complete FGA1 and FGA8 were calculated by graphing acceleration data from the sternal inertial measurement unit that was processed through a second-order, lowpass, Butterworth filter with a cutoff of 6.0 Hz and zero phase shift and then capturing the time between the manually selected beginning and end of each walking period.

GDT:

A detailed description of the GDT is provided in supplemental appendix S2 (available online only at http://www.archives-pmr.org/). The GDT should be performed in a quiet area along a 10-m pathway that includes a 6.096-m by 1.067-m scoring section and using a stopwatch for timing each task.

SOT^56,57:

The SOT^b is used to evaluate sensory contributions to standing balance. Overall performance on the SOT is represented by the composite score (COMP). The manufacturer-calculated somatosensory score (SOM) provides an objective expression of Romberg’s sign.¹⁴

Statistical considerations

The reliability of the GDT, FGA1, and FGA8 was assessed with the intraclass correlation coefficient model 3,1⁵⁸ by comparing performances on these tasks at time points A and B. The minimal detectable change (MDC) values for the GDT, FGA1, and FGA8 were calculated using 1.96*√2*SEM,⁵⁹ where the SEM was estimated by taking the square root of the within-subjects variance from a 2-way, random-effects, analysis of variance. Spearman correlation coefficient method was used to assess concurrent validity between the GDT and ABCS, DHI, DVAT (horizontal dynamic visual acuity test [hDVAT], vertical dynamic visual acuity test [vDVAT]), FGA, and COMP.

The discriminative ability of the GDT was assessed using receiver operator characteristic (ROC) analysis.⁶⁰ Similarly, ROC analyses were completed for FGA1, FGA8, FTSTST, hDVAT, vDVAT, COMP, and SOM. The threshold for the hDVAT and vDVAT was a degradation in visual acuity of 3 lines when comparing static to dynamic visual acuity.⁸ Otherwise, Youden’s method was employed to calculate the optimal threshold. Subsequently, the sensitivity and specificity for these tests were computed based on contingency tables constructed by using their thresholds. Separately, contingency tables for the hHIT and HSN tests were constructed based on dichotomizing the results.

The diagnostic odds ratio (DOR) and likelihood ratios for a positive (LR+) and a negative (LR−) test were calculated.⁶¹ In addition, the positive and negative posttest probability was determined for the GDT using the LRs we calculated and published data regarding the prevalence of vestibular dysfunction in specific clinical settings⁶² and the general population.³

An ROC analysis was used to compute the area under the curve (AUC) (95% confidence interval [95% CI]) for the GDT from the external dataset. The optimal threshold for the GDT that was obtained from site 1 was used to create a contingency table of the external dataset for subsequent calculation of diagnostic performance measures.

All analyses were conducted post hoc using R version 3.5.^c The GDT data were not normally distributed.

Results

Twenty healthy adults and 20 adults with chronic (3mo to 20y) vestibular hypofunction (left: n=2, right: n=12, bilateral: n=6) participated at site 1. Vestibular-impaired participants were significantly older and had worse performance on the GDT than healthy participants (table 1). Based on post hoc modeling, age was not a significant factor in GDT times (P=.954) (supplemental appendix S3, available online only at http://www.archives-pmr.org/). One adverse event, a soleus tear that occurred while the participant was bicycling home from work, led a healthy adult to withdraw.

Table 1.

Study population characteristics

		Site 1: Main Study
Variable	All	Healthy (n=20)	Impaired (n=20)	P Value
Age (y)	49.9 (16.1)	38.7 (11.7)	61.1 (11.2)	<.001*
Sex (women)	30 (75.0%)	16 (80.0%)	14 (70.0%)	.7†
BMI	26.1 (4.4)	26.4 (4.7)	25.8 (4.3)	.71*
GDT	3.7 (1.4–7.7)	1.6 (1.1–2.7)	7.7 (4.5–12.6)	<.001‡
		Site 2: External Validation Study
	All	Healthy (n=16)	Impaired (n=13)	P Value
Age (y)	39.4 (14.8)	32.8 (12.3)	47.6 (13.7)	.006*
Sex (women)	18 (62.1%)	9 (56.2%)	9 (69.2%)	.7†
BMI	25.7 (5.0)	22.5 (2.3)	29.7 (4.5)	<.001*
GDT	2.3 (1.0–6.4)	1.1 (0.7–2.1)	7.3 (3.1–11.1)	<.001‡

NOTE. These data are presented as mean ± SD for age, percentages for sex, and mean ± SD for BMI (BMI: 703*weight/height²). The median (interquartile range) is presented for the GDT because these data were not normally distributed.

Abbreviation: BMI, body mass index.

^*The P values are from 2-tailed t tests.

^†The P values are from a chi-squared test.

^‡The P values are from Wilcoxon rank-sum tests.

Sixteen healthy adults and 13 adults with surgically induced vestibular loss (left: n=8; right: n=5) contributed to the external dataset at site 2. These vestibular-impaired adults were significantly older and had worse performance on the GDT than healthy participants. Healthy participants at both sites performed comparably on the DHI, FGA, and gait speed.

Across all participants at site 1, the test-retest reliability for the GDT, FGA1, and FGA8 ranged from good to excellent (table 2). Reliability for these measures was moderate to good within healthy participants and good for vestibular-impaired participants (table 2). The MDC values were 3.7, 0.74, and 3.6 seconds for the GDT, FGA1, and FGA8, respectively. The GDT was moderately and negatively associated with the ABCS; strongly and positively related to the DHI; moderately and positively associated with performance on the hDVAT, vDVAT, and FTSTST; and strongly and negatively associated with the FGA and COMP (table 3).

Table 2.

Test-retest reliability of the GDT, FGA1, and FGA8 in seconds

Measure	Time Point A	Time Point B	ICC3,1 (95% CI)
	All Participants
GDT	5.7±6.2	4.7±5.0	0.887 (0.815–0.932)
FGA1	4.7±0.8	4.7±0.9	0.818 (0.708–0.889)
FGA8	10.4±6.6	9.5±5.6	0.909 (0.850–0.946)
	Healthy Participants
GDT	2.0±1.4	1.8±1.3	0.449 (0.097–0.702)
FGA1	4.4±0.7	4.3±0.7	0.847 (0.696–0.927)
FGA8	6.4±1.6	6.1±1.6	0.591 (0.285–0.788)
	Vestibular-Impaired Participants
GDT	9.4±6.9	7.7±5.6	0.856 (0.711–0.931)
FGA1	5.1±0.9	5.2±0.9	0.731 (0.495–0.866)
FGA8	14.4±7.2	12.8±6.1	0.878 (0.753–0.942)

NOTE. These data are reported as the mean ± SD and ICC_3,1 (95% CI) for data from the main study. The ICC_3,1 was used because it performs well with data that are not normally distributed. The inclusion of imputed data in the calculation of reliability did not affect the strength of the ICCs; thus, only ICCs based on a complete data set are reported.

Abbreviation: ICC_3,1, intraclass correlation coefficient model 3,1.

Table 3.

Correlation of the GDT with the ABCS, DHI, hDVAT, vDVAT, FTSTST, FGA, and the COMP

	N	ρ (95% CI)	P Value
Patient-reported outcome measures
ABCS (average score)	40	−0.69 (−0.83 to −0.48)	<.001
DHI (total score)	40	0.74 (0.56–0.85)	<.001
Performance-based outcome measures
hDVAT (degradation in acuity)	40	0.61 (0.37–0.77)	<.001
vDVAT (degradation in acuity)	40	0.59 (0.34–0.76)	<.001
FTSTST (time)	40	0.65 (0.43–0.80)	<.001
FGA (total score)	40	−0.85 (−0.92 to −0.74)	<.001
COMP (from the SOT)	40	−0.74 (−0.86 to −0.56)	<.001

NOTE. These data are presented as Spearman ρ (95% CI) for data collected at time point A in the main study. The Spearman correlation coefficient method was used to analyze all of these data due to their nonnormal distribution. The P value indicates the probability of finding a value for ρ within the associated 95% CI.

The optimal threshold for discriminating between healthy and vestibular-impaired adults using the GDT was 4.5 seconds at site 1 (AUC [95% CI]=0.910 [0.822–0.998]) (table 4). The only other measures with an AUC >0.900 were FGA8 and the COMP. The sensitivities of each test ranged from poor to excellent, while the specificities ranged from moderate to excellent (table 5). When the times for FGA8 and the GDT were analyzed in parallel and with either result exceeding the optimal threshold, the sensitivity and specificity were 0.98 and 0.81, respectively. The DORs for all measures except the FTSTST and SOM were significant given that their 95% CIs excluded 1.0 (table 5). The GDT had a high DOR and a strong LR+ value (table 5). In addition, the positive, posttest probability for the GDT was projected to be high across clinical practice settings and the general population (table 6).

Table 4.

Summary of ROC analyses

Test	Threshold	AUC (95% CI)	P Value
GDT	4.5 s	0.910 (0.822–0.998)	NA
FGA1	4.4 s	0.740 (0.584–0.896)	.039
FGA8	7.9 s	0.925 (0.844–1)	.391
FTSTST	8.7 s	0.846 (0.720–0.973)	.353
hDVAT	3 lines lost	0.831 (0.693–0.97)	.302
vDVAT	3 lines lost	0.861 (0.729–0.993)	.560
COMP	75.7	0.932 (0.861–1)	.676
SOM	98.5	0.406 (0.22–0.593)	<.001

NOTE. These data are based on the statistical considerations for data gathered at time point A in the main study. The P values compare the AUC of the GDT to all the other tests using a ROC comparison test.

Abbreviation: NA, not applicable.

Table 5.

Comparison of the diagnostic performance of all screening tests

Test	TP	FP	FN	TN	Sn (95% CI)	Sp 95% CI	DOR (95% CI)	LR+ (95% CI)	LR− (95% CI)
GDT	15	1	5	19	0.75 (0.51–0.91)	0.95 (0.75–1)	57 (6–541.47)	15 (2.18–103.0)	0.26 (0.12–0.9)
GDT (V)	9	3	4	13	0.69 (0.39–0.91)	0.81 (0.54–0.96)	9.75 (1.74–54.52)	3.69 (1.25–10.9)	0.38 (0.16–0.88)
FGA1	17	8	3	12	0.85 (0.62–0.97)	0.60 (0.36–0.81)	8.5 (1.86–38.82)	2.12 (1.2–3.75)	0.25 (0.08–0.8)
FGA8	18	3	2	17	0.90 (0.68–0.99)	0.85 (0.62–0.97)	51 (7.57–343.73)	6 (2.09–17.21)	0.12 (0.03–0.4)
FTSTST	5	0	15	20	0.26 (0.1–0.5)	0.98 (0.8–1)	14.6 (0.75–283.4)	11 (0.65–186.6)	0.76 (0.6–0.98)
hHIT	16	0	4	20	0.79 (0.55–0.93)	0.98 (0.8–1)	150.3 (7.5–2998)	33 (2.11–515.0)	0.2 (0.1–0.5)
HSN	9	0	11	20	0.45 (0.24–0.68)	0.98 (0.8–1)	33.87 (1.8–636.9)	19 (1.18–305.9)	0.56 (0.38–0.8)
hDVAT	14	2	6	18	0.70 (0.46–0.88)	0.90 (0.68–0.99)	21 (3.66–120.37)	7 (1.82–26.89)	0.33 (0.17–0.7)
vDVAT	13	0	7	20	0.64 (0.41–0.84)	0.98 (0.8–1)	73.8 (3.9–1401.6)	27 (1.71–425.4)	0.37 (0.21–0.7)
COMP	17	2	3	18	0.85 (0.62–0.97)	0.90 (0.68–0.99)	51 (7.57–343.73)	8.5 (2.25–32.06)	0.17 (0.06–0.5)
SOM	10	12	10	8	0.50 (0.27–0.73)	0.40 (0.19–0.64)	0.67 (0.19–2.33)	0.83 (0.47–1.47)	1.25 (0.63–2.5)

NOTE. These data are presented as the appropriate value (95% CI) for data collected at time point A in the main study unless otherwise noted. Calculations for the DOR and LR+ were adjusted by adding 0.5 to each cell in the contingency tables for the FTSTST, hHIT, HSN, and vDVAT, because these values would be undefined if they were calculated with 0 values for FP.⁶¹

Abbreviations: FN, false negative; FP, false positive; Sn, sensitivity; Sp, specificity; TN, true negative; TP, true positive; V, values are from the validation data set.

Table 6.

Projected posttest probabilities for the gait disorientation test

Setting (Population)	Pretest Probability (%)	Positive Posttest Probability (%)	Negative Posttest Probability (%)
Primary care*	43	92	17
Emergency department*	34	89	12
Dizziness clinic*	46	93	18
Neurology*	49	94	20
General population†	35	89	12

NOTE. Likelihood ratios from the main study were used in these analyses.

^*Prevalence based on data from Kroenke et al.⁶²

^†Prevalence based on data from Agrawal et al.³

The AUC (95% CI) for the GDT was 0.894 (0.783–1.0) at site 2. The sensitivity, specificity, DOR, LR+, and LR−, as calculated from the validation cohort, differed somewhat from those values obtained in the main analysis (table 5). Chi-squared tests were not statistically significant when comparing the sensitivity (P>.99) and specificity (P=.441) of the GDT across datasets.

Discussion

The GDT provides an objective comparison of a person’s ability to walk within a standardized path with eyes open versus their ability to perform the same task with eyes closed. The result is a composite measure of several determinants of spatial navigation, for example, walking speed,³³ spatial memory,²⁷ and cognitive-perceptual processes.³⁴ In this study, GDT times >4.5 seconds identified persons with vestibular disorders.

When deciding whether to use a new test, one must first consider whether the test accurately and consistently measures what it proposes to measure.⁶³ Our finding of excellent inter-rater reliability for the GDT is consistent with prior studies that found good to excellent reliability for measuring self-selected walking speed with a stopwatch in persons with locomotor impairments⁶⁴ and for assessing spatial navigation in persons with depression⁶⁵ and healthy, older adults.⁶⁶ Low variability within the healthy group may explain the discrepancies in intraclass correlation coefficient values across groups.

The criterion validity of the GDT is excellent. We attribute the discrepancies in the diagnostic performance of the GDT across studies to differences in the vestibular-impaired groups. Adults with chronic unilateral and bilateral vestibular dysfunction were enrolled in the main study, whereas adults with surgically induced, subacute, unilateral involvement comprised the validation cohort.

The GDT has strong convergent validity. Similarly, virtual navigation⁶⁷ and self-selected walking speed⁶⁸ have been found to be valid measures of spatial memory and gait performance, respectively, in adults with and without vestibular dysfunction. The construct validity of the GDT must be corroborated through correlations with established spatial navigation assessments.^{30,32,37,42,69,70}

One must also consider feasibility when deciding whether to use a clinical test.⁷¹ Fritz et al⁷¹ suggest that clinicians need to know whether the test is safe, how easy it is to administer, how easy is it to interpret the results, and if the test is cost effective.

Most patients with vestibular dysfunction can walk short distances over level, indoor surfaces with eyes open unaided. Although walking in the dark (or with eyes closed) is more challenging, participants at site 1 completed FGA8 96% of the time. Thus, the GDT appears to be a safe method of evaluating gait disorientation when conducted by providers who are trained in appropriate guarding techniques.

The GDT could be performed by individuals with a variety of technical and professional backgrounds who have minimal to extensive experience with vestibular-impaired persons. Also, the distance required for the GDT is consistent with the recommended feasible distance for measuring walking speed.⁷² Stopwatch and instrumented methods for measuring self-selected walking speed are shown to be highly correlated and not significantly different.⁷³ We found that diagnostic accuracy for the GDT was similar when stopwatch and instrumented methods were used to measure performance. Thus, using a stopwatch for timing provides a readily available and accurate alternative to instrumented gait testing. Finally, we constructed the GDT using subtraction because we surmised this method would be more easily integrated into clinical practice than calculating a quotient with times.

Because scoring of the GDT is accomplished with the start and stop of a stopwatch, interpretation is objective. Determining whether the result is abnormal in this population is uncomplicated and is accomplished by comparing the GDT time to the 4.5-second threshold.

Because the GDT is inexpensive, fast, and poses minimal risk when patients are appropriately guarded, whether the GDT proves to be cost-effective will depend largely on how clinicians use it to help guide their decision making and whether patients benefit from interventions offered after testing positive. Recent evidence suggests that early vestibular rehabilitation results in reduced dizziness and more complete functional recovery than the standard of care alone.⁷⁴ If the GDT is used as an outcome measure, a change of >3.7 seconds would exceed the measurement error for the test.

Findlay et al⁷⁵ provide the only other publication in which similar methods were used to screen patients. Their walking Romberg sign was based on comparing the ability of 50 persons with a diagnosis of cervical myelopathy to walk 5 m with eyes open versus their ability to do this task with eyes closed. The criteria for a positive walking Romberg sign were observed swaying, subjective reports of feeling unstable, or the inability to complete the walk with eyes closed. Although just 34% of patients with radiographically confirmed cervical myelopathy had a positive, traditional Romberg sign, 75% of patients were found to have a positive, walking Romberg sign. Our results extend those of Findlay and colleagues by demonstrating that comparing gait with eyes open to eyes closed can discriminate healthy persons from those who have gait disorientation resulting from a loss of sensory input to the central nervous system.

The sensitivity of the GDT (75%) is like that of the walking Romberg sign.⁷⁵ Because the criteria for an abnormal GDT is objective, we anticipate that it would have superior reliability compared to the walking Romberg sign. The sensitivity and specificity of detecting vestibular dysfunction with the GDT are comparable to the ability of the Walking Trail-making Test to detect mild cognitive dysfunction,⁷⁶ but less robust than the ability to identify impaired spatial navigation in concussed collegiate athletes using a computerized neuropsychological evaluation.³⁷

Though walking, eyes closed (FGA8) had 90% sensitivity, there is strong rationale for including walking, eyes open (FGA1). Vestibular loss effects spatial navigation with eyes open and eyes closed.^28,33,46 Conducting the eyes open task may help patients develop the spatial memory needed to complete the eyes closed task^{34,43,77–81} and provides clinicians with an objective basis for comparing performance under normal and challenging sensory conditions. Analyzing both the eyes closed task and the difference between the eyes closed and eyes open tasks in parallel yields 98% sensitivity for detecting gait disorientation due to vestibular dysfunction.

Ultimately, the clinical utility of a test relates to its diagnostic accuracy and whether the results lead to large changes from the pretest probability of the outcome of interest to the posttest probability of that outcome.⁸² The diagnostic performance of the GDT is comparable to tests that require specialized training and/or costly equipment (hHIT, HSN, DVAT, SOT). The DOR for the GDT suggests that those who had a GDT time >4.5 seconds were 57 times more likely to have peripheral vestibular dysfunction. Although this suggests that the GDT may be clinically valuable,⁸² diagnostic accuracy varies by setting and population. Our projections of positive posttest probability suggest that it may be reasonable to expect a ~90% probability of the diagnosis of vestibular dysfunction based on VNG and/or RC tests for a patient who presents to primary care, the emergency department, a dizziness specialty clinic, or neurology; is suspected to have vestibular dysfunction; and has a positive GDT.

Study limitations

Our findings are applicable to adults with and without peripheral vestibular dysfunction; however, there are numerous reasons why a person might have difficulty walking with their eyes closed. Prospective studies of the GDT in persons with visual, somatosensory, auditory, and cognitive-perceptual conditions are needed to assess whether it is broadly applicable. To address potential effects of age on GDT results, future studies should enroll age-matched participants. Finally, though the GDT is feasible, the space requirements for the test may hinder its adoption in some settings.

Conclusion

Our findings demonstrate that the GDT has promising clinical utility. The GDT is a simple metric that health care professionals could easily integrate into the examination of persons with dizziness and imbalance to screen for gait disorientation. Providers may use the threshold of 4.5 seconds for the GDT to help them determine if further workup, referral to specialty providers, and/or vestibular rehabilitation may be indicated.

List of abbreviations:

95% CI

95% confidence interval

ABCS

Activities-specific Balance Confidence Scale

AUC

area under the curve

COMP

composite score

DHI

Dizziness Handicap Inventory

DOR

diagnostic odds ratio

DVAT

dynamic visual acuity test

FGA

Functional Gait Assessment

FGA1

Functional Gait Assessment, item 1

FGA8

Functional Gait Assessment, item 8

FTSTST

Five-times Sit-to-stand Test

GDT

Gait Disorientation Test

hDVAT

horizontal dynamic visual acuity test

hHIT

horizontal head impulse test

HSN

head shaking nystagmus

LR

likelihood ratio

LR+

likelihood ratio for a positive test

LR−

likelihood ratio for a negative test

MDC

minimal detectable change

RC

rotational chair

ROC

receiver operator characteristic

SOM

somatosensory score

SOT

Sensory Organization Test

vDVAT

vertical dynamic visual acuity test

VNG

videonystagmography