The ability to attend to, and perform, two tasks at the same time, described as dual-task performance, has been shown to decline in people as they age.1 Clinical tests that require dividing attention between two tasks have been developed to characterize age-related changes in the effects of cognition on balance and safety. The inability to walk and talk, or to walk while carrying an object at the same time has been associated with falls risk in frail older adults.2,3 Older adults with a history of falls also have more difficulty maintaining balance in challenging sensory conditions while performing a concurrent cognitive task.4,5 Given the public health implications of falls in older adults, the association between dual-task performance and falls risk is an important area for investigation.
Existing clinical measures that assess dual-task performance with a cognitive task during walking include: the “stops walking while talking” test (SWWT),2 the Timed Up and Go cognitive (TUG cognitive),4 the Walking While Talking Test (WWTT),6 and the Multiple Tasks Test (MTT)7,8 (Table 1). These tests have important limitations for identifying dual-task performance problems in community-dwelling older adults. The SWWT, TUG cognitive test, and the WWTT present cognitive tasks as distracters. No effort is made to characterize the performance of the cognitive task prior to walking or during walking.
Task difficulty is an important consideration in dual-task performance.9 All of the published clinical tests of dual-task walking performance require walking at a self-selected speed, a task that for community-dwelling older adults may leave significant “reserve capacity” for the performance of other tasks. Furthermore none of the current clinical tests attempt to provide a level of cognitive challenge that takes into account the individual's capacity for the task. Education or literacy level may affect the ability to perform these tasks, making them simple for some but very challenging for others. For community dwelling older adults, the ease of talking (SWWT), responding to simple questions (MTT), reciting the alphabet (WWTT), or performing simple subtraction (TUG cognitive) may allow a reserve capacity so that dual-task costs are not observed in clinical settings.
There is a need for a clinical measure that provides sufficient challenge on both tasks to elicit dual-task costs in a way that can be easily observed by a clinician. The Walking and Remembering Test was designed to meet this need by characterizing dual-task performance in a manner similar to Lindenberger et al,10 but for use in clinical settings. Lindenberger et al devised a protocol for experimental assessment of the single and dual-task demands during walking in active older adults.10 In this study, adult subjects in three age groups (20-30, 40-50, and 60-70 years) walked on 19 cm narrow tracks while performing a verbal memory task. Walking trials were 170 seconds long to allow presentation of 16 item word lists via headphones. Electrically conductive walking tracks and shoes equipped with one graphite sole and one aluminum tape sole allowed instrumented assessment of the number of steps off the tracks as subjects walked as quickly and as accurately as possible. Significant dual-task costs were demonstrated in the oldest age group in the dual-task condition, as reflected by greater relative dual-task costs in walking speed, poorer step accuracy, and reduced word recall. These laboratory methods were effective in illuminating relationships between dual-task costs in mobility and cognition, but are not feasible for use in the clinical setting. In this paper, we describe development of the WART and provide evidence for its reliability and preliminary validity.
Literature review, consultation with experts in physical therapy and neuropsychology, and pilot testing with community dwelling older adults influenced selection of cognitive and motor tasks that could be feasibly combined in a clinical test. For clinical purposes, the test needed to be brief, use standard administration, require little equipment or training, and be challenging for community-dwelling older adults. In order to present an equivalent challenge to older adults with a wide spectrum of abilities, we chose tasks that could be readily adjusted to account for variations between subjects in single task performance.
Reaction time and visual tasks were ruled out because of difficulty in using such tasks during walking. Pilot testing was conducted with several cognitive tasks including verbal learning tasks, serial subtraction tasks, and forward/backward digit span. With consultation from a neuropsychologist, a single flexible task, forward digit span, was selected as the most appropriate task. Difficulty could be customized to the participant's baseline working memory function with simple standardized testing using the Wechsler Adult Intelligence Scale-Revised protocol.11 The digit span task can be performed during walking, does not require special equipment to present stimuli or record responses, and presents no bias against subjects with limited levels of literacy. Judging accuracy of recall is straightforward. By presenting digits at the beginning of each trial and asking for recall at the end of each trial, articulation during walking was eliminated to prevent verbal pacing of walking.
Walking was selected as the most functionally relevant task for community-dwelling older adults rather than standing or other balance activities. Walking is a critical function for daily life that requires mobility in the community. Following Lindenberger,10 subjects walked along a 19 cm wide, 6.1 meter long path marked in tape on a portable roll of carpet that was secured to the floor (Figure 1). The length was a distance that could be tested in a variety of environments, including many home settings. Subjects were able to walk along the path at self-selected and speeded rates without assistance, but the path presented a slight balance challenge.
Subjects were 25 college students (≥ 18 years old) and 25 community dwelling older adults (≥ 65 years old) who participated in a community-based cardiac rehabilitation program. Both groups were recruited by flyer or announcement prior to class or exercise session. Inclusion criteria for both groups required independent ambulation with or without an assistive device, independent community-dwelling, and no history of neurological conditions or orthopaedic deformity that would prevent being able to walk along a narrow path on the floor. Subjects were required to have functional vision and hearing with or without aids, and be English-speaking.
The study protocol was approved by the UNC-CH Biomedical Institutional Review Board. Informed consent was obtained from all subjects prior to testing.
The younger adults were comprised of 7 males and 18 females aged 22 to 35 years (mean 24.2 ± 3.0 years). The older adults were 15 male and 10 female community-dwellers aged 65 to 86 years (mean 75.9 ±. 5.7 years). Self-reported characteristics of this group are described in Table 2. Older adults reported exercising 3 to 7 times per week (mean 4.2 sessions/week).
TEST ADMINISTRATION AND PROTOCOL
The WART test sequence (Table 3) began with narrow path walking, a novel task with a balance component. Subjects first walked at a self-selected speed followed by 3 trials of single task walking at their fastest comfortable speed. An appropriate cognitive challenge was determined by seated single task testing of forward digit span using the WAIS-R protocol. Then the cognitive and walking tasks were tested in combination for 4 trials in the dual-task condition. The test ended with a second set of 2 single task walking trials. These single task walking trials were added based on observation in pilot testing that subjects seemed to improve in single task performance after dual-task practice. The examiner walked alongside each subject to provide assistance if needed for loss of balance. A second rater observed testing and simultaneously recorded walking time, step accuracy, and digit span accuracy for all younger adult subjects and 15 older adult subjects. Ratings were not discussed or compared prior to analysis.
All subjects were tested using the WART protocol on 2 occasions by the same examiner. Young adult testing took place in a classroom with minimal distractions. Retesting sessions were conducted 7 to 9 days after the initial test. Older adult testing took place in an area adjoining a track in a multilevel community fitness center where ambient noise and occasional distractions were present. If a subject appeared distracted at any point during the testing, test components were repeated. Older adult subjects were scheduled at times just prior to or immediately following their cardiac rehabilitation or maintenance program. Rest was provided if necessary prior to or during testing. Testing intervals varied to a greater degree in the older adult group of very active retirees. Because of the schedule of the cardiac rehabilitation class, subject vacations, and holidays, retest intervals ranged from 2 to 7 days, with the exception of one subject who was retested 21 days after the first test because of an extended vacation.
Inter-rater reliability was examined using the intraclass correlation coefficient, [ICC (2,1)] for walking time and accuracy of digit span responses. The number of steps off the path was compared for percent agreement and absolute differences, following determination that the distributions were not of sufficient range to be appropriate for other analyses. Test-retest reliability was also examined with the intraclass correlation coefficient, [ICC (2,1)] for walking time during select single and dual task walking trials.
Group Differences and Practice Effects
Differences in baseline performance for older adults relative to younger adults were expected. Practice effects associated with performing the test a second time were examined using repeated measures ANOVA with factors of time: test or retest (within subjects), and group: younger or older (between subjects). Separate ANOVA analyses were conducted for walking time, steps off the path, and digit span length and digit span accuracy.
Walking time and steps off the path in the single and dual task conditions were calculated by averaging 4 timed trials from each single and dual-task walking test (single task - last 4 trials; dual-task - all 4 trials). Consistency in steps off the path was compared by averaging the number of steps off the path in the single task and dual task conditions (same trials for each condition) across the 2 tests. Digit span length and digit span accuracy were examined for group differences. Average digit span accuracy in the 4 dual-task test conditions from each test session were compared with repeated measures ANOVA.
Relative dual-task costs for walking were calculated using single task performance as a comparison condition. The average single task performance of the last 4 single task trials was subtracted from the average dual-task performance (all 4 trials) and divided by average single task performance. This represented dual-task cost as a percentage change in performance due to the addition of the second task. Absolute dual-task costs for steps off the path were calculated by subtracting the average number of steps off the path in (single task, last four trials) from the average number of steps off the path in dual task trials. Relative dual-task costs were not calculated for steps off the path because there were so few stepping errors in the single task condition. Baseline digit span recall was 100%, as each subject had to recall a string of digits with complete accuracy for that span length to be used in the dual-task condition. Digit span accuracy of less than 100% in the dual-task condition represented a relative dual-task cost.
Inter-rater Reliability Walking time
Inter-rater reliability was based on scores by 2 raters. Intraclass correlation coefficients were calculated for the third single task trial, the first dual task trial, and the last single task trial for both groups. ICC (2,1) values for walking time were .98, with absolute differences ranging from .08-.11s for the younger adults; and .99 for the older adults, for with absolute differences ranging from .15-.16s.
Steps off the path
In the young adult group, the highest number of steps off the path for any subject was 2. Raters had 93% agreement on the number of steps off the path, with an absolute difference of .06 steps over 225 trials.
Older adults were less accurate walkers with steps off the path ranging from 0 to 5. Raters had perfect agreement on the number of steps off the path in 76% of trials of the inter-rater reliability sample (n=15), with an average absolute difference of .28 steps over 135 walking trials. Raters agreed whether or not a subject stepped off the path in 86% of trials.
Digit span accuracy
Rater assessments of digit span accuracy on all dual-task trials were assessed for reliability in each of the samples. Inter-rater reliability was excellent with ICC (2,1) values of .97 for the young adults and .99 for older adults.
ICC values for test-retest comparisons were lower than for inter-rater reliability. For young adults, ICC (2,1) values for the third single task trial, the first dual-task trial, and the last single task trial were .83, .76 and .92, respectively. Values for the older adults for the same trials were .65, .83, and .86.
Group Differences and Practice Effects
Mean values and 95% confidence intervals for variables related to walking time, steps off the path for subjects on both tests and digit span accuracy are provided in Table 4.
Mean walking times decreased over the course of 2 test administrations for both groups, but to a larger degree for the older adults (Figure 2). Repeated measures ANOVA indicated significant effects of time and group in both single and dual-task conditions but no interaction effects [single task condition: time (test vs. re-test), F=10.72, p=.002; group (younger vs. older), F=19.36, p < .001; time x group F=2.43, p=.13; dual-task condition: time, F=6.18, p=.016; group, F=22.25, p < .001; time x group, F=.11, p=.74].
Both groups had faster average walking times on the second test, although younger adult differences were of no clinical significance (mean differences of < .28 s in single or dual-task conditions). Given the improvement in walking time on the second test, an average difference in walking time of greater than one second in either the dual-task or single task trial on a second test may indicate a significant difference in performance as opposed to a practice effect for an active community dwelling older adult [mean changes in time (95% C.I.); third single task trial .78s (.07, 1.49); first dual-task trial .51s (.07, .93); last single task trial .27s (-.08, .62)].
Steps off the path
Young adults had minimal steps off the path in either condition and this pattern was stable across test administrations. Steps off the path increased significantly in the dual-task condition for older adults (Figure 3). ANOVA results confirm group differences between single and dual-task conditions (single task, F=13.03, p=.001, dual task, F=17.44, p < .001), but no significant effects for time (single task F=.17, p=.68, dual task F=.01, p=.94) and no interaction effects (single task F=.17, p=.68, dual task F=.05, p=.82).
Digit span accuracy
Both groups demonstrated dual-task costs for digit span recall with reductions from 100% baseline accuracy. Young adults were accurate with digit recall an average of 92%, whereas older adults were 85% accurate in digit span recall for each test session. Mean values of digit span accuracy differed by group but not by time (group F=4.91, p=.03, time F=.02, p=.90) with no interaction effect.
Digit span length
Younger adults demonstrated longer digit spans on both test administrations than the older adult group (F=22.133, p < .001). Since the entire test protocol was repeated on the second test, it was possible for the digit span length to change from that used on the first test. On the second test administration, differences in digit span length were as follows: younger adults: 20%-shorter, 36%-unchanged, and 44% - longer; older adults: 12%-shorter, 48%-unchanged, 40%-longer. These variations in digit span testing on the second test were not statistically different between groups (t= -.153, p=.879).
Strategies for Digit Recall
With the exception of 2 older adult subjects (on the first test only) all subjects described a cognitive strategy to remember digit spans while walking. Mental rehearsal was the most common strategy described by both groups of subjects, with various chunking patterns that were related to length of the digit span.
Dual Task Costs
Both groups demonstrated small but clinically insignificant changes in walking time in the dual task condition relative to the single task condition. Younger adults demonstrated relative dual-task costs for walking time averaging 2-3% (Table 5). Unexpectedly, older adults demonstrated slightly faster times in the dual-task condition on the first test resulting in an average dual-task benefit of 1%. On the second test relative dual-task costs averaged 4%.
Digit span dual-task costs were on average 8% to 9% for the younger adults and 15% for older adults, reflecting significant decrements in performance for both groups from baseline digit span accuracy of 100%. Average absolute dual-task costs for steps off the path were higher for older adults in both test sessions (Table 5).
A measure of dual-task performance should detect greater dual-task costs in older adults as compared to a younger adult sample. Results from the WART support this expectation, but not as strongly as anticipated. During the walking task relative dual-task costs for time did not differ between the groups on either test administration; however, older adults had significantly more missteps. Both groups demonstrated dual-task costs for digit span, but older adults had significantly greater costs than the younger adults. A test protocol that only measured walking time would miss decrements in performance observed with more careful data collection.
The WART is an easily administered clinical measure of dual-task performance that is feasible for community-dwelling older adults. The test uses readily available materials and is scored easily. Unlike other clinical tests of dual-task performance, the WART adjusts cognitive task difficulty to the person being tested, so that a similar level of challenge is presented for each subject. The speeded walking task with a narrowed base of support presented a detectable challenge to the older adults, as evidenced by greater difficulty maintaining steps on the path than the young adult group.
Reliability of judging steps off the narrow path was excellent for young adults, who rarely had missteps. Although older adults had more frequent missteps, absolute agreement between raters was still good. Step accuracy was the most difficult aspect of the test to judge, particularly for subjects who walked very quickly. In administering the WART, testers must walk alongside the subject, watch for steps off the path, provide guarding if necessary, time the subject's walking performance, and hold a clipboard to record digit span responses. The role of experience and training may influence the ability of raters to reliably observe and record information during testing. These effects were not examined in this study.
Practice Effects and Motor Learning Considerations
Performing a novel balance and mobility task such as the walking portion of the WART represents a motor learning task for many older adults. The presentation of 10 trials of the walking portion of the WART represents many more task attempts than are used in most clinical measures in rehabilitation. Practice effects have not been well studied in such clinical measures. Understanding change associated with practice would aid identification of clinical change. Older adult subjects appeared to stabilize in their performance by the second test, with walking speeds for the final two single task trials being fairly consistent with many fewer steps off the path.
Older adults appeared to use a speed-accuracy trade-off in the dual-task condition, as they continued to walk quickly, with little to no dual-task cost for walking time; however, they took more steps off the path than younger adults. An unexpected result was the faster walking time as older adult subjects transitioned from the first set of single task trials to the dual-task condition. A hypothesis of faster walking to minimize the time over which digits had to be held in memory was supported by anecdotal comments from participants. Despite this strategy, dual-task costs for digit span recall were consistent with older adults. This pattern of performance is consistent with the “posture-first” theory proposed by Shumway-Cook and Woollacott.12
Dual-task performance is sensitive to cognitive changes with aging.1 One theory of why dual-task costs increase with aging relates to slowed perceptual speed,13 and the increased complexity of a dual-task situation requiring more processing.14 Others theorize reduced capacity of working memory, attention or perceptual-motor ability, such that performance of two tasks is inherently more difficult for older adults.15 Yet another theory of age-related decrement in dual-task performance is difficulty with coordination and allocation of attention to multiple tasks. Kramer et al16 cite support for multiple mechanisms responsible for the dual-task costs observed with aging.
Interpreting differences in dual-task costs for older adults with a clinical measure is complex. Younger and older adult groups had baseline differences in working memory capacity. The difficulty of the cognitive task was manipulated to take this baseline difference into account by using each subject's maximal digit span for the cognitive task. This method elicited significant dual-task costs for the cognitive task for both groups of subjects, but older adult digit span dual-task costs were greater even though the cognitive task difficulty was adjusted for each subject. Had digit span accuracy not been measured in this study, significant group differences would have been obscured.
Baseline group differences were also present for the walking task, but task difficulty was not modified. A mathematical adjustment was made instead, by computing relative dual-task costs based on single task performance. This method is important when making comparisons across subjects who begin with different baselines if task difficulty is not adjusted. Dual-task costs for the walking task were most significant in step accuracy. However, interpretation of this component of dual-task cost is not intuitive, since stepping errors are not easily represented as a percentage without also counting the number of steps per trial.
The cause of the dual-task costs observed with the older adults is unclear. Older adults could have demonstrated greater dual-task decrements in digit span and walking accuracy than the younger adult group because the walking task was more difficult for them from a balance standpoint, and thereby required more attention. Alternatively, the results could be explained by reduced attentional capacity in the older adults that was manifest in poorer dual-task performance. Another possible confounder was the difference in testing environment between the two groups. Older adults were tested in a setting where ambient noise and occasional visual distractions were present. Younger adults were tested in a classroom with few environmental distractions. These environmental factors may have contributed to the difference in dual-task performance observed in the two groups. The results of this study can not distinguish between these potential causes of dual-task costs.
Practice of this novel walking task seemed necessary for the older adults in order to achieve more stable responses (decreased mean times with lower variance and improved step accuracy) in the final single task trials during re-testing. The tendency to improve on repeated administrations of the WART must be taken into account if the test is used to judge the effects of clinical intervention. The extent of practice effects observed in this walking task also raises questions about other physical performance measures that incorporate novel tasks of balance.
Single task trials administered at the end of the test were associated with improved performance. The possibility that practice in dual-task conditions could result in single task improvement could provide avenues for intervention. The effect of the dual-task condition may be interpreted more clearly if single and dual-task trials are alternated in future test administrations.
Dual-task costs for walking time were not significant in this older adult sample, a finding inconsistent with previous studies.10 The older adult group was composed of very active community-dwelling individuals, many of whom had participated in a community-based cardiac rehabilitation maintenance program for years. Older adult average walking speed on the narrow path, with a slight balance challenge, was 1.38 m/s on the first test (95% CI 1.22-1.54) and 1.49 m/s on the second test (95% CI 1.31-1.65). These values represent an older adult group functioning at a very high level, and may not be representative of clinical populations where dual-task performance is a greater concern. Generalization of these results to those with more limited mobility is uncertain.
The WART demonstrates many important qualities of a clinical measure, being administered safely and easily in a brief period of time with commonly available materials. The WART provides an alternative to existing clinical tests of dual-task performance with the following benefits: (1) tests at a higher mobility level than walking alone; (2) adjusts cognitive task difficulty in accordance with a person's working memory capacity; (3) provides methods for collection of single and dual-task performance data on both cognitive and motor tasks; and (4) allows for division of attention between two tasks that may be more representative of how people function in daily life.
The WART has excellent inter-rater reliability for walking time and digit span accuracy. Step accuracy is judged with acceptable reliability, but could be managed alternatively with a criterion level of step accuracy that requires a trial be repeated if more than 2 steps are taken off the path, for instance. Test-retest reliability is acceptable, although a practice effect was observed in this older adult sample.
The results of this study provide preliminary support for the construct validity of the WART as a measure of dual-task performance because greater dual-task costs were observed in older adults than a young adult comparison group; however, additional study is needed to validate the test. The older adults in this study were high functioning, and not representative of the broad range of mobility limitations that may be observed in community dwelling older adults or older adults seeking physical therapy. The relationship of dual-task performance on the WART to standard indices of falls risk is a critical unknown factor. The flexible nature of the cognitive task may also allow use of the test with populations that have cognitive deficits, such as persons with Alzheimer's disease, Parkinson's disease, brain injury, or stroke. Further examination of the validity and responsiveness of the WART is necessary to guide clinical use of the measure.
1. Craik FIM, Byrd M. Aging and cognitive deficits: The role of attentional resources. In: Craik FIM, Trehub SE, ed. Aging and Cognitive Processes
. New York, NY: Plenum Press; 1982:191-211.
2. Lundin-Olsson L, Nyberg L, Gustafson Y. “Stops walking when talking” as a predictor of falls in elderly people. Lancet
3. Lundin-Olsson L, Nyberg L, Gustafson Y. Attention, frailty and falls: the effect of a manual task on basic mobility. J Am Geriatric Soc.
4. Shumway-Cook A, Brauer S, Woollacott M. Predicting the probability for falls in community-dwelling older adults
using the Timed Up and Go Test. Phys Ther.
5. Shumway-Cook A, Woollacott M. Attentional demands and postural control: the effect of sensory context. J Gerontol
6. Verghese J, Buschke H, Viola L, et al. Validity of divided attention tasks in predicting falls in older individuals: a preliminary study. J Am Geriatri Soc.
7. Bloem BR, Valkenburg VV, Slabbekoorn M, Willemsen MD. The multiple tasks test: development and normal strategies. Gait Posture
8. Bloem BR, Grimbergen YAM, Cramer M, Valkenburg VV. “Stops walking when talking” does not predict falls in Parkinson's disease. Ann Neurol
9. Abernethy B. Dual-task methodology and motor skills research: some applications and methodological constraints. J Human Movement Studies
10. Lindenberger U, Marsiske M, Baltes PB. Memorizing while walking: increase in dual-task costs from young adulthood to old age. Psych Aging
11. Wechsler D. Wechsler Adult Intelligence Scale-Revised
. San Antonio, Tex: Psychological Corporation; 1981.
12. Shumway-Cook A, Woollacott M, Kerns KA, Baldwin M. The effects of two types of cognitive tasks on postural stability in older adults
with and without a history of falls. J Gerontol.
13. Salthouse TA, Rogan JD, Prill KA. Division of attention: age differences on a visually presented memory task. Mem Cognit.
14. McDowd JM, Craik FIM. Effects of aging and task difficulty on divided attention performance. J Exp Psychology: Hum Percept Perform.
15. Crossley M, Hiscock M. Age-related differences in concurrent-task performance of normal adults: evidence for a decline in processing resources. Psych Aging.
16. Kramer AF, Larish JF, Strayer DL. Training for attentional control in dual task settings: a comparison of young and old adults. J Exp Psychol.
17. Nebes RD, Butters MA, Houck PR, et al. Dual-task performance in depressed geriatric patients. Psychiatry Res