Dual Task Interference During Gait in People With Parkinson Disease: Effects of Motor Versus Cognitive Secondary Tasks

Subjects

Fifteen subjects with idiopathic PD (12 men and 3 women; mean age=68.33 years, SD=6.59, range=52–76) and 15 comparison subjects, matched for age, sex, and height (mean age=67.73 years, SD=6.97, range=52–79) were recruited from the Elsternwick Private Hospital, Kingston Centre Movement Disorders Clinic, and local community groups. Forty-two subjects were screened to obtain this sample, as 12 people were not able to be recruited. To be included in the study, subjects needed to be able to walk unassisted a distance of 14 m at least 15 times and have no other neurological, orthopedic, or cardiovascular conditions that affected their walking. They also had to score greater than 20 out of 38 on the Short Test of Mental Status (STMS)¹⁴ and be able to provide informed consent. The STMS is a cognitive impairment scale, and scores less than 21 indicate dementia. In addition, for the group with PD, a diagnosis of idiopathic PD needed to be confirmed by 1 of 3 neurologists (RI, AC, or OW), and subjects needed to score less than or equal to 20 out of 36 on the Modified Webster Scale¹⁵ when they were at peak dose of their levodopa medication cycles. The Modified Webster Scale has 12 items that rate impairment and functional capacity. Higher scores indicate greater impairment. Subjects were excluded if they were taking major tranquilizers, which are antipsychotics used for the treatment of mental illness; if they had neurological conditions other than PD; or if they had a visual disturbance that impaired their ability to walk or read. Table 1 summarizes subject characteristics for the group with PD. There were no differences between groups for age, height, weight, or sex.

Apparatus

A clinical stride analyzer (CSA) (version 6.0*) was used to measure the spatiotemporal (time and distance) variables of footstep patterns, because it has been shown to provide some reliable measurements when repeat tests are performed with a 30-minute interval between tests (intraclass correlation coefficients were .96 for gait speed and .98 for stride length) for subjects with PD.¹⁶ The CSA consisted of a set of inner soles with footswitches attached via leads to a data recorder worn around the subject's waist. The inner soles had 4 pressure-sensitive footswitches (one each for the heel, great toe, and first and fifth metatarsal heads) that were activated as the subject walked. Data from the recorder were downloaded to an IBM personal computer.^† All gait trials also were videotaped using a Panasonic WVCL350 video camera^‡ mounted on a tripod.

For the secondary motor task (coin transference), 2 pockets made of calico (denim-like) material (16.5 × 14.5 cm) were attached by calico loops to a black leather belt (size: 44 in/112 cm). The midpoints of the pockets were situated over the anterior aspect of the right and left hip joints, respectively. This required the subjects' arms to cross from the dominant side to the nondominant side repeatedly when transferring the coins. A total of 12 Australian 20-cent coins were used.

For the secondary cognitive task condition, a digit subtraction task was used. Subjects were asked to count backward aloud by threes from a starting number, which was determined by selecting a card with a randomly generated number from 125 to 250 written on it. For each gait trial in this condition, a different card was randomly selected from a pack of 15 cards. The card was shown to the subject for 5 seconds before walking began. This task was created to diminish the impact of response synchronization, whereby a person times his or her footsteps with the spoken word, as observed during standard digit span (forward/backward) tests used in previous dual task studies (N Georgiou, personal communication, 2000). To minimize the likelihood that subjects would learn the secondary task, no practice or familiarization trials of the additional tasks were permitted.

Procedure

All testing was conducted in the Gait Laboratory at Kingston Centre. Prior to testing, all subjects were interviewed about their medical history and had the research procedure explained to them. The STMS¹⁴ was completed, and measurements of height, weight, and leg length were taken. Subjects with PD also were examined using the Modified Webster Scale, which provides a measure of disability.¹⁵

Gait trials were performed on a 14-m-long gray linoleum walkway, with the middle 10 m used for data collection. Subjects with PD were tested during the self-determined peak or “on” phase of their medication cycle. We did this because greater consistency of gait performance has been demonstrated for people with PD when medication levels are optimal.¹⁶ The walking patterns of all subjects were tested under 3 conditions: (1) free walking (“free”), (2) coin transference while walking (“coin”), and (3) digit subtraction while walking (“digit”). Subjects walked 3 times under each condition, and the order of conditions was randomly allocated.

For all 3 conditions, subjects were instructed to walk at their preferred pace. During the coin condition, the belt with the pockets was fitted to the subject and the coins were placed in the pocket on the side of their dominant hand. Subjects reported their dominant hand by indicating which hand they would use to catch a small ball. Subjects were then instructed to use their dominant hand to transfer as many coins as they could, one at a time, from the starting pocket to the pocket on the opposite side. For the digit subtraction condition, subjects were asked to count backward aloud by threes as described earlier.

In order to establish that the secondary cognitive and motor tasks were comparable in difficulty and to gain information about the coin and digit tasks without the subject walking, data also were collected for the rate of coin transference, the digit response rate, and the number of errors committed during digit subtraction while subjects were standing (Tab. 2). One trial of these baseline data was collected on the same day, immediately before gait analysis and confirmed using video analysis.

Data Analysis

Because all spatiotemporal gait data were normally distributed and did not violate the assumptions of homogeneity of variance, a series of 2 (groups) × 3 (tasks) analysis of variance tests for repeated measures were used to analyze the gait data. Selected post hoc planned comparisons were made using paired t tests with Bonferroni adjustments.

Stride Length

Table 3 and Figure 1 show that during walking without an additional task, subjects in the group with PD walked with shorter strides than the comparison group (F=32.16; df=1,28; P<.0001). Both groups demonstrated a decrease in their stride length with the coin and digit tasks. However, the group with PD demonstrated a much greater decline in stride length in both dual task conditions, as indicated by the group × task interaction effect for stride length (F=5.44; df=2,56; P=.007). No difference for mean stride length was detected in the group with PD or the comparison group between the coin and the digit conditions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Mean stride length for preferred walking (free condition) and walking with an additional motor task (coin condition [ie, coin transference]) or a cognitive task (digit condition [ie, digit subtraction]) in the group with Parkinson disease (PD) and the group of comparison subjects (Comp). Error bars indicate the standard deviation.

Walking Speed

Table 3 and Figure 2 illustrate that subjects with PD demonstrated a reduction in walking speed for the coin (t₁₄=−7.28, P<.0001) and digit conditions (t₁₄=−5.09, P<.0001). For the group with PD, no difference in walking speed was found between the coin and digit trials (Tab. 3). The comparison subjects showed a decline in their walking speed with the secondary coin (t₁₄=−6.45, P<.0001) and digit tasks (t₁₄=−5.25, P<.0001); however, there was no difference between the coin and digit conditions. Although the comparison group demonstrated a reduction in their walking speed when shifting to the dual task conditions, the magnitude of decline was smaller than for the subjects with PD (F=5.25; df=2,56; P<.01).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Mean walking speed for preferred walking (free condition) and walking with an additional motor task (coin condition [ie, coin transference]) or cognitive task (digit condition [ie, digit subtraction]) in the group with Parkinson disease (PD) and the group of comparison subjects (Comp). Error bars indicate the standard deviation.

Cadence

Table 3 and Figure 3 show the cadence values for the 2 groups for each of the 3 walking conditions. In this sample, there was a difference in cadence between the 2 groups for the unitask condition (F=15.39; df=2,56; P<.0001), with the group with PD showing a slower stepping rate from the outset (Tab. 3). The table also shows that cadence decreased to a greater extent in people with PD, resulting in a group × condition interaction effect (F=3.29; df=2,56; P<.05). In the group with PD, the decline in mean cadence from unitask performance to dual task performance occurred for both the coin (t₁₄=−4.23, P<.05) and digit (t₁₄=−3.60, P<.005) conditions (Tab. 3). No differences in cadence were found for the comparison group between the unitask and dual task conditions. In addition, neither group demonstrated differences in cadence between the 2 secondary task conditions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Mean cadence for preferred walking (free condition) and walking with an additional motor task (coin condition [ie, coin transference]) or cognitive task (digit condition [ie, digit subtraction]) in the group with Parkinson disease (PD) and the group of comparison subjects (Comp). Error bars indicate the standard deviation.

Double Support Duration

For the unitask walking condition, there was a difference in percentage of the gait cycle in double-limb stance (DS) between the group with PD and the comparison group, with higher mean percentages of the gait cycle spent in DS in the group with PD (Tab. 3). For the control group, there were increases in DS in dual task walking for both coin transference (t₁₄=4.39, P<.01) and digit subtraction (t₁₄=2.90, P<.05). For the group with PD, however, there were no changes in percentage of the gait cycle spent in DS from unitask to dual task conditions. No differences in percentage of the gait cycle spent in DS were found for either group between the coin and digit conditions.

Secondary Tasks

Table 2 presents the means and standard deviations for performance on the coin (motor) and digit (cognitive) tasks in both standing and walking conditions. A difference between the groups was found for the coin transference rate for standing trials (t₂₈=−3.45, P<.005) and walking trials (t₂₈=−6.21, P<.0001). Subjects with PD also showed a reduction in coin transference rate between the standing and walking conditions (t₁₄=−3.60, P<.05).

For the digit subtraction task, subjects with PD responded at slower rates than the comparison subjects during both standing trials (t₂₈=−3.65, P< 0.01) and walking trials (t₂₈=−4.41, P<.0001). The response rate during standing was no different from the response rate during walking trials for subjects with PD (Tab. 2), whereas comparison subjects demonstrated an increase in response rate with walking trials (t₁₄=2.96, P<.05). During walking trials, subjects with PD committed more errors in digit subtraction than did the comparison subjects (t₂₈=3.19, P<.005).

Dual Task Interference

Camicioli et al,¹ Morris et al,² and Bond and Morris³ noted that people with PD experienced marked difficulties when they were instructed to perform a complex secondary task while walking. Older people without pathology or impairments (aged 65–74 years) exhibit diminished performance when performing attention-demanding activities at the same time as walking.^17,18 Dual task interference during locomotion is also problematic for people with neurological conditions such as Alzheimer disease¹⁹ and Huntington disease.²⁰ In our study, subjects with PD may have demonstrated interference in their walking performance because central nervous system processing mechanisms were being used to perform the coin and digit tasks. In theory, this required gait to be controlled by impaired basal ganglia. When gait is controlled by the defective basal ganglia, reduction in step size and walking speed occurs.²¹

Models of Dual Task Interference

The main theoretical models accounting for dual task interference in people with PD are: (1) the capacity- or resource-sharing model, (2) the bottleneck model, and (3) the cross-talk model (see Pashler²² for a detailed review). These are “attentional” models, with the term “attentional” referring to the focus of mental activity on a task. Capacity-sharing models are based on the assumption that attention resources are limited. Therefore, when people perform 2 tasks simultaneously, attention must be divided between the tasks. How attention is divided between the 2 tasks relies on several factors, including task complexity, familiarity, and importance.²² According to the capacity-sharing model, dual task interference will occur only if the available resource capacity is exceeded, resulting in a decline in performance on one or both of the tasks.²²

The bottleneck and cross-talk models assume that dual task interference is affected by the type of tasks performed simultaneously, rather than the amount of attention needed to sustain performance.²² According to the bottleneck model, similar tasks performed concurrently cause “bottleneck” interference because they compete for the use of the same pathways.²² In contrast, cross-talk models assume that task similarity reduces dual task interference, because the use of the same pathway increases the efficiency of processing by using less attentional resource capacity.²²

The results of our investigation lend support to the capacity-sharing model of dual task interference. For both elderly people with PD and those without PD, a large proportion of the attentional capacity appeared to be directed toward the coin and digit subtraction tasks at the expense of walking performance. Because the secondary tasks were relatively novel compared with walking, we believe that they would require more attentional resources.

In our opinion, gait changes occurring during dual task situations may be the result of compensations undertaken by people with PD to reduce the risk of falling. Fast walking speeds require greater balance control because of the rapidly changing accelerations of the center of mass and the reduction in double support time.²³ We argue that, by slowing walking speed and reducing stride length during secondary tasks, people with PD may be attempting to decrease the balance requirements for gait. Paradoxically, slow walking speeds also can increase balance demands because greater time must be devoted to balancing the head, arms, and trunk over the stance leg.²³ Increases in double support time are thought to negate this effect during slow walking.²³ In our study, comparison subjects who were instructed to walk at their preferred speed demonstrated an increase in their double support time in the dual task conditions (Tab. 3), which may indicate that they were able to accurately compensate for the reductions in stride length and walking speed. By contrast, subjects with PD in our study and in other studies^2,3 did not increase double support times during dual task performance when walking at their preferred speed. We believe that these results may indicate that people with PD have an impaired ability to modulate double support to compensate for the reductions in stride length and walking speed. This impaired ability may increase the risk of falls.

Clinical Implications and Limitations

A major goal of physical therapy for people with PD is to help them walk with normal step size and speed in order to reduce the risk of trips and falls.²⁴ One strategy is to teach people with PD to avoid simultaneous tasks whenever possible,²⁴ to prevent attention being directed away from generating long strides or responding to unexpected perturbations. This way people can perform tasks in isolation when they need to walk with long, fast strides, such as when they cross a busy road. Doing one task at a time, however, is not always practical and, in our opinion, carries a high cognitive demand by necessitating continuous conscious attention to the task. It is thus likely that people with PD will sometimes revert to dual task performance. Therefore, we believe that it is advisable to teach people with PD about the safety risks associated with doing more than one task at a time. Some therapists might argue that teaching people with PD about the safety risks associated with simultaneous task performance should include engaging them in other tasks during gait training, while they are under close supervision. Whether people with PD have the capacity to learn how to perform dual tasks during walking safely and independently has not been established. Research is also needed to determine whether people with PD can learn how to safely and independently switch from doing several tasks to only walking when needed.

Our study had several limitations. First, the findings cannot be generalized to all people with PD, because only subjects with gait hypokinesia and mild to moderate impairments were included. Further research is needed to examine the effects of dual task performance during walking for subjects with other movement disorders such as akinesia, dyskinesia, and postural instability. The secondary tasks used were what we call “semifunctional” tasks. Investigation of the effects of functional tasks during gait in more real-world settings during activities of daily living is needed. In addition, all patients were tested at peak dose in the levodopa medication cycle, and it is not clear whether the results could be generalized to “off” phase performance.