What is the Evidence behind using Technology to Deliver Treatment?
There have been several efforts to apply various forms of computer technology to rehabilitation
of brain-injured patients. Given the focus of this special issue, we will limit our
review of the evidence on the application of technology (available via Internet connection
either on a computer or mobile application) for rehabilitation to language and communication
deficits in the spirit of retraining disordered function, and hence we will refrain
from discussion about alternative/augmentative technologies (and text-to-speech devices)
that serve as communication aids. These include programs such as Lingraphica (Lingraphica
Inc., Princeton, NJ) Sentence-Shaper (Psycholinguistic Technologies, Jenkintown, PA)
and Touchspeak (Touchspeak™, London, England). Internet-based software treatments
have been increasingly available for individuals with brain damage. For instance,
a few studies have examined the effectiveness of CogMed (Pearson Company, Scandinavia,
Sweden), a software targeted at improving working memory abilities in individuals
with brain injury.[8]
[9] These studies found improvements in working memory skills on the trained CogMed
software as well as on other working memory tasks and functional settings. Likewise,
Barnes et al examined the effectiveness of the software Posit Science (Posit Science,
San Francisco, CA) in improving auditory processing speed in individuals with mild
cognitive impairment (MCI).[10] Although differences between the experimental and control groups were not statistically
significant, verbal learning, and memory measures were higher in the experimental
group than the control group. In another study, also examining cognitive training
in MCI, Finn and McDonald used Lumosity software (Lumos Lab, San Francisco, CA) to
target attention, processing speed, visual memory in experimental and wait-listed
controls.[11] Results showed experimental participants improved on the training exercises more
than the controls. Software solutions such as Lumosity and Posit Science provide a
wide range of primarily cognitive therapy tasks that include attention, memory, and
visuospatial processing, but they are not necessarily targeted toward individuals
with brain damage. They appear to be geared more toward healthy older adults interested
in delaying age-related cognitive decline. It should also be pointed out that these
software platforms are primarily delivered on a computer (PC/MAC) over the Internet
and are not specifically tablet-based delivery systems.
There are other software solutions targeted specifically at aphasia therapy. Notwithstanding
communication aids such as Sentence Shaper and Lingraphica, recent studies have demonstrated
that use of “virtual speech therapists” in delivering remote language therapy to replace
face-to-face patient-clinician sessions. Thus, Sentactics[12] (Sentactics Corporation, Concord, CA) provides therapy for sentence production deficits
and Oral Reading for Language in Aphasia with Virtual Therapist (ORLA-VT)[13] provides oral reading practice for patients with aphasia. These technology-based
rehabilitation solutions are limited in their functionality and range of available
therapy tasks but provide a promising alternative to traditional flash card–based
therapy.
Individualizing Rehabilitation Options for Patients
A relatively unanswered question in most of the broad software platforms that are
currently available is the nature and specificity of treatment provided. Given that
all these software platforms can be used with non-brain-damaged individuals and brain-damaged
individuals, is the treatment targeting a specific impairment that the patient exhibits
or is it primarily a brain exercise? For both the software-based platforms and the
virtual-therapist platforms, can therapy be individualized based on the patients'
demographic profile and level of impairment severity? What are the pretreatment profiles
of participants enrolled in treatment? How does performance improve, stabilize, or
plateau as a function of this training? These issues are all very difficult to address
as no two patients are alike in their demographic profile and, more importantly, no
two patients are alike in their treatment outcomes. Nonetheless, the burden of evidence
for technology-based treatment applications is no different than traditional treatment
approach for rehabilitation after brain damage. Until we begin to address these questions,
it is difficult to recommend technology-based applications for rehabilitation as replacements
for traditional flash card–based therapy to improve clinical care for patients with
brain damage.
In this article, we describe the development of an impairment-based, individualized
treatment plan for patients that can be delivered through an iPad (Apple Inc., Cupertino,
CA) software platform. Because of the flexibility that iPads provide to patients and
the accessibility to free/paid apps that provide variable levels of exercises, it
is important to standardize the nature and form of treatment that is provided to patients
using iPads. Second, because patients have access to iPads at home, it provides a
unique opportunity to examine the extent of compliance when patients are provided
with a homework regimen, something that is very difficult to assess when providing
traditional paper-and-pencil homework tasks. This article describes a portion of a
broader project examining the effectiveness of specific impairment-based treatments
for patients with brain damage in which a large group (n = 55) patients entered into a prospective clinical efficacy study.
This article is organized into two sections. We will first describe the general methods
of the treatment workflow for patients involved in this study. The second section
provides four case studies that provide a representative sample of participants progressing
through their individualized treatment plan. Because the therapy is delivered on an
iPad, patients can either practice the therapy tasks assigned to them in the clinic
or at home on a regular basis. Patient performance (both accuracy and latency) is
analyzed on a periodic basis and patients progress to the next level of hierarchy
after a certain level of performance criteria is met.
Methods
Treatment Workflow
In this section, we describe the treatment workflow for each individual patient participating
in our ongoing clinical efficacy study.
Assessment of Patient Impairment
Each patient is given a battery of standardized language tests, including the Revised-Western
Aphasia Battery (R-WAB)[14] to establish the type and severity of aphasia, the Boston Naming Test (BNT)[15] to determine confrontation naming ability, the Pyramids and Palm Trees[16] to determine overall soundness of the semantic system, and the Cognitive Linguistic
Quick Test (CLQT)[17] to determine the relative contribution of cognitive deficits such as attention and
memory to language dysfunction and to rule out dementia. Based on the performance
on each of these tests, a general profile is built for each patient based on whether
they present with primary language deficits or primary cognitive deficits, and within
each category, whether they present with low performance in that domain (more than
half test scores in that domain are below 50%) or high performance (more than half
test scores in that domain are above 50%). Clearly, patients present with a continuum
of language/cognitive deficits, and therefore, generally, WAB Language Quotient (LQ),
Aphasia Quotient (AQ), BNT, and Pyramids and Palm Trees (PPTT) tests are used to evaluate
language skills and WAB Cortical Quotient (CQ) and Cognitive Linguistic Quick Test
(CLQT) is used to evaluate cognitive skills. Performance on these tasks serve as a
starting point for the therapy assignment; for instance, if a patient's letter discrimination
score is 40% on WAB, this patient would likely benefit from working on sound-to-letter
matching exercises.
Assign Individualized Tasks for each Patient
The next step in the treatment workflow is to assign specific therapy tasks for each
patient based on their relative language and cognitive profile. We implemented the
Constant Therapy iOS platform (www.constanttherapy.com) to deliver the therapy. The choice of therapy tasks to be assigned came from a set
of 30+ therapy tasks broadly divided in language and cognitive therapy. Language therapy
tasks were divided into (1) naming therapy: (a) rhyme judgment, (b) syllable identification,
(c) phoneme–sound identification, (d) category matching, (e) feature matching, and
(f) picture naming; (2) reading therapy: (a) spoken word–to–written word identification,
(b) written word category identification, (c) reading passages, (d) long passage reading
comprehension, and (e) reading maps, (3) writing therapy: (a) word copy completion,
(b) word copy, (c) word spelling completion, (d) word spelling, (e) picture spelling
completion, (f) picture spelling, (g) sound-to-letter matching, and (h) letter-to-sound
matching; (4) sentence planning: (a) active sentence completion, (b) passive sentence
completion. In addition, cognitive therapy tasks were divided into (1) visuospatial
processing: (a) symbol cancellation, (b) telling time/analog clock; (2) memory: (a)
visuospatial picture/word memory matching, (b) visuospatial auditory memory, and (c)
voicemail task; (3) attention: (a) response inhibition, (b) symbol cancellation; (4)
problem solving: (a) analytical reasoning with subtasks such as (i) alphabetical word
ordering, (ii) alphabetical picture ordering; (b) arithmetic with subtasks such as
(i) addition, (ii) multiplication, (iii) subtraction, and (iv) division; and (c) quantitative
reasoning with subtasks such as (i) time estimation task, (ii) word math problems;
(5) executive function: (a) sequencing a set of steps/instructions. Details regarding
the development of the stimuli for each of these tasks are not provided here due to
space limitations. The range of therapy tasks, their scientific and clinical rationale,
and specific evidence regarding their efficacy are provided in [Appendix 1].
The general procedures for assignment of treatment were as follows. During the initial
session, a subset of potential therapy tasks was assigned as baselines. As long as
performance on a task was below 80% accuracy, that task was assigned for therapy.
Because we examined both accuracy and latency as our dependent measures, we chose
80% accuracy as a cutoff to allow for examination of a maximum of 20% change in accuracy
and corresponding decreases in latencies. If a patient performed higher than 80% on
a particular task, the next level of difficulty of that task was assigned to the patient,
or a different task examining the same skill (e.g., rhyming instead of sound identification)
was assigned to the patient. For each patient, five to six tasks with up to 10 items
in each task were assigned as that week's “therapy schedule.” Participants were incrementally
assigned tasks during the course of the 10-week treatment program. Therefore, each
patient was provided with an individualized treatment plan, which was modified frequently
during the course of their participation in the study.
Participant Action at Home/Clinic
Participants were provided with usernames and passwords to log into the Constant Therapy
app and were then asked to practice the therapy up to 6 days a week for 1 hour each
week. Participants were assured that there was no penalty whether they did or did
not log in every day and complete the therapy, but that their therapy practice time
would be recorded by the software. Participants were seen in the clinic on a weekly
basis to review and monitor progress.
Analysis of Patient Performance and Progress
One of the key aspects of the software platform is that the clinician can remotely
analyze each patient's progress from his or her clinician account. During the weekly
clinic sessions, the clinician would decide to continue the participant on the same
task or to modify the treatment plan based on the patient's performance. If the participant
achieved 95% or higher accuracy two times in succession, the clinician would either
progress the next level of difficulty (e.g., addition level 1 to addition level 2)
or would progress to a different task (e.g., assign category identification after
category matching). If participants performed at low accuracies (40% or lower) over
several sessions, that therapy task was replaced with another task from the task list.
Data for each patient's accuracy and latency were plotted over for every session they
performed the therapy task for every week up to 10 weeks. [Fig. 1] shows a representative sample of participant 84's overall treatment plan in a snapshot,
indicating how therapy tasks were assigned during the period of treatment and the
overall accuracy for that task each session. As can be seen, therapy tasks are introduced
and removed from the patient's therapy schedule based on performance. In addition,
for each patient, for each treatment task, a trend line was computed on the time series
data and the coefficient of determination (R
2) was computed. R
2 > 0.2 indicated a linear trend in data (for both accuracy and latency).
Figure 1 Overall performance of patient 84 on accuracy for all the therapy tasks assigned.
The x-axis indicates the schedule for each week and whether the session was at home
or in the clinic. The y-axis indicates all the therapy tasks assigned. Cells indicate
the patient accuracy (e.g., 89%) and cell colors indicate the population mean: green
indicate patient performance is above the population mean, red, yellow and orange
indicate the patient performance in below the population mean.
Case Studies Implementing This Treatment Workflow
In this section, we review four cases from our ongoing data collection efforts.
Example 1: Low Language Profile, Low Cognitive Profile
The first patient, No. 84, is a 66-year-old man who suffered a stroke ~14 months prior
to participation in this study. This participant presented with low performance on
both language and cognitive measures. Therefore, as seen in [Table 1], this participant's LQ was 38 and AQ was 50.2. This participant also presented with
a low CQ (45.05) and generally low scores on all aspects of cognitive processing including
attention, memory, executive functions, and visuospatial processing on the CLQT. Oral
naming was severely impaired. This patient was assigned the following tasks in treatment:
category matching, feature matching, picture naming, rhyming, sound identification,
word identification, sound-to-letter matching, and word copy to work on various aspects
of naming, reading, and writing within language processing; and picture matching and
symbol cancellation to work on attention, memory, and visuospatial aspects of cognitive
processing. Note that these tasks were assigned to this participant over the course
of the 10-week period, and for some tasks, such as symbol matching and word copy,
the patient progressed to the next level. [Fig. 2A] shows a representative sample of this participant's progression through word copy
(level 2) where improvements in accuracy are accompanied by increased latency on the
task. This patient completed 48 therapy sessions (i.e., logged in to complete the
therapy assignment) over the course of 10 weeks. At the end of treatment, this participant
was tested on all the standardized measures and showed improvements in LQ (38 to 50.1)
CQ (45.0 to 61.5), AQ (50.2 to 67.6), as well as improvements on all subtests of CLQT
including attention (8 to 36.2%), memory (57.8 to 64.8%), executive function (EF)
(7.5 to 15%), visuospatial skills (16.1 to 28.5%) and on the BNT (0% to 3%).
Table 1
Standardized Test Performance for Four Patients, Pre- and Posttreatment
|
|
Patient 84
|
Patient 108
|
Patient 25
|
Patient 05
|
|
n
|
Pre (%)
|
Post (%)
|
Pre (%)
|
Post (%)
|
Pre (%)
|
Post (%)
|
Pre (%)
|
Post (%)
|
|
Fluency total
|
20
|
45.00
|
75.00
|
95.00
|
100.00
|
60.00
|
65.00
|
95.00
|
95.00
|
|
Auditory comprehension total
|
200
|
53.00
|
69.00
|
96.50
|
93.00
|
66.50
|
79.50
|
98.00
|
94.00
|
|
Repetition total
|
100
|
84.00
|
85.00
|
92.00
|
96.00
|
66.00
|
75.00
|
92.00
|
94.00
|
|
Naming and word finding total
|
100
|
24.00
|
34.00
|
90.00
|
93.00
|
63.00
|
58.00
|
85.00
|
93.00
|
|
Reading total
|
100
|
13.00
|
33.33
|
84.00
|
84.00
|
52.00
|
52.00
|
86.0
|
84.0
|
|
Writing total
|
100
|
25.00
|
42.50
|
73.00
|
89.00
|
34.5
|
37.5
|
46.0
|
74.0
|
|
Apraxia
|
60
|
90.00
|
98.33
|
98.33
|
100.00
|
95.00
|
88.33
|
98.33
|
100.00
|
|
Constructional, visuospatial total
|
100
|
18.50
|
42.00
|
53.00
|
62.00
|
69.00
|
65.00
|
84.00
|
91.00
|
|
WAB Language Quotient
|
|
38
|
50.12
|
90.3
|
92.1
|
61.9
|
66.5
|
86.8
|
89.7
|
|
WAB Cortical Quotient
|
|
45.05
|
61.58
|
88.53
|
91
|
66.45
|
69.68
|
89.78
|
92.35
|
|
WAB Aphasia Quotient
|
|
50.2
|
67.6
|
93.7
|
96.4
|
63.2
|
68.5
|
93
|
94.2
|
|
Cognitive Linguistic Quick Test
|
|
Attention
|
215
|
8.84
|
36.28
|
33.49
|
45.58
|
78.60
|
79.07
|
87.44
|
92.56
|
|
Memory
|
185
|
57.84
|
64.86
|
68.65
|
80.00
|
45.41
|
54.05
|
83.24
|
87.57
|
|
Executive functions
|
40
|
7.50
|
15.00
|
25.00
|
27.50
|
47.50
|
55.00
|
65.00
|
70.00
|
|
Language
|
37
|
40.54
|
43.24
|
72.97
|
78.38
|
40.54
|
45.95
|
67.57
|
78.38
|
|
Visuospatial skills
|
105
|
16.19
|
28.57
|
31.43
|
37.14
|
74.29
|
80.95
|
85.71
|
91.43
|
|
Composite severity
|
20
|
25.00
|
35.00
|
60.00
|
75.00
|
75.00
|
80.00
|
90.00
|
100.00
|
|
Clock drawing
|
13
|
0.00
|
0.00
|
61.54
|
76.92
|
46.15
|
61.54
|
100.00
|
92.31
|
|
Boston Naming Test
|
60
|
0.00
|
3.33
|
93.33
|
95.00
|
13.33
|
36.67
|
73.33
|
80.00
|
|
Pyramids and Palm Trees test
|
52
|
63.46
|
65.38
|
82.69
|
90.38
|
78.85
|
78.85
|
96.15
|
98.08
|
Results are provided for the WAB, Cognitive Linguistic Quick Test, Boston Naming Test,
and Pyramids and Palm Trees test. Test scores are provided for the total number of
items. See text for details. Abbreviation: WAB, Western Aphasia Battery.
Figure 2 (A) Performance on accuracy and latency for patient 84 on one task: word copy level
2; (B) performance on accuracy and latency for patient 108 on one task: naming pictures;
(C) performance on accuracy and latency for patient 25 on one task: letter-to-sound
matching; and (D) performance on accuracy and latency for patient 05 on one task:
multiplication level 1. For all graphs, accuracy is indicated on the left y-axis,
latency is indicated on the right y-axis. the x-axis indicates number of sessions
the task was completed. Linear slopes andR
2 values are plotted for accuracy and latency.
Example 2: High Language Profile, Low Cognitive Profile
The next patient, No. 108, was a 75-year-old man who suffered a stroke 29 months prior
to participation in this study. This participant presented with high performance on
language subtests but low performance on cognitive measures. That is, this participant's
LQ was 90.3 and AQ was 93.7. Patient 108 presented with a lower CQ (88.5) and generally
low scores on attention, executive functions, and visuospatial processing on the CLQT.
This patient was assigned the following tasks in treatment: picture spelling and picture
naming to address language processing and clock reading and instruction sequencing,
picture ordering, sound matching, symbol matching, voicemail to address attention,
visuospatial processing, and executive function deficits. Note that these tasks were
assigned to this participant over the course of the 10-week period, and he progressed
multiple levels of difficulty on picture spelling (levels 1 to 3) and symbol matching
(levels 1 to 2).[Fig. 2B] shows a representative sample of this participant's progression on a picture naming
task.[*] As accuracy on the task improved over the course of 11 weeks, latency also decreased.
At the end of treatment, this participant was tested on all the standardized measures
and showed improvements in LQ (90.1 to 92.1) CQ (88.5 to 91), AQ (93.7 to 96.4), as
well as improvements on all subtests of CLQT including attention (33.4 to 45.58%),
memory (68.6 to 80.0%), EF (25 to 27.5%), visuospatial skills (31.4 to 37.14%) and
on clock drawing (61.5 to 76.9%).
Example 3: Low Language Profile, High Cognitive Profile
Our third patient, No. 25, is a 77-year-old man, who suffered a stroke 168 months
and a secondary traumatic brain injury ~60 months prior to participation in the study.
Based on standardized tests, this participant presented with low performance on language
subtests but relatively higher performance on cognitive measures. That is, this participants'
LQ was 61.9, CQ was 66.4, and AQ was 63.2. This participant presented with moderate
scores on memory and language subtests on the CLQT and was within normal limits for
performance on attention, EF, and visuospatial skills. This patient was assigned the
following tasks in treatment: category identification, category matching, feature
matching, letter-to-sound matching, reading passage, sound identification, sound-to-letter
matching, word copy, and word spelling to address language processing and addition
and word ordering to address working memory. Over the 10 weeks of the study, this
patient completed 73 sessions (i.e., logged in to complete the therapy assignment).
During this period, the patient progressed to the next levels for several tasks including
such as addition (level 1 to 5), word copy (level 1 to 5).[Fig. 2C] shows a representative sample of this participant's progression through letter-to-sound
matching; as accuracy on the task improved, latency also decreased. At the end of
treatment, this participant was tested on all the standardized measures and showed
improvements on the LQ (61.9 to 66.5%) CQ (66.4 to 69.6%), AQ (63.2 to 68.5), as well
as improvements on all subtests of CLQT.
Example 4: High Language Profile- High Cognitive Profile
Patient 05 was a 56-year-old man who suffered a stroke 147 months prior to his participation
in the study. This participant presented with high language and cognitive skills,
based on standardized tests. Thus, this participants' LQ was 86.8, CQ was 89.7, and
AQ was 93. This participant shows high performance on all subtests of the CLQT including
attention, memory, EF, and visuospatial skills. This patient was assigned the following
tasks in treatment: category matching, feature matching, letter-to-sound matching,
sound-to-letter matching, map reading, picture spelling, reading passage, rhyming,
sound identification, syllable identification, word spelling to address language processing
and addition, subtraction, multiplication, division, picture ordering, word ordering,
and word problems to address cognitive processing. During the course of 10 weeks,
this participant completed 89 therapy sessions. Again, this patient progressed through
several levels for various tasks including addition (level 1 to 4), map reading (level
1 to 3), multiplication (level 1 to 2), picture spelling (level 1 to 3), subtraction
(level 1 to 2), word spelling (level 1 to 5) during the course of treatment.[Fig. 2D] shows a representative sample of this participant's progression through multiplication
level 1. At the end of treatment, this participant was tested on all the standardized
measures and showed improvements on the LQ (86.8 to 89.7), CQ (89.7 to 92.3), AQ (93
to 94.2, BNT (73 to 80%), as well as improvements on all subtests of CLQT with the
exception of clock drawing.
Discussion
The present study was aimed at developing an impairment-based, individualized treatment
plan for patients that can be delivered through an iPad. Using research evidence published
for rehabilitation of various aspects of brain damage, we first developed a series
of tasks targeted at addressing specific aspects of language and cognitive processing.
These tasks were implemented into a software platform (Constant Therapy) to provide
individualized therapy for individual patients enrolled in the study. The four case
studies described describe how therapy can be targeted for an individual with low
cognitive and language performance, high language and low cognitive performance, high
cognitive and low language performance, and high language and cognitive performance.
These examples are simply four benchmarks; however, most patients obviously fall somewhere
along these two dimensions. Importantly, the range of tasks described in this article
illustrates the breadth of impairment-based language and cognitive rehabilitation
that can be provided to patients.
Preliminary results from four patients provide some interesting observations. First,
all participants improved in their iPad-based therapy tasks in general, both terms
of accuracy and well as latency on the tasks (however, see exception about patient
84 who increased in his latency). A complete description of the progression of each
therapy task for each patient is out of the scope of this article; however, all participants
had at least one or two tasks that they did not improve during the course of the study.
Nonetheless, participants showed changes on standardized tests, although the amount
of change was variable: patient 84 showed improvements on both language and cognitive
measures, patient 108 showed more improvements on cognitive measures, patient 25 did
not show much improvement on most tasks excepts naming on the BNT, and patient 05,
who had high performances before treatment, showed some subtle changes in cognitive
measures and in reading. There are two important aspects that likely contribute to
the improvements observed. First, therapy was targeted at specific aspects of language
and cognitive processing that were deemed to be impaired during the standardized tests.
As an example, patient 84 presented with severe language impairments; therapy targeted
at copying words of different lengths improved as a function of treatment. This patient
also showed notable improvements on the writing subtest of the WAB (which measures
copying words/sentences among other things). Likewise, patient 05, who presented with
relatively high cognitive and language skills, practiced multiplication among other
therapy tasks and demonstrated improvements not only on that task, but also showed
improvements on the subtests of the WAB that assessed calculation. Therefore, impairment-based
therapy when targeted toward the right language/cognitive skill can be improved, even
in seemingly chronic patients.
A second and noteworthy observation was that when participants were encouraged to
complete homework practice in addition to their clinic sessions, they appeared to
be very diligent and motivated in completing therapy at home. Thus, patient 84 completed
48 therapy sessions, patient 25 completed 73 therapy sessions, and patient 05 completed
89 therapy sessions. Patient 108 was a control participant (thus he only completed
the therapy during his weekly therapy session), but even this individual was very
motivated to complete therapy and showed improvements on the standardized tests. Therefore,
participant motivation to complete the therapy was likely another factor that facilitated
the positive findings in this study. Clearly, a caveat to these results is that all
four cases discussed here have shown remarkable improvements. A larger study underway
will allow us to systematically examine the amount of improvements as a function of
therapy practice.
In the meantime, there are some tentative preliminary conclusions we can draw from
this study. There is a huge need to continue long-term therapy for individuals with
chronic brain damage; however, insurance limits to therapy reimbursement and physical
limitations are huge barriers for these individuals to effectively obtain continued
rehabilitation services. Recent technological advances, especially smart tablets and
Internet-based applications, provide a unique way to empower these individuals to
take control of their rehabilitation, by providing them access to these technologies.
The key part of access to tablet- and mobile-based technologies is the collaborative
and interactive aspect that allows patients to continue therapy outside the traditional
clinical setting, such as at their home, and stay connected with their clinician to
manage their rehabilitation program. These advances in technologies have the potential
to reshape the way rehabilitation is conducted for individuals who require ongoing
communication therapy but struggle to find practical and financially viable options
to continue their rehabilitation. As Van de Sandt-Koenderman notes, “The role of the
clinician will then shift to one of an advisor and orchestrator of the rehabilitation
process. Based on careful diagnostics at all three levels of aphasia rehabilitation,
the clinician can choose which treatment approach is needed and offer relevant treatment
programs that enable the client to work on his or her own rehabilitation, independently
and at his or her own pace.” [5]
(p. 26) For researchers in rehabilitation research, the ability to use these emerging technologies
provides new and exciting opportunities to examine the effectiveness of different
rehabilitation approaches.
