Effect of Action Observation Therapy in the Rehabilitation of Neurologic and Musculoskeletal Conditions: A Systematic Review

Objective To investigate the effect of action observation therapy (AOT) in the rehabilitation of neurologic and musculoskeletal conditions. Data Sources Searches were completed until July 2020 from the electronic databases Allied and Complementary Medicine Database (via OVID SP), Cumulative Index to Nursing and Allied Health Literature, Cochrane Library, EMBASE, MEDLINE, and the Physiotherapy Evidence Database. Study Selection Randomized controlled trials comparing AOT with standard care were assessed. Musculoskeletal (amputee, orthopedic) and neurologic (dementia, cerebral palsy, multiple sclerosis, Parkinson disease, stroke) conditions were included. There were no age limitations. Articles had to be available in English. Data Extraction Two reviewers independently screened titles, abstracts and full extracts of studies for eligibility and assessed the risk of bias of each study using the Cochrane Risk of Bias Tool. Data extraction included participant characteristics and intervention duration, frequency, and type. Results The effect of AOT in different outcome measures (OMs) was referenced in terms of body structures and functions, activities and participation, and environmental factors as outlined by the International Classification of Functioning, Disability, and Health (ICF). Of the 3448 articles identified, 36 articles with 1405 patients met the inclusion criteria. Seven of the 11 meta-analyses revealed a significant effect of intervention, with results presented using the mean difference and 95% CI. A best evidence synthesis was used across all OMs. Strong evidence supports the use of AOT in the rehabilitation of individuals with stroke and Parkinson disease; moderate evidence supports AOT in the rehabilitation of populations with orthopedic and multiple sclerosis diagnoses. However, moderate evidence is provided for and against the effect of AOT in persons with Parkinson disease and cerebral palsy. Conclusions This review suggests that AOT is advantageous in the rehabilitation of certain conditions in improving ICF domains. No conclusions can be drawn regarding treatment parameters because of the heterogeneity of the intervention. AOT has been considerably less explored in musculoskeletal conditions.

the International Classification of Functioning, Disability, and Health (ICF). Of the 3448 articles identified, 36 articles with 1405 patients met the inclusion criteria. Seven of the 11 metaanalyses revealed a significant effect of intervention, with results presented using the mean difference and 95% CI. A best evidence synthesis was used across all OMs. Strong evidence supports the use of AOT in the rehabilitation of individuals with stroke and Parkinson disease; moderate evidence supports AOT in the rehabilitation of populations with orthopedic and multiple sclerosis diagnoses. However, moderate evidence is provided for and against the effect of AOT in persons with Parkinson disease and cerebral palsy. Conclusions: This review suggests that AOT is advantageous in the rehabilitation of certain conditions in improving ICF domains. No conclusions can be drawn regarding treatment parameters because of the heterogeneity of the intervention. AOT has been considerably less explored in musculoskeletal conditions. ª 2021 The Authors. Published by Elsevier Inc. on behalf of the American Congress of Rehabilitation Medicine. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).
In recent years, rehabilitation interventions have evolved to reflect new understandings of neuroscience. 1 Neuroplasticity refers to the ability of the nervous system to adapt in response to environmental or physiological changes and experiences. 2 These changes can present within the structure, function, or organization of the nervous system and may occur centrally or peripherally. Cortical reorganization can result from structural lesions within the brain and from periods of disuse or pain. 3,4 This ability to reorganize can be considered adaptive or maladaptive depending on whether it is associated with an increase or decrease in function. Restoration of maladaptive neuroplasticity may need to be actively targeted in rehabilitation programs to have the greatest chance of restoring functional abilities. 5 Neurophysiological findings in recent times have led to the emergence of novel treatment strategies that address cortical reorganization. The discovery of the mirror neuron system (MNS) is one such advancement, 6 which has led to the development of action observation therapy (AOT).
The MNS refers to a series of neurons distributed throughout the brain. This particular set of neurons activate both when one observes an action being performed or when one physically performs the action themselves. 6 The core locations of the MNS lie within the inferior frontal gyrus, dorsal premotor and inferior parietal cortex, supplementary motor area, and the supplementary temporal gyrus. 6 The MNS was first discovered in macaque monkeys when they observed another monkey or an experimenter perform an action. 7 This prompted the exploration for a similar system within humans, which was subsequently discovered in the early 1990s. 8 The presence of this cortical network is supported by brain imaging, electroencephalography, magnetoencephalography, and transcranial magnetic stimulation studies. 9 Over the past 2 decades, AOT has become a wellsubstantiated therapeutic treatment in the field of neurorehabilitation but has been minimally investigated in patients with musculoskeletal conditions. 10 AOT, which is the systematic observation of movements, facilitates engagement of the motor system as attention and is directed toward the central mechanisms that influence movement quality, promoting the reorganization of cortical changes and the restoration of cognitive references. 1 Thus, AOT can lead to motor learning and the building or rebuilding of a motor memory via the MNS. AOT can be performed in isolation (observing the movement only) but more commonly is followed by the physical practice of the observed movements. Individuals with limited motor ability can participate in AOT, and so adaptive plasticity can still be promoted despite physical limitations. 10 Additionally, AOT can be performed independently by patients and so maximizes the Physiotherapists time.
Despite the widespread use of AOT across a range of conditions and environments, a consensus has not yet been formulated on the optimal parameters in the implementation of this technique. The aims of this systematic review are therefore to (1) systematically review the effectiveness for AOT in improving impairment and functional outcomes in patients with neurologic and musculoskeletal conditions and (2) establish whether optimal parameters for the administration of AOT exist.

Methods
The protocol of this review was registered and published at PROSPERO, https://www.crd.york.ac.uk/prospero/ registration number CRD42018116029.

Search strategy
A literature search was performed with the assistance of a medical librarian using the following electronic databases: Allied and Complementary Medicine Database (via OVID SP), Cumulative Index to Nursing and Allied Health Literature, Cochrane Library, EMBASE, MEDLINE, and the Physiotherapy Evidence Database. The search strategy was limited from 2008 to July 2020 and the English language only. Previously identified search terms were used; additionally each database was analyzed for predefined Medical Subject Headings of the National Library of Medicine terms. To ensure relevancy, a proximity search of 5 words was used. The following are examples of the search terms used: "action observation," "visual feedback," "action simulation," "motor simulation," and "mirror neuron * ."

Study identification
Articles retrieved in the initial search strategy were imported into EndNote, the reference management software. After the cross-referencing and removal of duplicates, the remaining articles were screened by title and abstract by 2 independent researchers. The references were selected following the inclusion and exclusion criteria (box 1). Eligible articles were sourced in full text and independently read by the same 2 researchers. The final number of articles that fulfilled the criteria was selected through discussion ( fig 1). No disagreements arose in the selection process, and so no third party was consulted. Data detailing participant characteristics along with the duration, frequency, and type of intervention were extracted from the included studies.

Risk of bias
The Cochrane Risk of Bias (RoB) 2.0 tool (table 8.5a in the Cochrane Handbook for Systematic Reviews of interventions) 11 was used by the 2 independent researchers to assess the RoB of each study. Any disagreement encountered was resolved through discussion. The RoB was classified as high, low, or some concerns in accordance with the criteria. The domains assessed are outlined in fig 2. Results are displayed using the robvis tool. 48

Data synthesis
As a particular strength of the International Classification of Functioning, Disability, and Health (ICF) is its focus on the functioning abilities of the individuals, recognizing the interaction between an individual's health condition, personal factors, and environmental factors, the ICF will be referenced as a framework to articulate the findings of this review. 49 A best evidence synthesis was used across the outcome measures (OMs). This qualitative analysis was performed based on a modified version of the 5 levels of evidence as outlined by van Tulder (box 2). 50 For this synthesis, studies with a low RoB were considered high quality, while studies with some concerns or a high RoB were considered low-quality studies. Where studies provided sufficient homogeneity, a meta-analysis was performed in RevMan 5.3 using a random effects model. Treatment effect was calculated using mean difference (MD) with 95% CIs. The MDs were calculated using the reported pre-and post means, selecting the most comparable time point in cases where there were multiple follow-up time points. SDs for the mean change were calculated using the following formula: where s is the reported SD and r is the Pearson correlation coefficient between pre-and postscores. As these correlations are very rarely reported, where they were not provided, a conservative estimate of rZ0.5 was used. Forest plots were created using this information, and the I 2 statistic was used to assess heterogeneity. Treatment effect was compared with the minimal detectable change (MDC) or the minimum clinically important difference (MCID) values where these values are available.

Outcome measures
A wide range of outcomes were reported throughout the studies and are considered under ICF framework. 11 A total of 52 OMs are listed (table 2): activities and participation (nZ31), body structure and function (nZ24), and 2 evaluated environmental factors (nZ2). Ten of the OMs assessed more than 1 domain.

Musculoskeletal conditions Amputees
One study with some concerns of bias evaluated the effect of AOT in the rehabilitation of bilateral amputees with phantom limb pain 44 (see table 1).   table 3). With respect to the information provided, it was possible to estimate the MD in both OMs. Significant between-group differences emerged for the McGill Questionnaire in favor of the AOT group, with scores decreasing more than the smallest detectable change of 5 points in this group only. 52 Similarly, VAS score estimations revealed a between-group MD, with 73% of the AOT group demonstrating an MCID (!20mm decrease) vs none in the mental visualization groups.

Orthopedic surgery
Three studies investigated the effect of AOT post total knee or hip replacements, 2 studies had a RoB with some concerns, 36 Level 3 evidence does not support AOT as an effective  intervention to improve functional status as assessed by the  Barthel Index and Lequesne Index but does support motor  recovery in the Short Form-36 Health Survey (SF-36) and the  function scale of the WOMAC (see table 3). A single lowquality study found no between-group differences for the Barthel Index and Lequesne Index but did find a significant effect of time (P<.001) for motor recovery in the SF-36, with moderate between-group effect sizes at the end of treatment (dZ0.76). 45 A low-quality study, 36 reported in participants with knee arthroplasty secondary to degenerative gonarthritis, significant between-group differences in the function scale of the WOMAC, again in favor of the AOT group with a between-group difference of À13.32, exceeding the MCID of 9.1 for the WOMAC function scale. 53 Level 2 evidence supports functional improvements in the FIM as positive results are seen in a high-quality study, with FIM absolute functional efficiency score changes being significantly different, with a between-group MD of 6.4. 15 Level 2 quality evidence supports AOT in positively influencing gait and balance measures as evaluated by the Tinetti scale and FIM motor scores (see table 3). In the Tinetti Scale, a lower quality study found no between group differences, 45 whilst a high-quality study found significant differences in changes in the Tinetti scale in favor of the AOT groups. 15 A pooled analysis of these scores from a total of 91 patients revealed a significant positive effect size of 1.45 (95% CI, 0.93-1.97) in favor of the AOT group (fig 3), with a low heterogeneity (I 2 Z0%), exceeding the MDC of 0.97 as referenced in the literature. 54 Belleli et al 15 also reported a significant change in the motor component of the FIM (PZ.01) in the AOT group, with a clinically significant change in the absolute functional gain score (MCID>22), 55 along with a reduction in the number of the walking aids needed (PZ.01). Despite more patients in the AOT group being prescribed a walker at baseline, 96.7% were mobilizing with a single crutch at discharge vs the 73.3% in the control group (PZ.01). Level 3 evidence is not in support of selecting AOT in improving balance or quality of life, as assessed by the TUG and SF-36, respectively (see table 3). A lower-quality study reported no significant between-group differences in the TUG, 36 with both groups exceeding the MCID of 2.27 seconds. 56 A separate lowerquality study reported no significant effect in the mental component of the SF-36. 45

Neurologic conditions
Cerebral palsy Six studies examined the effect of AOT in improving upper limb function in the rehabilitation of children with cerebral palsy; 4 studies had a low RoB, 16,29,41,42 1 with some concerns of bias, 17         Level 3 evidence does not support AOT in improving spasticity scores or ankle stiffness, as examined by the Modified Tardieu Scale and an electronic goniometer,  respectively (see table 3). A single low-quality study demonstrated no significant between-group MD in children with diplegia in either measure. 24 Level 2 evidence shows AOT to have no effect on spasticity in the Modified Ashworth Scale (MAS) or strength (see table 3). A single high-quality study found no significant between-group difference in MAS scores or grip strength assessed by the Jamar dynamometer in children with unilateral cerebral palsy. 42 The other 2 studies did not demonstrate a significant between-group differences in children with unilateral cerebral palsy. 29,42 The MUUL or Melbourne Assessment Scale was assessed in 1 low-quality 17 and 3 high-quality studies. 16,29,41 Significant between-group changes in MUUL scores were reported in 1 study 17 but not in the second study 41 ; the MD did not exceed the clinically significant threshold of 8.9% in either study. 58 One study found that functional score gain in the Melbourne Assessment Scale was significantly different in favor of AOT, with an estimated 15-score difference, 16 while the fourth study reported no between-group difference in the Melbourne Assessment Scale 2. 42 Level 3 evidence supports AOT's effectiveness in improving reach performance (see table 3). The mean values of the pediatric reaching test increased significantly more in the AOT group in a single low-quality study. 24 Level 2 evidence does not support AOT in improving bimanual abilities improvements or manual dexterity in children with unilateral cerebral palsy as evaluated by the ABILHAND-Kids, Jebsen- Taylor Hand Function Test, and Tyneside  Pegboard Test, respectively (see table 3). Two high-quality studies demonstrated no significant between-group difference in the ABILHAND-KIDS. 29,42 A single high-quality study found no between-group differences for the Jebsen-Taylor Hand Function Test and the Tyneside Pegboard Test. 42 Four studies which implemented a long-term follow-up found that the positive results seen post intervention continued in the long-term. 17,29,41,42 simulate the action with their affected limbs.    Level 2 evidence supports the use of AOT in improving walking, running, and jumping activities as captured in significant between-group difference for the walking, running, and jumping abilities in the Gross Motor Function Measure part E. 42

Dementia
One study with some concerns of bias examined the effects of observing hand function on cognition in older individuals with dementia 19 (see table 1).
(a) Body structure and function (i) Neuropsychological tests Level 3 evidence was not supportive of AOT for cognitive gains in populations with dementia (see table 3). No significant results were found in any of the memory function or cognition domains. Further analyses showed an improvement in face recognition tasks only.

Multiple sclerosis
A single study with a low RoB investigated the effects of AOT in adults with multiple sclerosis 39 (see table 1).
(a) Body structure and function (i) Handgrip strength Level 2 evidence supports the implementation of AOT in improving hand strength in persons with multiple sclerosis (see table 3). The right Jamar dynamometer score was significantly better in the AOT group vs the control group (PZ.04), with only the AOT group exceeding the MCID value of 2.7 kg as reported for immune-mediated neuropathies. 59 Parkinson disease Five studies with a low RoB investigated the effect of AOT in patients with idiopathic Parkinson disease, stage 1-3 on the Hoehn and Yahr scale 12,23,31,37,38 (see table 1). Four studies examined the effect of AOT on freezing of gait (FOG). 12,31,37,38 The fifth study examined gait patterns, assessing spatiotemporal walking variables. 23 (a) Activities and participation (i) Balance (ii) Walking ability Level 1 evidence supports the use of AOT in improving static and dynamic balance in patients with Parkinson disease (see table 3). The BBS and Tinetti part 2 were selected to assess balance. Three studies favored the AOT group in BBS scores at either short-term 31 or long-term, 12,38 revealing a significant effect for time (P<.001). A fourth study found no significant between-group difference in both the Tinetti and BBS. 37 A meta-analysis of the BBS was only possible with 3 of the studies (fig 4), revealing a positive but nonsignificant effect size of 0.56 (95% CI, À1.65 to 2.76) in 89 participants and a low heterogeneity (I 2 Z0%), with the MCID for this OM (1.9) falling within the limits of CIs. 60 Level 1 evidence supports the use of AOT in patients with Parkinson disease in improving FOG as evaluated by the FOG Questionnaire (see table 3). All 4 studies favored the AOT group, 12,31,37,38 with significant between-group differences being reported immediately post intervention 12,31 or in the long-term 4-week follow-up assessment. 37,38 Additionally, Pelosin et al 37 also found the number of FOG episodes in the AOT group to be significantly lower in the follow-up period 4 weeks post intervention (P<.001). A meta-analysis of the 3 studies pooling results from 107 participants revealed a low heterogeneity (I 2 Z13%) and a significant positive effect size, with the intervention group decreasing in score by 1.38 times that of the control group (95% CI, À2.79 to 0.03) ( fig 5). Level 4 evidence is found for the use of AOT in improving functional gait and mobility as assessed by the TUG or 10MWT in 4 of the studies (see table 3). 12,31,37,38 No between-group differences were found in the TUG in 2 studies, 31,37 while Pelosin et al 38 found the improvements to be maintained only in the AOT group at the 4-week follow-up. A meta-analysis was possible on the TUG scores in 2 studies; the pooled results from 82 participants revealed a nonsignificant effect of À0.75 (95% CI, À3.62 to 2.11) and a low heterogeneity (I 2 Z0%)( fig 6). The lower  value in the MCID range of 2-5 seconds falls within the CI range. 61 Two studies found no between-group differences in the 10MWT, 37,38 while 1 study 12 found between-group improvements presented at an earlier time point in the AOT group, exceeding the MDC of 0.18m/s. 61 Level 2 evidence supports the 6-minute walk test but does not support the Tinetti part 1 scale (see table 3) because 1 study found significant between-group differences in the 6-minute walk test at the second follow-up, 31 while the second study, which had no physical practice of AOT, found no betweengroup difference in the Tinetti part 1. 37 (b) Combined body structure and function, activities and participation, or environmental (i) Disease-specific health (ii) Functional abilities Level 1 evidence supports the PDQ-39, which assesses Parkinson diseaseespecific health, as indicated with favorable results in the AOT groups (see table 3). Three studies found significant improvements in the AOT group only, either in the short-term or at the 1-or 3-month follow-up. 12,23,31 A fourth study, found no between-group differences. 37 A metaanalysis for 3 of the 4 studies revealed low heterogeneity of the pooled studies (I 2 Z0%) (fig 7). While results from the included 66 participants revealed a nonsignificant effect of À1.04 (95% CI, À7.99 to 5.90), the MCID (À4.72) for this OM does fall within the range of the CI. 62 Level 1 evidence supports AOT in improving Unified Parkinson Disease Rating Scale (UPDRS) scores in individuals with Parkinson disease (see table 3). Two studies assessed motor and nonmotor abilities using the UPDRS. 12,31 Performance improvements in the UPDRS II presented immediately post intervention in the AOT groups in both studies, with these being significant in the first and second follow-up (P<.05) in 1 study. 31 Similarly, the positive findings in the UPDRS III were reported in both studies; one study reported a great effect size for AOT training over the control group, 31 while the second study reported between-group MDs, with only the AOT group exceeding the MCID of À3.25 for this OM. 61 These significant changes were maintained in the final follow-up assessment in both studies. Level 2 evidence does not support the modified Parkinson Assessment scale because a single study found no significant between-group difference 31 (see table 3).

Stroke
Nineteen studies examined the effect of AOT within this population (see table 1). The effect of AOT was examined in terms of upper limb function (nZ9), including 5 studies with a low RoB, 20,22,25,40 (see table 3). A single lower-quality study assessed strength via the Motricity Index and reported no between-group differences. 18 Two high-quality studies measured spasticity using the MAS in patients with subacute first-time stroke. 25,46 Conflicting results were found. The MD in 1 study showed no significant between-group difference (P>.05; 95% CI, À0.402 to 0.624), 25 64 Overall, level 1 evidence supports the use of AOT in improving upper limb function in patients with stroke. All 7 studies found positive improvements in the Fugl-Meyer Assessment in patients with subacute stroke, ranging from 30 days to 6 months post event. Significant between-group changes in favor of the AOT group were reported in 5 studies: 4 studies were high quality and 1 study was low quality. 21,25,40,46,47 Two high-quality studies found no significant between-group differences. 20,22 A meta-analysis was conducted on 6 studies. Unfortunately, because the results in 1 study were presented as percentages of maximum recovery potential, it was not possible to deduce an effect size for this study. 40 The meta-analysis pooled results from a sample size of 271 participants and revealed both low heterogeneity (I 2 Z0%) and a positive significant large effect size of 3.42 (95%, 1.02-5.81) in favor of the AOT group, (fig 9) with the MCID (5.2 points) falling within the CI margin. 65 Level 2 evidence supports the use of AOT in improving upper limb motor ability assessed by the Wolf Motor Function Test (see table 3). Two low-quality studies selected the Wolf Motor Function Test and reported significant between-group differences. 21,26 The meta-analysis results from 65 participants revealed a nonsignificant effect size of 2.15 (95%, À3.15 to 7.46) and a low heterogeneity (I 2 Z0%) ( fig 10). The MCID of this OM (4.36), falls within the limits of the CI. 66 Level 2 evidence does not support the use of AOT in improving reach test scores in the Frenchay Arm Test; and level 3 evidence does not support improvements in the Action Research Arm Test (see table 3). No between-group difference in individuals with an acute hemiplegic stroke was found for the Frenchay Arm Test in 1 high-quality study 20 or the Action Research Arm Test in a separate low-quality study. 18 Within the latter study, 18 participants were recruited early after the stroke (3-31 days), and so the authors suggest perhaps the benefits from AOT are to be found in interventions introduced later on.
Level 1 evidence supports the use of AOT in improving activities of daily living (ADL) in patients with subacute hemiplegic stroke (see table 3). Four studies, 2 high-quality 25,46 and 2 of low-quality, 21,47 used the Modified Barthel Index to assess ADL. After intervention, the changes in scores between the intervention and control groups were significantly different in all 4 studies (P<.05). The meta-analysis pooled results revealed a significant positive effect size of 7.48 (95% CI, 5.18-9.77) and a low level of heterogeneity (I 2 Z0%) (fig 11), far exceeding the MCID of 1.85 for this OM. 67 Level 2 evidence does not support AOT in improving disability and quality of life scores in patients with stroke (table 3). A single high-quality study selected the Stroke Impact Scale to investigate disability and quality of life, reporting no between-group difference. 22 Stroke: walking ability and balance (a) Activities and participation (i) Walking ability (ii) Balance  Level 1 evidence supports the use of AOT in improving functional mobility and combined balance and falls risk, as assessed by the TUG and the Dynamic Gait Index, respectively. Four studies, 2 high-quality and 2 low-quality, which used the TUG, reported significant between-group differences (P<.05) in favor of the AOT group in chronic (>6 months) hemiplegic stoke. 27,33,34,43 The fifth study, 28 which was a crossover trial, reported TUG times significantly decreased in the AOT group between pretraining and post training 1 (P<.05) in chronic stroke (>6 months). While the sixth study reported a significant improvement in both groups, the AOT group demonstrated a more significant improvement in patients with chronic stroke (!12 months). 32 A meta-analysis revealed a significant effect size in the TUG, with the 72 experimental patients decreasing in scores by 1.96 seconds (95% CI, À2.89 to À1.03) greater than the 71 participants in the control group ( fig 12). This score is below the MDC of 3.2 for this OM. 68 Two high-quality studies reported significant betweengroup differences in favor of the experimental group in the Dynamic Gait Index, 32,35 and a third low-quality study found no between-group difference. 28 Only the score change in the intevention groups in both of these studies exceeded the MDC value (1.9) for this OM. 68 Level 1 evidence supports the use of AOT in improving walking speed in individuals with chronic stroke (>6-12 months) as assessed by the 10MWT. Four of the 6 studies reported significant between-group differences in favor of the AOT group in the 10MWT; 3 studies were highquality and 1 study was low-quality. 14,33,35,43 The fifth 34 and sixth studies, 32 both high-quality, found significant improvements in both groups, with a significant betweengroup difference favoring the intervention group (P<.05). The meta-analysis possible within 3 of the studies using the 10MWT pooling 81 participants revealed low levels of heterogeneity (I 2 Z0%). Overall, there was a significantly greater decrease in time in the experimental group, with a large effect of À1.75 (95% CI, À2.55 to À0.95) (fig 13), exceeding the MDC values of 0.1-0.2 depending on speed for this OM in patients with chronic stoke. 69 Level 2 evidence supports the use of AOT in chronic stroke (>6-12 months) to improve motor planning in gait and walking distance as assessed via the Figure-of-8 Test and 6-minute walk test, respectively (see table 3). Authors reported significant between-group difference in 2 high-quality studies for the Figure-of-8 Test 35 and the 6minute walk test, 14 with only the intervention group exceeding the MCID of 34.4m. 70 Contrastingly, level 3 evidence does not support AOT in improving ambulation status because no signicant beween-group differences were found for the functional ambulation category 28 or the modifed functional ambulation profile 30 in 2 lowquality studies.

Discussion
This systematic review included 1045 participants across 36 studies and examined the effect of AOT in rehabilitation of neurologic and musculoskeletal conditions. Level 1 and level 2 evidence supports the use of AOT in populations with orthopedic conditions, cerebral palsy, multiple sclerosis, Parkinson disease, and stroke. Level 1, representing strong evidence, supports of the use of AOT to improve OMs in Parkinson disease and stroke (see table 3). Within Parkinson disease, AOT therapy has been shown to result in improvements in balance scores, FOG, disease-specific health, and motor and nonmotor abilities. Similarly, consistently strong level 1 evidence demonstrated the effect of AOT in populations with subacute and chronic stroke  in manual dexterity, upper limb function, balance, and walking ability. Level 2, representing moderate-quality evidence, advocates the implementation of AOT into rehabilitation to improve pain, stiffness, functional efficiency, gait, and balance in persons with orthopedic conditions and to improve grip strength in persons with multiple sclerosis. Moderate evidence shows walking, jumping, and running improvements in cerebral palsy, bimanual activities, dexterity, and spasticity in this population are not supported. Similarly, while AOT is supported for walking ability and aerobic capacity in Parkinson disease, improvements in spatiotemporal variables, mobility, and balance are not supported by moderate levels of evidence. Walking skills and aerobic capacity are also supported by moderate evidence in persons with stroke, as is upper limb motor ability, while reaching ability, quality of life, and disability go unsupported.
The OM used in the studies included cognitive, motor, and nonmotor assessments, including both functioning and disability components as outlined by the multidimensional ICF model. An excellent retention rate of improvements in the medium-to long-term was seen in 11 of the 12 studies that included a follow-up period, ranging from 1-6 months. 12,17,20,22,31,37,38,[40][41][42]47 This is a clear indication of the effect AOT has in promoting neuroplasticity and subsequent motor control improvements in rehabilitation.
Twenty-two studies presented with a low RoB, while the remaining 14 scored an uncertain or high RoB. Sufficient homogeneity of the studies allowed for 11 meta-analyses to be performed, the results of which revealed a significant effect in 7 of the OMs. Differing units in the reporting of OMs or insufficient information provided were the main limitations in performing further meta-analyses. While the meta-analyses of the BBS, TUG, and PDQ-39 in Parkinson disease and the Wolf Motor Function Test in stroke failed to show a significant effect, the MCID for these values did fall within the bounds of the CIs, thus illustrating that results can be statistically insignificant but clinically significant and so the clinician must not disregard the potential positive effect of treatment too hastily. 71 Similarly, while the meta-analysis of BBT and TUG in stroke showed a significant effect of AOT, the effect size was below the reported MCID values for these OMs, again requiring judicious deliberation on the clinician's behalf. Because effect sizes and sample sizes are interrelated, it is important to judiciously consider the sample sizes.
While not an aim of this review, strong psychometric properties are associated with the listed level 1 and level 2 OMs, further validating the positive results found within these measures. The BBS is the most widely and validated OM used to asses balance in populations with neurologic conditions and is associated with high reliability, validity, and responsiveness. 60 The PDQ-39 is associated with good construct validity and meets the standard for acceptable reliability. 72 The UPDRS and FOG Questionnaire are sensitive and reliable OMs for assessing treatment intervention. 73 The BBT has excellent reliability in assessing hand function in individuals with stroke. 64 Similarly, the interrater reliability of the Fugl-Meyer Assessment to assess motor recovery after stroke is high. 74 The Barthel Index is a reliable, valid, and responsive OM to assess ADL in stroke. 75 Excellent reliability is associated with the TUG and Dynamic Gait Index, with a significant correlation found between the 2 measures. 68 Equally, the 10MWT is established as a reliable measure to assess walking speed in stroke. 73 Both the WOMAC and the FIM are valid and reliable OMs in populations with orthopedic conditions. 76,77 A wide range of AOT parameters were implemented across the studies, rendering it not possible to outline specific optimal parameters in the implementation of AOT. The length of sessions ranged from 10-60 minutes, the frequency varied from daily to twice a week, and the  duration of studies spanned 8 days to 12 weeks. Within the 7 studies that demonstrated no positive changes in OMs in the AOT groups, 3 of these studies assessed children with cerebral palsy. It is evident that AOT is not supported within this population. Factors of consideration are the participants' age and the length of the sessions. Age ranged from 3-10 years, with the mean age being 5 years. It has been postulated that the development of the MNS runs parallel with the motor experience of the observer 78 ; it is possible that the combination of the reduced motor experience in children along with the reduced attention span may had led to the lack of progress with AOT.
Similarly, AOT was not supported in improving cognitive function in participants with dementia who observed hand movements. 19 Hand movements stimulate cortical areas that are involved in sensorimotor and cognitive processes, 79 but no physical practice post observation was incorporated into the study protocol, perhaps explaining the lack of any notable progress within the cognitive domains. A lack of activation of the MNS in individuals with Alzheimer disease, which accounts for the leading cause of dementia in older persons, has been found in functional magnetic resonance imaging studies. 80 It is reasonable to suggest that if there is a lack of presence of the MNS, then therapies targeting this neuron system are unlikely to be beneficial. Interestingly, 1 study found cognitive functions in patients with multiple sclerosis to improve in the AOT group. 39 Perhaps indicating AOT can have varying effects on cognitive function, depending on the underlying neurologic condition.
Motor imagery (MI) has been found to be effective in improving motor skills. 2 The case for incorporating MI into AOT lies in the shared neural regions within the brain that activate during both forms of therapy. 81 However, conflicting views exist regarding the benefit of incorporating MI in AOT programs. A single study directly compared the effect of combined MI and physical practice vs AOT and physical practice vs physical training alone. 27 The authors reported that only the AOT group demonstrated significant improvements in OMs. A potential explanation for this is the fact that MI is dependent on an individual's inherent capability for imagining movements. 4 AOT, however, provides the clear motor representation of the task.
Individuals do not need to have an underlying neurologic or musculoskeletal disorder to benefit from AOT. Athletes and members of the general population have benefited from this form of therapy. 10,51,82 While AOT has been widely applied in the field of neurologic rehabilitation, the question emerges why is it underexplored in musculoskeletal rehabilitation? We know that neurophysiological changes occur across the central and peripheral nervous systems in chronic musculoskeletal disorders, including sensorimotor cortical areas. 83 Strategies known to optimize neuroplasticity in the rehabilitation of musculoskeletal conditions have been called for in the literature. 5 Could AOT potentially offer the solution to the current inconsistency seen in the rehabilitation of chronic musculoskeletal conditions? The answer lies within further investigation of AOT in musculoskeletal conditions.

Study limitations
The main limitations of this systematic review are the lack of large samples sizes, the medium to high RoB identified in a number of the studies, and the risk of selection bias because only English studies published within the last 12 years were included.

Conclusions
AOT is suggested to be an effective tool in promoting neuroplasticity and motor learning, making it an important and valid consideration for the clinician. The benefit of incorporating AOT training into rehabilitation programs where motor and nonmotor improvements are a desired outcome is strongly supported in populations with Parkinson disease and stroke and moderately supported in populations with orthopedic conditions and multiple sclerosis. AOT has been considerably less explored in musculoskeletal conditions. No conclusions can be drawn regarding optimal parameters of implementation for AOT.