Measuring Outcomes for Children with Cerebral Palsy Who Use Gait Trainers

Gait trainers are walking devices that provide additional trunk and pelvic support. The primary population of children using gait trainers includes children with cerebral palsy (CP) functioning at Gross Motor Function Classification System (GMFCS) levels IV and V. A recent systematic review found that evidence supporting the effectiveness of gait trainer interventions for children was primarily descriptive and insufficient to draw firm conclusions. A major limitation identified was the lack of valid, sensitive and reliable tools for measuring change in body structure and function, activity and participation outcomes. Twelve different clinical tools were identified in the systematic review and in this paper we review and discuss the evidence supporting their reliability, validity and clinical utility for use with children using gait trainers. We also describe seven additional clinical measurement tools that may be useful with this intervention and population. The Pediatric Evaluation of Disability Inventory (PEDI) rated highest across all areas at this time. Individualized outcome measures, such as the Canadian Occupational Performance Measure (COPM) and Goal Attainment Scaling and measuring user satisfaction with tools, such as the Quebec User Evaluation of Satisfaction with assistive Technology, show potential for gait trainer outcomes research. Spatiotemporal measures appear to be less useful than functional measures with this intervention and population. All tools would benefit from further development for use with children with CP functioning at GMFCS levels IV and V.


Introduction
Cerebral Palsy (CP) is the most prevalent diagnosis for children with significant motor impairments who are followed by physical and occupational therapists [1]. This diagnosis results in lifelong progressive deforming secondary sequellae, most likely as a direct result of immobility. One of the newest revealed pressing issues is that change in muscle physiology including the loss of stem (satellite) cells and increase in collagen and muscle stiffness may be a result of non-use [2].
The Gross Motor Function Classification System (GMFCS) [3] classifies children with CP according to their mobility level and need for assistive devices. Children classified as GMFCS levels IV and V are unable to walk with hand-held walkers due to lack of trunk/postural control, strength, range of motion (ROM) and balance. These children require more supportive walkers that provide additional trunk and pelvic support, as well as unweighting. These types of walkers are often referred to as gait trainers, but the term support walker is also found in the literature [4]. Following onset of independent ambulation, children at GMFCS levels I-III are unlikely to use gait trainers except for activities that are enhanced by a device that is hands-free and/or offers increased postural support with or without body weight support. The use of a gait trainer may be one methodology to increase activity and muscle use and thus decrease contracture and disuse weakness for children at GMFCS levels IV and V.
A recent systematic review of outcomes of gait trainer use in children [5] identified mainly descriptive evidence supporting a positive impact on walking distance, mobility level and ability to take steps. Some studies suggested a positive impact on bone mineral density and bowel function along with a variety of activity and participation outcomes, but overall the evidence was low level and strong conclusions could not be reached. This review identified a limited number of clinical measurement tools that have been used to evaluate impact of gait trainer interventions for children. The review suggests that lack of reliable sensitive outcome measures may be a factor contributing to lack of stronger intervention research in this area. However, outcome measurement was not the focus of this previous paper and no critical appraisal of the tools was conducted.
The purpose of this review is to summarize the clinical measurement tools available to assess outcomes relevant to children with CP who are functioning at GMFCS IV and V as these are primarily the population of children who are using gait trainers; to critically appraise the psychometric properties of these tools and to identify which International Classification of Functioning, Disability and Health (ICF) [6,7] domains they address. We defined a gait trainer as a walker that provides additional trunk and pelvic support and is designed to be used in typical home or school settings. Our specific clinical question was "For children aged 18 years or younger, with CP, what clinical measurement tools have been used to assess gait trainer outcomes; what additional clinical measurement tools may be appropriate for this intervention and population; what are their psychometric properties; and which ICF domains do they address?"

Materials and Methods
We built on the search that was completed for the systematic review [5] and updated it in January 2016. Databases included CINAHL; Medline; EMBASE; and EBM Reviews. Search terms included "gait trainer", "supported walking", "support walker", "walking device", "body weight support gait trainer", "walker", "supported ambulation" and "David Hart Walker". Articles from the systematic review [5] were included if they provided a description of the clinical tools used and information on reliability, validity and/or clinical utility. We defined a clinical tool as one that could be used by a physical or occupational therapist in a family home or community, school or typical therapeutic setting. We specifically searched for tools that could be used for evaluative purposes rather than those designed to classify level of disability or discriminate between children with and without disabilities or delays. Outcome measures that require specialized medical or hospital-based equipment and are not able to be completed in a typical clinical environment were excluded, e.g., X-ray, bone mineral density scans, specialized gait analysis, etc. We included questionnaire-type assessments describing typical performance and designed to be completed by child or parent-proxy report. Capacity assessments completed by professional observation could include observation of physical abilities in real time or reviewing later via video-taping. Tempero-spatial measures could include use of a measuring tape or wheel, and stop-watch or video-camera, but would exclude specialized gait analysis that would require a gait lab setting or specialized camera or computer technology to complete. We excluded studies that did not provide detail on the clinical measurement tools or information on their psychometric properties.
In order to identify additional tools that might be appropriate for use with children with CP who use gait trainers, the search was expanded to include non-peer reviewed articles and conference proceeding submissions regarding children using gait trainers. We also searched for intervention studies involving other types of mobility interventions for children classified as GMFCS level IV and/or V and clinical tool development articles including children classified as GMFCS level IV and/or V. Search terms included "cerebral palsy" and "gait" or "walk* (* includes other word endings e.g. walker, walking, etc.)" in combination with terms such as "measure", "tool" or "assess* (* includes other word endings such as assessment or assessed)". We also hand searched for further articles including each of the tools identified in the electronic search by searching for them by name. We contacted known researchers and presenters on this topic in case there were other studies on these tools.
Knowing that a limited number of clinical tools have been used with gait trainer interventions, we also hand searched the reference lists of several systematic reviews of mobility and other motor interventions for children classified as GMFCS IV and V for potentially relevant tools. We purposely selected reviews of interventions that we considered to be commonly used by community-based therapists with our selected population, and that had some characteristics related to use of gait trainers. These included power mobility, adaptive seating, as well as supported walking and movement interventions that did not meet the inclusion criteria for the gait trainer systematic review [5], such as treadmill training or larger institutional or powered walking devices. We did not include reviews of general therapy interventions (e.g., Neurodevelopmental Therapy, Constraint induced movement therapy, Intensive motor therapies, etc.) or reviews of hippotherapy or aquatic therapy.
Reviews selected for hand-searching of reference lists were identified through the electronic search process using the search terms "cerebral palsy", "power mobility", "motor intervention", "treadmill training", "adaptive seating", "participation" and "review". All electronic searches were completed using the databases listed above. Any reviews identified in the electronic search were read full text to determine if they provided information on the clinical tools used in their included studies. If the studies included additional tools that had not previously been identified as having been used with gait trainer interventions, the individual articles were then read full-text.
The reviews selected from this process included seating interventions [8], power mobility interventions [9], motor interventions [10], and treadmill training [11], as well as reviews of measurement tools suitable for children with CP [12], children using seating interventions [13], and children using power mobility [14]. The additional tools were only included based on agreement of the two reviewers if at least one study evidencing psychometric properties of these clinical tools and their use in measuring intervention outcomes with children with CP classified as GMFCS level IV or V was identified.
Data on included tools were extracted independently from identified articles by reviewers using the McMaster Outcome Measures Rating Form [15]. This tool was used to rate reliability and validity results as well as to provide information on clinical utility of the tools and to provide an overall utility rating. The McMaster Outcome Measures Rating Form provides a standardized form for recording the reliability, validity, responsiveness and clinical utility of tools. It also provides a means of recording ICF domains. Ratings of excellent, adequate and poor take into account the conduct of the studies as well as the psychometric results. Individual reliability coefficients are considered excellent if above 0.8, adequate at 0.6-0.79 and poor if less than 0.6.
Tools are considered to have excellent reliability if supported by more than 2 well-designed studies with adequate to excellent reliability values; adequate reliability if supported by 1-2 well-designed studies with adequate to excellent reliability ratings; and poor reliability if only supported by poorly designed studies or poor reliability ratings. Similarly, validity would be rated excellent if supported by more than two well-designed studies supporting the measure's validity, adequate if supported by 1-2 well-designed studies and poor if the studies were poorly completed or did not support the measure's validity. Responsiveness would be rated as excellent if more than two well-designed studies showed strong hypothesized relationships between changes on the measure and other measures of change on the same attribute; adequate if supported by 1-2 well-designed studies and poor if the studies were poorly completed or did not support the measures responsiveness. Clinical tools were rated adequate or excellent if these psychometric properties had been demonstrated with children functioning at GMFCS levels IV and/or V. Overall ratings also take into account clinical utility with this population.

Results
Eleven of the original 17 articles identified in the previous systematic review of gait trainer outcomes [5] were included in this new review specific to outcome measurement [16][17][18][19][20][21][22][23][24][25]. These articles described 12 different clinical measurement tools that have been used in gait trainer outcomes research with children functioning at GMFCS levels IV and V and provided information on the tools psychometric properties and clinical utility.
In the expanded search, methodological studies providing further information on the Top Down Motor Milestone Test [33,36], use of the 10-m walk test with children with CP GMFCS levels I-III [37,38], and the PAMS [39] were identified. Only one gait trainer intervention study not included in the systematic review was identified [40]. It used the WeeFIM mobility scale [34,35] to discriminate between mobility function with and without a specialized gait trainer and orthotic device.
In the expanded search seven other clinical tools that have been used in mobility-related interventions with children with CP and appeared relevant to review regarding potential for use in gait trainer outcomes research were identified and agreed by two reviewers. All articles and tools were identified through the expanded electronic search and the hand search of the reference lists in the previously identified systematic reviews. These include: 6-min walk test [41]; Canadian Occupational Performance Measure (COPM) [42]; Early Activity Scale for Endurance (EASE) [43]; Goal Attainment Scaling (GAS) [44]; Matching Assistive Technology and Child (MATCH) [45]; Quebec User Evaluation of Satisfaction with assistive Technology 2.0 (QUEST 2.0) [46]; and the Supported Walker Ambulation Performance Scale (SWAPS) [47]. Figure 1 shows a flow diagram of the search. outcomes research with children functioning at GMFCS levels IV and V and provided information on the tools psychometric properties and clinical utility. The tools include the 10-meter (10-m) walk test [26,27], 10-minute (10-min) walk test [27], Directional Mobility Assessment (DMA) [24,25], Gross Motor Function Measure (GMFM) [28], Pediatric Evaluation of Disability Inventory (PEDI) [29], Physical Abilities and Mobility Scale (PAMS) [30], School Function Assessment (SFA)-Travel subscale [31], Step/leg movement count [16,20,21,32], Step length and velocity footprint analysis [24,25], Top Down Motor Milestone Test (TDMMT) [33], Videoing indices of happiness [19,21], and the Functional Independence Measure for children (Wee FIM) [34,35].

Tool Identification
In the expanded search, methodological studies providing further information on the Top Down Motor Milestone Test [33,36], use of the 10-m walk test with children with CP GMFCS levels I-III [37,38], and the PAMS [39] were identified. Only one gait trainer intervention study not included in the systematic review was identified [40]. It used the WeeFIM mobility scale [34,35] to discriminate between mobility function with and without a specialized gait trainer and orthotic device.
In the expanded search seven other clinical tools that have been used in mobility-related interventions with children with CP and appeared relevant to review regarding potential for use in gait trainer outcomes research were identified and agreed by two reviewers. All articles and tools were identified through the expanded electronic search and the hand search of the reference lists in the previously identified systematic reviews. These include: 6-min walk test [41]; Canadian Occupational Performance Measure (COPM) [42]; Early Activity Scale for Endurance (EASE) [43]; Goal Attainment Scaling (GAS) [44]; Matching Assistive Technology and Child (MATCH) [45]; Quebec User Evaluation of Satisfaction with assistive Technology 2.0 (QUEST 2.0) [46]; and the Supported Walker Ambulation Performance Scale (SWAPS) [47]. Figure 1 shows a flow diagram of the search.   Table 1 provides a summary of the measurement properties of the clinical tools included in this review. Readers are directed to the published systematic review [5] for evaluation of the evidence provided in the gait trainer intervention studies. A number of the included tools merited poor McMaster Outcome Measure [15] ratings, as psychometric properties have not been established for children with CP functioning at GMFCS levels IV and V.  [29] Performance measure involving structured parent interview. Measures functional abilities and caregiver assistance in self-care, mobility and social function.

Tool Characteristics
Good reliability in CP [50,51] Good validity in children with CP reported in multiple studies.
MCID-between 6 and 15 points or an 11% change on all scales [52] Excellent Excellent Adequate SFA Travel subscale [31] Performance measure involving parent or caregiver report. Criterion referenced judgment-based questionnaire with 3 domains. The travel subscale is part of the Activity domain. Rated on 4-point scale: 1 (does not perform) to 4 (consistently performs).
Content validity from expert panel and factor analysis. Discriminative validity: Detects differences between children with/without CP [12]. Activity domain difficult for children with CP [53].

Adequate Adequate Not established
Step/leg movement counting [16,20,21,32] Participant observation capacity measure. Involves counting number of independent steps or leg movements within a set time period.

Description Reliability Validity Sensitivity and Responsiveness
Step length and velocity footprint analysis [24,25] Participant observation capacity measure.
Step length is analyzed from footprint analysis taken from pressure sensitive paper.
Good inter-rater reliability in school-aged children with CP (ICC > 70) [54] Excellent test-retest reliability for children GMFCS I-III ICC = 0.86 SEM 0.14 m/s [54] Face validity: Used with children using hands-free gait trainers in two studies [24,25] Poor Poor Not established Top Down Motor Milestone Test (TDMMT) [33] Performance measure completed by school and therapy team. Assesses sitting, standing and walking skills in individuals with severe and profound multiple disabilities.  [42] Performance level individualized outcome measure-completed by child or parent proxy in semi-structured interview. Clients identify and rate importance of up to five goals in the areas of self-care, productivity and leisure. Performance and satisfaction of these goals are rated from 1-10 and difference from baseline to follow-up (change score) is used as an outcome. The manual recommends parent-proxy rating for children below 8 years of age.  [44] Performance level individualized outcome measure completed by child and/or parent proxy. Clients select goals and rate these on a 5 point scale where 0 = expected level of achievement; +1 = somewhat more than expected; +2 = much more than expected; 1 = somewhat less than expected;´2 = much less than expected. Overall score calculated by incorporating goal outcome scores into a single aggregated T score.

Internal consistency
Inter-rater reliability Good to excellent ICC 0.82 (95% CI 73-91) Children with CP GMFCS I-V [63] Excellent inter and intra-rater reliability in children with CP at a range of GMFCS levels ICC = 0.96 (95% CI 93-97 and 94-98, respectively) [64] Content validity established for use with children with CP at a range of GMFCS levels [64] Convergent validity with COPM for children with hemiplegic CP [59] Responsive to change in activity goals in children CP at all GMFCS levels [65] Likert scale GAS more sensitive than weighted GAS goals or COPM in hemiplegic CP [59]. Responsive to change in gross motor goals at a range of GMFCS levels [ Abbreviations in Table 1 See Appendix A for a brief summary of the six intervention and three methodological or tool development studies providing information on additional clinical tools that have been used with children functioning at GMFCS levels IV and/or V and may have utility for evaluating outcomes of gait trainer interventions.
Functional measures, such as the PEDI [29] and GMFM [28], are reported to have excellent inter, intra and test-retest reliability with children with CP, but evaluation of all the articles published on these widely known measures is beyond the scope of this paper. The COPM [42] was rated as having adequate reliability due to the limited number of studies relevant to the specific population being considered in this paper whereas GAS [44] is reported to have excellent reliability with children at all GMFCS levels.

Validity
Face and content validity is evident for most of the included measures but only the most relevant examples are listed in the table in the interests of space. Some measures used factor analysis (e.g., SFA [31]) while others (e.g., SWAPS [47]) were developed through expert consensus. The SWAPS [47] and DMA [24,25] demonstrate construct (convergent) validity with a moderate correlation with the GMFM [28]. The PAMS [30,39] demonstrates moderate to strong correlation with the WeeFIM [34,35]. Discriminative validity is demonstrated by some measures (e.g., EASE [43]) being able to discriminate between GMFCS levels [43] and others (e.g., WeeFIM [34,35]) being able to discriminate between mobility levels with and without a specific mobility device [40]. However, this does not indicate that a measure is suitable for evaluative purposes with this population.

Responsiveness and Sensitivity to Change
Limited information was available for the majority of tools identified. Responsiveness indicates that a measure is able to detect clinically meaningful change and sensitivity or the ability to detect any change is an important first step. Although the GMFM [28] is reported to have good responsiveness to change in children with CP regardless of GMFCS level, many of the items cannot be completed by children using gait trainers and therefore it is limited in its utility for assessing change related to gait trainer interventions. The PEDI [29] has been found to be sensitive to change in children with CP and a scaled score change of 11 points is thought to represent a clinically important change in an inpatient population [71]. The COPM [42] has been shown to be sensitive to change in children with mild CP (GMFCS I and II equivalent) although parents have rated a significant increase in independence in children with severe CP (GMFCS IV and V equivalent) using power mobility [61].
The 10-m walk test [26,27] has been heavily studied with adult populations and a change of more than 0.5 s is thought to be greater than measurement error [70]. However, studies with children with CP suggest much greater variability between assessments with differences as great as 17 s [37] and 10 s [38] being measured in children functioning at GMFCS level III. Since no studies have compared measurement properties of the 10-m walk test [26,27] with children at GMFCS level IV and V, it is unclear whether or not greater variability between performances would be present in this population. Table 2 provides a summary of the clinical utility of all the measures identified in this review. Clinical utility was rated using the McMaster Outcome Measures rating form and includes consideration of the clarity of the instructions, format, time needed to complete, need for specialized training or equipment, availability of the measure and instruction manual as well as cost. Consideration of usefulness for children using gait trainers would also include whether or not the measure can be used with the child in the gait trainer or not. Excellent overall utility indicates that the tool has adequate to excellent clinical utility, is easily available and has excellent reliability and validity. Adequate overall utility indicates that the tool has adequate to excellent clinical utility, is easily available and has adequate to excellent reliability and validity. Poor overall utility indicates that the tool has poor clinical utility for this population or intervention, is not easily available or has poor reliability and validity. A poor score in any of these factors would result in an overall score of poor overall utility.

Clinical Utility
Several outcome measures identified are freely available and require little specialized equipment other than a tape measure or stop watch. Counting steps or positive affect instances from video or at the time are easily accomplished in clinical practice and ways to increase reliability are included in the relevant articles [16,[19][20][21]32]. Detailed instructions for administration of the 10-m [26,27] and 10-min [27] walk tests are provided in the citing article [23]. In addition, detailed instructions for use with adult patients with stroke or spinal cord injury and video support materials are available online [72].
The GMFM [28] and PEDI [29] are commonly available in pediatric therapy settings, but the SFA [31] is costly and its purchase in order to use only one subscale may be prohibitive. The PAMS [30,39] appears to be a useful measure for tracking smaller changes in positioning and mobility function and tolerance in an inpatient population, but is not freely available. The DMA [24,25] was developed specifically for children with CP using gait trainers but has had limited development on its psychometric properties and is also not freely available for use. In contrast, the SWAPS scoring and administration detail, is provided in an appendix to the article [47].
The GMFM-88 [28] is most sensitive to change for children at GMFCS levels IV and V because of the increased number of test items in domains A and B. If clinicians choose to use the GMFM-66 [73] they should note that orthotics and assistive devices cannot be used during testing. The updated Gross Motor Ability Estimator (GMAE-2) software must be used (available for free, download at www.canchild.ca) and the GMFM-66 and percentiles can never be "hand-scored". However, the usefulness of the GMFM [28] for gait trainer outcomes is questionable given that function in the gait trainer cannot be directly assessed.

Measure (ICF) Clinical Utility and Usefulness for Children Using Gait Trainers Overall Utility
Tools suitable for use in clinical practice or research

PEDI [29] (A & P)
Appears sensitive to changes in mobility level in short-term intervention studies. Sensitivity to changes in caregiver assistance and impact on self-care and social domains require further exploration.

Excellent
Tools that need further development but show potential for use in clinical practice/research

COPM [42] (A & P)
Manual available for purchase. Time efficient and easy to use in OT practice. Responsiveness with children at GMFCS IV and V needs further development.

GAS [44] (A & P)
Freely available. Can be time consuming and difficult to score. Training is strongly recommended prior to use in effect studies with children [74]. Reliability, validity and responsiveness for children functioning at GMFCS IV and V need further development.

SWAPS [47] (A)
Freely available with scale and scoring descriptions available as an appendix to the original article [47]. Designed to measure changes in gait in children transitioning from dependent to independent walking. Limited documentation of use to measure gait changes in children at GMFCS levels IV or V. May need adaptation for use with gait trainers (Francine Malouin, personal communication 18, April 2016).

TDMMT [33] (A)
Can administer a shortened set of items increasing clinical utility. Designed to be used with the Rifton gait trainer. However not all skills included can be accomplished with a gait trainer e.g., sitting skills, or higher level walking skills such as stairs. Validity and responsiveness need further development.

Adequate
Further development of psychometric properties needed

DMA [24,25] (A)
Not available. Not included as an appendix to the articles. Only two sample items are included in the follow-up study [25]. Reliability, validity and responsiveness for children using gait trainers need further development.

EASE [43] (A)
Feasible and low burden indirect measure of a child's endurance for physical activity. Reliability, validity and responsiveness with children at GMFCS IV and V need further development.

Poor
Indices of happiness [19,21] (BSF) Simple to carry out in clinical practice. Poor QUEST 2.0 [46] (EF) Freely available. Low administrative and time burden. Reliability, validity and responsiveness for children at GMVCS levels IV and V need further development.

(A)
Simple to carry out in clinical practice. Demands full attention of assessor unless videotaped for later analysis Poor Table 2. Cont.

Measure (ICF) Clinical Utility and Usefulness for Children Using Gait Trainers Overall Utility
May be useful for children who are more active walkers or require less body weight support 6-min walk test [41] (A) Free and easy to use in clinical practice. Younger children and children with reduced cognitive abilities can be difficult to motivate to complete the test reliably [38] as can young children who use walking aids [43].
Poor 10-m walk test [26,27] (A) Testing protocol freely available and utility with children who use gait trainers documented. Simple to set up in a school or community setting. However, reliability, validity and responsiveness not yet established for children with CP functioning at GMFCS IV and V.
Poor 10 min walk test [27] (A) Testing protocol described in Willoughby [23]. Able to be used by children who use gait trainers. Relatively simple to conduct in a school or community setting. Reliability, validity and responsiveness not yet established for children functioning at GMFCS IV and V.

Poor
Utility for children using gait trainers questionable GMFM-88 [28] (A) Only 6/14 items in domain D and 8/14 items in domain E can be completed by a child using a gait trainer [24]. May show secondary changes due to increased activity in gait trainer but does not directly assess gait trainer function.

MATCH [45] (EF & PF)
The entire questionnaire is extensive and too lengthy for routine use in clinical practice. Impact on the reliability and validity from non standard use e.g., administration of selected items only [75] unclear. Poor

PAMS [30] (A)
Developed for an inpatient rehabilitation setting. Not validated for children with CP. Not freely available. Poor

SFA [31] Travel subscale (A)
Reflects typical performance rather than based on a single trial therefore not suitable for blinded assessment. Reliability, validity and responsiveness with children at GMFCS IV and V need further development.

Poor
Step length and velocity footprint analysis [24,25] (A) May be challenging to complete in the clinical settingrequires pressure sensitive paper or some type of gait recognition mat. Reliability, validity and responsiveness with children at GMFCS levels IV and V who use gait trainers need further development.

Poor
WeeFIM [34,35] (A) Able to discriminate mobility function in children with and without functional mobility aids, but may not be useful for detecting change over time in children at GMFCS levels IV and V.

Poor
Abbreviations for Table 2

Discussion
This review has identified 12 different clinical measures that have been used to measure outcomes with children diagnosed with CP functioning at GMFCS levels IV and V who use gait trainers. An additional seven measures have been used in assessment of mobility interventions with children functioning at GMFCS levels IV and V and have been evaluated regarding their potential for use in future gait trainer intervention research. The majority of tools identified address the ICF domain of activity despite the recent emphasis on maximizing participation for children with motor impairments [76].
Only one measure (PEDI [29]) has demonstrated reliability, validity and responsiveness to change in children with CP at GMFCS levels IV and V and appears to have utility for use with gait trainer outcomes research. While the GMFM [28] has excellent psychometric properties, its usefulness in gait trainer outcome research is limited. In the short term, significant change in PEDI [29] mobility level was demonstrated from use of the Hart walker in two studies [17,24]. However, it was only over the longer-term [25] that reduced need for caregiver assistance in social function and self-care skills were measured. Similarly, one study investigating effects of 12 sessions of treadmill training in children aged 2.5-3.9 years at GMFCS levels I-IV [77] also measured change in mobility level and need for caregiver assistance in mobility activities, but not in self-care skills. However, Jones and colleagues [78] measured significant change in mobility skills and reduced need for caregiver assistance in mobility and self-care tasks in subjects versus controls following 12 months power mobility intervention in children at GMFCS levels IV and V aged 14-30 months. This suggests that the PEDI [29] may be more useful as a measure of change in mobility level in shorter-term intervention studies. Impact on participation or environmental factors may only be measured in more longitudinal research and requires further exploration.
Walking (and supported/assisted stepping) remains a primary long-term goal for parents of children with CP, despite the fact that their children often see walking as exercise rather than functional mobility [79]. Many of the outcomes measured in gait trainer research focus on increasing number of steps, increasing walking speed or endurance, or quality of stepping. Most methodological articles exploring psychometric properties of gait and walking assessment tools are focused on children functioning at GMFCS levels I-III. Walking endurance tests, such as the 10-min walk test [27], have not established validity, reliability or responsiveness for children functioning at GMFCS levels IV and V.
Clinical experience suggests that many children who use gait trainers would not be able to complete this task, and perhaps the 6-min walk test [41] maybe a more appropriate length of time. However, this measure has not been used in gait trainer research and use with only a few children at GMFCS level IV has been demonstrated in studies investigating body-weight support treadmill training [77] and robotic step training [80]. One study [43] states that the 6-min walk test [41] is not valid for young children who walk with walking aids following pilot testing with children at GMFCS III. For children who are just learning to walk with assistive device, their movements are slow and labored in comparison with floor mobility and therefore may not be a valid representation of tolerance for physical activity.
This same study described the development of the Early Activity Scale for Endurance (EASE) [43], a parent-report measure of endurance for physical activity developed for use with children at all GMFCS levels. The study measured a moderate correlation with the 6-min walk test [41] for children at GMFCS I and II. They state that walking tests such as the 6-min walk test [41] are not appropriate for children under three years of age due to difficulties in following the directions. However, Mattern-Baxter [77] used the 6-min walk test [41] and the 10-m walk test [26,27] with children aged 2.5-3.9 years and report children with gait trainers being able to complete both tests at follow-up. Children needed directional locks on gait trainers as they were unable to steer, but could step forward. The EASE [43] was found to have construct validity and test-retest reliability for children at all GMFCS levels and to be a feasible and low burden method of assessing tolerance of and endurance for physical activity. However, responsiveness has not been established so it is unknown if it could be used to measure increase in endurance for physical activity following an intervention. In addition, it has a wide minimal detectible change of eight points, due to wide variability in scores.
The DMA [24,25] directly assesses functional use of the gait trainer and has initial documentation of face, content and construct validity, however it is unfortunately not available for use in practice. The SWAPS [47] can be adapted to assess level of support and stepping in a gait trainer, although in the initial development studies children were supported by an adult. Other studies [68,69] have used this measure to assess function of children with CP at GMFCS III or possibly IV using a walker or gait trainer. Inter-rater reliability testing has been completed and it is freely available for use in clinical practice. For children whose gait velocity is less than 50% of typical, clinical measures are more appropriate and sensitive to change than spatiotemporal gait measures [81,82]. There are no published gait lab studies including children at GMFCS IV and V although gait lab studies, with EMG and 3D motion analysis are considered to be the gold standard in gait analysis [83].
Improving participation in meaningful life activities is being stressed as a major focus in current pediatric rehabilitation [76]. For many children functioning at GMFCS levels IV and V, gait trainers are primarily a means of increasing activity and participation rather than as a means of training gait or progression towards independent walking. A lack of measures sensitive to measure change in functional activity and participation goals has been identified for children using gait trainers [5]. In a review of appropriate participation measures for children with CP, Sakzewski [12] suggests that the COPM [42] and GAS [44] are worth considering. The COPM [42] has been successfully used to measure significant improvement in self-selected participation goals following robot-assisted gait training [84] in a study that included one child at GMFCS IV. The COPM [42] has also been used to measure increase in independence or progress in self-selected activity goals in effectiveness research for adaptive seating [60] and power mobility [61].
Environmental and personal factors are other important influences on use of assistive technology [85]. The QUEST 2.0 [46] has been widely used to measure user satisfaction with a variety of different assistive technologies in many different populations. Although it does not have reliability, validity or responsiveness established for children with CP functioning at GMFCS IV or V, it has been previously used with parents of children with CP whose children were using a power mobility intervention [75,86]. Device acceptability is thought to be an important factor influencing device adoption and abandonment. Parents generally rated a high level of acceptability of power mobility device features other than size, weight and transportability [75]. When considering acceptability of gait trainers environmental factors, such as location of use may be influential. If the device is primarily being used in a therapeutic or educational setting, impressions of care staff may be more influential, but parent views may be more important to collect if the device is primarily going to be used in the home setting.
There are a number of limitations to be considered in this review. Although we expanded beyond peer-reviewed journals to grey literature, it was challenging to search conference proceedings and studies may have been missed. Although the term gait trainer is widely used and known particularly in the United States, other terms are used in other countries and this made the electronic search challenging. Using more generic terms such as "walker" generates a wide range of citations and was difficult to narrow down.
Since the number of tools that have been used in gait trainer interventions is limited, we attempted to identify other measures that might be useful in future gait trainer intervention research. However, searching for all potential measures that could be used with this population, would have made the search unmanageable. Instead, we chose to limit our search to an electronic search for systematic reviews of interventions that we considered likely to include children with CP, GMFCS IV and V. We have identified the specific reviews that we used as a source to hand search for additional measures and studies in order to increase replicability of our methods, but recognize that some studies or measures may have been missed.
However, a large-scale comprehensive review of interventions for children with CP [87] that did not meet our inclusion criteria also identified the same reviews on treadmill training [11] and adaptive seating [8] as providing the highest level of evidence published on these interventions. In addition, another review of gait related outcomes [88] and a review of walking ability measures published since our search took place [89], failed to suggest any additional measures that had been used in intervention studies with children functioning at GMFCS IV and V.
Some of the clinical tools identified in this review (e.g., GMFM, PEDI, COPM, GAS) have been widely used with other populations and interventions and it was not within the scope of this paper to evaluate all the studies published on these tools. As a result some of the table results are listed as a summary of ratings based on the data retrieved from the articles we included. We attempted to retrieve the studies relevant to children with CP GMFCS levels IV and V but had to rely on evidence related to children classified as GMFCS I-III, other disabilities or adult research in some instances.
The McMaster Outcome Measures Rating Form [15] was used with all studies but ratings take into account whether the studies had included our identified population or an alternate population. Reliability, validity and responsiveness can only be established in regard to a specific population or setting [90,91] and therefore only studies including children functioning at GMFCS levels IV and V were considered when rating properties as excellent, adequate or poor. Tools may have demonstrated good psychometric properties with another population but could be rated as poor for CP, or poor for GMFCS IV and V.
We considered using an additional quality rating such as the Consensus-based Standards for the selection of health status Measurement Instruments (COSMIN) [92,93], but the sample size requirement for this tool means that almost all the studies included in this review would automatically receive a poor or fair rating, regardless of the conduct of the study. As a result, we determined that this rating would not provide any additional information to assist readers in discriminating between measures. Instead, we chose to divide Table 2 into sections to clarify the tools most useful for gait trainer research, for which specific populations, and which tools required further development, but recognize that this is a qualitative rating based on the consensus of the two reviewers.

Conclusions
Lack of sensitive outcomes measures has been identified as a factor in low evidence quality ratings in gait trainer outcomes research. In this study we identified 17 potential tools with the PEDI [29] attaining the highest ratings in all categories. The COPM [42] and GAS [44] show potential for measuring progress on individually selected activity and participation goals while the SWAPS [47] may be adapted to measure progress in mobility or gait function within the gait trainer. The QUEST 2.0 [46] may be beneficial for exploring user satisfaction in comparing different gait trainer features and function with different children. Gait trainer interventions have a multi-dimensional impact while, at this point, most tools have been used to measure activity level outcomes. A more comprehensive outcomes framework [94] should be used to assist in measuring and understanding contextual and functional factors, as well as including activity and participation outcomes that are meaningful to families and children. Further development of all potential tools used with children functioning at GMFCS levels IV and V is warranted.
Author Contributions: Roslyn Livingstone conceived and designed this review; Roslyn Livingstone and Ginny Paleg equally performed the review and analyzed the data; Roslyn Livingstone and Ginny Paleg both wrote the paper. Roslyn Livingstone was responsible for the bibliography and formatting. Both authors have read and approved the final manuscript.
Conflicts of Interest: Roslyn Livingstone declares no conflicts of interest. Ginny Paleg is a paid educational consultant for Prime Engineering, manufacturer of the KidWalk gait trainer. No funding was received for this manuscript and this relationship did not influence the paper in any way.