Cost Effectiveness of an Electromechanical Gait Trainer for Ambulation Training after Stroke in A Singaporean Community Hospital: A Single Blind Randomised Trial

Special Article – Gait Rehabilitation

Phys Med Rehabil Int. 2016; 3(6): 1102.

Cost Effectiveness of an Electromechanical Gait Trainer for Ambulation Training after Stroke in A Singaporean Community Hospital: A Single Blind Randomised Trial

Joyce Chua¹*, Alexander Haines² and Mark Perry³

¹Principal Physiotherapist, Inpatient Therapy Services, St. Andrew’s Community Hospital, Singapore

²Senior Health Economist, National Guideline Centre, London

³Assistant Professor, Singapore Institute of Technology, Singapore

*Corresponding author: Joyce Chua, Principal Physiotherapist, Inpatient Therapy Services, St. Andrew’s Community Hospital, Singapore

Received: October 05, 2016; Accepted: October 27, 2016; Published: November 01, 2016


Background and Purpose: Electromechanical gait trainers (GT) combined with conventional physiotherapy may have equivalent or better efficacy than conventional physiotherapy alone when retraining ambulation in sub-acute stroke patients. However, no studies have measured effects on quality of life or health status, or evaluated cost effectiveness.

Methods: This randomised controlled trial involved 106 non-ambulant individuals recruited approximately one month post-stroke. Both groups received treatment 6 times per week for 8 weeks. The GT group received 20 minutes of GT training and 5 minutes of stance/gait training in contrast to 25 minutes of stance/gait training for the conventional physiotherapy group, and both groups completed 10 minutes of standing and 10 minutes of cycling. Health status was measured with the Stroke Impact Scale (SIS) at baseline, 4, 8, 12, 24 and 48 weeks. Relative cost effectiveness of one treatment over the other was also assessed.

Results: There were no significant group x time or group differences for any outcomes. Given this equivalence, a cost-minimisation analysis was conducted. GT combined with conventional physiotherapy was S$4.63 less than conventional therapy alone per session per patient, and remained costsaving across 99.45% of 10,000 simulations when the analysis was conducted probabalistically. Sensitivity analyses showed that this result depended on the number of times equipment was used across its lifetime in combination with therapist and therapy assistant (TA) pay ratios.

Conclusion: GT combined with conventional physiotherapy is as effective as conventional physiotherapy applied alone for sub-acute stroke survivors, and can be considered an efficient use of healthcare resources.

Keywords: Stroke; Walking; Gait trainer; Quality of life; Cost-effectiveness


GT: Gait Trainers; SACH: St Andrew’s Community Hospital; FAC: Functional Ambulation Category; BI: Barthel Index; SIS: Stroke Impact Scale; ANOVA: Analysis of Variance; MMSE: Mini Mental State Examination; TA: Therapy Assistant; ICER: Incremental Cost Effectiveness Ratio.

Ethics Approval

St. Andrew’s Community Hospital Research Ethics Committee approved this study. Participants gave written informed consent before data collection began.

Source(s) of Support

Funding from St. Andrew’s Community Hospital for all research costs, including S$10 for each participant for transport at each followup.


Strokes can be devastating, and one of the most disabling effects can be the loss of the ability to walk. Approximately 80% of stroke patients are left with ambulation difficulties, which may be both severe and persistent [1]. It is also a common condition, affecting at least one in six middle-aged Americans [2]. Improving ambulation is therefore a major goal in stroke rehabilitation [3].

Conventional gait-retraining methods, where therapists facilitate normal movements, are effective in improving ambulation [4]. However, the patient’s weight needs to be supported by therapists during the process, making this very labour intensive. Harnessassisted treadmill training can alleviate the weight-bearing problem, but the complex movements of gait are often difficult to control using this method [5]. This may explain why a Cochrane review [6] showed that such treadmill training is no more effective than conventional physiotherapy. To address the limitations of both of these approaches, ‘robotic assisted locomotor trainers’ that can both support weight and directly simulate the complex patterns of the gait cycle have been developed.

Of the two main types of robotic assisted locomotor trainers, ‘gait-trainers’ (GT) may be more effective than ‘exoskeleton’ devices, based on their relative effects compared to conventional therapy [7-20]. Athough the GT appears to work best when combined with conventional therapy, and applied to non-ambulant individuals in the acute or sub-acute stages [21], overall findings in comparison to conventional therapy have been conflicting. Some studies have found positive effects for GT approaches [8,9,11,13,21,22], but others have not noted a difference [7,10,12,23].

No studies, apart from our clinical efficacy study [23], have monitored quality of life or health status as an outcome. Such an outcome should encapsulate all aspects of a patient’s return to health [21]. In addition, no previous study has examined the cost effectiveness of a GT approach compared to a conventional physiotherapy approach, despite a Cochrane review call for a costeffectiveness study [21]. Cost effectiveness is of particular relevance in the context of stroke, due to the large economic burden associated with it. The aim of this study was to evaluate the cost effectiveness of a combined GT approach relative to conventional therapy, using a health status variable.


This cost-effectiveness study is based on the same participants and interventions as a published clinical effectiveness study [23]. For brevity the combined GT and conventional physiotherapy group will be referred to as the ‘GT approach group’ where appropriate.


Patients were recruited from all inpatients admitted for stroke at St Andrew’s Community Hospital (SACH), Singapore. All patients gave informed consent and the study was approved by the hospital ethics research committee, conforming to the Helsinki Declaration.Patients were recruited from all inpatients admitted for stroke at St Andrew’s Community Hospital (SACH), Singapore. All patients gave informed consent and the study was approved by the hospital ethics research committee, conforming to the Helsinki Declaration.

Inclusion criteria were unilateral hemorrhagic/ischemic stroke, age between 18 and 80 years, and independent ambulation pre-stroke. Exclusion criteria were >8 weeks post-stroke, Functional Ambulation Category (FAC) >4, cardiovascular instability, Mini Mental State Examination (MMSE) score <16, communication deficits and lower limb joint contractures.

The study was powered for the primary outcome measure as described in the clinical effectiveness study [23]. The target size for each group was 53 patients.

Randomisation and blinding

Randomisation to the two parallel groups (GT combined with conventional physiotherapy versus conventional physiotherapy) was performed in a 1:1 allocation ratio using computer randomisation. An independent department generated the random allocation sequence, and transferred the sequence to a series of serially numbered opaque envelopes, which were not opened and revealed until after acceptance into the study and the baseline tests. Data assessors were blinded to group allocation. An intention-to-treat approach was used, and participants failing to complete either intervention were asked to return for follow up.


In line with previous findings [8,9,11,13,21], it was decided that the group receiving GT should also receive some conventional physiotherapy to optimise effectiveness. This group was compared to an independent group only receiving conventional physiotherapy.

The GT approach group received 20 minutes of electromechanical gait training and 25 minutes of conventional physiotherapy (5 minutes of stance/gait, 10 minutes of cycling, 10 minutes of tilt-table standing) 6 days/week for 8 weeks. The conventional physiotherapy group received 45 minutes of conventional physiotherapy (25 minutes of stance/gait, 10 minutes of cycling, 10 minutes of tilt-table standing) 6 days/week for 8 weeks. Stance/gait training focused on postural alignment, lower limb stepping exercises in supported standing, and over-ground walking.

Patients in the GT approach group were strapped to the GT harness, which initially gave from 10 to 20% weight support, and was reduced as appropriate. The patients placed their feet into the two footplates of the electromechanical GT machine (Reha-Stim), and took step lengths of 48cm, with a ‘velocity’ of 1.4-1.8 km/h.

During the 8-week training both groups also received occupational therapy and optional acupuncture. No study-based intervention was provided for participants after the 8-week training, but all were encouraged to continue community ambulation. A diary was provided to record any walking and leg exercises after discharge.

Demographic data and potential confounders

Data were collected on age, time from stroke, gender, weight, height, side of involvement, spasticity at baseline, depression and the use of ankle supports and knee gaiters.

Outcome measures

All outcome measures were measured at baseline, and then at 4, 8, 12, 24 and 48 weeks in the inpatient treatment area at SACH. Health Status was measured by the Stroke Impact Scale (SIS) version 3.0. This is a valid and reliable measure of functioning and social wellbeing for stroke survivors [24,25], full details of which are provided in our previous work [23]. Other outcomes included the FAC, Barthel Index (BI), gait velocity and gait endurance, but the methodology and results for these are described elsewhere [23].

Statistical analysis

Because the sample size was considerably larger than 30, normality of sampling distributions was implied by the Central Limit theorem [26].

Baseline analysis: To evaluate baseline equivalence, the two groups were initially compared for demographic variables, other potential confounders such as spasticity levels, and the baseline values of the outcome variables. If any potential confounders differed across groups to a degree that could potentially influence the outcome [26] then they were later added to the Generalised Linear Measures analysis as a covariate or factor.

Follow–up analysis: Any missing data were imputed on the basis of the most recently available data as described in our previous work [23]. For all outcomes, data at 4, 8, 12, 24 and 48 weeks were compared across both groups using a Generalised Linear Measures approach (Generalised Estimating Equations), assuming an AR(1) correlation model. For each outcome, time was entered as the withinsubject variable, with group entered as the factor and time entered as the covariate. As previously stated, any suspected confounders were also added as covariates or factors. A type III full log quasi-likelihood function analysis was adopted for all outcomes. For the SIS outcomes a linear scale response model was used. Beta co-efficients with 95% confidence intervals were derived. For the group x time analysis these represented the difference in slope between the GT and control groups.

Cost-effectiveness analysis

The cost effectiveness of one treatment, relative the other, is assessed by dividing the difference in costs between treatments by the difference in outcomes between treatments [27-29]. The cost of each treatment was calculated by identifying and estimating the resource use required to undertake each procedure, using figures derived from the Human Resource department at SACH. The physical scale of the SIS was chosen as the health status outcome as the SIS domains cannot be combined to produce an overall score, and the physical subscale was deemed to be the subscale most relevant to improvements in ambulation.

An incremental cost effectiveness ratio was calculated using the following formula:

ICER= Cos t GT Cos t CT SIS (physical) GT SIS (physical) CT [email protected]@[email protected]@+=feaaguart1ev2aqatCvAUfeBSjuyZL2yd9gzLbvyNv2CaerbuLwBLnhiov2DGi1BTfMBaeXatLxBI9gBaerbd9wDYLwzYbItLDharqqtubsr4rNCHbGeaGqiVu0Je9sqqrpepC0xbbL8F4rqqrFfpeea0xe9Lq=Jc9vqaqpepm0xbba9pwe9Q8fs0=yqaqpepae9pg0FirpepeKkFr0xfr=xfr=xb9adbaqaaeGaciGaaiaabaqaamaabaabaaGcbaGaamysaiaadoeacaWGfbGaamOuaiabg2da9maalaaabaGaam4qaiaad+gacaWGZbGaamiDamaaBaaaleaacaWGhbGaamivaaqabaGccqGHsislcaWGdbGaam4BaiaadohacaWG0bWaaSbaaSqaaiaadoeacaWGubaabeaaaOqaaiaadofacaWGjbGaam4uaiaacIcacaWGWbGaamiAaiaadMhacaWGZbGaamyAaiaadogacaWGHbGaamiBaiaacMcadaWgaaWcbaGaam4raiaadsfaaeqaaOGaeyOeI0Iaam4uaiaadMeacaWGtbGaaiikaiaadchacaWGObGaamyEaiaadohacaWGPbGaam[email protected]@

To account for uncertainty around input point estimates the analysis was run probabilistically. A probability distribution was defined for each input and a Monte Carlo simulation was run whereby a value for each input was randomly selected from its respective distribution, and mean costs and outcomes were calculated by averaging across all simulations. Probability distributions were defined for each parameter by fitting distributions using the package R Risk distributions in R software.

The following probabilistic sensitivity analyses were conducted to test the robustness of the analyses’ conclusion in relation to changes in key parameters, and also to test the applicability to different settings.

Sensitivity analysis 1: In Singapore the ratio between TA and therapist pay is about 0.44, which may be low compared to western countries. Hence the cost analyses were repeated at the ratio of 0.6.

Sensitivity analysis 2: As salary costs are likely to be slightly lower in Singapore than in western countries the cost analysis was repeated with 20% increases in salary costs for both therapists and therapy assistants (TAs).

Sensitivity analysis 3: In the base case it was assumed the equipment would be used 8 times a day. In smaller practices this number may not be possible due to resource constraints. Therefore the analysis was re-run assuming only 4 uses per day.

Sensitivity analysis 4: A two-way sensitivity analysis was conducted by simultaneously changing TA to therapist pay ratio and number of equipment uses by the same levels as mentioned in sensitivity analyses 1 and 3 above.

Sensitivity analysis 5: Finally a three-way sensitivity analysis was conducted that simultaneously changed: wages, TA to therapist pay ratio and the number of equipment uses per day.


Recruitment occurred from June 2011 to July 2014, with the last follow up occurring in June 2015. Recruitment was stopped once the sample target size had been reached. 106 patients were randomly allocated to the groups. Seven patients were lost to follow up in the GT approach group and 13 were lost to follow up in the conventional physiotherapy group at 12 months (Figure 1). Imputation of data meant that analysis was performed on a full data-set.