Cost-effectiveness analysis of robotic exoskeleton versus conventional physiotherapy for stroke rehabilitation in Singapore from a health system perspective

Abstract

Objectives

This study conducted a comprehensive probabilistic cost-effectiveness analysis comparing robotic exoskeleton therapy to conventional physiotherapy for stroke rehabilitation in Singapore, focusing on three patient groups categorised by their Functional Ambulation Category (FAC) scores.

Design

A probabilistic cost-effectiveness analysis was conducted alongside a non-randomised controlled study. Costs and Quality-Adjusted Life Years (QALYs) for both interventions were calculated and compared over a 6 month period.

Setting

The study was carried out at Alexandra Hospital, Jurong Community Hospital and St Luke’s Hospital in Singapore.

Participants

Individuals requiring inpatient gait rehabilitation from acute to subacute stages of stroke recovery, with FAC scores of 0–1, were included in the analysis.

Primary outcome measure

The primary outcome measure was QALYs, a composite measure combining the length and quality of life into a single value.

Results

Robotic exoskeleton therapy was found to be cost-effective compared with conventional physiotherapy across all patient groups, with Group 2 (FAC 0) showing the most favourable cost-effectiveness profile (incremental cost-effectiveness ratio (ICER): US$ 28 259.62 per QALY gained). The probabilistic sensitivity analysis demonstrated the robustness of the results, with QALY gains and the cost of the robotic exoskeleton having the largest impact on the ICER.

Conclusion

The findings suggest that robotic exoskeleton therapy is likely to be cost-effective for stroke rehabilitation in Singapore, particularly for patients with severe mobility impairments (FAC 0). The results have important implications for clinical practice, resource allocation and future research in the field of stroke rehabilitation in Singapore.

Trial registeration number

NCT05659121.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The study employs a comprehensive probabilistic cost-effectiveness analysis, using Monte Carlo simulations to account for parameter uncertainty and enhance result reliability.

• It examines cost-effectiveness across three patient groups categorised by Functional Ambulation Category scores, providing nuanced insights for different mobility impairment severities.

• The analysis is tailored to Singapore’s healthcare system, using local cost data and considering implementation within national healthcare priorities.

• The non-randomised design introduces a significant risk of selection bias due to patient self-selection for treatment groups, potentially leading to systematic differences affecting results.

• The 6 month follow-up and 12-session treatment duration may not capture the full recovery trajectory and long-term costs, potentially affecting long-term cost-effectiveness estimates.

Introduction

Stroke is a leading cause of long-term disability worldwide, with a significant impact on individuals, families and healthcare systems. In Singapore, stroke is the leading cause of adult disability and the fourth leading cause of death.^{1 2} The high incidence of stroke in Singapore highlights the critical importance of effective rehabilitation strategies to improve functional outcomes, quality of life, and independence for stroke survivors.¹

Rehabilitation plays a crucial role in improving functional outcomes for stroke patients, with the primary goal of enhancing their quality of life and independence. Traditional rehabilitation approaches, including conventional physiotherapy, have been the mainstay of stroke rehabilitation for decades. However, these methods can be labour-intensive and may have limitations in terms of the intensity and repetition of movements that can be achieved during therapy sessions.^{3 4}

In recent years, robot-assisted gait training (RAGT) has emerged as a promising approach to stroke rehabilitation, offering the potential to deliver high-intensity, repetitive and task-specific training in a more efficient and standardised manner compared with conventional physiotherapy.^{5 6} A recent meta-analysis by Mehrholz et al⁵ showed that RAGT improved the odds of recovering independent walking by approximately two times compared with conventional gait training. The study also found that RAGT was more effective when initiated early post-stroke, particularly within the first 3 months after the event.

However, the cost-effectiveness of this new technology remains a subject of debate, particularly given the significant investment required for acquisition and maintenance. In the context of Singapore’s healthcare system, which emphasises cost-effectiveness and value-based care, it is crucial to evaluate whether the potential benefits of robotic exoskeleton therapy justify its costs.⁷

This study aims to provide a comprehensive probabilistic cost-effectiveness analysis comparing robotic exoskeleton therapy to conventional physiotherapy for stroke rehabilitation in the Singaporean context. We focus on three patient groups:

1. Group 1: FAC 0 and 1, stroke acute and subacute, inpatient.

2. Group 2: FAC 0, stroke acute and subacute, inpatient.

3. Group 3: FAC 1, stroke acute and subacute, inpatient.

By calculating and comparing the costs and quality-adjusted life years (QALYs) for both interventions using a probabilistic approach, we aim to provide a robust assessment of the cost-effectiveness of robotic exoskeleton therapy.

The results of this study will contribute to the growing body of evidence on the cost-effectiveness of advanced rehabilitation technologies. It will help inform policy decisions, resource allocation and clinical practice guidelines in the field of stroke rehabilitation, particularly in the Singaporean healthcare system.

Methodology

Study design and participants

This cost-effectiveness analysis was conducted alongside the Improving Mobility Via Exoskeleton (IMOVE) programme, a non-randomised controlled study comparing robotic exoskeleton training (RET) to conventional physiotherapy for gait rehabilitation. The IMOVE study was retrospectively registered with ClinicalTrials.gov (trial registration: NCT05659121). The main clinical outcomes from this programme have been published separately.⁸ The programme was carried out at Alexandra Hospital, Jurong Community Hospital and St Luke’s Hospital, targeting individuals requiring inpatient gait rehabilitation from acute to subacute stages of stroke recovery, from tertiary rehabilitation centre to community hospitals in Singapore, with participants recruited over a 5 year period. The multi-centre recruitment approach across tertiary care (Alexandra Hospital) and community hospitals (Jurong Community Hospital and St Luke’s Hospital) aimed to ensure a diverse, representative sample of the stroke rehabilitation population in Singapore. This reflects the typical care pathway where patients transition from acute to rehabilitation settings. While the three hospitals serve diverse catchment areas with varying demographic profiles, we did not observe statistically significant differences in patient characteristics or outcomes between sites. Our programme captured approximately 50% of eligible stroke patients requiring rehabilitation at these facilities during the recruitment period. Ethical approval for both the IMOVE clinical trial and this cost-effectiveness analysis was obtained from the Domain Specific Review Board of the National Healthcare Group, Singapore (reference: 2018/00368).

Participants were eligible for inclusion if they met the following criteria: (1) age between 21 and 90 years; (2) FAC score of 0–3, indicating varying degrees of assistance required for ambulation; and (3) ability to follow instructions and safely use the robotic exoskeleton. In this study, the acute phase was defined as less than 3 months post-stroke, while the subacute phase was defined as 3 to 9 months post-stroke. Exclusion criteria included severe osteoporosis, uncontrolled medical conditions, terminal illness with life expectancy <1 year, pressure sores or wounds at the exoskeleton contact points, severe lower limb contractures or spasticity, unstable lower limb fractures, significant cognitive impairment, severe lower limb pain and pregnancy.

Eligible participants were offered the choice to undergo RET. Those who declined were invited to participate in the control group, receiving conventional physiotherapy as part of their standard care. Written informed consent was obtained from all participants or their legal representatives prior to study enrolment.

To address potential confounding from the non-randomised design, we implemented analytical approaches appropriate for our sample size. Given the small subgroup samples (particularly in Groups 2 and 3), we performed direct covariate adjustment using regression models that controlled for key baseline differences, including age, sex, time since stroke and baseline utility values. This approach allowed us to estimate treatment effects while accounting for observed confounders within the constraints of our sample size.

Additionally, we conducted sensitivity analyses using the E-value approach to quantify how strong an unmeasured confounder would need to be to nullify the observed treatment effects. We also performed stratified analyses by baseline characteristics to assess whether treatment effects were consistent across different patient subgroups. These methods provide a more robust assessment of the intervention’s effectiveness while acknowledging the limitations inherent in our study design and sample size.

A subset of participants with acute or subacute stroke (less than 9 months from stroke onset), FAC 0–1 (non-functional ambulator to dependent ambulator requiring continuous assistance), receiving inpatient rehabilitation were included in data analysis. The final sample sizes for each patient group were as follows:

• Group 1 (FAC 0 and 1): control (10 patients) intervention (21 patients).

• Group 2 (FAC 0): control (5 patients), intervention (8 patients).

• Group 3 (FAC 1): control (5 patients), intervention (13 patients).

Table 1 presents the baseline characteristics of the study participants.

Table 1. Baseline characteristics of study participants (mean±SD or count).

Characteristic	Group 1 (FAC 0 and 1)		Group 2 (FAC 0)		Group 3 (FAC 1)
	Intervention	Control	Intervention	Control	Intervention	Control
Sample size	21	10	8	5	13	5
Age (years)	59.3±11.7	58.8±14.1	62.6±10.5	55.8±21.7	57.3±12.3	61.8±8.8
p value	0.915		0.462		0.449
Sex (M/F)	16/5	6/4	6/2	3/2	10/3	3/2
p value	0.420		0.596		0.595
Time since stroke (months)	0.2±0.5	0.2±0.4	0.5±0.8	0.4±0.5	0.1±0.3	0.0±0.0
p value	0.971		0.831		0.514
Baseline utility score	0.135743914	0.226633167	0.022875935	0.238658853	0.233356129	0.271256186
p value	0.068		0.002		0.397
Ethnicity (%)
Chinese	95.2	80.0	87.5	80.0	100.0	80.0
Malay	4.8	10.0	12.5	20.0	0.0	0.0
Indian	0.0	10.0	0.0	0.0	0.0	20.0
Comorbidities (%)
Hypertension	80.0	80.0	87.5	80.0	75.0	80.0
Hyperlipidaemia	60.0	40.0	62.5	40.0	58.3	40.0
Diabetes	20.0	10.0	25.0	0.0	16.7	20.0
Atrial fibrillation	5.0	0.0	0.0	0.0	0.0	0.0

FAC, Functional Ambulation Category.

Interventions

Participants in the intervention group received 12 sessions of RET incorporated into their conventional physiotherapy session. Typical total duration of physiotherapy was 45–60 min, comprising 30 min of RET and 15–30 min of conventional physiotherapy. The RET training schedule coincided with the participant’s own conventional physiotherapy schedule, with frequencies ranging from five times a week for inpatients to once every 2 weeks for outpatients. If a participant’s FAC score remained at one or lower after the initial 12 sessions, additional six RET sessions were provided.

The EksoGT from Ekso Bionics, a Conformité Européenne-marked and Food and Drug Administration-approved device, was used for the RET intervention. Trained physiotherapists handled the exoskeleton, helping subjects wear and take off the device and performing the physiotherapy training with the exoskeleton. The EksoGT is fitted with firm handles at the back, allowing therapists to hold onto patients. It can support patients up to 100 kg (220 lbs) while enabling them to fully bear their own weight with minimal support. The device also provides trunk support, allowing patients with impaired trunk control to use it safely.

Participants in the control group continued with their conventional physiotherapy, with the frequency and type of activity recorded. They did not receive any RET sessions.

Outcome measures

The primary outcome measure for this cost-effectiveness analysis was QALYs, a composite measure that combines the length and quality of life into a single value.⁹ QALYs were calculated based on the utility values provided for each patient group at baseline and after 6 months of intervention. The formula for QALY gain was:

QALY gain = (Utility value after 6 months − Utility value at baseline) * 0.5

The multiplication by 0.5 was done because the time horizon was 6 months (half a year).

Utility values for QALY calculations were derived using the EQ-5D-5L instrument, a standardised measure of health status developed by the EuroQol Group. Patients completed the EQ-5D-5L questionnaire at baseline and at the 6 month follow-up. Responses were converted to utility values using the Singapore value set developed by Wang et al,¹⁰ which provides country-specific preference weights reflecting the sociocultural context of Singapore. This approach ensures that the QALY calculations incorporate health state valuations that are contextually appropriate for the Singaporean population. The EQ-5D-5L was administered by trained research assistants who were blinded to group allocation to minimise assessment bias.

Cost analysis

The cost analysis adopted a healthcare system perspective, considering direct costs associated with RET and conventional physiotherapy. Costs were categorised as equipment costs, personnel costs and consumables.

For conventional physiotherapy equipment, the 6 month cost per patient was calculated using the formula:

6 month cost per patient = [(Cost of device on purchase / Lifespan in years / 2) + (Annual maintenance cost / 2) + (Yearly depreciation cost / 2)] / (No. of sessions/year / 2) * (Proportion of patients using the device / 100)

This formula ensures that the cost is distributed over the actual number of patients who may benefit from the equipment. The cost of a part-time therapy assistant was also included for conventional physiotherapy, based on typical salary rates in Singapore.

For the robotic exoskeleton, the key cost parameters were:

• Purchase cost: US$ 209 642 (converted from $S 283 300).

• Annual maintenance: US$ 10 360 (converted from $S 14 000).

• Lifespan: 8 years.

The 6 month cost per patient for the robotic exoskeleton was calculated using the same formula as above, adjusted for the total number of sessions provided to all patients over 6 months (the detailed cost calculations are provided in online supplemental file 1).

All costs were converted from $S to US$ using an exchange rate of 1 $S=0.74 US$ (as of June 2024 from https://www.xe.com).

Probabilistic sensitivity analysis

To account for uncertainty in the input parameters and assess the robustness of the results, a probabilistic sensitivity analysis (PSA) was conducted. In the PSA, probability distributions were assigned to key model parameters, such as utility values and costs. The analysis involved 1000 Monte Carlo simulations, each drawing a random value from the specified distributions for each parameter.

The results of the PSA were presented using cost-effectiveness planes, cost-effectiveness acceptability curves (CEACs), and a tornado diagram. Cost-effectiveness planes plot the incremental costs and QALYs for each simulation, providing a visual representation of the uncertainty surrounding the incremental cost-effectiveness ratio (ICER). CEACs show the probability of RET being cost-effective compared with conventional physiotherapy at various willingness-to-pay (WTP) thresholds. The tornado diagram illustrates the impact of individual parameters on the ICER, ranking them from the most to the least influential.

In addition to the PSA, we conducted deterministic scenario analyses to evaluate the impact of specific assumptions on the cost-effectiveness results. These included: (1) extending the time horizon to 12 months with conservative assumptions about sustained benefits, (2) varying the robotic exoskeleton lifespan from 5 to 10 years, (3) assuming a 20% reduction in equipment costs to reflect potential future price changes, and (4) increasing the number of sessions per patient from 12 to 18. These scenarios were chosen to reflect real-world implementation variables that might influence decision-making in different healthcare settings.

WTP threshold

A WTP of US$ 50 000 per QALY gained was used to assess the cost-effectiveness of robotic exoskeleton therapy. This threshold is based on common practice in health economic evaluations in the United States and was used to facilitate international comparison.^{11 12}

Patient and public involvement

This study did not directly involve patients or the public in its design, conduct, reporting, or dissemination plans. The cost-effectiveness analysis was based on existing clinical data and economic models. Future research in this area could benefit from direct patient and public involvement to ensure that cost-effectiveness analyses reflect the priorities and experiences of individuals undergoing stroke rehabilitation.

Results

Cost calculations

The 6 month cost per patient for conventional physiotherapy equipment was US$ 11.78, with an additional US$ 2.85 per patient for a part-time therapy assistant, resulting in a total cost of US$ 14.63 per patient (Detailed calculation in online supplemental annexure A).

For the robotic exoskeleton, the 6 month cost per patient was US$ 2560.59, based on a purchase cost of US$ 209 642, annual maintenance of US$ 10 360, a lifespan of 8 years and 300 total sessions over 6 months (Detailed calculation in online supplemental annexure A).

QALY calculations

QALY gains per patient over the 6 month study period were as follows:

• Group 1 (FAC 0 and 1): Control: 0.134071523, Intervention: 0.203058345

• Group 2 (FAC 0): Control: 0.133648261, Intervention: 0.223740997

• Group 3 (FAC 1): Control: 0.133232620, Intervention: 0.199046108

Table 2 presents the utility values and QALY gains for each patient group.

Table 2. Utility values and quality-adjusted life year gains by patient group.

Group	Timepoint	Utility value		QALY gain
		Intervention	Control	Intervention	Control
FAC 0 and 1	Baseline	0.135743914	0.226633167	–	–
	6 months	0.541860603	0.494776212	0.203058345	0.134071523
FAC 0	Baseline	−0.022875935	0.238658853	–	–
	6 months	0.424606058	0.505955375	0.223740997	0.133648261
FAC 1	Baseline	0.233356129	0.214607481	–	–
	6 months	0.631448345	0.481072721	0.199046108	0.133232620

FAC, Functional Ambulation Category.

ICERs

ICERs for each patient group were:

1. Group 1 (FAC 0 and 1): US$ 36 905.64 per QALY gained.

2. Group 2 (FAC 0): US$ 28 259.62 per QALY gained.

3. Group 3 (FAC 1): US$ 38 684.21 per QALY gained.

ICERs for each patient group with 95% confidence intervals from the PSA were:

1. Group 1 (FAC 0 and 1): US$ 36 905.64 per QALY gained (95% CI: 28 430.22 to 48 672.18).

2. Group 2 (FAC 0): US$ 28 259.62 per QALY gained (95% CI: 21 347.55 to 39 105.84).

3. Group 3 (FAC 1): US$ 38 684.21 per QALY gained (95% CI: 30 105.73 to 52 241.96).

In all groups, the ICERs were below the WTP threshold of US$ 50 000 per QALY, indicating that RET was cost-effective compared with conventional physiotherapy. The most favourable ICER was observed in Group 2 (FAC 0), at US$ 28 259.62 per QALY gained.

PSA results

The PSA provides valuable insights into the robustness and variability of the cost-effectiveness estimates. The box plot of ICER distributions (figure 1) shows that Group 2 (FAC 0) has the lowest median ICER and smallest IQR, suggesting it is the most cost-effective group with the least uncertainty. Group 1 (FAC 0 and 1) has a higher median ICER and larger spread, while Group 3 (FAC 1) demonstrates the highest median ICER and largest spread.

Figure 1. Box plot of ICER distributions by patient group.ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life years.

The density plots of ICER distributions (figure 2) reinforce Group 2 as the most consistently cost-effective, with Group 1 showing more variability and Group 3 having the least favourable and most variable cost-effectiveness profile.

Figure 2. Density plots of ICER distributions by patient group. FAC, Functional Ambulation Category; ICER, incremental cost-effectiveness ratio.

The cost-effectiveness plane (figure 3) shows that most points for all groups fall below the WTP threshold of US$ 50 000 per QALY, indicating that the intervention is generally cost-effective. Group 2 clusters furthest from the threshold line, confirming its superior cost-effectiveness, while Group 3 has many points closer to or above the line, suggesting lower cost-effectiveness. The cost-effectiveness acceptability curves (figure 3) show that at the US$ 50 000/QALY threshold, Group 2 has the highest probability of being cost-effective (approximately 98%), followed by Group 1 (about 87%) and Group 3 (roughly 80%).

Figure 3. Cost-effectiveness plane and cost-effectiveness acceptability curves by patient group. FAC, Functional Ambulation Category; QALY, quality-adjusted life year; WTP, willingness to pay.

The tornado diagram (figure 4) illustrates that QALY gains for both robotic and conventional therapies have the largest impact on the ICER, highlighting the importance of accurate utility measurements in determining cost-effectiveness. The cost of the robotic exoskeleton also significantly influences the ICER.

Figure 4. Tornado diagram. FAC, Functional Ambulation Category; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year.

The scenario analyses revealed that the cost-effectiveness findings were most sensitive to the time horizon assumptions. Extending the time horizon to 12 months with conservative assumptions about sustained benefits improved the ICERs by approximately 15–20% across all groups. Reducing equipment costs by 20% improved ICERs by 12–15%, while extending the exoskeleton lifespan from 8 to 10 years improved ICERs by approximately 8–10%. These findings suggest that the cost-effectiveness of robotic exoskeleton therapy could be substantially enhanced through longer-term benefits, economies of scale and extended equipment use (details in online supplemental file 1).

Discussion

The results of this comprehensive probabilistic cost-effectiveness analysis suggest that robotic exoskeleton therapy is likely to be cost-effective compared with conventional physiotherapy for stroke rehabilitation in the Singaporean context, across all three studied patient groups. However, there are notable differences in the cost-effectiveness profiles between the groups.

Group 2 (FAC 0) showed the most favourable cost-effectiveness, with the lowest ICER of US$ 28 259.62 per QALY gained and the highest probability of being cost-effective at the US$ 50 000 threshold. This suggests that for patients with the most severe mobility limitations, robotic exoskeleton therapy could provide substantial clinical benefits at a highly favourable cost-effectiveness ratio. Group 1 (FAC 0 and 1) demonstrated improved effectiveness compared with conventional physiotherapy but with less favourable cost-effectiveness than Group 2, while Group 3 (FAC 1) had the least favourable but still potentially cost-effective profile. These findings are consistent with previous studies that have shown the potential benefits of robotic exoskeleton therapy for patients with severe impairments.13,15 Additionally, it is worth noting that robotic exoskeleton therapy can serve as a motivating factor for some patients, potentially improving adherence to therapy regimes.^{16 17} This increased motivation and adherence could contribute to better outcomes and, consequently, enhanced cost-effectiveness.

These findings have important implications for clinical practice in Singapore. They suggest prioritising the use of robotic exoskeleton therapy for patients with severe mobility impairments (FAC 0), where the cost-effectiveness is most favourable. Developing clear clinical guidelines for patient selection could help optimise the use of this technology.¹⁸ The inclusion of acute patients in all groups and the favourable results also suggest potential benefits of early implementation of robotic exoskeleton therapy, which aligns with recent literature emphasising the importance of early rehabilitation interventions.¹⁹

Investing in robotic exoskeleton technology for rehabilitation units that treat a high proportion of severely impaired stroke patients could provide consistent value for money, based on the cost-effectiveness ratios found in this study. However, it is important to integrate robotic exoskeleton therapy into comprehensive rehabilitation programmes rather than using it as a standalone intervention.⁵ To fully realise the benefits demonstrated in this analysis, comprehensive training for physiotherapists and other healthcare professionals in Singapore will be necessary.⁶ Establishing robust monitoring and evaluation systems will also be crucial to ensure that real-world outcomes and cost-effectiveness align with these analytical predictions.

The findings of this study provide new insights when compared with existing literature on the cost-effectiveness of robotic exoskeleton therapy in stroke rehabilitation, particularly in the context of Singapore’s healthcare system. Unlike some previous studies that found mixed results,¹⁰ this analysis shows consistent cost-effectiveness across different patient groups, although with varying degrees of certainty. The ICERs calculated in this study are generally lower than those found in many previous studies,¹⁰ which could be due to factors such as the specific utility values used, the Singaporean healthcare context and potentially more efficient use of the technology.

Despite the comprehensive nature of this analysis, several limitations should be considered when interpreting the results. The focus on acute and subacute patients limits generalisability to later recovery stages, and the healthcare system perspective, considering only direct costs, may not capture the full economic impact. The derivation of QALY utility values, particularly the discrepancies in baseline values between groups, could potentially bias the results. The relatively small sample size, particularly in the subgroup analyses (n=5 for control groups in both FAC 0 and FAC 1 categories), reduces statistical power and increases the uncertainty of our estimates. This limitation is partially addressed through our PSA, but nevertheless affects the precision and reliability of our findings, especially for the subgroup comparisons. The small control groups also limited our ability to perform more extensive confounder adjustment or stratification. Larger, multicentre studies would help validate these preliminary findings across more diverse stroke populations. The short-term follow-up of 6 months and relatively short treatment duration of 12 sessions may not capture the full trajectory of recovery and costs associated with rehabilitation.¹⁸ A key limitation of this study is the non-randomised controlled design where patients could choose their treatment group. This self-selection may have introduced significant bias, as patients who opted for robotic exoskeleton therapy might have had different characteristics, motivation levels or rehabilitation expectations compared with those who chose conventional therapy. These systematic differences could have influenced recovery trajectories and outcomes independently of the intervention itself. While we attempted to account for baseline differences in our analysis, residual confounding remains a possibility. Future studies should employ randomised controlled designs with proper allocation concealment to minimise such biases and strengthen causal inference regarding the effectiveness of robotic exoskeleton therapy.

Our focus on acute and subacute stroke patients (<9 months post-stroke) provides valuable insights for early intervention strategies but limits generalisability to the chronic stroke population. The effectiveness and cost-utility of robotic exoskeleton therapy may differ significantly in patients with long-standing mobility impairments, where recovery patterns and neuroplasticity mechanisms operate differently.

These limitations highlight the need for future studies to address these issues. Randomised controlled trials with larger sample sizes would minimise selection bias and enable more robust statistical analyses. Long-term follow-up studies and the inclusion of patients in chronic phases would provide a more comprehensive understanding of RET’s cost-effectiveness across different stages of recovery. Adopting a societal perspective in future economic evaluations, including indirect costs and long-term economic impacts, would provide more comprehensive insights for policymakers.²⁰ Research on the implementation of robotic exoskeleton therapy in various clinical settings across Singapore could provide insights into real-world effectiveness, resource utilisation and potential barriers to adoption. Conducting qualitative research on patient experiences and preferences would provide a more comprehensive understanding of the economic and social impact of robotic exoskeleton therapy in Singapore.

While our findings suggest favourable cost-effectiveness of robotic exoskeleton therapy, particularly for patients with severe mobility impairments, it is important to acknowledge methodological constraints beyond the research design limitations. The derivation of utility values, while using validated instruments, may not fully capture all dimensions of post-stroke quality of life that are affected by mobility interventions. Additionally, the healthcare system perspective adopted in this analysis excludes indirect costs such as productivity losses, caregiver burden and long-term social care needs, which may significantly influence the societal value of these interventions.

Furthermore, the technological landscape for rehabilitation robotics is rapidly evolving, with newer, potentially more cost-effective devices entering the market. The cost-effectiveness ratios calculated in this study are specific to the EksoGT device and may not be generalisable to other robotic exoskeletons with different acquisition and maintenance costs. Healthcare systems considering the implementation of robotic rehabilitation technologies should conduct context-specific economic evaluations that reflect their local pricing structures, healthcare delivery models and patient populations.

The PSA demonstrates the robustness of our findings across a range of parameter values, but cannot eliminate the fundamental uncertainty introduced by the study design and sample size limitations. The high probability of cost-effectiveness at the US$ 50 000/QALY threshold is encouraging but should be interpreted cautiously given these constraints. Implementation decisions should ideally be supported by additional evidence from larger, randomised studies that can provide more precise estimates of both clinical effectiveness and cost-utility.

Conclusion

This comprehensive probabilistic cost-effectiveness analysis provides a favourable view of robotic exoskeleton therapy in stroke rehabilitation within the Singaporean healthcare context. The results suggest that this technology may not only improve patient outcomes but also provide good value for money across different patient groups, particularly for those with severe mobility impairments. However, the decision to adopt this technology should be made carefully, considering factors such as implementation challenges, training requirements and long-term sustainability.

The findings of this study provide a strong rationale for increased investment in robotic exoskeleton technology and further research to optimise its use in stroke rehabilitation in Singapore. Key recommendations include prioritising the implementation of robotic exoskeleton therapy for patients with severe mobility impairments (FAC 0), developing tailored approaches to patient selection, integrating the technology into comprehensive rehabilitation programmes, conducting further research to address the limitations of this study, implementing robust monitoring and evaluation systems and investing in workforce development.

As the field of stroke rehabilitation continues to evolve, ongoing economic evaluation will be crucial to ensure that promising technologies like robotic exoskeletons are implemented in ways that maximise patient benefit while maintaining the financial sustainability of Singapore’s healthcare system. The results of this study contribute to the growing body of evidence supporting the use of advanced rehabilitation technologies and inform the development of policies and practices to promote their successful integration into clinical practice in Singapore and beyond.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.