You are here

  • Home
  • Archive
  • Volume 9, Issue 5
  • Economic evaluation of a phase III international randomised controlled trial of very early mobilisation after stroke (AVERT)

Article Text

Article menu

Download PDFPDF

XML
Neurology
Research
Economic evaluation of a phase III international randomised controlled trial of very early mobilisation after stroke (AVERT)
  1. Lan Gao1,
  2. Lauren Sheppard1,
  3. Olivia Wu2,
  4. Leonid Churilov3,
  5. Mohammadreza Mohebbi1,
  6. Janice Collier3,
  7. Julie Bernhardt3,
  8. Fiona Ellery3,
  9. Helen Dewey4,
  10. Marj Moodie1
  11. The AVERT Trial Collaboration Group
  1. 1 Deakin University, Burwood, Victoria, Australia
  2. 2 University of Glasgow, Glasgow, UK
  3. 3 Florey Institute of Neuroscience and Mental Health - Austin Campus, Heidelberg, Victoria, Australia
  4. 4 Eastern Health, Box Hill, Victoria, Australia
  1. Correspondence to Dr Lan Gao; lan.gao{at}deakin.edu.au

Abstract

Objectives While very early mobilisation (VEM) intervention for stroke patients was shown not to be effective at 3 months, 12 month clinical and economical outcomes remain unknown. The aim was to assess cost-effectiveness of a VEM intervention within a phase III randomised controlled trial (RCT).

Design An economic evaluation alongside a RCT, and detailed resource use and cost analysis over 12 months post-acute stroke.

Setting Multi-country RCT involved 58 stroke centres.

Participants 2104 patients with acute stroke who were admitted to a stroke unit.

Intervention A very early rehabilitation intervention within 24 hours of stroke onset

Methods Cost-utility analyses were undertaken according to pre-specified protocol measuring VEM against usual care (UC) based on 12 month outcomes. The analysis was conducted using both health sector and societal perspectives. Unit costs were sourced from participating countries. Details on resource use (both health and non-health) were sourced from cost case report form. Dichotomised modified Rankin Scale (mRS) scores (0 to 2 vs 3 to 6) and quality adjusted-life years (QALYs) were used to compare the treatment effect of VEM and UC. The base case analysis was performed on an intention-to-treat basis and 95% CI for cost and QALYs were estimated by bootstrapping. Sensitivity analysis were conducted to examine the robustness of base case results.

Results VEM and UC groups were comparable in the quantity of resource use and cost of each component. There were no differences in the probability of achieving a favourable mRS outcome (0.030, 95% CI −0.022 to 0.082), QALYs (0.013, 95% CI −0.041 to 0.016) and cost (AUD1082, 95% CI -$2520 to $4685 from a health sector perspective or AUD102, 95% CI -$6907 to $7111, from a societal perspective including productivity cost). Sensitivity analysis achieved results with mostly overlapped CIs.

Conclusions VEM and UC were associated with comparable costs, mRS outcome and QALY gains at 12 months. Compared with to UC, VEM is unlikely to be cost-effective. The long-term data collection during the trial also informed resource use and cost of care post-acute stroke across five participating countries.

Trial registration number ACTRN12606000185561; Results.

  • stroke
  • rehabilitation
  • avert
  • economic evaluation
  • cost-effectiveness analysis
  • cost-utility analysis

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/bmjopen-2018-026230

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Strengths and limitations of this study

    • This is the first economic evaluation assessing the cost-effectiveness of a very early rehabilitation intervention within the largest phase III randomised controlled trial in patients with stroke.

    • The study assessed the long-term cost and cost-effectiveness of this very early rehabilitation intervention at 12 month.

    • The difficulty posed by the multi-country design of the trial and the percentage of missing data may undermine the confidence in the results.

    Introduction

    Stroke is one of the biggest killers and a leading cause of disability worldwide.1 2Sixty five per cent of stroke survivors live with some degree of disability that impedes their ability to carry out daily living activities unassisted.3 Therefore, ways of improving the outcomes of patients after stroke is an important focus of research.4 5 Early mobilisation after stroke is believed to contribute to better patient outcomes and clinical trials have been conducted globally.6–9

    The short-term efficacy and safety of a very early rehabilitation trial after stroke (AVERT) has been evaluated in a phase III randomised controlled trial (RCT) with 2104 patients enrolled from Australia, New Zealand, UK, Singapore and Malaysia.10 The evidence from this trial indicated that at 3 months after stroke, very early mobilisation (VEM) of patients was associated with a reduction in the probability of a favourable outcome as defined by a modified Rankin Scale (mRS) score of 0 to 2 compared with that in the usual care (UC) group.10 In the research field of stroke, primary endpoint is usually assessed at month 3 after stroke,11–14 which means there is a paucity of data in terms of long-term resource use and cost of care for patients with stroke. Given AVERT provided a longer-term (ie, 12 months) comprehensive measurement of costs relating to stroke care (ie, direct medical, direct non-medical and indirect costs), and the broader representativeness of patients across countries and regions (>2000 patients were recruited from both developing and developed world), together with the implications of stroke economic burden sustained beyond the acute phase (ie, 3 months), holistically examining the cost of stroke care that falls within health and non-health sectors could potentially advance understanding of pattern of resource use post stroke and identify any gaps to improve care for stroke and chances to curb the increasing economic burden of disease. This examination also benefits healthcare funders (ie, governments, insurance companies) and the public with addition of substantial knowledge of long-term rehabilitation cost for stroke.

    This economic evaluation, which was part of the registered trial protocol (Australian New Zealand Clinical Trials Registry, ACTRN12606000185561) and planned prior to knowledge of outcomes, was conducted alongside the phase III RCT,10 The aim of this paper is to assess the cost-effectiveness of very early mobilisation within 24 hours after stroke in terms of improving patient outcomes at 12 months, in comparison to usual care, with a particular focus on examining the resource use and cost of care after stroke.

    Methods

    The economic analysis was undertaken following the previously published plan.15 It also conforms to the Consolidated Health Economic Evaluation Reporting Standards checklist.16 Ethics approval was granted by relevant institutions.

    Intervention and comparator

    The trial design has been reported in detail elsewhere.10 In brief, patients with confirmed stroke who were admitted to a stroke unit within 24 hours of stroke onset were randomised to receive usual stroke-unit care alone or VEM in addition to UC in a multinational phase III trial.

    Outcomes

    The mRS at 12 months, a secondary outcome of the trial, and quality-adjusted life years (QALYs) derived from the assessment of quality of life-4 dimension (AQoL-4D)17 were used as the effectiveness measures in the economic evaluation. The AQoL-4D instrument is a multi-attribute utility scale used to assess the health-related quality of life (HRQoL)18; it was administered at 3 and 12 months.

    Outcome of mRS was dichotomised into ‘favourable’ (mRS 0 to 2) and ‘poor’ (mRS 3 to 6) based on patients outcomes at 12 month follow-up.10 The difference in the probability of patients achieving a favourable mRS outcome (mRS 0 to 2) was used to estimate the incremental benefits between treatment groups for the primary efficacy outcome.

    Due to the inherent difficulties of administering the AQoL instrument to acute stroke patients (ie, most of patients were not able to respond to these questions at baseline), the mRS score at baseline10 was used as a surrogate measure of patient utility during the acute phase. The detailed methods of this work are reported elsewhere19 and a brief description is supplied in the online supplementary document 1.

    Supplemental material

    Costs

    A societal perspective with a key focus on the health sector was adopted.

    Intervention delivery

    Intervention delivery costs consisted of the time costs of physiotherapists (PT) and nurses delivering VEM (or UC) to patients. The mean of the total physiotherapist time (recorded by a log documented by each participating PT across whole hospital stay) per patient was calculated. Given insufficient data, physiotherapist’s mean time per session was used as a proxy for nurse time spent on delivering either VEM or UC.

    Resource use

    All resource use during the study period was electronically collated using a validated cost case report form (Cost CRF) administered and recorded by trained staff at 3 and 12 months using face-to-face assessments with patients and carers, and medical records. Cost CRF used in Australia is supplied as an example (online supplementary document 2). Cost CRF from other participating countries could be requested from corresponding author.

    Healthcare resource use

    The quantity of resources used for the following healthcare resource items was recorded: number of ambulance transfers (emergency and non-emergency), acute hospitalisation (including length of stay (LoS)), rehospitalisation (number of occasions and LoS for each occasion), rehabilitation hospital admission (number of occasions and LoS for each occasion), outpatient rehabilitation programme (number of occasions and number of days for each occasion), rehabilitation provided at home/nursing facility (number of occasions and number of sessions for each occasion), private physiotherapy (number of sessions), respite care (number of sessions) and individual outpatient (including physiotherapy, occupational therapy and speech and language therapy) visits (service type and number of sessions) for patients from UK, Singapore and Malaysia only.

    Non-healthcare resource use

    The quantity of resources used was recorded for the following non-healthcare resource items: accommodation move due to stroke (location moved to and date of move), community service (type of service use and number of service used both for prior to and post stroke), home modification (type of modification, supplier and cost), special equipment and aids (type of equipment/aids and quantity consumed), informal care (purpose of the care and hours used), live-in maids (number of maids prior to and post stroke) (for Singapore and Malaysia only), changes to employment (employment status and weekly hours of working both prior to and post stroke).

    Resource use reported at 3 (ie, resources used between 0 and 3 months) and 12 (ie, resources used between 4 and 12 months) months was used to calculate the total annual resource use for each participant. Generally, where patients were still using a particular resource at the time of 12 month data collection, the last day of 12 months’ follow-up (calculated from the day of index stroke) was used to estimate the duration of that resource utilisation. In the event of a patient dying, resource use data for the period prior to death was ascertained from their carer and medical records, wherever possible.

    Unit costing

    Costs were computed by applying country-specific unit costs to each resource item utilised. Therefore, five sets of unit costs (one for each of the participating countries) were compiled from the most up-to-date and reliable source (online supplementary document 3). Unit costs from a country with a similar economic status and healthcare system were used where local country-specific unit costs were unavailable.

    All costs are expressed in Australian dollars (AUD) for the 2015 reference year value and can be converted to US dollar (USD) using the purchasing power parity rate 1 USD=1.463 AUD20 (December 2015). The currency of other countries was converted to AUD using the corresponding exchange rate. The country-specific consumer price index from the health sector was employed to adjust costs not valued in the year of 2015.

    The details of unit cost for acute stroke hospitalisation, rehospitalisation, rehabilitation (inpatient and outpatient), non-health sector costs (home modifications, community services, aids, etc) and productivity cost are provided in online supplementary document 3.

    Statistical analysis

    All the costs that were attributable to stroke including healthcare costs, non-healthcare costs and productivity costs were accounted for in the economic analysis. Since a 12 month economic evaluation was undertaken, no discounting was applied to either costs or benefits.

    Quantity of resource use and costs were summarised using medians and IQRs due the skewness of the raw data. Means and SDs were also reported. Base case analysis of the economic evaluation was performed based on the intention-to-treat (ITT) population21 with an assumption for the main analysis that data were missing at random (MAR). The difference in costs was analysed using generalised linear regression model (GLM) with gamma family and a log link, with treatment groups as an independent variable, including baseline National Institutes of Health Stroke Scale (NIHSS), baseline mRS15 and age as treatment covariates.

    For the primary outcome, the mRS score at 12 months was compared following the method detailed in the statistical analysis plan.22 While for the secondary effectiveness outcome (ie, the QALY gains at 12 months), a linear regression model with treatment group as the factor variable and 12 months AQoL-4D utility value as the dependent variable, adjusted for age, baseline mRS was utilised to estimate the difference in QALY gains over 12 months. Non-parametric bootstrap simulations with 2000 replications were used to calculate 95% CIs around mean difference in costs and effects for cost-effectiveness analysis. To examine the cost-effectiveness of VEM measured against UC, incremental cost-effectiveness ratios (ICERs) were calculated where applicable. For the ICER from a societal perspective, all the costs from health and non-health sector were summed together, including the productivity cost; for ICER of a health sector perspective, all the costs borne by healthcare system were counted (ie, excluding non-healthcare costs and productivity cost). The differences between groups in terms of costs and benefits (ie, QALYs) were compared regardless of the statistical significance of the difference.23 Cost-effectiveness acceptability curves were plotted to show the probability of VEM being the optimal choice. The ICERs were compared with a common benchmark in Australia of ≤AUD50,000 per QALY.24 All the analyses were performed using the STATA V.14.0 statistical package.

    Sensitivity analyses

    To investigate the impact of using country-specific costs, a country dummy variable was added to the GLM analysis to adjust for country effect.25 Subgroup analysis on the basis of individual countries were also conducted to explore the difference in costs and benefits across countries.

    Multiple imputation was performed to test the sensitivity of results to the missing data assumption. The missing patterns were explored with the use of logit regression to investigate if any of the other variables predicted whether a given variable was missing26 (online supplementary document 4).

    Secondary analyses were undertaken to assess the robustness of the base case results. Subgroup analyses were performed at the country-specific level to test for differences in efficacy and costs.

    Patient and public involvement

    No patient and public were involved.

    Results

    Between July 2006 and October 2014, 2104 patients (VEM 1054, UC 1050) were recruited across 58 sites from Australia (n=1054, 24 sites), New Zealand (n=189, one site), UK (n=610, 29 sites), Singapore (n=128, one site) and Malaysia (n=123, one site). At recruitment, over 80% of patients had no prior history of stroke; NIHSS was greater than 7 points (indicating a moderate-to-severe stroke) for around 45% of patients; 26% aged over 80 years and 24% had received recombinant tissue plasminogen activator prior to randomisation.10 Baseline characteristics were similar between the two treatment groups.10

    Outcomes

    There was no difference between VEM and UC groups in terms of favourable mRS outcome and quality of life (as measured by AQoL-4D) at month 12. Specifically, a comparable percentage of patients from both treatment groups achieved a favourable outcome at 12 months after stroke (between-group difference in probability: 0.030, 95% CI −0.021 to 0.082, adjusted for baseline age and NIHSS). Likewise, for the outcome of AQoL-4D at 12 months, no between-group difference was observed (−0.013, 95% CI −0.043 to 0.017). The detailed mRS outcomes are presented in online supplementary document 5: table I.

    Resource use and costs

    The proportion of patients reporting use of a specific resource varied from item to item (table 1). In relation to the healthcare resource items, nearly half of the patients experienced rehabilitation hospital admission and more than a quarter of the patients had a stroke-related rehospitalisation, rehabilitation service use (outpatient/provided at home or nursing facility) and ambulant transfers whereas only a small proportion of the patients (less than 10%) recorded the use of private physiotherapy and/or respite care. Regarding non-health-related resource use, the majority of patients (>50%) used some form of special aids or equipment during the 12 months after their index stroke, while nearly 40% of patients received informal care and around 27% reported the use of community services and home modifications. Only 16% (VEM) and 17% (UC) of patients respectively, experienced accommodation changes due to the index stroke. For maid’s service use in the home in Singapore and Malaysia, a small proportion (less than 10%) of the patients hired a maid both before and after the index stroke.

    View this table:
    Table 1

    Quantity of resource use over 12 months (ITT) (median, IQR)

    With respect to productivity, nearly one in four patients were employed prior to their stroke; this proportion fell to only one in eight patients at 12 months follow-up. Generally, resource use was comparable between VEM and UC groups (p>0.05) across all items (table 1).

    The median total medical cost was marginally higher in the UC group ($20 411, IQR $7238 to $63 835) than in the VEM group ($19 271, IQR $6294 to $52 637), primarily due to the higher rehabilitation admission cost in UC. In both groups, the major cost component was acute hospitalisation which accounted for around 30% of medical costs. The median non-medical cost was also marginally higher in the UC group ($438, IQR $0 to $4561) than in the VEM group ($358, IQR $0 to $3334). The median productivity cost was zero for both treatment groups given that less than one quarter of patients were in paid employment before the index stroke. Overall, the median total cost (including productivity cost and non-medical costs) were nominally higher in the UC group ($27 042, IQR $7257 to $63 824) compared with the VEM group ($25 675, IQR $6766 to $63 617). The detailed costs of each resource item and summary costs are presented in table 2. The costs for VEM and UC interventions are summarised in online supplementary document 5: table II.

    View this table:
    Table 2

    Cost of all the resources used over 12 months (AUD)

    Generally, the cost from VEM and UC groups were comparable: the differences between VEM and UC groups was $1082 (95% CI  -$2399 to $4563) for the total medical cost (online supplementary document 6: table I) and $3 (95% CI -$5 to  $12) for the productivity cost per person at 12 months; the between-group difference in the total non-healthcare cost was -$1300 (95% CI -$3361 to  $760) over the same period of time.

    Cost-effectiveness analysis

    The between-group difference in both efficacy and cost outcomes generated from the GLM model are presented in online supplementary document 6: table I.

    In the base case health sector perspective analysis, the VEM yielded comparable total medical costs ($1082, 95% CI -$2520 to $4685, p=0.544) and QALY gains (−0.013, 95% CI −0.041 to 0.016) at 12 months. When a societal perspective was adopted, the VEM entailed, again, similar costs with the UC group ($102, 95% CI -$6907 to $7111, p=0.982, including productivity costs) or (-$6, 95% CI -$5476 to $5463, p=0.933, excluding productivity costs) (table 3).

    View this table:
    Table 3

    Baseline cost-utility analysis ITT

    The cost-effectiveness planes and cost-effectiveness acceptability curves from the two perspectives are shown in online supplementary document 7: figures I to V.

    Sensitivity analyses

    Inclusion of a country dummy variable in the analysis produced similar results to the base case (online supplementary document 6: table II).

    The analysis from imputed data including all randomised participants produced consistent results with regard to the incremental cost and effectiveness between treatment groups. For example, from a health sector perspective, VEM was associated with similar costs ($940, 95% CI $−4622 to $4682) and QALY gains (−0.019, 95% CI−0.044 to 0.005) over 12 months. (online supplementary document 6: table III and online supplementary document 7: figures VI-VIII)

    The country-specific analysis showed similar results in the between-group differences for both costs and QALYs, indicating that VEM and UC yielded comparable results within each participating countries (table 4).

    View this table:
    Table 4

    Results of country-specific analysis of costs and benefits

    When a societal perspective was assumed, again, the point estimate of difference in costs between groups across countries varied substantially, with the 95% CIs mostly overlapping (table 4).

    Discussion

    The 12 months within-trial cost-effectiveness analysis showed that VEM was unlikely to be cost-effective than UC in patients with stroke. Between-group differences in costs and benefits (probability of achieving a favourable outcome of mRS and differences in QALYs) over the 1 year study period were comparable from a health sector perspective. The findings from this economic evaluation is also underpinning an adapted version of trial under way to investigate the effectiveness of optimal rehabilitation in patients with mild-to-moderate stroke (ie, AVERT-DOSE, National Health and Medical Research Council Australia, project grant #1139712).

    Our earlier economic evaluation of the phase II AVERT trial which consisted of only 71 patients (38 VEM and 33 UC) from two Australian centres reported that VEM was likely to be a cost-effective intervention with both less cost and more benefit when compared with UC.27 Since it was a national pilot study with a limited sample, the direct comparison between the results from this and our current economic evaluation is problematic. In addition, inconsistent with the pilot study, no service shifting was observed in the current study. Across all resource use components, the proportion of patients consuming specific types of resources were comparable between the two groups in this study. On the contrary, in the phase II AVERT trial, patients from VEM group were more likely to be discharged earlier from hospital than their UC counterparts; those discharged early tended to use more care provided in the outpatient setting, which incurred lower costs and informal care was not costed. In the current study, the LoS for acute hospitalisation and rehabilitation were similar between treatment groups (median: VEM 16 vs UC 17 days). These differences between the two studies highlight the importance of large, adequately powered studies to inform healthcare policy.

    In this study, resources used were valued on the basis of country-specific unit costs sourced for each participating country. To counteract any concern arising from the adoption of this approach, extensive sensitivity analyses were performed to test the robustness of the results. The conduct of incorporation of a country dummy variable into the model or country-specific analysis did not alter the outcomes substantially, with the resultant 95% CIs overlapping to a great extent. Ramsey et al 2015 suggest that a country-specific costing approach is likely to yield few qualitative differences in summary measures of cost-effectiveness among countries with similar levels of economic development.25 Therefore, it was believed that any differences in economic status of the participating countries (as reflected by the unit costs applied in our study) are unlikely to bear a major influence on the results of the cost-effective analysis.

    This multinational trial also revealed that in managing patients post-stroke, practice of stroke care varied from country to country. Although 100% of patients with stroke were hospitalised for the initial acute care, the LoS differs significantly greatly, ranging from 4 days (Malaysia) to 25 days (New Zealand), which might be attributable to the different severity of stroke and/or differences in clinical practice care processes. Moreover, in Malaysia, patients tended to receive rehabilitation services in an outpatient (ie, up to 52% of patients received the outpatient rehabilitation programme services) rather than inpatient (ie, only up to 2% patients were admitted to rehabilitation hospital) setting; and patients were less likely to use ambulant transfer and apply home modifications, as compared with participants from other countries. This might be a signal for future study around stroke care in Malaysia, research potentially could be helpful to improve the service delivery for outpatient rehabilitation programme. Patients from western countries consumed more community services and rehabilitation services that provided at home/nursing home than their Asian counterparts, which reflects the difference in social welfare and healthcare systems.

    Economic evaluations have been conducted for other types of stroke rehabilitation interventions including early-supported discharge service, community- or home-based rehabilitation.28–36 Generally, these interventions trended towards being cost-saving measured against usual practice. In regards to HRQoL outcomes measured by a series of quality of life instruments (including SF-36, WHOQoL-Bref, Nottingham Health Profile, Sickness Impact Profile and EQ-5D), most studies did not detect an overall significant effect.28–32 34 35 37 Only one study reported a significant difference improvement in the overall HRQoL score.36 The conclusions drawn from these economic evaluations of stroke rehabilitation interventions were fairly consistent; the interventions were likely to cost less,33 34 37–42 although the difference in costs was statistically significant in only one study.41 None of these studies evaluated the costs and benefits, particularly benefits measured in terms of QALYs, in an aggregated manner, and all were limited by small sample sizes. Another study using a Markov model explored the increased intensity of physiotherapy for stroke patients from a health system perspective, concluding that increased physiotherapy could be cost-effective by improving health outcomes and reducing costs due to the resultant shorter stay in rehabilitation facilities.43

    Given that it is not practical to obtain a baseline utility value from patients with stroke, in this study, the baseline AQoL value was mapped from mRS score at baseline.15 While the mapping exercise was carried out using the baseline mRS score and AQoL values at 3 and 12 months, the significant variation in the mapped baseline utility values for patients falling within the same category of mRS hampered its application to the current economic evaluation. Instead, only the 12 month utility values were compared with to approximate the difference in QALY gains over 1 year between the two treatment groups. Comprehensive sensitivity analyses were undertaken surrounding this assumption. It was observed that there was no noticeable difference among approaches examining the annual QALY gain difference between VEM and UC.

    While the results from the clinical study showed that there were no significant differences in either costs or effects between treatment groups, the cost-effectiveness analysis was still performed to investigate the possible ICER of the VEM intervention. It is possible to have greater confidence in the joint outcome of costs and QALYs than looking at them individually.44

    To the best of our knowledge, this study evaluated the cost-effectiveness of the largest international acute stroke rehabilitation trial ever conducted. The cost-effectiveness analysis was performed alongside the randomised controlled trial, where the costs and benefits data were collected prospectively. Moreover, the Cost CRF was completed by trained and blinded assessors via interviews with individual patients/carers and accessing medical records, which provides for greater accuracy than resource use questionnaires or diaries completed by participants themselves. Since the trial was designed in a pragmatic manner, with close resemblance to real clinical practice, it is believed that the assessment of its cost and cost-effectiveness under this setting reflects the real-life resource use (health and non-health).

    This study provides some insights for future economic evaluation alongside multi-country, multi-centre clinical trials. It is important to note that given the large number of centres involved (56 stroke units across five geographical jurisdictions), it was not practical or reasonable to collect centre-specific unit costs which probably leads to huge variations even within a single country. Country-level unit costs were therefore applied to the valuation of resource uses across the trial sites. However, the heterogeneity in the resource utilisation and unit cost among the included countries undermines confidence in the conclusion. A country-specific economic evaluation might be more appropriate in this regard but the lacking of statistical power poses another concern. The current study made a trade-off between them both approaches by presenting both the aggregated (ie, base case of pooling all countries) and disaggregated (ie, sensitivity analysis of individual countries) form of results. The resource utilisation, costs and benefits were also tabulated across all sites and individually to allow close scrutiny from various perspectives.28 It is believed that this practice can be recommended to other multi-country studies.

    A couple of limitations of the study are acknowledged. First, the missing data on total costs from a societal perspective was around 24%, and related mainly to the missing information on community services (10.9%) and productivity loss (10.7%). The base case analysis was based on the ITT population with an assumption of missing pattern being MAR. To account for this, the sensitivity analysis using multiple imputation was undertaken and yielded the identical conclusion (ie, comparable results in costs and benefits between treatment groups). Second, unit costs originating from individual countries were assigned to value resource use. The differences in healthcare systems and cost structures among the five participating countries may potentially confound the cost comparisons between groups. However, analysis by country produced results consistent with the base case, which overcomes any concern that the latter were heavily weighted towards Australia, the largest sample country.

    Conclusions

    This economic evaluation alongside a phase III RCT evidenced that based on the ITT population, the VEM intervention for patients with stroke was unlikely to be cost-effective compared with UC. The sensitivity analyses based on the multiple imputation and subgroup analyses by each country separately yielded fairly consistent results. Despite substantial differences observed, in resource use and unit costs across the countries, the marginal differences between VEM and UC were consistent. Overall, the VEM intervention was demonstrated to be comparable with UC in terms of both benefits and costs at 1 year, however given its poorer outcomes at 3 months, VEM cannot be recommended to clinicians, patients or policymakers.

    Acknowledgments

    The authors wish to acknowledge all of the in-country staff who administered the cost and resource use data collection. Thanks are also extended to Gina Harris who undertook the preliminary work to modify the Cost CRF for each country. Finally, our gratitude is expressed to all persons who participated in the AVERT trial.

    References

    1. 1.↵
      2. Feigin VL ,
      3. Norrving B ,
      4. Mensah GA
      . Global burden of stroke. Circ Res 2017;120:439–48.doi:10.1161/CIRCRESAHA.116.308413
      OpenUrlAbstract/FREE Full Text
    2. 2.↵
      2. Johnston SC ,
      3. Mendis S ,
      4. Mathers CD
      . Global variation in stroke burden and mortality: estimates from monitoring, surveillance, and modelling. Lancet Neurol 2009;8:345–54.doi:10.1016/S1474-4422(09)70023-7
      OpenUrlCrossRefPubMedWeb of Science
    3. 3.↵
      Deloitte Access Economics. The economic impact of stroke in Australia. National Stroke Foundation. 2013 https://www.deloitteaccesseconomics.com.au/uploads/File/Stroke%20Report%2014%20Mar%2013.pdf.
    4. 4.↵
      2. Forster A ,
      3. Young J ,
      4. Chapman K , et al
      . Cluster randomized controlled trial: clinical and cost-effectiveness of a system of longer-term stroke care. Stroke 2015;46:2212–9.doi:10.1161/STROKEAHA.115.008585
      OpenUrlAbstract/FREE Full Text
    5. 5.↵
      2. Winstein CJ ,
      3. Wolf SL ,
      4. Dromerick AW , et al
      . Effect of a task-oriented rehabilitation program on upper extremity recovery following motor stroke: the ICARE Randomized Clinical Trial. JAMA 2016;315:571–81.doi:10.1001/jama.2016.0276
      OpenUrlCrossRefPubMed
    6. 6.↵
      2. Indredavik B ,
      3. Bakke F ,
      4. Slordahl SA , et al
      . Treatment in a combined acute and rehabilitation stroke unit: which aspects are most important? Stroke 1999;30:917–23.
      OpenUrlAbstract/FREE Full Text
    7. 7.↵
      2. Govan L ,
      3. Langhorne P ,
      4. Weir CJ
      . Does the prevention of complications explain the survival benefit of organized inpatient (stroke unit) care?: further analysis of a systematic review. Stroke 2007;38:2536–40.doi:10.1161/STROKEAHA.106.478842
      OpenUrlAbstract/FREE Full Text
    8. 8.↵
      2. Dean SG ,
      3. Poltawski L ,
      4. Forster A , et al
      . Community-based rehabilitation training after stroke: results of a pilot randomised controlled trial (ReTrain) investigating acceptability and feasibility. BMJ Open 2018;8:e018409.doi:10.1136/bmjopen-2017-018409
    9. 9.↵
      2. Norris M ,
      3. Poltawski L ,
      4. Calitri R , et al
      . Acceptability and experience of a functional training programme (ReTrain) in community-dwelling stroke survivors in South West England: a qualitative study. BMJ Open 2018;8:e022175.doi:10.1136/bmjopen-2018-022175
    10. 10.↵
      2. Bernhardt J ,
      3. Langhorne P ,
      4. Lindley RI , et al
      . Efficacy and safety of very early mobilisation within 24 h of stroke onset (AVERT): a randomised controlled trial. Lancet 2015;386:46–55.doi:10.1016/S0140-6736(15)60690-0
      OpenUrlCrossRefPubMed
    11. 11.↵
      2. Ilyas A ,
      3. Chen CJ ,
      4. Ding D , et al
      . Endovascular mechanical thrombectomy for acute ischemic stroke under general anesthesia versus conscious sedation: a systematic review and meta-analysis. World Neurosurg 2018;112:e355–67.doi:10.1016/j.wneu.2018.01.049
      OpenUrl
    12. 12.↵
      2. Lambrinos A ,
      3. Schaink AK ,
      4. Dhalla I , et al
      . Mechanical thrombectomy in acute ischemic stroke: a systematic review. Can J Neurol Sci 2016;43:455–60.doi:10.1017/cjn.2016.30
      OpenUrl
    13. 13.↵
      2. Koh JS ,
      3. Lee SJ ,
      4. Ryu CW , et al
      . Safety and efficacy of mechanical thrombectomy with solitaire stent retrieval for acute ischemic stroke: a systematic review. Neurointervention 2012;7:1–9.doi:10.5469/neuroint.2012.7.1.1
      OpenUrlCrossRefPubMed
    14. 14.↵
      2. Papanagiotou P ,
      3. White CJ
      . Endovascular reperfusion strategies for acute stroke. JACC Cardiovasc Interv 2016;9:307–17.doi:10.1016/j.jcin.2015.11.014
      OpenUrlAbstract/FREE Full Text
    15. 15.↵
      2. Sheppard L ,
      3. Dewey H ,
      4. Bernhardt J , et al
      . Economic Evaluation Plan (EEP) for A Very Early Rehabilitation Trial (AVERT): An international trial to compare the costs and cost-effectiveness of commencing out of bed standing and walking training (very early mobilization) within 24 h of stroke onset with usual stroke unit care. Int J Stroke 2016;11:492–4.doi:10.1177/1747493016632254
      OpenUrlCrossRefPubMed
    16. 16.↵
      2. Husereau D ,
      3. Drummond M ,
      4. Petrou S , et al
      . Consolidated Health Economic Evaluation Reporting Standards (CHEERS)--explanation and elaboration: a report of the ISPOR Health Economic Evaluation Publication Guidelines Good Reporting Practices Task Force. Value Health 2013;16:231–50.doi:10.1016/j.jval.2013.02.002
      OpenUrlCrossRefPubMedWeb of Science
    17. 17.↵
      2. Hawthorne G ,
      3. Richardson J ,
      4. Osborne R
      . The Assessment of Quality of Life (AQoL) instrument: a psychometric measure of health-related quality of life. Qual Life Res 1999;8:209–24.doi:10.1023/A:1008815005736
      OpenUrlCrossRefPubMedWeb of Science
    18. 18.↵
      2. Hawthorne G ,
      3. Richardson J ,
      4. Day NA
      . A comparison of the Assessment of Quality of Life (AQoL) with four other generic utility instruments. Ann Med 2001;33:358–70.doi:10.3109/07853890109002090
      OpenUrlCrossRefPubMedWeb of Science
    19. 19.↵
      Mapping the modified Rankin Scale (mRS) measurement to the Assessment of Quality of Life (AQoL) utility values. 38th Annual Conference of the International Society for Clinical Biostatistics. Vigo, Spain, 2017. (9-13 Jul 2017).
    20. 20.↵
      OECD. StatsExtracts 4. PPPs and exchange rates. http://stats.oecd.org/Index.aspx?DataSetCode=SNA_TABLE4 (Accessed 5 Apr 2017).
    21. 21.↵
      2. White IR ,
      3. Horton NJ ,
      4. Carpenter J , et al
      . Strategy for intention to treat analysis in randomised trials with missing outcome data. BMJ 2011;342:d40.doi:10.1136/bmj.d40
      OpenUrlFREE Full Text
    22. 22.↵
      2. Bernhardt J ,
      3. Churilov L ,
      4. Dewey H , et al
      . Statistical analysis plan (SAP) for A Very Early Rehabilitation Trial (AVERT): an international trial to determine the efficacy and safety of commencing out of bed standing and walking training (very early mobilization) within 24 h of stroke onset vs. usual stroke unit care. Int J Stroke 2015;10:23–4.doi:10.1111/ijs.12423
      OpenUrlCrossRefPubMed
    23. 23.↵
      2. Claxton K
      . The irrelevance of inference: a decision-making approach to the stochastic evaluation of health care technologies. J Health Econ 1999;18:341–64.doi:10.1016/S0167-6296(98)00039-3
      OpenUrlCrossRefPubMedWeb of Science
    24. 24.↵
      2. Shih ST ,
      3. Carter R ,
      4. Heward S , et al
      . Economic evaluation of future skin cancer prevention in Australia. Prev Med 2017;99:7–12.doi:10.1016/j.ypmed.2017.01.013
      OpenUrl
    25. 25.↵
      2. Ramsey SD ,
      3. Willke RJ ,
      4. Glick H , et al
      . Cost-effectiveness analysis alongside clinical trials II-An ISPOR Good Research Practices Task Force report. Value Health 2015;18:161–72.doi:10.1016/j.jval.2015.02.001
      OpenUrlCrossRefPubMed
    26. 26.↵
      2. Garson G
      . Missing Values Analysis & Data Imputation. Statistical Associates Blue Book Series. 2015 http://www.statisticalassociates.com/missingvaluesanalysis_p.pdf (Accessed 14 May 2018).
    27. 27.↵
      2. Tay-Teo K ,
      3. Moodie M ,
      4. Bernhardt J , et al
      . Economic evaluation alongside a phase II, multi-centre, randomised controlled trial of very early rehabilitation after stroke (AVERT). Cerebrovasc Dis 2008;26:475–81.doi:10.1159/000155984
      OpenUrlCrossRefPubMed
    28. 28.↵
      2. Craig LE ,
      3. Wu O ,
      4. Bernhardt J , et al
      . Approaches to economic evaluations of stroke rehabilitation. Int J Stroke 2014;9:88–100.doi:10.1111/ijs.12041
      OpenUrlCrossRefPubMed
    29. 29.↵
      2. Donnelly M ,
      3. Power M ,
      4. Russell M , et al
      . Randomized controlled trial of an early discharge rehabilitation service: the Belfast Community Stroke Trial. Stroke 2004;35:127–33.doi:10.1161/01.STR.0000106911.96026.8F
      OpenUrlAbstract/FREE Full Text
    30. 30.↵
      2. Gladman J ,
      3. Whynes D ,
      4. Lincoln N
      . Cost comparison of domiciliary and hospital-based stroke rehabilitation. DOMINO Study Group. Age Ageing 1994;23:241–5.doi:10.1093/ageing/23.3.241
      OpenUrlCrossRefPubMedWeb of Science
    31. 31.↵
      2. Roderick P ,
      3. Low J ,
      4. Day R , et al
      . Stroke rehabilitation after hospital discharge: a randomized trial comparing domiciliary and day-hospital care. Age Ageing 2001;30:303–10.doi:10.1093/ageing/30.4.303
      OpenUrlCrossRefPubMedWeb of Science
    32. 32.↵
      2. Widén Holmqvist L ,
      3. de Pedro Cuesta J ,
      4. Möller G , et al
      . A pilot study of rehabilitation at home after stroke: a health-economic appraisal. Scand J Rehabil Med 1996;28:9–18.
      OpenUrlPubMedWeb of Science
    33. 33.↵
      2. Anderson C ,
      3. Rubenach S ,
      4. Mhurchu CN , et al
      . Home or hospital for stroke rehabilitation? results of a randomized controlled trial : I: health outcomes at 6 months. Stroke 2000;31:1024–31.
      OpenUrlAbstract/FREE Full Text
    34. 34.↵
      2. Harrington R ,
      3. Taylor G ,
      4. Hollinghurst S , et al
      . A community-based exercise and education scheme for stroke survivors: a randomized controlled trial and economic evaluation. Clin Rehabil 2010;24:3–15.doi:10.1177/0269215509347437
      OpenUrlCrossRefPubMed
    35. 35.↵
      2. Young J ,
      3. Forster A
      . Day hospital and home physiotherapy for stroke patients: a comparative cost-effectiveness study. J R Coll Physicians Lond 1993;27:252–8.
      OpenUrlPubMedWeb of Science
    36. 36.↵
      2. Sritipsukho P ,
      3. Riewpaiboon A ,
      4. Chaiyawat P , et al
      . Cost-effectiveness analysis of home rehabilitation programs for Thai stroke patients. J Med Assoc Thai 2010;93(Suppl 7):S262–70.
      OpenUrl
    37. 37.↵
      2. Anderson C ,
      3. Ni Mhurchu C ,
      4. Brown PM , et al
      . Stroke rehabilitation services to accelerate hospital discharge and provide home-based care: an overview and cost analysis. Pharmacoeconomics 2002;20:537–52.doi:10.2165/00019053-200220080-00004
      OpenUrlCrossRefPubMedWeb of Science
    38. 38.↵
      2. Beech R ,
      3. Rudd AG ,
      4. Tilling K , et al
      . Economic consequences of early inpatient discharge to community-based rehabilitation for stroke in an inner-London teaching hospital. Stroke 1999;30:729–35.doi:10.1161/01.STR.30.4.729
      OpenUrlAbstract/FREE Full Text
    39. 39.↵
      2. Larsen T ,
      3. Olsen TS ,
      4. Sorensen J
      . Early home-supported discharge of stroke patients: a health technology assessment. Int J Technol Assess Health Care 2006;22:313–20.doi:10.1017/S0266462306051208
      OpenUrlPubMed
    40. 40.↵
      2. McNamee P ,
      3. Christensen J ,
      4. Soutter J , et al
      . Cost analysis of early supported hospital discharge for stroke. Age Ageing 1998;27 345–51.doi:10.1093/ageing/27.3.345
      OpenUrlCrossRefWeb of Science
    41. 41.↵
      2. Teng J ,
      3. Mayo NE ,
      4. Latimer E , et al
      . Costs and caregiver consequences of early supported discharge for stroke patients. Stroke 2003;34:528–36.doi:10.1161/01.STR.0000049767.14156.2C
      OpenUrlAbstract/FREE Full Text
    42. 42.↵
      2. von Koch L ,
      3. de Pedro-Cuesta J ,
      4. Kostulas V , et al
      . Randomized controlled trial of rehabilitation at home after stroke: one-year follow-up of patient outcome, resource use and cost. Cerebrovasc Dis 2001;12 131–8.doi:10.1159/000047692
      OpenUrlCrossRefPubMedWeb of Science
    43. 43.↵
      2. Chan B
      . Effect of increased intensity of physiotherapy on patient outcomes after stroke: an economic literature review and cost-effectiveness analysis. Ont Health Technol Assess Ser 2015;15:1–43.
      OpenUrlPubMed
    44. 44.↵
      2. Glick HA ,
      3. Briggs AH ,
      4. Polsky D
      . Quantifying stochastic uncertainty and presenting results of cost-effectiveness analyses. Expert Rev Pharmacoecon Outcomes Res 2001;1:25–36.doi:10.1586/14737167.1.1.25
      OpenUrlPubMed

    Footnotes

    • Contributors Conception/design, revising and final approval: LG, LS, LC, OW, MMohebbi, JC, FE, JB, HD, MM. Methodology: LG, MMohebbi, LS, LC, MM, OW. Data analysis: LG, LS, MMohebbi, LC. Drafting: LG, MM.

    • Funding The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: The economic evaluation component of the AVERT study was funded by the National Health and Medical Research Council (NHMRC) of Australia (grant numbers 386201 and 1041401). Additional funding was received from Singapore Health (SHF/FG401P/2008), Chest Heart and Stroke Scotland (Res08/A114), Northern Ireland Chest Heart and Stroke and the Stroke Association UK (TSA2009/09). NHMRC fellowship funding was provided to HD (336102) and JB (1058635). JB was also supported by an Australia Research Council Future Fellowship (IFT0991086) and the National Heart Foundation. The Florey Institute of Neuroscience and Mental Health acknowledges the support received from the Victorian Government via the Operational Infrastructure Support Scheme. The content of the publication is solely the responsibility of the authors and does not reflect the views of the funders.

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement No additional data are available.

    • Collaborators The AVERT Trial Collaboration Group.

    • Patient consent for publication Not required.