NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
National Clinical Guideline Centre (UK). Stroke Rehabilitation: Long Term Rehabilitation After Stroke [Internet]. London: Royal College of Physicians (UK); 2013 May 23. (NICE Clinical Guidelines, No. 162.)
K.1. Introduction
The GDG identified the comparison of more intensive programmes of rehabilitation for people with stroke with less intensive programmes as a high priority area for economic analysis.
The review question linked to this high priority area was: “In people after stroke what is the clinical and cost-effectiveness of intensive rehabilitation versus standard rehabilitation?” The literature review is described in the full guideline. No cost-effectiveness analysis was found which addressed this question.
More intensive rehabilitation may be more costly to deliver than less intensive rehabilitation because it may require additional staff time. However, additional costs may be offset by an improvement in outcomes for the patient (such as independency in activities of daily living), leading to increased QALYs and potentially a reduction in future healthcare and social care costs.
There is no definition of what ‘intense’ and ‘standard’ rehabilitation is in any current UK standard. There is an existing NICE stroke quality standard for on-going rehabilitation23, which specifies that: “patients with stroke are offered a minimum of 45 minutes of each active therapy that is required, for a minimum of 5 days a week, at a level that enables the patient to meet their rehabilitation goals for as long as they are continuing to benefit from the therapy and are able to tolerate it”. The GDG considered however that actual delivery of rehabilitation is very variable across the country and that often this level of rehabilitation is not achieved. As far as they are aware there is not yet data about what the actual level of provision of rehabilitation is currently across the UK NHS.
No economic analyses were identified in the published literature that evaluated the cost effectiveness of different levels of intensity of rehabilitation. The review of the clinical literature (described in the full guideline) identified limited evidence although what was found generally suggested that more intensive rehabilitation had more favourable outcomes. However, to calculate QALYs a particular measure of quality of life is required known as utility; commonly assessed using the EuroQol 5 dimension (EQ-5D) instrument. Only the Ryan study28 reported utility data and the GDG agreed that an economic analysis should be undertaken based on this study.
It was noted, that over time the expectations have changed and the amount of therapy delivered is now much greater than it was, say, 10 years ago. Thus older studies (like the Ryan study) may describe ‘high’ levels of input which are actually lower than described in the stroke strategy and quality standard11,23.
It was also noted that rehabilitation is a complex intervention, that is, the outcome does not vary linearly with inputs. One possibility is that there is a critical threshold for improvement. For example, if one leg is weak the patient will be unable to walk. The strength may increase linearly for 6 weeks, but only in week 7 will the patient walk. If a functional outcome is used, the patient will appear to plateau for 6 weeks and then may show a significant change in functional status.
The following general principles were adhered to in developing the cost-effectiveness analysis:
- The GDG was consulted during the construction and interpretation of the model.
- Model inputs were based on the systematic review of the clinical literature supplemented with other published data sources where possible.
- When published data was not available expert opinion was used to populate the model.
- Model inputs and assumptions were reported fully and transparently.
- The results were subject to sensitivity analysis and limitations were discussed.
- The model was peer-reviewed by another health economist at the NCGC.
K.2. Methods
K.2.1. Model overview
A cost-utility analysis was undertaken to evaluate the cost-effectiveness of more intensive versus less intensive stroke rehabilitation. Lifetime quality-adjusted life years (QALYs) and costs were estimated from a current UK NHS and personal social services perspective. As is standard practice in economic evaluation, costs and QALYS were discounted to reflect time preference; a rate of 3.5% per annum was used in line with NICE methodological guidance22. The cost effectiveness outcome of the model was cost per QALY gained.
The analysis was based on data from the UK randomised clinical trial (N=89; 75% drop out) reported by Ryan and colleagues, 200628. This study compared two different intensities of rehabilitation in the community setting: less intensive rehabilitation was three or less face-to-face contacts per week, for 12 weeks maximum; more intensive rehabilitation was six or more face-to-face contacts per week, for 12 weeks maximum. Outcomes were assessed at 12-weeks.
K.2.1.1. Population
The population for the cost-effectiveness analysis was adults and young people 16 or older who have had a stroke and required rehabilitation.
K.2.1.2. Comparators
The comparators in the model were:
- Less intensive multidisciplinary rehabilitation
- More intensive multidisciplinary rehabilitation
Following Ryan et al (2006)28, the intervention was assumed to be delivered at home. Less intensive rehabilitation was three or less face-to-face contacts per week, for 12 weeks maximum. More intensive rehabilitation in the study was six or more face-to-face contacts per week, for 12 weeks maximum.
K.2.2. Approach to modelling
K.2.2.1. Model structure
A life table approach was taken to the analysis. Life tables for England and Wales were adjusted for the increased mortality in people who have had a stroke. This estimated the number of people alive after each 3 month period (each cycle) and this was used to estimate life years for people in the model. It was assumed that mortality is not impacted by the type of rehabilitation received and so life expectancy did not vary by comparator.
A quality of life (utility) value was attributed to people who were alive in the model that depended on the type of rehabilitation received (‘more intensive’ or ‘less intensive). This resulted in differences in QALYs between patients.
Differences in total costs between the groups were due to differences in the cost of delivering rehabilitation – this cost was incurred in the first 3 month cycle. It was assumed in the base-case analysis that in the post-rehabilitation period costs did not vary between the more intensive and the less intensive rehabilitation groups as data was not identified on which to base a difference.
K.2.2.2. Uncertainty
The model was built probabilistically to take account of the uncertainty around input parameter point estimates. Probability distributions were defined for model input parameters. When the model was run, a value for each input was randomly selected from its respective probability distribution simultaneously; mean costs and mean QALYs were calculated using these values. The model was run repeatedly – 1000 times – and results were summarised. Probability distributions in the analysis were parameterised using error estimates from data sources.
In addition, various sensitivity analyses were undertaken to test the robustness of model assumptions and data sources. In these, one or more inputs were changed and the analysis rerun to evaluate the impact on results.
Threshold sensitivity analyses were also performed to explore different cost and QALY differences.
K.2.3. Model inputs
K.2.3.1. Summary table of model inputs
Model inputs were based on clinical evidence identified in the systematic review undertaken for the guideline, supplemented by additional data sources as required. Model inputs were validated with clinical members of the GDG. A summary of the model inputs used in the base-case (primary) analysis is provided in Table 30 below. More details about sources, calculations and rationale can be found in the sections following this summary table.
Table 30
Summary of base-case model inputs.
K.2.3.2. Initial cohort settings
As the model was developed using outcomes from the Ryan study, initial cohort setting were based on the mean baseline characteristics from this study (N=89)28. The population therefore had an age of 77 years on entry to the model and 61% were female.
K.2.3.3. Mortality
Mortality was incorporated into the model using life-tables for England and Wales adjusted to reflect the increased mortality rates in people who have had a stroke. Standardized mortality ratios (SMR) for all-cause mortality after stroke compared with age/sex adjusted rates for the general population reported by Bronnum-Hansen et al. 20018 were used. These were from a large Danish study of people who had a stroke 1982–1991 (n=4162) with up to 15 years follow-up. For females this was 2.85 (CI: 2.66, 3.05) over the course of the study and for males 2.58 (CI: 2.43, 2.75).
It was assumed that mortality does not differ with more or less intensive rehabilitation.
K.2.3.4. Quality of life (utility)
Utility values were based on the Ryan et al (2006)28 study that reported EQ-5D scores for people undergoing more intensive rehabilitation (n=35) and less intensive rehabilitation (n=32) using the UK tariff.
Quality of life before rehabilitation for both groups in the model was based on an average of the median values observed in the Ryan study at baseline (0.54). This input was not incorporated into the probabilistic analysis.
Quality of life following rehabilitation was based on the 12 weeks follow-up in the Ryan study. Change from baseline in the less intensive arm was 0 (SE 0.04). So after less intensive rehabilitation in the model quality of life was still 0.54. The difference at 12 weeks with more intensive rehabilitation compared with less intensive was 0.14 (SE 0.05). So after more intensive rehabilitation in the model quality of life was increased to 0.68. Inputs were incorporated into the probabilistic analysis using normal distributions.
As longer term follow-up was not available in the Ryan study but we wished to consider the impact over the whole lifetime, in the model we applied a series of different assumptions regarding what happens to the difference in utility between groups observed at 12 weeks over time:
- Scenario 1: the difference remains the same over the remaining lifetime of the patient (Figure 413)
- Scenario 2: the difference disappears over time (Figure 414)
- it was assumed the difference disappears over 3 months, 1 year, or 5 years (analysis 2a, 2b and 2c respectively).
Figure 413
Quality of life (utility) over time – scenario 1, difference maintained.
Figure 414
Quality of life (utility) over time –scenario 2(b), difference disappears.
The GDG noted that these were the two extremes of what could happen to differences in quality of life over time and that in reality it may be somewhere between the two with the difference diminishing over time but not completely disappearing. However, for analysis purposes these were considered the most useful scenarios to model as they consider the best and worst case scenarios.
The GDG also noted that plausibly both groups’ quality of life may improve over time (if functional status independently improves with time) or worsen over time (quality of life decreases with age). However, in the model it is the difference between the groups that drives the results rather than the absolute values and so it was not considered necessary to explicitly model these scenario as they would not impact conclusions.
K.2.3.5. Resource use and cost
Rehabilitation
The cost of the less and more intensive rehabilitation programmes was calculated based on the resource use from the Ryan study supplemented by assumptions where required and relevant UK unit costs10,28. The average total cost per person with less intensive rehabilitation was £634 and with more intensive rehabilitation was £865.
The number of rehabilitation sessions reported was 17.9 (SE 1.19) with less intensive rehabilitation sessions with an additional 6.5 (SE 1.76) with more intensive rehabilitation. The number of rehabilitation sessions with less intensive rehabilitation was incorporated in the probabilistic analysis using a gamma distribution as this is bounded by 0 and so reflects the plausible range. The difference in session with more versus less intensive rehabilitation was incorporated using a normal distribution. Information was provided by AW Ryan (email January 2011)regarding the duration of rehabilitation sessions in the study (between 30 and 60 minutes) and the professionals who delivered the work (physiotherapist, occupational therapist, speech and language therapist, and rehabilitation assistant). Data was not available about the average length of sessions from the study. It was therefore assumed that typically the length of a session would be 45 minutes (midpoint of the range) and that the length of sessions did not vary between groups. Information was not available about the proportions of sessions carried out by different professionals, and so it was assumed that 75% were carried out by rehabilitation professionals and 25% by rehabilitation assistants. These inputs were not incorporated into the probabilistic analysis. Unit costs per hour home visit for a band 6 rehabilitation professional and a band 3 rehabilitation assistant were £54 and £27 respectively – these were calculated using the PSSRU 2010 approach and data adjusted for the salary band indicated as most appropriate by the GDG10. Costs include qualifications, overheads and travel expenditure. The cost per hour of home visiting takes account of the proportion of time spent on travel and non-contact time. These inputs were not incorporated into the probabilistic analysis.
Post-rehabilitation
In the base-case analysis it was assumed that there was no difference in costs post-rehabilitation as data was not identified on which to base a difference.
K.2.4. Computations
The model was constructed in Microsoft Excel and was evaluated by cohort simulation.
People started in cycle 0 in the alive health state. Patients moved to the dead health state each 3 month cycle as defined by the mortality rate. Life years for the cohort were computed each cycle. To calculate QALYs for each cycle, Q(t), the time spent (i.e. 0.25 years) in the alive state of the model was weighted by a utility value that was dependent on the cycle, the long-term utility assumption being employed and the treatment group. A half-cycle correction was applied. QALYs were then discounted to reflect time preference (discount rate = r). QALYs during the first cycle were not discounted. The total discounted QALYs was the sum of the discounted QALYs per cycle. The total discounted QALYs were the sum of the discounted QALYs per cycle.
Costs per cycle, C(t), were calculated in the same way as QALYs. Rehabilitation costs were applied in cycle 1 only. If a difference in post-rehabilitation costs was being included, this was applied in cycle two and beyond. Costs were discounted to reflect time preference (discount rate = r) in the same way as QALYs.
Discount formula for costs and QALYs:
|
| Where: r = discount rate per annum n = time (years) Total = total costs or QALYs |
The widely used cost-effectiveness metric is the incremental cost-effectiveness ratio (ICER). This is calculated by dividing the difference in costs associated with two alternatives by the difference in QALYs (formula below). The decision rule then applied is that if the ICER falls below a given cost per QALY threshold the result is considered to be cost effective. If both costs are lower and QALYs are higher the option is said to dominate and an ICER is not calculated.
|
|
|
It is also possible, for a particular cost-effectiveness threshold, to re-express cost-effectiveness results in term of net monetary benefit (NMB). This is calculated by multiplying the total QALYs for a comparator by the threshold cost per QALY value (for example, £20,000) and then subtracting the total costs (formula below). The decision rule then applied is that the comparator with the highest NMB is the most cost-effective option at the specified threshold. That is the option that provides the highest number of QALYs at an acceptable cost. For ease of computation NMB was used to identify the optimal strategy in the probabilistic analysis simulations.
| Netmonetarybenefit(X) = (QALYs(X)× D)−Costs(X) Where: Costs/QALYs(X) = total discounted costs/QALYs for option X; D = threshold |
|
The probabilistic analysis was run for 1000 simulations. Each simulation, total discounted costs and total discounted QALYs were calculated for each diagnosis option. Net benefit was also calculated and the most cost-effective option identified (that is, the one with the highest net benefit), at a threshold of £20,000 per QALY gained. The results of the probabilistic analysis were summarised in terms of mean costs, mean QALYs and mean net benefit for each treatment option, where each was the average of the 1000 simulated estimates. The option with the highest mean net benefit (averaged across the 1000 simulations) was the most cost-effective at the specified threshold. The percentage of simulations where each strategy was the most cost-effective gives an indication of the strength of evidence in favour of that strategy being cost-effective.
K.2.5. Sensitivity analyses
Sensitivity analyses were performed using the probabilistic analysis unless otherwise specified. Below is a description of the sensitivity analyses that were undertaken.
Rehabilitation costs
The length of the rehabilitation sessions for both less and more intensive rehabilitation in the base-case was based on the midpoint of the range used in the study (45 minutes). In sensitivity analysis we varied this input in both groups to the minimum length (30 minutes) and the maximum length (60 minutes) to explore how sensitive results were to this assumption.
Impact of age
In the base-case analysis the cohort had an initial age of 77 years in line with the mean age in the Ryan et al study28. We explored the impact of age in sensitivity analysis running the model for age 40, 50, 60, 70, 80 and 90 year olds. All other inputs were kept constant.
Threshold analyses
The GDG noted that the intensity level in the Ryan study more intensive rehabilitation arm was likely to be lower than that now specified by the stroke rehabilitation quality standard23. We therefore undertook threshold analyses to provide information help inform GDG decision making.
Costs
An analysis was undertaken to determine the cost difference where intensive rehabilitation was no longer cost-effective (given a £20,000 per QALY gained cost-effectiveness threshold) using the quality of life inputs from the Ryan study as in the basecase analysis.
QALYs
We also undertook a threshold analysis where we varied the difference in the number of rehabilitation sessions between the more and less intensive groups and then calculated what QALY difference would be required for it to be considered cost effective.
The GDG estimated that in current UK practice a level of input in line with the current NICE quality standard23 would be 45 minutes of each relevant therapy at least 5 days a week as long as they are continuing to benefit from it. Thus over 6 weeks an individual might receive 60 – 90 sessions of input. The GDG recognised that the recent Stroke Sentinel audit highlighted that about a third of patients received less than this while in hospital19. No data is available for community-based rehabilitation services. The GDG estimated that a typical level of input would be three physiotherapy sessions per week, one occupational therapy session per week, and one speech and language therapy session per week (that is 30 sessions). This would be a difference of 60 sessions total between ideal and typical input. The difference in number of sessions was therefore varied between 6.5 (from the Ryan study28) and 60 (based on the GDG estimate).
We then also calculated the number of months different quality of life (utility) gains would need to be maintained for in order to achieve these QALY gains. The quality of life (utility) gain explored in this analysis was 0.02 to 0.24 (Ryan study reported a 0.14 gain28).
Discount rate for health outcomes
A sensitivity analysis was undertaken to look at the effect of changing the discount rate for outcomes from 3.5% to 1.5%. Costs remained discounted at 3.5%.
Impact of making resource use and costs probabilistic
There is methodological debate as to whether it is necessary and/or appropriate to incorporate inputs based on standard national cost sources or resource use assumptions into a probabilistic analysis. In the base-case analysis we used fixed values for such inputs but we undertook a sensitivity analysis to assess the impact of this decision. Probability distributions were parameterised by assuming that point estimate used represented the mean and that the standard error was 20% of the mean.
Table 31
Sensitivity analysis – probability distributions.
K.2.6. Interpreting results
NICE’s report ‘Social value judgements: principles for the development of NICE guidance’ sets out the principles that GDGs should consider when judging whether an intervention offers good value for money21,22.
In general, an intervention was considered to be cost effective if either of the following criteria applied:
- The intervention dominated other relevant strategies (that is, it was both less costly in terms of resource use and more clinically effective compared with all the other relevant alternative strategies), or
- The intervention cost less than £20,000 per quality-adjusted life-year (QALY) gained compared with the next best strategy.
As the analysis is based on a single RCT28 with a selected population and specified intervention it had limited applicability to the overall stroke population and current UK practice. The GDG felt that the Ryan study compares two levels of intensity which are likely to be below that of the current quality standard (even the high intensity arm only received 17 therapy sessions in 60 days, far below the current standard). It was felt that the analysis could help evaluate the likelihood that more intensive rehabilitation was cost-effective and provide useful information to feed into decision making; however, it was also noted that it would not be able to provide definitive conclusions given these limitations.
K.2.7. Validation
The model was developed in consultation with the GDG; model structure, inputs and results were presented to and discussed with the GDG for clinical validation and interpretation.
The model was systematically checked by the health economist undertaking the analysis; this included inputting null and extreme values and checking that results were plausible given inputs. The model was peer reviewed by a second experienced health economist from the NCGC; this included systematic checking of many of the model calculations.
K.3. Results
K.3.1. Base case results
The analysis found that more intensive rehabilitation was cost effective compared to less intensive rehabilitation, based on levels of intervention and outcomes from the Ryan et al. 2006 study28. There was an additional cost associated with more intensive rehabilitation as more rehabilitation sessions were provided; however this was offset by the additional improvement in quality of life. This conclusion was maintained for all long-term utility scenarios. There was low within analysis uncertainty about this conclusion.
Table 32
Base case results – more intensive versus less intensive rehabilitation (probabilistic analysis).
People entering the model were aged 77 years and the undiscounted life expectancy generated by the model was 5.8 years. To validate the model this was compared against a paper suggested by a GDG member that reported expected life expectancy for 1-week survivors after stroke according to age at stroke onset from another, more recent, Danish cohort (n=392) who had a stroke between 1998–20017. This reported an estimated life expectancy for women of 8.8 years for those aged 70 years at time of stroke and 4.9 years for those aged 80 years. The corresponding figures for men were 7.8 and 4.3 years. This was considered consistent with the life expectancy estimated by the model.
K.3.2. Sensitivity analysis
Rehabilitation costs
The length of the rehabilitation sessions in the base-case was based on the midpoint of the range used in the study (45 minutes). In sensitivity analysis we varied this input to the minimum length (30 minutes) and the maximum length (60 minutes) to explore how sensitive results were to this assumption. This did not impact conclusions.
Table 33
Sensitivity analysis results – varying cost of rehabilitation based on change in session length: (probabilistic analysis).
Impact of age
In the base-case analysis the cohort had an initial age of 77 years in line with the mean age in the Ryan et al study28. We explored the impact of age in sensitivity analysis running the model for age 40, 50, 60, 70, 80 and 90 year olds. All other inputs were kept constant.
Table 34
Sensitivity analysis results – varying cohort initial age (scenario 1 - probabilistic analysis).
Threshold analysis: costs
An analysis was undertaken to determine the cost difference threshold where intensive rehabilitation was no longer cost-effective (using a £20,000 per QALY gained cost-effectiveness threshold). Under the most conservative long-term utility assumption (where the utility difference observed at the end of rehabilitation had disappeared over 3 months), more intensive rehabilitation would no longer be cost effective if the difference in rehabilitation cost was more than £685 (equivalent to a difference of about 17 sessions, of 45 minutes, with a rehabilitation professional). Under the most favourable utility assumption (where the difference observed at the end of rehabilitation was maintained indefinitely), more intensive rehabilitation remained cost effective until the difference in rehabilitation costs exceeded £13,433 (equivalent to a difference of over 300 sessions with a rehabilitation professional).
Table 35
Threshold analysis results: cost threshold where more intensive rehabilitation is no longer cost effective at £20,000 per QALY gained threshold.
Threshold analysis: QALYs
We also undertook a threshold analysis where we varied the difference in the number of rehabilitation sessions between the groups (difference of 6.5 to 60) and then calculated what QALY difference would be required for it to be considered cost-effective. Table 36 shows the resulting cost differences and the lifetime QALY gain that would be required in order for the higher level of intervention to be cost-effective (using on a £20,000 per QALY gained threshold). The lifetime QALY gain required for more intensive rehabilitation to be cost effective ranged from 0.01–0.11 when the difference in number of rehabilitation sessions was varied between 6.5 and 60.
Table 36
Sensitivity analysis results – rehabilitation cost and QALY gain threshold.
We then also calculated the number of months different quality of life (utility) gains would need to be maintained for in order to achieve these QALY gains – see Table 37. With a difference of 60 rehabilitation sessions with more intensive compared to less intensive rehabilitation, it was found that a utility gain of 0.14 would need to be maintained for 9 months in order for more intensive rehabilitation to be cost effective. When utility gain was varied between 0.02 and 0.24, this varied from 5 months to 64 months.
Table 37
Sensitivity analysis results – number of months that quality of life difference must be maintained for intensive rehabilitation to be cost effective under different resource use and quality of life gain scenarios.
Discount rate for health outcomes
A sensitivity analysis was undertaken to look at the effect of changing the discount rate for outcomes from 3.5% to 1.5%. Costs remained discounted at 3.5%. This did not impact conclusions.
Table 38
Sensitivity analysis results - discount rate for health outcomes 1.5% (probabilistic analysis).
Making costs probabilistic
This sensitivity analysis essentially did not change any results.
K.4. Discussion
K.4.1. Summary of results
The analysis found that more intensive rehabilitation was cost effective compared to less intensive rehabilitation, based on levels of intervention and outcomes from the Ryan et al. 2006 study28. There was an additional cost associated with more intensive rehabilitation as more rehabilitation sessions were provided; however this was offset by the additional improvement in quality of life. This conclusion was maintained for all long-term utility scenarios. There was low within analysis uncertainty about this conclusion. It was also robust to various one-way sensitivity analyses.
Due to concerns about the generalisability of the Ryan study to current UK practice, threshold analysis was used to explore under what scenarios more intensive rehabilitation would not be cost effective and these are described below.
Keeping the utility gain with more intensive rehabilitation constant we first looked at the cost threshold where more intensive rehabilitation was not cost effective. Under the most conservative long-term utility assumption (where the utility difference observed at the end of rehabilitation disappeared over 3 months), more intensive rehabilitation would no longer be cost effective if the difference in rehabilitation cost was more than £685 (equivalent to a difference of about 17 sessions, of 45 minutes, with a rehabilitation professional). Under the most favourable utility assumption (where the difference observed at the end of rehabilitation was maintained indefinitely), more intensive rehabilitation remained cost effective until the difference in rehabilitation costs exceeded £13,433 (equivalent to a difference of over 300 sessions with a rehabilitation professional).
Secondly, we systematically varied the difference in number of rehabilitation sessions (and thus cost) between more and less intensive rehabilitation and calculated the QALY difference that would be required for each difference to be considered cost-effective. We then calculated how long different utility gains would need to be maintained for in order to achieve these QALY gains. Assuming a difference of 60 sessions between more and less intensive rehabilitation: a utility difference of 0.14 would need to be maintained for 9 months for more intensive to be cost effective; a difference of 0.24 for 5 months; and a difference of 0.02 for 64 months (about 4 years).
K.4.2. Limitations & interpretation
Ryan study generalisability
The key limitations of this analysis are the limitations of the clinical effectiveness data for the comparison of more and less intensive rehabilitation. Only one study (Ryan 200628) reported utility data that could be used to calculate QALYs and the amount of rehabilitation received in this study compared with the current quality standard, and even current UK practice is very different. In Ryan more intensive rehabilitation was a total of 17 sessions on average per person and less intensive was 11. The GDG estimated that a level of intervention similar to that recommended by the current NICE quality standard23 would be more like 90 rehabilitation sessions per patient (spread across specialities), and that typical levels of input in the UK would be around 30 sessions.
It was noted that rehabilitation is a complex intervention, that is, the outcome does not vary linearly with inputs. One possibility is that there is a critical threshold for improvement. For example, if one leg is weak the patient will be unable to walk. The strength may increase linearly for 6 weeks, but only in week 7 will the patient walk. If a functional outcome is used, the patient will appear to plateau for 6 weeks and then may show a significant change in functional status. This again makes it difficult to extrapolate from the Ryan study.
Stratification
The GDG noted that younger patients also often have the capacity to participate in more sessions of rehabilitation as this is linked to cardiovascular fitness, frailty and co-morbidity, all of which tend to be worse in older patients. They also often have a greater range of needs (education, work, parenting). Yet often younger patients do not get more rehabilitation. It was not possible to undertake subgroup analysis on this basis in the model as not clinical studies had examined this.
Quality of life assumptions
The Ryan study28 reported EQ5D quality of life data at 3 months but did not have any longer term follow-up and so assumptions were made regarding what happens to the difference in quality of life over time between the groups. However both conservative and more favourable assumptions were explored in the model to test the impact on results.
The analysis does not include any impact on carer quality of life as there was no evidence available. It is plausible that greater functional ability for the person who has had a stroke, may also mean less burden on their carer and this may lead to an improvement in the carer’s quality of life as well. If this were the case this would increase the QALY gain with more intensive rehabilitation, making it more cost effective.
Post-rehabilitation costs
In the base-case analysis we assumed no difference in post-rehabilitation costs however greater functional ability could plausibly result in lower dependency and potentially lower social care costs. This would further favour more intensive rehabilitation.
Rehabilitation setting
The Ryan study28 was based on community rehabilitation and so costs in the model are also based on community rehabilitation. The GDG considered that the amount of rehabilitation should be the same whether delivered in the community or in hospital. In addition if rehabilitation was taking place in hospital the intensity of rehabilitation would most likely not change the length of stay but would just impact the amount of input from different professionals whilst in hospital. Therefore in either setting the cost impact would largely be about people’s time rather than changes in hospital capacity, overheads or hotel costs and so this was not considered likely to greatly impact the results. It was noted that potentially more intensive rehabilitation during the initial hospitalisation may even reduce hospital stay as patients become more functionally able more quickly.
K.4.3. Comparisons with published studies
No published cost-effectiveness studies were identified that compared more versus less intensive rehabilitation after stroke.
K.4.4. Conclusion = evidence statement
More intensive rehabilitation was found to be cost effective compared to less intensive rehabilitation, based on a modelled analysis using levels of intervention and outcomes from the Ryan et al. 2006 study (24 versus 18 rehabilitation sessions; EQ5D difference 0.14 at 3 months) and a range of long-term utility assumptions. However, these conclusions are limited by concerns regarding applicability of the Ryan study28 to current UK practice. Exploratory threshold analyses found:
- Under the most conservative long-term utility assumption (where the utility difference observed at the end of rehabilitation had disappeared over 3 months), more intensive rehabilitation would no longer be cost effective if the difference in rehabilitation cost was more than £685 (equivalent to a difference of about 17 sessions, of 45 minutes, with a rehabilitation professional).
- Under the most favourable long-term utility assumption (where the difference observed at the end of rehabilitation was maintained indefinitely), more intensive rehabilitation remained cost effective until the difference in rehabilitation costs exceeded £13,433 (equivalent to a difference of over 300 sessions with a rehabilitation professional).
- Assuming a difference of 60 sessions between more and less intensive rehabilitation: a utility difference of 0.14 would need to be maintained for 9 months for more intensive to be cost effective; a difference of 0.24 for 5 months; and a difference of 0.02 for 64 months (about 4 years).
K.4.5. Implications for future research
This analysis provides evidence to suggest that more intensive rehabilitation may be cost effective. However, it is limited by the generalisability of the study it is based on28 to current UK practice, in particular the low levels of intervention in both arms of the trial, and the lack of follow-up of the impact of different levels of rehabilitation on long-term quality of life for people with stroke. Further research in these areas and associated cost-effectiveness analysis should therefore be undertaken.
- Cost Effectiveness Model - Stroke RehabilitationCost Effectiveness Model - Stroke Rehabilitation
- Delphi Methodology Report - Stroke RehabilitationDelphi Methodology Report - Stroke Rehabilitation
- Guideline summary - Stroke RehabilitationGuideline summary - Stroke Rehabilitation
- Support and information - Stroke RehabilitationSupport and information - Stroke Rehabilitation
- Results - Image directed redesign of bladder cancer treatment pathways: the Blad...Results - Image directed redesign of bladder cancer treatment pathways: the BladderPath RCT
Your browsing activity is empty.
Activity recording is turned off.
See more...