What are the current and projected future cost and health‐related quality of life implications of scaling up cognitive stimulation therapy?

Cognitive stimulation therapy (CST) is one of the few non‐pharmacological interventions for people living with dementia shown to be effective and cost‐effective. What are the current and future cost and health‐related quality of life implications of scaling‐up CST to eligible new cases of dementia in England?


| INTRODUCTION
Responses to dementia are of three kinds: upstream prevention, efforts to find disease-modifying treatments, and better care and support. Current evidence suggests that about 40% of dementia cases can be prevented with better education and health behaviours in early or mid-life 1,2 and multidomain lifestyle interventions for slightly older people at known risk of dementia. 3 It is imperative that such preventive efforts are made, but benefits will be slow to materialise. There have been few breakthroughs in the search for disease-modifying interventions; in any case, prices of new medications and associated biomarkers could present challenging obstacles to widespread adoption. 4 An important aim must therefore be to deliver the best care to all people living with dementia and their carers.
Looking forward, if today's care arrangements and risk factors remained unchanged, expected growth in dementia prevalence will increase costs in England over a 25 year period by 249%. 5 This figure includes the costs of health and social care services that employ paid staff, as well as the estimated costs of unpaid support from family and other carers. It also takes into account the expected growth in other long-term conditions. Whether future governments or other funding bodies would be willing to increase their spending by this amount is uncertain, given that it would require higher taxes or funding to be diverted from other programmes. There must also be doubts about the future availability of family and other carers, given broad societal shifts such as reduced family size and labour force participation rates. 6 But even if this projected future cost was viable, it would still only fund a system that replicates today's diagnostic, treatment and care gaps and other failings. There is an obvious need to find and scale-up affordable, cost-effective interventions that support people with dementia and their carers to enjoy good quality lives.
Unfortunately, few such interventions are currently available.
Of nine non-pharmacological interventions for promoting cognition, independence and wellbeing reviewed by the National Institute for Health and Care Excellence (NICE) for their 2018 clinical guideline on dementia, only two had sufficient evidence for broad recommendation and only group cognitive stimulation therapy (CST) was considered specific enough to replicate widely. 7 What, then, are the current and projected future cost and healthrelated quality of life implications of scaling-up CST to eligible new cases of dementia in England? This is the question answered in this paper.

| MATERIALS AND METHODS
We combine evidence on effectiveness and cost-effectiveness with projections of numbers of people with dementia to estimate potential overall costs and health-related quality of life outcomes for the period 2015 and 2040 (in 5 year intervals). We start from 2015 because our baseline calculations of dementia costs and future projections relate to that year.
This study was part the MODEM project 6 and draws on its other components: � estimates of current and future dementia prevalence in England over the period 2015-2040 using microsimulation modelling; 8 � estimates of current and future care, treatment and support costs under today's arrangements using macrosimulation modelling, 5,9 including new data collected from a sample of people with dementia and carers to fill evidence gaps; 10 � systematic mapping of evidence on interventions in dementia care which identified, as did NICE, 7 that CST is one of few interventions considered to be effective and cost-effective.
Estimates are combined in a cell-based model (MS Excel).

| Intervention
Group CST is a manualised intervention targeted at people with mild or moderate dementia. It is a brief intervention involving small groups of people with dementia engaging in activities and discussions to improve cognitive and social functioning. It is based on the biopsychosocial model of dementia and disuse theory, emphasising enjoyment and information-processing. Session themes include physical games, sound, childhood, food, current affairs, faces/scenes, word association, being creative, categorising objects, orientation, using money, number games and word games. 11 Trials demonstrate that CST has beneficial effects on general cognition and quality of life, [11][12][13] without significantly increasing total care costs. 14 Such benefits are achieved over and above any medication effects. 13 A trial of additional CST over a longer period (maintenance CST (MCST)), based on the same theory of cognitive stimulation, demonstrated modest outcome gains over 6 months for people with mild-to-moderate dementia, 15 and reasonable likelihood of costeffectiveness based on cognition scores and both self-rated and proxy-rated quality of life. 16 Although lack of longer-term follow-up data from England or elsewhere 17 leaves ongoing uncertainty about long-term impacts, there is nevertheless potential to extend shortterm benefits by making CST available to a wider population.
We assume the same implementation scenarios as in relevant trials 11,15 : fourteen group-sessions of standard CST (45 min each, twice per week for 7 weeks, groups of approximately five persons) delivered within community-based care; followed by 24 groupsessions of MCST delivered weekly in community or residential settings. We assume every person accessing CST is offered MCST shortly after CST (except for those expected to die between those events).

| Target population
Our estimates are for a target population of diagnosed eligible new cases from 2015 onwards (without any catch-up delivery for previously identified new cases). Our starting point is the number of people expected to develop mild-to-moderate dementia, derived by applying incidence rates 7 to population projections 18 by age and sex.
We then make three sequential subtractions, at a constant annual rate: 10% of the incident population would never be diagnosed; 35% of the remainder would be ineligible for CST because of severe disability; and 41% of the remainder would decline participation when offered CST (sources in Table 1).
To estimate numbers receiving MCST, we further subtract: 1% expected to die between receiving CST and being offered MCST; 6% too disabled to remain eligible; and 6% declining MCST. 14

| Costs
Comprehensive costs for delivering the intervention average £464 per person for CST and £655 for MCST, based on groups of five persons. 7,16 Impacts on other health and social care costs are estimated from previous cost-effectiveness studies. 14,16 To account for any differences in baseline costs in these studies, we calculate cost differences between intervention and control groups in terms of difference in change in costs between baseline and follow-up. CST reduces both NHS and social care costs, while MCST reduces NHS costs but increases social care costs. We assume that these changes occur in the same year in which CST/MCST is provided. There is no available evidence to account for longer-term cost impacts.

| Outcomes
A Cochrane review concluded that the most consistently shown benefit of CST across studies is short-term improvement in cognitive function. 13 There are also benefits across other outcome domains: Annual incidence of dementia, base value for 2015 124,700 Cognitive function and ageing study (CFASII) incidence rate. 8 incidence for subsequent years was estimated by applying office for national statistics (ONS) population projections for England, by age and sex. 18 Proportion never diagnosed 0.10 NHS data analysis show that 70% of prevalent cases of dementia have been diagnosed. Since we know that there is often a delay before people receive a diagnosis, we assumed the proportion of people who ultimately receive a diagnosis must logically be higher than 70%.
Proportion ineligible for CST due to severe disability 0.35 Average of estimates obtained for North East London NHS foundation trust (NELFT; personal communication) and pan-London services. 19 for NEFLT, we excluded from our calculations the proportion of service users for whom it was unclear whether CST was offered at all (undocumented). For pan-London, we excluded the proportion for whom no service was available. Rate assumed to remain constant each year.
Proportion not using memory clinic or day centre 0 Assumption that everyone has access to the facilities in which CST is provided.  16 and applying proportional differences to costs reported by Knapp et al., 14 as described in the methods section. A real cost of care inflation index (modelling based on office for budget responsibility data) is applied to estimates for subsequent years.
QALY gain linked to CST, per person, in first year only 0.03 NICE 7 -from meta-analysis of various interventions that were considered under CST; refers to QALY gain derived from benefits on outcomes measured with clinical scales using the following formula (health utility = 0.359 + 0.00745 x MMSE + 0.00394 x DAD -0.0054 x NPI); health utility is then applied to mean times over which benefit is assumed to last (this includes intervention, follow up and convergence periods); after this period (in NICE base case about 600 days) the health state of person who participated in CST is assumed to return to the one of a person in control group; whilst NICE takes a period of 600 days over which benefits last, we assume for simplicity that those changes occur in the first year.

Maintenance cognitive stimulation therapy (MCST)
Mortality between CST and time for MCST 0.05 Assumption based on orrell et al. 15 -5% of people in MCST trial died over period of 6 months; Newcomer et al. 21  For MCST, average QALY gain of 0.026 over 6 months has been reported, 16 based on proxy EQ-5D assessments for the sub-group of participants taking acetylcholinesterase inhibitors (ACHEIs).
Given lack of evidence on QALY outcomes beyond 6-month follow-up for either CST or MCST, QALY gains are conservatively assumed to occur in the first year and to be achieved only once per lifetime.

| Sensitivity analyses
We examine alternative assumptions for four key parameters through one-way sensitivity analyses. Our base-case analyses assume Finally, there is uncertainty surrounding uptake rates for CST.
Orrell et al. 15 reported that, of the 13% of participants who withdrew  from their study, about half did so due to not liking CST groups. Given the potential to address this barrier (e.g., through individual delivery models), 24 we explore the impact of reducing the rate who decline CST by 50% (from 41% to 20.5%). We assume that average costs and QALY gains per person remain unchanged: we have no way of knowing whether they might alter if 'decline rates' fell.

| Incidence and target population
Between 2015 and 2040, the number of people with mild-tomoderate dementia is projected to grow from 124,700 to 203,200 (

| Costs and QALYs
Scaled-up implementation of CST for eligible people is estimated to cost £24.2 million in 2020 (    In return, there would be improvements in general cognitive functioning, language comprehension and production, and quality of life. 17 We followed the approach used by NICE to summarise this effectiveness in terms of QALYs gained: for 2025, for example, these would be 1633 (CST alone) or 2821 (CST + MCST). Cost per QALY in 2025 would be £14,686 (CST alone) or £23,191 (CST + MCST), the former clearly below the usual NICE cost-effectiveness threshold, and the latter meaning that CST followed by MCST may be less likely to be recommended.

| Strengths and limitations
We used a well-constructed microsimulation model to project future dementia prevalence, and a macrosimulation model to estimate current and future costs. Evidence on cost and QALY consequences of CST and MCST was drawn from robust randomised trials. However, the CST trial did not collect data on unpaid care costs and so we estimated them from MCST trial findings; our assumption of the same proportional change figures is a limitation.
CST uptake data came from current practice in a few English localities, and our assumptions on uptake may therefore not be generalisable: there is, for example, variation in CST delivery across England, Wales and Scotland. 25 There are also uncertainties about CST eligibility, and we could only obtain routine practice data from a few local services. In using those data, we adopted conservative assumptions throughout.
In their modelling of cost-effectiveness, NICE emphasised that their estimates were sensitive to changes in all parameters, with variations in their plausible ranges generating estimates on either side of the £20,000 per QALY cost-effectiveness threshold; and, at this threshold, probabilities of cost-effectiveness were estimated at around 50% (i.e., equivalent to the control group). 7 While several studies demonstrate effectiveness of CST/MCST based on specific outcomes, including quality of life, a Cochrane review noted methodological limitations associated with the evidence base and the need for research into potential benefits of longer-term CST programmes. 13 We assumed that outcomes (QALY gains) occur over a 1-year period, which probably underestimates true impacts.
There have been very few previous projections of expected future numbers of people living with dementia or their needs; our microsimulation model is the best currently available, but can obviously only generate estimates subject to error.
In calculating the target population, we assumed that everyone accessed memory clinics, day centres or residential care, so that there was no further reduction in numbers due to restrictions in access to facilities where CST/MCST is provided, and that investment for such facilities increases commensurately with incidence.
CST is now available in at least 29 countries, including some low-

| Implications
CST offers one of the few evidence-based opportunities to achieve health and wellbeing improvements for people with mild-tomoderate dementia. There is emerging evidence that it may also counteract the progression of neuropsychiatric symptoms. 28 Scalingup CST to the full eligible population, even allowing for those who decline the intervention or are too ill to join a group, has the potential to improve the lives of many more people than currently benefit from this intervention. Are there ways to bring down the cost of CST or improve qualityof-life benefits? Our estimates are based on an implementation model that is staff-intensive, even when delivered in groups. It is therefore worth further exploration of ways to reduce staff costs, for example through exploration of whether there are (more) 'active ingredients', 31 through greater integration into routine care as shown to be feasible elsewhere, 22,23 or through online delivery. 26 There may also be scope to reduce the number of people who are eligible for, but decline CST. Reducing that rate by a half could substantially increase total annual QALY gains, as we showed, but would also raise implementation costs.
Even though there is evidence that dementia incidence rates might be declining slightly, 32 the total number of people with dementia is not expected to reduce substantially over the coming decades. Symptomatic medication can help slow down cognitive decline for a time, 7 and to do so cost-effectively, 33 but CST is one of very few non-pharmacological interventions known to improve health or quality of life whilst also being cost-effective.

| CONCLUSIONS
Drawing on microsimulation projection modelling, new cost data and evidence from previous trials, we find that scaling-up group CST across England for people with incident dementia and mild-tomoderate symptoms offers an excellent opportunity to improve lives in an affordable, cost-effective manner. Adding maintenance CST would further improve health-related quality of life, although the economic case is less compelling.