FormalPara Key Points for Decision Makers

This review found that maintenance cognitive stimulation therapy showed the strongest evidence of being cost-effective, but with a small number of economic evaluations; case management, occupational therapy and dementia care management also showed good evidence of cost-effectiveness.

More economic evidence about the cost-effectiveness of dementia care interventions is needed, and researchers need to be consistent with study methods and outcome measures to allow comparison across interventions. Increased economic information and consistency would increase confidence of policy and local and national decision makers when they are planning and implementing dementia care interventions in the future.

1 Introduction

The worldwide economic burden of dementia care is high at US$815 billion [1]. The total annual cost of dementia in the UK is estimated at £24.2 billion [2]. With the increase in numbers of people being diagnosed with dementia and the high costs of dementia care, economic evaluations are needed to ensure that non-pharmacological interventions which are offered are cost-effective; however, economic evidence of these interventions remains limited [3].

By 2050, the number of people with dementia (PwD) is projected to rise to 152 million due to population growth and an increasingly ageing population [4]. There is currently no cure for dementia; existing drug therapies are either symptomatic therapies to relieve symptoms of dementia or are primarily indicated for Alzheimer’s disease [5]. Drug therapies have the potential for serious side effects including mortality [6]. Non-pharmacological therapies may be considered as complements to pharmacological treatments.

The National Institute for Health and Care Excellence (NICE) guideline for dementia care [7] recommends four non-pharmacological interventions: group cognitive stimulation therapy (CST), group reminiscence therapy and cognitive rehabilitation or occupational therapy. The main aim of these types of dementia interventions is to reduce symptoms including cognitive decline, promote independence and wellbeing and improve quality of life.

Existing systematic reviews of economic evidence commonly focus on a particular intervention [8, 9] or dementia symptom [10, 11]. Previous reviews have also included interventions to improve the quality of life of carers as well as PwD [3, 12]. The term ‘carer’ here refers to anyone supporting a family member, partner or friend and not receiving payment for providing this care [13].

The aim of this review was to provide a comprehensive summary of existing economic evaluations of non-pharmacological interventions delivered in the community and nursing homes, evaluating a wide range of dementia symptoms and interventions that measured the impact on the PwD and not solely their carer. The intended audience of the review is policy makers and key decision makers, including healthcare providers and managers at a local and national level.

2 Methods

The protocol for this systematic review was established before work commenced and was registered on PROSPERO (CRD42021252999). Preferred Reporting Items for Systematic review and Meta-Analysis (PRISMA) guidelines were followed throughout [14].

2.1 Eligibility Criteria

Inclusion criteria (detailed in Table 1) stated that papers should be trial-based economic evaluations, observational studies or modelling simulations; the population under observation was PwD or those with mild cognitive impairment (MCI). People with MCI were included as a high percentage go on to later develop dementia [15]. To be eligible, interventions needed to aim to delay progression of the disease or improve quality of life. Papers could have evaluated dementia interventions throughout the dementia pathway, ranging in severity from recent diagnosis to advanced dementia, but prevention/early detection of dementia or end-of-life care interventions were excluded. Both narrative and systematic reviews of economic studies were also eligible for inclusion.

Table 1 Eligibility criteria
Full size table

2.2 Search Strategy

The databases searched were Academic Search Premier, Google Scholar, Web of Science, Cochrane Database of Systematic Reviews, MEDLINE, Cumulative Index of Nursing and Allied Health Literature (CINAHL), PsycInfo, Psychology and Behavioural Sciences Collection, PsycArticles, Business Source Premier and Regional Business News. Papers published between 1 January 2011 and 11 May 2023 were included to search only recent articles. Articles published in any languages were eligible.

The search terms used within the databases were the disease-specific terms ‘dementia’, ‘Alzheimer’s’ and ‘mild cognitive impairment’ combined with the economic terms ‘cost*’ or ‘econ*’, with additional search terms to identify interventions: ‘intervention’ or ‘therapy’. The reference lists of primary studies and review articles that met the inclusion criteria were manually searched for other relevant articles for inclusion.

2.3 Study Selection

Titles and abstracts were screened according to inclusion/exclusion criteria by author GE, and results were verified by EG/HJ. Any disputes were resolved by PM. Full texts of selected articles were retrieved and reviewed by GE and EG/HJ, and any disputes were resolved by PM.

2.4 Data Extraction

Extracted data included intervention description, participant numbers, follow-up period, study design, economic evaluation type, main economic outcome measure, primary outcome (PwD only) and perspective (see Table 2). Data extraction was performed by GE, with EG/HJ independently undertaking data extraction for 40% of the included articles. Any disagreement was resolved by PM.

Table 2 Data characteristics of economic evaluation
Full size table

Separate data extraction was undertaken for the review of reviews by GE (see Table 3). Extracted data included interventions reviewed, number of studies included in the review and databases used.

Table 3 Review characteristics
Full size table

2.5 Data Synthesis

A narrative approach to data synthesis was undertaken to summarise and allow comparison of the methods and results of the included evaluations whilst demonstrating heterogeneity. A narrative reporting approach was used as meta-analysis could not be carried out due to the context-specific nature of the economic evaluations and the numerous outcomes used in studies.

Interventions were classified according to the following categories representing distinct forms of care: physical activity, cognitive interventions, training interventions, multicomponent interventions, assistive technology and other interventions (specialist dementia care, group living, home care vs care home).

2.6 Quality Appraisal

Quality appraisal was undertaken using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement [16] (see Table 7). CHEERS is designed to assess reporting quality rather than the quality of the evaluation. The statements relate to the following aspects: title, abstract, introduction, methods, results, discussion and disclosure. Statements that related only to modelling evaluations were excluded for non-modelling evaluations and resulting scores adjusted accordingly. Each was assigned a score, based on the number of statements met on the CHEERS checklist (0 = unmet, 0.5 = partially met, and 1 = met); the total was then translated into a percentage of items met. Quality appraisal was carried out by GE, EG/HJ undertook an independent appraisal of 30% of the articles, and any disagreement was resolved by PM.

In the review of reviews, A Measurement Tool to Assess Systematic Reviews (AMSTAR 2) [17] (see Table 4) was used to critically appraise the quality of the reviews. GE undertook the initial assessment, and EG independently undertook assessment for 60% of the included articles. The tool includes an assessment of potential bias, assessing each area or domain of the review, with certain domains being defined as critical. An overall rating of confidence in quality was calculated based on the total number of weaknesses in critical domains. These weaknesses were defined by unmet/partial met statements on the AMSTAR 2 checklist according to the following criteria:

  • High confidence None or one non-critical weakness

  • Moderate confidence More than one non-critical weakness but no critical flaws

  • Low confidence One critical flaw with or without non-critical weaknesses

  • Critically low confidence More than one critical flaw with or without non-critical weaknesses

Table 4 Quality assessment of systematic reviews using AMSTAR 2 [17]
Full size table

2.7 Usefulness of Economic Evaluations to Decision Making

A score of usefulness of the economic evaluations to decision making was calculated for each included evaluation. The scoring system was based on an existing method of assessing usefulness of economic evaluations based on data extraction and assessment of reporting quality [18]. Usefulness was then categorised according to these scores: limited ≤ 4, moderate 4.5–5, strong 5.5–6 (details in Table 5).

Table 5 Usefulness of economic evaluations to decision making
Full size table

3 Results

The systematic literature review identified 769 publications, duplicates were manually removed, and 489 articles were screened. Forty-one papers were included in the final review. The articles selected for inclusion comprised 37 single economic evaluations and four reviews. The search strategy has been reported using a PRISMA [14] flow diagram (Fig. 1). The literature search revealed a lack of economic evaluations on the cost-effectiveness of creative therapies such as art, music, drama, creative writing and dance and also sensory therapies such as aromatherapy and massage.

Fig. 1
figure 1

PRISMA flow diagram showing inclusion process of articles in the systematic review [14]. MCI mild cognitive impairment, PRISMA Preferred Reporting Items for Systematic review and Meta-Analysis

Full size image

3.1 Study Characteristics

The specific conditions being studied included PwD (n = 29), Alzheimer’s disease (n = 2), MCI (n = 2) and mixed populations of MCI/Alzheimer’s disease/dementia (n = 4). Types of interventions reviewed were exercise (n = 6), cognitive (n = 9), multicomponent (n = 12), training (n = 8) and assistive technology (n = 2). Studies recruited participants living in a variety of settings: 27 studies were for community dwelling participants, nine were for those in nursing homes, and two were for people living either in a nursing home or the community. Studies used varying criteria to define dementia/MCI in their inclusion criteria, ranging from having symptoms of dementia (n = 1) to a formal diagnosis of dementia (n = 11). A number of studies defined specific Clinical Dementia Rating scores (n = 3) and/or used Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) assessment criteria in their inclusion criteria (n = 8).

The majority of studies evaluated were randomised controlled trials (n = 30), followed by non-randomised studies (n = 5) and two modelling studies. Sample size varied greatly, ranging from 50 to 3269 (10,000 in a modelling evaluation). The international reach of the studies was as follows: the UK (n = 14) [6, 19,20,21,22,23,24,25,26,27,28,29,30], Germany (n = 4) [31,32,33,34], the Netherlands (n = 4) [35,36,37,38], the United States (n = 5) [39,40,41,42,43] Finland (n = 3) [44,45,46], Denmark (n = 2) [47, 48], Australia (n = 1) [49], Japan (n = 1) [50], Canada (n = 1) [51], Singapore (n = 1 [52], Sweden (n = 1) [53] and a multi-national study involving Italy, Poland and the UK [54].

A range of interventions were identified, which focused on improving the behavioural and psychological symptoms of dementia (BPSD), as well as interventions to prevent decline in cognitive function and mobility. In total, 27 different primary outcome measures were used across the included 37 studies. Outcomes included generic, dementia-specific and utility-based quality-of-life scales.

Over half of the evaluations employed a cost-utility analysis (n = 20), followed by cost-effectiveness analysis (n = 10) and cost-benefit analysis (n = 4); quality-adjusted life years (QALYs) were the most frequently used measure of benefit (n = 20). Costs and outcomes are reported in Table 6. It was noted that where studies compared an intervention to treatment as usual (TAU), the meaning and content of TAU varied between studies, potentially affecting the relative effectiveness and cost-effectiveness of interventions.

Table 6 Economic evaluation results
Full size table

Fourteen economic evaluations took a societal perspective. A societal perspective takes into account all costs and effects of the intervention on the whole of society and is often favoured by health economists [55]. The perspective taken when conducting an economic evaluation could affect the analyses and results [55].

3.2 Quality Appraisal of Economic Evaluations

Individual evaluations met between 62 and 98% of criteria items in assessment of CHEERS reporting quality (mean 81%) (Table 7). Twenty-one evaluations met over 80% of total assessed items. In terms of scores for categories of interventions, the highest was for the ‘other’ category, which scored 89% (mean); however, it should be noted that this category only evaluated two interventions. Exercise interventions ranked next highest with 84% (mean); training interventions scored the lowest overall, with 76% (mean).

Table 7 Quality assessment of studies using CHEERS [16]
Full size table

3.3 Usefulness of Economic Evaluations to Decision Making

Scores for level of usefulness were as follows: evaluations rated as having ‘strong’ usefulness, n = 8; ‘moderate’, n = 17; and ‘limited’, n = 12. The results showed that a high CHEERS quality assessment score did not necessarily translate to a high usefulness score for aiding decision making (see Table 5).

3.4 Intervention Specific Results

3.4.1 Cognitive Interventions

Nine evaluations explored interventions focused on cognition (Table 2). Study populations comprised community dwelling PwD [19, 22, 23, 44, 47], nursing home residents [49, 50] and a combination of both populations [20, 26].

CSTs were evaluated in three papers: one evaluating cognitive stimulation for participants new to CST [22] and two evaluating maintenance cognitive stimulation therapy (MCST) [20, 26]. The D’Amico et al. evaluation reported that MCST dominated TAU, with a 40% probability of being cost-effective at a willingness-to-pay (WTP) threshold of £20,000 for cost per QALY using proxy EQ-5D ratings [20]. In the Brown et al. evaluation, MCST dominated TAU in terms of QALYs for a subgroup of people living alone [26]. However, this evaluation used a small secondary dataset that may not have been representative of the general population and was not powered to detect subgroup changes; therefore, results should be interpreted with caution. The CST intervention was not evaluated as cost-effective though; it did not improve cognition or quality of life for PwD and was dominated by TAU [22].

Cognitive multicomponent interventions were evaluated in two papers [47, 50]. Sado et al. evaluated a combination of cognitive training and stimulation in a nursing home population [50], while Sogaard et al. assessed a community counselling/education and support intervention [47]. Both followed up participants for significant periods of time [47, 50]. Although the intervention Sado et al. evaluated showed significant cost savings compared to TAU [50], the non-randomised matched control design was not gold standard, and such designs risk introducing bias and confounding [56]. The intervention Sogaard et al. evaluated was not cost-effective and was dominated by TAU in terms of QALYs [47].

The remaining four cognitive papers all evaluated different types of cognitive interventions. Laakkonen et al. evaluated cognitive rehabilitation with promotion of self-management skills for people recently diagnosed [44]. Participants in the intervention group showed significantly less decline in verbal fluency scores compared with TAU, without increasing total costs [44]. Spector et al. evaluated cognitive behavioural therapy (CBT) to reduce anxiety in PwD [23]. Significant improvements in depression were reported by the intervention group; improvements in anxiety scores were also reported, but this failed to reach statistical significance [23]. Mervin et al. evaluated an intervention to improve levels of agitation using soft toys and interactive toys with artificial intelligence; no statistically significant between-group differences in agitation were found between the soft toy and the interactive toy groups, and the intervention was not cost-effective [49]. Woods et al. evaluated joint reminiscence therapy groups with carers and PwD [19]. They reported no significant difference in outcomes or service use between the intervention group and TAU [19].

3.4.2 Training Interventions

There were eight evaluations of specialised dementia training for staff caring for PwD (Table 2). Seven of the eight interventions involved nursing home staff training, with one community-based training intervention for general practitioners (GPs). The training intervention evaluation with the longest follow-up period was evaluated by Williams et al. at 36 months (mean = 15.4, SD 9.83) [39].

Four training interventions were evaluated as cost-effective in terms of QALYs [24, 27, 36, 39]. Ballard et al. reported the staff training intervention demonstrated benefits in terms of quality of life, agitation and neuropsychiatric symptoms, as well as cost savings [24]. The staff training intervention Livingston et al. evaluated was cost-effective in terms of QALYs with a 62% probability of cost-effectiveness at a WTP of £20,000/QALY [27]. The multicomponent training intervention for people with severe dementia evaluated by El Alili et al. was reported to dominate TAU in terms of QALYs [36]. Williams et al. evaluated staff communication with residents, and this intervention was found to be cost-effective; however, the evaluation lacked appropriate statistical analysis, as no incremental cost-effectiveness ratio (ICER) in terms of QALYs was reported; direct comparison with other studies is difficult [39]. A strength of this evaluation was its long time horizon of 36 months; it was limited by the small sample size, and generalisability was also limited due to the small geographical location of the nursing homes.

An evaluation of the cost-effectiveness of a programme that provided GP training in dementia care, carer counselling and promotion of support groups was carried out by Menn et al. [31]. Despite having the longest of all follow-up periods at 48 months, it showed no significant reduction in time to institutionalisation [31]. Meads et al. and van de Ven et al. both evaluated dementia care mapping interventions in nursing homes; they were evaluated as not cost-effective and cost-neutral, respectively [6, 38].

3.4.3 Multicomponent Interventions

Twelve papers evaluated a range of multicomponent dementia care programmes.

Dementia care management evaluations by Jennings et al., Michalowsky et al. and Radke et al. looked at multicomponent community dementia care for assessment, management and support [33, 40, 57].

Only two of the three dementia care management interventions evaluated showed evidence of cost-effectiveness, and TAU was dominated in terms of QALYs [33, 57]. Radke et al. undertook subgroup analysis to determine which group of participants benefitted most [57]. A higher probability of cost-effectiveness was reported in those aged < 80 years compared to > 80 years; females compared to males; PwD living alone compared to those not living alone; and in people with more co-morbidity than less co-morbidity [57]. The authors suggest that the high probability of cost-effectiveness in females and those living alone could be attributed to these groups having fewer relatives or carers to provide care and support and therefore having a higher number of unmet needs, meaning that they are more likely to benefit from a multicomponent management programme [57].

Case management, supporting the person with dementia and their carer in conjunction in a multicomponent team approach, was evaluated by MacNeil Vroomen et al. and Mostardt et al. [32, 35]. Mostardt et al. reported a significant difference in average additional months spent in the home environment for intervention group participants, although no limitations were discussed in the evaluation, which also used a non-randomised matched controlled study design. MacNeil Vroomen et al. evaluated an intensive care management model that dominated TAU in terms of QALYs, with 99% WTP at €30,000/QALY [35]. However, the observational, non-randomised design may have introduced selection bias and more heterogeneity into the study population.

Three occupational therapy-based interventions were evaluated that involved PwD assessment and carer/PwD education [30, 41, 42]. While cost-benefit analysis showed cost savings for two of these interventions [41, 42], the third intervention was dominated by TAU [30].

Rosenvall et al. and Jutkowitz et al. each undertook modelling simulation evaluating interventions to delay transition to long-term care [43, 45]. Rosenvall et al. demonstrated potential cost-effectiveness through care management, family support and rehabilitation interventions to delay cognitive decline and transition to long-term care [45]. The Jutkowitz et al. modelling simulation also demonstrated potential cost-effectiveness; they modelled four interventions, including two targeting caregivers’ education/support and two targeting assessment/management of PWD, and found that all showed small QALY improvements and increased time spent at home [43].

Saxena et al. undertook an evaluation comparing a primary care dementia clinic with a hospital-based memory clinic in Singapore [52]. QALYs were higher for the primary care clinic, and the ICER at 12 months was S$29,042 (Singapore dollars) per QALY (less than the assumed threshold of S$78,690). The authors concluded that the care provided by the primary care clinic had similar effectiveness to that provided by a hospital clinic, suggesting that these clinics could be cost-effectively set up elsewhere in primary care.

Henderson et al. conducted an evaluation on a non-randomised study of meeting centres providing day support [54]. The evaluation had a short 6-month time horizon; it was rated highest for reporting quality by CHEERS at 98%, but the intervention was not cost-effective in terms of QALYs.

3.4.4 Exercise Interventions

Six RCTs on the effect of physical activity on PwD were evaluated. One intervention included people living in nursing homes [21]; the rest focused on people living in the community [25, 37, 46, 48, 51]. The majority of evaluations were of group-based exercise outside of the home.

None of the interventions demonstrated cost-effectiveness in terms of QALYs; however, two showed improvements in primary outcomes [46, 51]. Davis et al. reported that resistance and aerobics dominated balance and toning in terms of seconds gained/lost on Stroop test of cognitive function and were less costly [51]. In Pitkäl̈a et al., both home exercise and group exercise intervention groups demonstrated a significantly slower decline in functioning measured by Functional Independence Measure (FIM) (−7.1 and −10.3 FIM change, respectively) than the control group (−14.4 FIM change) without increasing the total costs of health and social services [46].

3.4.5 Assistive Technology Interventions

Howard et al. evaluated a telecare and assistive technology intervention, which was not cost-effective in terms of QALYs and did not enable PwD to live safely at home for longer [29]. Ghani et al. evaluated an app for people with MCI and their carers [53]. The results of the cost-effectiveness analysis were inconclusive but suggest the intervention may be more beneficial for carers than PwD.

4 Review of Reviews

Alves et al. undertook a systematic review of five RCTs on the efficacy and feasibility of cognitive interventions for those with Alzheimer’s disease [8]. The review included cognitive stimulation, cognitive training and cognitive rehabilitation. It was rated as critically low for confidence in quality using AMSTAR 2, with several critical domains not being fully met. Only one relevant economic evaluation of cost-effectiveness was identified: in the cognitive intervention category, a programme of cognitive stimulation for PwD living either at home or in nursing homes (participant numbers not stated). The intervention was not reported as cost-effective.

Home support interventions for PwD living in the community were systematically reviewed by Clarkson et al. [12]. The review included 14 economic evaluations; six of these evaluated only carer outcomes and as such were excluded here. The economic evaluation by Pitkäl̈a et al. [46] was reviewed by Clarkson et al.; it has already been reviewed in this paper (see Sect. 3.4.4) and will not be discussed here again to avoid duplication. The remaining seven relevant economic evaluations looked at interventions across several categories: training (dementia care mapping), cognitive (activity sessions), multicomponent (occupational therapy, dementia care management n = 2) and other (specialist dementia day care, home care compared to care home). Only occupational therapy was reported to show cost-effectiveness. Confidence in the Clarkson et al. review was rated as critically low for confidence in quality using AMSTAR 2 and did not fully meet all critical domains.

As part of a systematic review of interventions to reduce agitation in older adults with dementia in any setting, Livingston et al. [10] identified an economic evaluation by Norman et al. evaluating a comparison of dementia care mapping and person-centred care [60]. Economic outcomes were measured using Cohen-Mansfield Agitation Inventory (CMAI) scores. Norman et al. reported that for person-centred care relative to usual care there were costs of A$6.43 (Australian Dollars) per CMAI point averted, for the dementia care mapping intervention, the costs were higher at A$46.89 [60]. The summary economic measure for this evaluation was cost per CMAI score change; further data on the effectiveness of this intervention were not available within the review.

Additionally, as part of the review, Livingston et al. created a simulation model using the most effective strategies identified from an effectiveness review of 30 studies of wide-ranging dementia therapies and existing patient cohort data [10]. Modelling of a multicomponent intervention for participants with mild to moderate dementia revealed 82% probability of cost-effectiveness at a maximum WTP threshold of £20,000/QALY. However, due to the multicomponent nature, it was not possible to determine which particular component of the intervention was most effective. This review rated high for confidence in quality using AMSTAR 2.

The Nickel et al. review [3] of interventions for PwD and their carers contained three evaluations in our physical activity intervention category [21, 46, 51] and three cognitive interventions [19, 20, 22] that were relevant to this systematic review. The evaluations contained in this review have already been individually identified through the literature review, and methods and results are discussed above. The review rated critically low for confidence in quality using AMSTAR 2.

5 Discussion

5.1 Summary of Main Findings

This paper reviewed economic evaluations of a wide range of interventions for dementia and MCI. These interventions took place around the world in community and nursing home settings.

Of the 37 evaluations and four reviews evaluated in this paper, 16 interventions demonstrated evidence to favour interventions, although some of these had limitations. The category with the greatest number of interventions showing evidence of cost-effectiveness was multicomponent interventions. Eight of the 16 multicomponent interventions were cost-effective in terms of QALYs gained compared to TAU. Case management showed evidence of cost-effectiveness in both evaluations; each of these evaluations rated ‘moderate’ for usefulness [32, 35]. However, one of these was rated ‘low’ in the CHEERS assessment of reporting quality and did not report cost-effectiveness in terms of QALYs [32]. Occupational therapy and dementia care management showed evidence of cost-effectiveness in over 60% of the evaluations [40, 45].

MCST demonstrated strongest evidence of cost-effectiveness, and results were reported using QALYs. Both interventions were cost-effective in terms of QALYs compared to TAU [20, 26]. They both scored ‘moderate’ for usefulness [20, 26], and D’Amico et al. scored highly for CHEERS reporting quality [20]. Overall conclusions should be interpreted with caution, however, as only two evaluations of MCST were available. As previously mentioned, group CST is already recommended as an intervention [7] and has been proven to be cost-effective in a large-scale study outside of the timeframe of this review [61]. However, the economic evaluation of CST was not found to be cost-effective [22], and overall, the cognitive interventions category did not demonstrate good evidence of cost-effectiveness.

There was also evidence for care home and nursing home staff training interventions. Of the training evaluations evaluated, 62% showed evidence of cost-effectiveness in terms of QALYs or showed significant patient benefits [24, 27, 28, 36, 39], and a sixth intervention was cost neutral [38].

Limitations were identified in some of the evaluations, and evidence was weakened by the small number of evaluations per intervention, small sample sizes, short timeline or reliability of evidence. Four of the evaluations reporting cost-effectiveness did not use an RCT design, and this could have led to biased estimates of effect [32, 35, 50, 52]. The generalisability of the sample to the wider population was a potential issue in two evaluations as the study population was limited to a rural community [33, 34]. Five of the cognitive evaluations took place in the same country, which may lead to a geographical bias [19, 20, 22, 23, 26].

5.2 Limitations

Despite the broad search terms used, it is possible that some economic evaluations may have been missed. There may be an element of publication bias, as it is acknowledged that authors may be less willing to publish evaluations that do not demonstrate cost-effectiveness or cost savings.

Synthesising evidence from evaluations and reviews evaluating a wide range of interventions presented challenges. Due to the variety of outcome measures used, the heterogeneity of the study methods and the variety of different interventions, it was difficult to compare the cost-effectiveness of different interventions. The lack of WTP thresholds for different countries made it difficult to compare the ICERs between countries.

The review was also hampered by a lack of robust economic evidence generally for non-pharmacological studies. This in turn resulted in limited evidence for each category of intervention.

It is acknowledged that the revised AMSTAR 2 tool that evaluated the systematic reviews was not published until after a number of the reviews had been published, and this may have negatively impacted on the scores received.

5.3 Recommendations for Future Research

Our review excluded carer-targeted interventions; it is acknowledged that there may be interventions that improve health or quality of life of PwD in ways considered to be cost-effective, even though those consequences would be secondary to the impact on carers. Also, if carer health and costs are taken into account, this could change results.

It was noted that evaluations tended to not analyse the distribution of costs, effects and cost-effectiveness across population subgroups. Given the wide inequalities in the experiences of PwD in terms of time to diagnosis and access to dementia services, for example, future research could focus on these areas.

The results of the review demonstrate gaps in the economic evidence on non-pharmacological interventions that could benefit from further research. There was limited evidence shown for multicomponent cognitive evaluations and CBT. Additionally, economic evaluations of creative and sensory therapies were not found in the literature review. Further high-quality research would evaluate whether these interventions are cost-effective.

6 Conclusion

MCST was evaluated as having the strongest evidence of cost-effectiveness, although the number of economic evaluations was small. Case management, occupational therapy and dementia care management also showed good evidence of cost-effectiveness. More economic evidence on the cost-effectiveness of dementia care interventions is needed, with consistency of study methods and outcome measures, to inform policy and local and national decision makers future decision making. This could improve decision makers’ confidence to promote cost-effective dementia interventions in the future.