Which factors increase informal care hours and societal costs among caregivers of people with dementia? A systematic review of Resource Utilization in Dementia (RUD)

Background Nearly 19 million people across OECD countries are living with dementia, and millions of family caregivers are affected by the disease. The costs of informal care are estimated to represent 40–75% of the total dementia cost exceeding formal care time and medical costs. Objective To conduct a systematic review to evaluate the methodological quality and factors associated with high informal care hours per month that increase societal costs, and to identify what type of interventions may alleviate the entire burden of informal and formal caregiving. Methods The systematic review was registered at PROSPERO (15.12.2020). A search in Medline, Embase, PsycINFO, and web of science for observational studies, cost-effectiveness, and cost of illness (COI) analyses on resource utilization in dementia (RUD) was conducted on 1 December 2020. Our inclusion criteria included a requirement that studies had to use the original RUD, RUD-FOCA or RUD lite in terms of hours or days per month, and costs as primary or secondary outcome, OECD countries, within the last 20 years and a sample population comprising persons with dementia (PwD) ≥65 years and their caregivers. We followed the PRISMA, GRADE, PICO guidelines and Drummond criteria to assess the methodology and quality of the studies. Results Of 307 studies, 26 cross-sectional and 3 longitudinal cohort studies were included in the analyses. Two studies had a randomized controlled trial (RCT) design. The methods and cost categories in each study varied widely. Disease severity, caregiver factors, and behavioural and psychological symptoms of dementia (BPSD) were associated with high informal care hours and societal cost. One RCT found no effect of a non-pharmacological intervention on informal care hours, yet another RCT found a cost-effective impact of an in-home respite care programme reducing informal care burden and costs. Conclusion The divergent use of the RUD components within included studies encourage more harmonized analyses. There are only two RCTs on RUD, one of which shows a significant treatment effect. Larger sample sizes and longer follow-up periods are required in future RCTs with dedicated focus on cost-enhancing and resource intensive factors such as disease severity and BPSD. Novel interventions must diversify between caregiver and PwD groups. PROSPERO registration CRD42021226388.


Introduction
Nearly 19 million people across OECD countries are living with dementia, and millions of family members who provide care and support throughout their lives are directly or indirectly affected by the disease. Total costs, i.e., costs related to informal care (unpaid care provided by family and others), direct costs of social care (provided by community care professionals and in residential home settings) and direct costs of medical care (treatment costs and other conditions in primary and secondary care), in dementia were estimated to be US$279.6 bn in 2000, US$ 604 bn in 2010 [1,2], $948 bn in 2016, with an annual growth rate of 15.9% [3]. Since 2010, the total costs have grown by 35% [4]. Globally, it was estimated that costs related to dementia passed USD 1 trillion in 2018, and the number was expected to triple the next 20 years [5]. As people not only live longer, but also spend a longer time span of their last years at home, the caregiver burden of dementia rises, as well as the need for family members to provide care 24 h a day [5].
Informal care is the care provided by family caregivers, and formal care by professional staff. The more physical disabilities and cognitive decline, the more formal and informal care complement each other [6][7][8]. Informal care costs constitute 40-75% of total dementia cost, exceeding formal care costs, and total medical costs [7]. Formal care is only the second largest cost component of the total cost [9,10].
A large number of informal care hours are associated with lost productivity or leisure time by informal care givers [11]. Worsening physical and mental health [12], poorer quality of life for caregivers [12], institutionalization [12], increased financial burden [13], and disease progression [14] are other factors associated with informal care. Thus, increasing caregiver hours affect informal care cost [1,15] in several different ways. The economic aspects of dementia are a major public health policy concern [16], particularly family relations that establish the infrastructure and supply of informal care of persons with dementia (PwD).
Resource Utilization in Dementia (RUD) is an assessment tool developed and validated to evaluate resource use for PwD. RUD has been proven comparative across countries with different health care systems [10,17] and provides the number of hours and days per month of care within activities of daily living (ADL), instrumental activity of daily living (IADL) and supervision by informal caregiver. RUD is used in trials, observational studies, and economic evaluations, and is thus a useful source for health policy makers, decision making and planning of dementia care services by making distributions of several cost components transparent [18].
Patient factors such as behavioural and psychological symptoms of dementia (BPSD), disease severity and progression, and caregiver attributes such as age, socioeconomic status, employment-and situation, and the ability of the caregiver to deal with the burden of caring are key factors influencing RUD. Economic evaluations using different valuation methods are one of the major concerns when estimating the cost of dementia [19].
In this systematic review of the literature, we investigate the resource utilization in dementia care with and without explicit economic evaluations, as well as publications with solely economic evaluations using RUD as cost input and primary outcome. Although, there are some studies reviewing the economic evaluation of dementia [10,19,20], this is the first to review both economic evaluations and research papers jointly. The benefits of focusing on RUD studies is the gain in comparability and in-depth methodological analyses. In previous economic evaluation studies, cost estimates vary depending on the method of valuation for informal care [21]. Following the PICO framework [22], the review aims to investigate the following questions.
1) What factors are more strongly associated with high caregiver burden and increased societal costs? 2) Are there specific health service interventions that are more beneficial in terms of reducing caregiver burden and societal cost levels? 3) What is the methodological quality of the economic evaluations of RUD studies?

Protocol and registration
This study is conducted in accordance with the PICO and PRISMA guidelines for reporting of systematic reviews and meta-analyses, [23] and is registered at PROS-PERO (CRD42021226388). The protocol can be accessed at https://www.crd.york.ac.uk/prospero/#recordDetails.

Search strategy
Electronic databases We did a systematic search in Medline, Embase, APA PsycINFO, and Web of Science for the period 2000-2020 on 1 December 2020. The following combined search terms were used: RUD OR "resource utility dementia" OR "resource utility in dementia" OR "RUD-FOCA" OR "RUD-formal care" OR "resource utility dementia-formal care" AND caregiver OR "care giver" OR carer OR "care burden" OR "formal care" OR "informal care" OR "nursing home*" OR "care institution", "health policy" OR "health care policy" OR HTA OR "health technology assessment" OR "health economics" OR "cost-benefit" OR "cost benefit*" OR "cost analysis" OR "cost assessment". The search terms were adjusted to the specific databases and consisted of free-text search.
Language We included studies published in English.

Study selection
The reference search and part of the selection method was done by the university librarian RKL and first author RCA. Search results were independently reviewed and screened by MHG and RCA at three levels: titles, abstracts, and full text papers. All studies including cost-estimation and observational analyses not using RUD, RUD-FOCA, RUD Lite were eliminated. RUD Lite is a shorter version of the RUD, which focuses mainly on the patient's resource use more than the caregiver's [24]. The RUD and RUD-FOCA instruments measure the informal care time for family caregivers in homedwelling, and nursing home PwD [25]. The selection criteria of cost-effectiveness and cost of illness (COI) studies were practised according to Drummond and Jefferson [26]. All published protocols were eliminated by RCA and MHG.
Study design and participants We included research papers on informal caregiving to PwD living either at home or in nursing homes. The inclusion and exclusion criteria are displayed in Table 1. We included papers with RCT design, observational analyses, and economic evaluations such as cost-effectiveness and COI studies. Studies that include both observational analyses and economic evaluations assess the relationship between demographic and clinical factors, and total costs of dementia. Reviews, editorials, and conference papers were excluded. Further, we included studies that recruited participants ≥65 years. Studies that do not report age criteria but nevertheless reported a mean sample age higher than 75 were included. We included studies within the OECD countries, including cross-country studies with non-OECD countries in the sample, studies from 2000, and studies that had RUD as primary or secondary outcome. As nursing home dementia patients have significantly lower informal care hours [27] research papers studying formal care with the RUD-FOCA instrument that did not assess informal care were excluded.
Outcomes The focus of this review was to identify factors affecting both formal and informal care hours per month measured by the original RUD and RUD-FOCA, as well as costs at different levels (individual, society). The RUD and RUD-FOCA instruments measure the informal care time in terms of hours and days per month for family caregivers in home-dwelling and nursing home PwD [25]. Both assessment tools comprise three

Type of analyses
RCT design, observational analyses, economic evaluations Descriptive

Type of paper
Research papers with the following design: longitudinal, crosssectional, cross-country, economic evaluations Review, editorial, conference papers

Geography OECD countries
Year 2000-2020 parts: instrumental activities of daily living (IADL) covering assistance with medication, financials, transportation, grocery shopping and cooking meals, personal activities of daily living (PADL or ADL) such as assistance with personal hygiene, and the extent of mobility [17,18,25,28,29]. The RUD instrument assesses informal care on monthly hours of advising and supervision, i.e., surveilling dangerous events.
Quality assessment To evaluate the quality of the observational studies and/or the economic evaluation, we followed the recommendation checklists of international health economic guidelines [26,30], and for COI studies we followed the COI evaluation checklist [31,32]. Further, we followed the quality assessment recommendation according to the GRADE [33]. GRADE provides criteria for rating the quality of evidence that includes study design, risk of bias, imprecision, inconsistency, indirectness, and magnitude of effect. Following Drummond et al. [34] and Molinier et al. [32], we use the following ten points to assess the studies including economic evaluations: clear definition of illness, comprehensive description of competing alternatives, disaggregation of costs, identification of all important and relevant costs and consequences for each alternative, costs and consequences measured accurately in appropriate physical unit, unit costs appropriately valued, careful explanation of the methods adopted, discounting of costs, performance of sensitivity analysis, inclusion of all issue of concerns in the presentation and discussion.

Number and type of studies
The database searches retrieved 307 records. (Fig. 1). We excluded 170 duplicates and eliminated 63 studies in which RUD in the title did not refer to Resource Utilization in Dementia. In addition, from the remaining 74 articles, we excluded protocols and conference abstracts, descriptive studies, including economic evaluations using other input than RUD, and ended up with 29 papers within the period 2000-2020 that were included in the systematic review ( Fig. 1). Three studies were cohort studies, 26 were cross-sectional and two had an RCT design. Nineteen studies combine observational analyses and economic evaluation based on RUD, two Non-cost evaluation studies  [41]. One study, by Sköldunger et al. [42], used the Global Deterioration Scale for assessment of primary degenerative dementia (GDS) (range: 1-7) [43]. Among the studies in a home-dwelling setting [35-38, 40, 44] there was no specified dementia disease in all 7 studies. There was one RCT study by Luttenberger et al. [39] (n = 119) that followed caregivers for persons accommodated in nursing homes within a six-month period. This study found no impact of a non-pharmacological multicomponent therapy including 24 h assistance to ADL and IADL care, on informal caregiver hours within the intervention period. Despite the null finding, there were some limitations that increased the risk of bias. The lack of blinding may have increased distortion. Despite the strength of a control group, the follow-up period was six-months.
Hajek et al. [45] is one of the three longitudinal studies assessing caregiver time for 126 community-dwelling PwD at 6-month intervals over a 1.5-year period, in total 4 assessments. The magnitude of the effect of dementia severity on total caregiving time (formal and informal care) increased by 39.4% per month at 1% level of significance; 68% of the effect was informal caregiving time, which was deemed to have high clinical relevance. Married patients received more total caregiving time (+ 33%), and 72.5% of the effect was informal caregiving hours. Neubauer et al. [36] found a significant effect of the PwD's functional level for both primary caregiver and secondary caregivers, but the difference in the effect between the groups was rather small (0.07 and 0.10, p < 0.05). The t-test showed that informal care time by a secondary informal caregiver was underestimated by 14%. A strength of the study was the comparison with formal care. Parrota et al. [40] found a 1.7 increase in supervision hours per day when PwD had depressive symptoms. Depression was defined with the Cornell Scale for Depression in Dementia (CSDD) [46]. The strength of the study is the magnitude of the findings, Regression 160 MMSE The 6-month non-pharmacological intervention improved dementia symptoms in nursing home residents, especially in social behavior and IADL capabilities, but no effect was seen on informal care time.
(2020) Crosscountry f [40] Crosssectional Community-dwelling; 81.9 (62); MMSE 12.5 Regression 1223 MMSE Depression in PwD is associated with an increased burden and distress of informal caregivers and a reduction of their quality of life.
also including the explicit discussion of confounders, and the comparison with other caregiver burden measurements such as the Zarit caregiver burden scale [47], while a limitation was that it did not include other dimensions of RUD. Two studies investigated disease progress by MMSE using nursing home population comparing people with and without dementia. One of them, by Nordberg et al. [38] found no effect on the level of cognitive functioning on informal care time, while the other, by Nordberg et al. [37] found a significant difference between formal care and informal care users. A major limitation was that these bivariate analyses are not adjusted for other factors associated with both MMSE and care times, such as ADL function and neuropsychiatric symptoms. Sköldunger et al. [42] found that both dependency and disease severity increased informal care hours by 1.5 h per week which was of low magnitude. Finally, a cohort study by Teipel et al. [35] found a significant association between total score of the Neuropsychiatric Inventory (NPI) [48] and increased caregiver burden. This finding was robust to separate items in the NPI instrument, and by clustering within different GPs. A strength of the study was the use of all three parts of RUD assessing hours per month; supervision, IADL and ADL, and the analyses incorporated a comparison of the effect of different psychotropic medication to different NPI-symptoms (items) as an alternative analysis. Among the eight non-cost evaluation studies, one study used RUD-FOCA [39], five studies used all measurements of RUD [36][37][38]42], one study, by Teipel et al. [35] excluded supervision, and one study only, by Parrotta et al. [40], used supervision.
Combination studieseconomic evaluation, observational-and RCT studies Table 3 summarizes 21 studies with economic evaluations (cost-estimations) and provides information on the type of study, sample size, research methods, setting, mean age of the PwD and their caregivers, and conclusions. Vossius et al. [49] was the only study providing longitudinal data with a 2year period and a 6 month interval. Except for Boström et al. [50], Nakabe et al. [51,52], and Hojman et al. [53], all studies in Table 3 [32], Table 4 lists all cost studies providing information on the type of health care system, valuation year, currency, cost variables identified, perspective, the unit for which the costs are provided, total societal costs, and the share of informal costs to total costs.

Characterization of dementia
The cost studies included in Table 3 [71].

Perspective of the cost-analyses and disaggregation of costs
Eighteen of the 21 included economic evaluations in Table 3

Estimation of resource consumption and sources of activity data
Twelve of 21 studies listed in Table 3 [36] and Darba et al. [68] were the only studies that also included costs of the secondary caregiver. In a validation study by Neubauer [36], it was found that 93% of the caregivers yielded complete and plausible feedback on the number of hours of informal care [44]. Three studies [58,63,72] still emphasized possible errors of recall bias and the possibility of overlap between supervision on the one  hand and IADL and ADL based care on the other. These studies used a maximum threshold of number of hours of supervision or IADL and ADL based caregiver hours.

Discounting costs and sensitivity analyses
Since most of the economic evaluations in Table 3 are cross-sectional, there is no discounting of costs. Vossius et al. [49], however, had a follow up of 12-and 24 months. The cost-effectiveness study of the in-home respite care treatment of both caregivers and PwD [54,74] used the annual discount rate (beyond the first year of the model) of 3% for future costs and 1.5% for future QALYs recommended by the Belgian Healthcare Knowledge Centre. Only 10 of 21 economic evaluations listed in Table 3 performed sensitivity analyses [53-55, 58, 59, 63-65, 67, 72]. Hojman et al. [53] and Michalowsky et al. [72] used an alternative valuation method of informal care as sensitivity analysis.

Presentation of results and inclusion of all issues of concern
All 21 studies included in Table 3 provided a careful explanation of the methods adopted. However, one study fulfilled this only partly, i.e. Bostrom et al. [50]. Moreover, included studies also provided an analysis of issues of concern, while only partly fulfilled by the following studies [49,[58][59][60]62]. The cost-effectiveness analysis of the RCT study by Vandepitte et al. [54] used the control group to provide analysis of competing alternatives. The control group however, had treatment as usual and not alternative care which makes the results more indirect [33,75]. Twelve studies provided sources of previous studies, epidemiological approaches, or had a discussion of the problems regarding an absence of previous epidemiological studies. Five studies did this partly [49,[58][59][60][61][62]. Darba et al. [68] mentioned the limitation of not comparing PwD with older adults without dementia. A strength in the design of Boström et al. [50] is the comparison between AD and DLB, yet, the strength is limited by a small sample size of patients with DLB (n = 15). A strength of Carter et al. [55] is the comparison between public and private residential care. While Ersek et al. [58] provided national estimates, the large sample size of both Nakabe et al. [51,52] studies is due to an online survey design.

Discussion
Our primary aim was to identify factors related to PwD and their caregivers, that are strongly associated with high caregiver burden and increased societal costs. We also aimed to determine specific health service interventions that reduce caregiver burden and societal cost levels, and to assess the methodological quality of the economic evaluations of RUD studies. This systematic review illustrates that the functional dependency and severity of dementia are strongly associated with increased monthly hours of care and societal costs, lasting over time. It is further demonstrated that programmes aimed at alleviating caregivers' burden (e.g., the in-home respite care "Baluchonnage") may be more cost-effective and increase QALYs in an RCT study design. Since 2000, only two RCT studies on care burden assessed by RUD are conducted. However, low sample sizes and short followup period encourage researchers to conduct further explorations.
The results of this article are of key importance for society, politicians, and clinicians due to the expected increasing costs of dementia, and the pressure on family caregivers and the primary health care system, in general. While the need for care will continue to increase, the number of potential informal caregivers will stagnate [5]. The results are also crucial when designing new treatment and health care service innovation.
Michalowsky et al. [72] is the only economic evaluation study to fulfil all points of the Drummond criteria, and further, Darba et al. [68] fulfils 9/10. The two studies showed the importance of disease progression to caregiver burden and societal cost. Darba et al. [68] identified that the severity of dementia was not only associated with RUD outcomes in terms of hours per month, but also with all cost components including total costs for the society. The fact that functional dependency and severity of dementia are cost-intensive is also in accordance with previous research [76]. For instance, a study by Schwarzkopf et al. [77] investigated the cost differences between mild and moderate dementia and Information on the share of informal care costs to total cost is authors' own calculations. In Bostrom et al. [50] informal care added as lost production of workers and lost leisure of retired persons, and costs were given according to the following groups: Dementia Lewy Body (DLB) and Alzheimer's Disease (AD). In Gustavsson et al. [61] costs are divided regionally: North Europe (NE), Western Europe (WE) and Southern Europe (SE). In 5 studies costs were given annually, in 2 studies costs were given weekly and further, 2 studies reported costs by 6 months. All costs were transformed into monthly cost. Currency abbreviations: Euro (EUR), US dollar (USD), Great British Pound (GBP) found that informal care costs of moderate dementia exceed the costs of mild dementia by 69.9%, whereas costs for formal health care services differ by 14.3% between the two dementia stages. Furthermore, caregiver factors such as the level of education are also associated with high societal cost due to the alternative costs and thirdparty payers' cost (i.e., government agencies or employers). The total costs also varied significantly across socio-economic status of PwD, in which persons with lower status had higher costs [53]. Finally, single neuropsychiatric symptoms such as depression [40] as well as the total symptom burden [35] generated a high amount of caregiver hours per month. The findings of Hojman et al. [53] demonstrate the importance of differences in costs and resource use of the dementia disease across groups of PwD. However, what the review reveals is that few economic evaluation and resource use studies analyze costs and resource use across groups of caregivers. Two studies with an online sample, and thus a young population of caregivers have done so the last 20 years [51,52]. Moreover, among the evaluated studies, health status as well as comorbidities of caregivers and the influence on the total cost and resource use are absent, and in COI studies, medical care costs are mainly comprised by the medical care cost of PwD. Previous studies have shown a positive relation between caregiving intensity on the one hand and adverse health effects and increased health care utilization on the other [78]. As age is associated with higher medical care costs among PwD [76], different age groups among caregivers affect resource use and costs. Future economic evaluation studies and cost of illness studies should more thoroughly include health care expenses of caregivers.
The MAKS intervention elaborated in Luttenberger et al. [39] -a multicomponent non-pharmacological therapy comprising of motor stimulation, activities of daily living, and cognitive stimulation for nursing home residents (n = 71) which was delivered for 6 months, 6 days a week, 2 h a day [39] with an objective to measure informal care burden and dementia symptoms for PwD in nursing homes had no significant effect on informal care hours, compared to the control group. On the contrary, the in-home respite care intervention study by Vandepitte et al. [54], where the intervention group comprising of home-dwelling PwD and caregivers received in-home respite care support in personal and instrumental ADL every 6 month for 5 days, concluded that their care programme was cost-effective and generated increased QALYs compared with standard community-based dementia care. In the respite program, caregivers could be relieved from their caregiver's task for at least 5 days while a trained employee took their place. For the patient, all daily habits/activities and resource use remained unchanged.
According to our methodological analyses based on the GRADE guidelines and the Drummond criteria, some characteristics of dementia are more strongly associated with higher societal costs than others. The indirect costs and the total societal costs are higher for home-dwelling PwD compared to nursing home dwellers, especially those living alone. Importantly, the direct costs were higher when living in a nursing home. These cost differences were also found in previous non-RUD based studies [21,27]. High costs and resource use were associated with disease severity, functional dependency, and BPSD. It is unclear what type of dementia etiology demands more resources and generate higher costs, but persons who have had the diagnosis for a long period are especially resource intensive. Interventions aimed at home-dwelling PwD, in combination with standard care could relieve the burden for caregivers and increase safety for PwD.
The divergent use of the outcome measurement RUD may potentially under-or overestimate informal care hours, cost levels, and the cost-effectiveness. Although 17 of 26 of the included studies used all three measurements -ADL, IADL and supervision -inclusion of supervision may overlap with ADL and IADL due to recall bias [44]. Similarly, excluding supervision may underestimate care time and costs [29]. Only three studies used a maximum threshold of number of hours to solve potential recall bias. Although the RUD instrument has enhanced comparability of COI studies and other non-cost RUD studies, the review calls for a more harmonized approach.
The clinical and economic importance is to provide knowledge that can lower the hours of care per month along with the direct and societal costs it generates for the family and formal caregivers.
Future studies that aim to reduce the burden of carers and the related societal costs should include randomized controlled trials with longer follow-up and compare different treatment approaches to each other as well as standard care; they should also investigate the incremental cost-effectiveness ratio of the treatment among both PwD living in the municipality and those living in nursing homes. The intricacy of alleviating the entire burden of caregivers demands for health service innovation [79]. However, as organization of health care involves multiple stakeholders and layers of organizational levelsa multicomponent complex interventions could potentially mitigate caregiver burden and costs [79][80][81]. Multicomponent intervention is not a discrete package of separate components, but a process of changing what complex systems do, introducing new ways of how the system can work [82]. The treatment in multicomponent intervention should also target disease severity, functional dependency, and BPSD, and distinguish between different caregiver profiles including their socioeconomic status. The clinical and economic importance of the systematic review is to provide knowledge that can lower the hours of care per month and the direct and societal costs it generates by family and formal caregivers.

Limitation
The criteria we set in this systematic review (Table 1) may potentially exclude important information. First, we set the criterion of only reviewing studies with PwD aged 65 or older, mainly because the degree of resource utilization and cost levels may be very different between young (aged 55 or older) and older PwD [10].
Oure systematic review focused exclusively on the resource use (time) variables. Although this focus enhances comparability of the findings, there are multiple measurements of informal caregiver burden. For instance, we did not consider subjective well-being of caregivers, even though this is highly emphasized in the literature. The Relative Stress Scale (RSS), a self-rated 15-item scale that measures the caregiver's burden of care [83], offers an opportunity to differentiate between different patterns of caregivers distress, while Costa et al. [20] also show how multiple forms of informal care questionnaire are used.
The RUD questionnaire only allows loss of labour productivity information to be gathered from the perspective of the employee. Thus, its analysis requires the use of the human capital approach that may overestimate indirect cost. Alternative approaches exist when valuing productivity loss such as the friction period approach.

Conclusion
The divergent use of the RUD components within RUD studies shows that future economic evaluations and observational studies should be better harmonized. There are only two RCTs with RUD as primary or secondary outcome, and only one of these shows a significant treatment effect. This calls for future RCTs to be more methodologically sound with appropriate sample sizes and a representative follow-up period. Cost-enhancing and resource intensive factors such as disease severity, functional level, and BPSD should be specifically targeted when designing future treatment or intervention for innovate primary health care services. Furthermore, new treatments should diversify between caregiver groups, and between different living situation of the PwD person with dementia.