Projections of costs and quality adjusted life years lost due to dementia from 2020 to 2050: A population‐based microsimulation study

Efficient healthcare planning requires reliable projections of the future increase in costs and quality‐adjusted life years (QALYs) lost due to dementia.

projections, however, do not account for uncertainties about the natural history of dementia, established declining trends in age-specific dementia incidence, 5,6 or uncertainties about costs attributed to dementia.
The burden of disease for a society is captured in the combination of (healthy) life years lost and monetary cost.Quality-of-life reduction often extends beyond the person with dementia, also affecting caregivers.When applied to policymaking, quality of life is commonly expressed as quality-adjusted life years (QALYs).QALYs lost capture reductions in quality of life for years lived with disease, as well as loss of life years due to premature death.Projections of QALYs present useful insights for policymakers into how the burden of a disease will develop, thereby guiding policy and allocation of healthcare resources.However, the anticipated effect of the change in dementia prevalence on societal costs and QALYs is uncertain.
Societal costs and QALYs lost are highly dependent on dementia severity and living situation.Based on symptoms and dependence in (instrumental) activities of daily living, dementia is often divided into three severity stages: mild, moderate, and severe, 7 preceded by a stage of mild cognitive impairment (MCI) in which activities of daily living are generally unimpaired. 8With increasing dementia severity, the burden on formal and informal caregivers increases, either due to additional care needs for patients living at home or the need for institutional care.The living situation of people with dementia is typically divided into home dwelling (ie, living at home, possibly with an informal caregiver) and institutionalized care in nursing homes.In some countries, specialized healthcare professionals (ie, case managers) oversee and coordinate continued care of community-dwelling patients with dementia to facilitate independent living and quality of life for as long as possible.Costs due to a disease can be divided into separate cost types (eg, medical, formal and informal care, institution).Providing insight into the contribution of dementia severity, living situation, and cost type to the total costs and QALYs lost will provide tools for policymakers to prioritize interventions and ensure optimal care for patients with dementia despite increasing demands on the healthcare system.Therefore, the aim of this study was to provide projections of costs and QALYs lost due to dementia from 2020 to 2050 in the Netherlands from a population perspective and to describe the contribution of various severity, living situation, and cost types.

METHODS
To estimate the costs and QALYs lost due to dementia in the Netherlands for the coming decades, we conducted an analysis using the MISCAN-Dementia microsimulation model.MISCAN-Dementia is a microsimulation model based on the validated Microsimulation Screening Analysis (MISCAN) model. 9We have published details of the model elsewhere. 10

Data
Age-and sex-specific dementia data were derived from participants of the Rotterdam Study, an ongoing population-based cohort study of 14,926 individuals aged >40 years who reside in the Ommoord suburb of Rotterdam, the Netherlands.Details of the study design and of case ascertainment for dementia and mortality have been described elsewhere. 11,12Dementia prevalence and overall life expectancy at birth in the Netherlands lie in the middle of the European distribution. 13,14Compared to the United States, the Netherlands has slightly lower age-standardized dementia prevalence (USA: 6.77%, Netherlands: 6.44%) 3,13 and somewhat higher life expectancy (USA: 78.50, Netherlands: 81.79).Costs are considered from a societal perspective, including healthcare costs and informal care costs.Average annual cost data for home-dwelling dementia patients with and without case managers (Linkage model) was taken from a Dutch cost-effectiveness study. 15formal care costs consist of the hours of care provided by an informal caregiver. 15Other societal costs, such as productivity losses, are not included.7][18] Based on Dutch health insurance claims data from 2020, 19 the share of institutionalized patients was set at 30%.Based on case management research in the Netherlands in 2018, 20 the share of home-dwelling patients with a case manager was set at 30%.Consequently, the share of home-dwelling patients without a case manager was set at 40%.Finally, the annual cost of institutionalized patients was based on estimates from Dutch healthcare insurances. 21The costs in an institution are assumed to be independent of the severity stage due to the substantial amount of fixed costs (ie, accommodation, food, nurses) compared to severity-specific variable costs.One study that investigated costs by dementia severity stage and setting supported this assumption, finding no significant difference between severity stages in an institution but a significant increase in costs by severity stage in home care. 22Due to the high level of dependency and need for (informal) care in the severer dementia stages, home-dwelling costs can be higher than institutionalization costs.This is the case for severe dementia and moderate dementia without a case manager.For this study, the utility assumptions for home-dwelling persons with dementia were taken from a recent review on utility values by dementia severity. 23The weighted mean of the EQ-5D proxy-rated utilities from European and North American studies was used in the absence of severity-specific utilities for the Netherlands.The difference in utilities between home-dwelling and institutionalized patients was taken from a study investigating utilities in home and institutional care. 24Furthermore, average utility data for caregivers were taken from the same Dutch cost-effectiveness study as the cost data. 15e difference in utilities between severity stages and settings for caregivers was taken from a study from the United States 25 with a similar average caregiver utility.Table 1 summarizes the resulting utility assumptions by dementia severity and care setting.A utility of one equates to perfect health, and a utility of zero represents death.The age-specific norm utility values for the Dutch population were adjusted for dementia and caregiving utilities (for details see Supplement A).For this analysis, MISCAN-Dementia was further developed to account for differences in costs and QALYs by disease severity and care setting (home vs institution) for both patient and caregiver (for details see Supplements B and C). Figure 1 depicts the MISCAN-Dementia model structure used in this study.Furthermore, we explicitly modeled informal caregivers in terms of partner and child caregivers to account for caregiver loss of quality of life by age (for details see Supplement D).

Base case analysis
We estimated annual costs and QALYs lost in the Netherlands due to dementia from 2020 to 2050.The model output was scaled to reflect the observed number of people born in the birth cohorts. 26Based on a recent meta-analysis of North American and European cohorts, 27 we assumed a linear decline in the age-specific incidence of 13% per decade for the birth cohorts that contributed to the meta-analysis

Sensitivity and scenario analyses
In univariate sensitivity analyses, we varied several uncertain parameters to investigate their influence on the results.First, we varied the stage-specific cost estimates using the 95% confidence intervals of the estimates. 154][25] Third, we varied the proportion of home-dwelling individuals with a case manager using the 25th and 75th percentile of the estimate, as well as the extreme scenarios of no case managers and case managers for every home-dwelling individual. 20Fourth, we varied the institutionalization prevalence from 20% to 45%. 28 the scenario analyses, we assessed the impact of structural uncertainty on the results.First, we varied the distribution of time spent in the severity stages (ie, increasing and decreasing time spent in milder stages, equal time across all stages). 29Second, we investigated the effect of different age-specific incidence trends between birth cohorts (ie, no trend, linear trend of 13%, nonlinear trends averaging 13% 10 ).
Third, we investigated the contribution of MCI and mild dementia costs associated with institutionalization as well as MCI costs in general.Fourth, we investigated different sets of caregiver (dis)utilities.A detailed overview of the sensitivity and scenario analyses can be found in Supplement E and F.

Cost projection until 2050
Dementia costs in 2020 were estimated at €13 billion, in line with earlier reports of dementia care costs in the Netherlands in 2017. 30e total societal cost of dementia in the Netherlands is expected to increase up to 2050 (Figure 2).Compared to the year 2020, dementia costs in the Netherlands will have increased by 30% in 2030, to more than €17 billion annually, and will further increase to an annual €26 billion by 2050 (+97% compared to 2020; Table 2).
The majority of the costs are incurred by institutionalized individuals (34%), followed closely by care costs for home-dwelling individuals (31%).This can be explained by the high institution costs for the 30% of individuals living in an institution and lower costs for the 70% of individuals living at home (Supplement C).The third biggest contributor to total costs is informal care costs for home-dwelling individuals (20%).
Up to 2050, costs associated with dementia for those living in an institution will increase at a higher rate (108%) than costs associated with dementia for those living at home (92%).Despite the fact that people live only a relatively short time with severe dementia (Supplement G), total costs are as high as for the earlier stages of the disease due to the high costs for people with severe dementia.Projected increases in cost over time are highest for people with severe dementia, compared to earlier stages of the disease (Table 2).

QALYs projection until 2050
In the year 2020, 141,000 QALYs were lost due to dementia in the Netherlands.Compared to the year 2020, QALYs lost will have increased to 188,000 by 2030 (+34%) and will further increase to 311,000 by 2050 (+120%; Table 3).Life years lost due to dementia are the largest cause of QALYs lost, accounting for two thirds of all QALYs lost.While most of the life years with dementia are lived in the mild to moderate stages of the disease, it is the life years with severe dementia that account for most of the QALYs lost.By 2050, QALYs lost will increase at a faster rate for the more severe dementia stages (35% for MCI compared to 110% for severe dementia).However, life years lost show the greatest increase between 2020 and 2050 (139%).
As expected, QALYs lost by patients outweigh the QALYs lost by caregivers.QALYs lost by caregivers make up 2.5% to 3% of total QALYs lost and 7.5% to 8% of QALYs lost excluding life years lost.Between  3).
Hence, QALYs lost by caregivers represent an important aspect of the dementia burden.

Sensitivity and scenario analyses
A detailed overview of the results of the sensitivity and scenario analyses can be found in Supplements E and F, and the results are summarized in Figure 3.We found that the results were least influenced by changes in case manager prevalence, institutionalization preva-lence, severity stage distribution, and excluding MCI institution costs (−13% to +6% compared to the main analysis).In contrast, projections were highly dependent on changes to the cost and utility assumptions, such that incorporation of the wide confidence intervals of the cost and utility estimates resulted in projections that were 36% lower to 44% higher compared to the main analysis.The scenario analysis investigating trends in age-specific incidence showed that the effect of the trend assumption on the projections is increasing with time, hence the uncertainty around the projections.

DISCUSSION
The burden of dementia in the Netherlands, both in terms of costs and QALYs lost, is set to double by 2050.The biggest contributors to costs are institutionalization costs (34%), home-dwelling care costs (31%), and home-dwelling informal care costs (20%).Furthermore, between 2020 and 2050, costs will increase at a faster rate with increasing dementia severity (mild dementia 91% compared to severe dementia 111%).Life years lost is the single most important contributor to QALYs lost, accounting for two thirds of total QALYs lost.Also, QALYs lost will increase between 2020 and 2050 at a faster rate with increasing dementia severity.
In 2017, formal dementia care costs in the Netherlands were €9.1 billion. 30The difference between these costs and the projections in this study (€13 billion in 2020) can be explained by the inclusion of informal costs in the study (€3 billion in 2020).Hence, the cost estimates of this study align well with the observed dementia costs in the Netherlands.
Several studies, including the widely cited World Alzheimer Reports, 3 have reported on the basis of an aging population, that the costs attributable to dementia will increase dramatically in the coming decades (doubling every 10 years).2][33] An exception is a population-based microsimulation study from the UK 4 that reported projections of dementia care costs between 2019 and 2040.This study also found costs to almost double every 10 years, from £37 billion in 2020 to £59 billion in 2030 and £94 billion in 2040.
These projections, not accounting for secular trends in dementia incidence, are in line with the projection of the "stable age-specific incidence" scenario in the sensitivity analysis.Given observed declines in age-specific incidence in European and North American cohort studies, 27 the "stable age-specific incidence" scenario may not be the most realistic scenario for the future and could overestimate billion by 2040 4 ).These results underscore that redoubled efforts into the prevention of dementia over the life course are likely to be cost-effective in reducing dementia burden.
To our knowledge, there are no estimates of the population loss of QALYs due to dementia.Comparisons with other diseases such as COVID-19 can put the estimates of this study into context.We estimated that 141,000 QALYs were lost in 2020 due to dementia, including the burden of disease and dementia mortality, and 87,000 QALYs were lost due to dementia mortality alone.For 2020, it was estimated that 61,000 QALYs were lost due to COVID-19 mortality in the Netherlands. 34Another study estimated that 5,000 QALYs were lost per one million individuals in the United States and the UK due to COVID-19 deaths. 35Transforming the estimates of this study to the same denominator results in 5,400 QALYs lost per one million individuals due to dementia mortality.The loss in QALYs due to dementia mortality appears to be slightly higher than for COVID-19 mortality.
Caregivers of individuals with dementia are an important but often overlooked group affected by dementia.Caregivers of individuals with dementia experience high levels of burden, 36 resulting in lower levels of utility.Moreover, caregivers often spend significant amounts of time with caregiving duties and therefore need to reduce or give up their formal work. 37These often overlooked costs are the third biggest contributor to costs in our study.Although QALYs lost by patients make up the vast majority of the total QALYs lost in the population, it should be noted that utilities lost by caregivers are difficult to estimate using a generic measure like the EQ-5D, and a substantial share of the caregiver burden may therefore not be accounted for in available studies on caregiver utilities. 25Due to this measurement shortcoming, changes in utility by severity stages are often negligible in studies on dementia caregiving.Without reliable and (inter)nationally accepted specific dementia-caregiving measures, estimates from the literature that showed little impact of severity stages on caregiver utility had to be used.More sensitive utility estimates are needed to accurately reflect the overall perceived loss of quality of life for caregivers in future studies.Despite these challenges to account for caregiver burden and informal costs, we believe that by explicitly modeling caregivers we capture the effect of caregivers on costs and QALYs lost as best and detailed as possible.
Case managers reduce the costs of dementia patients by coordinating continued care.The estimates on case managers in this study are based on a Dutch cost-effectiveness study that investigated the effect of two different case manager forms on dementia outcomes, including costs and quality-of-life measures. 15The study found that case managers significantly lowered costs, especially informal care and day center costs.However, the study did not find an improvement in quality of life of either the patient or the caregiver.The sensitivity analysis on case manager prevalence showed that dementia costs can be lowered by 16% (Table E.3) if every dementia patient had a case manager.The use of case managers should therefore be promoted to lower costs due to dementia.
The introduction of disease-modifying treatments (DMTs) for Alzheimer's disease and other dementias is not modeled in this study and could impact the future costs and QALYs due to dementia greatly.
National governments and international agencies have adopted policies aiming to reduce the dementia burden and formulated goals to find a cure or DMTs for dementia by 2025. 38While it is not clear whether the 2025 goal is going to be reached, drug development in all phases of clinical trials has increased in recent years. 39,40In 2021, there were 152 ongoing clinical dementia drug trials, most of which investigated the safety and efficacy of DMTs (82.5%) and a significant portion were Phase 3 trials (18.5%). 41DMTs will most likely lead to more life years spent in earlier dementia stages and improvements in quality of life, resulting in fewer QALYs lost.Cost projections incorporating DMTs depend on the price of the drug but are also likely to be lower than current projections.DMTs might become available within the timeframe of this study and thus make the projections in the far future more uncertain.
Attitudes and expectations regarding family and caregiving responsibilities vary greatly between cultures. 42,43Utilization of formal care outside of the family is considered less often in cultures with strong intergenerational family ties such as Hispanic, Asian, and African cultures. 44Caucasian Americans and Northern and Western Europeans make more use of formal care opportunities.Hence, informalformal caregiving patterns influence costs and quality of life related to dementia.Furthermore, policies that encourage patients to stay at home as long as possible and limit access to nursing homes increase the burden on caregivers.In Northern European countries, informal care and indirect costs represent a third to two-fifths of the total costs. 18,43This study finds similar estimates for the Dutch population with roughly 30% of the costs being attributed to informal and welfare costs.The cost and QALY estimates used in this study represent the Dutch culture with high utilization of formal care opportunities, especially in the severe dementia stage.Projections for other cultures that have a strong intergenerational focus will inevitably have higher informal costs and lower formal costs.
Although mild dementia is the most prevalent dementia stage, severer stages are causing most of the costs and QALYs lost.This can be explained by the inverse relationship between both costs and quality of life and the dementia severity stages due to increasing levels of dependency. 23,45Interventions that target cost reductions and qualityof-life improvements in the severe dementia stage or interventions that prolong the duration of the less severe stages might have a large effect on the population level and should therefore receive more attention.
Although we believe the model provides valid projections of the future dementia burden, some limitations should be taken into account.
Any model is a simplification of reality, and its accuracy depends on the underlying model assumptions and quality of the data.While data on certain model inputs in this study were limited, we chose the best available data and investigated the uncertainty around model assumptions and inputs in various sensitivity and scenario analyses.First, the model was developed to provide projections for the Dutch population; however, not all input parameters were available for the Netherlands.
For example, severity-specific utility estimates were not available for the Netherlands, so European and North American estimates from a recent review were used. 23While patient populations and methodologies are likely to differ between countries, the best available utility data were chosen, and it was confirmed to be in accordance with average Dutch utility data included in the review. 23Age-specific EQ-5D utility data for the general Dutch population were taken from a report of the EuroQol Group, 46 an international, multidisciplinary group responsible for the development and maintenance of the EQ-5D measure.Average cost data were based on Dutch studies, providing estimates from the same healthcare system and time frame to ensure compatibility. 15,21e distribution between severity stages was based on estimates from three European reviews [16][17][18] in the absence of Dutch severity-specific estimates.While we believe that we chose the best available estimates for this analysis, better data, especially severity specific quality-oflife data for patients and caregivers, are needed.The aim of providing projections for the Dutch population also presents another limitation.
Although changes in population structure and relative reductions in dementia incidence are similar across high-income countries, 27,47 this may hamper the generalizability of the projections.Furthermore, cost and quality-of-life estimates are often country specific and depend on a variety of factors, such as policy recommendations, family values, and accessibility of the healthcare system.While the absolute projections will most likely differ per country, we believe the overall conclusions will be similar for most other high-income countries.
Second, the measurement and estimation of quality of life is important for the evaluation of the burden of a disease, but it is difficult to quantify.Although common measurements, such as QALYs and EQ-5D utilities, do not capture all relevant aspects of quality of life, we used the best estimates that are presently available.Third, a clear distinction of costs attributable to dementia and regular healthcare costs is difficult, especially in the less severe disease stages.Scenario analyses assuming no MCI and mild dementia (institution) costs, however, showed cost estimates that aligned well with the observed dementia costs in the Netherlands (Supplement F).Finally, we did not account for the potential interaction of chronic comorbidity on dementia-related care costs.

CONCLUSION
This analysis showed that costs and QALYs lost are set to double between 2020 and 2050 due to aging of the population and trends in dementia incidence, in comparison with other projections showing a doubling of costs every decade due to aging of the population alone.
The estimates of this study can be used by policymakers to allocate -Dementia, coded in Python 3.10, is a stochastic microsimulation model that predicts how dementia outcomes will develop by 2050.The model simulates the life histories of 10 million individuals, representative of the Dutch population in 10-year birth cohorts from 1910 to 1989.The model allows for unidirectional stage transitions between cognitively normal, MCI, mild dementia, moderate dementia, severe dementia, and death due to dementia.MISCAN-Dementia is a time-toevent model; hence, transitions are determined by the duration of the stages.In any stage, an individual may also die of other causes, informed by birth cohort-specific life tables.Elevated mortality of individuals with dementia is implicitly introduced by the duration of the dementia stage, which is often shorter than the remaining life expectancy without dementia.The model was calibrated to fit age-specific dementia incidence data of the Rotterdam Study.The model development and calibration were published previously. 10

(
1910 to 1940) and stable age-specific incidence for subsequent birth cohorts (1950 to 1980).To estimate the costs of dementia, we multiplied the time spent in the different dementia stages, as predicted by MISCAN-Dementia, by F I G U R E 1 MISCAN-Dementia model structure with severity stages and institutionalization. the corresponding cost estimates.In addition to estimating the total costs, we also estimated costs by severity stage (MCI, mild, moderate, and severe) and by the type of costs (institution, medical, care, informal, welfare, and case management).

F I G U R E 2
Annual costs (in billions of euros) and QALYs lost (in thousands) projections.(A) Costs by severity stage.(B) QALYs lost by severity stage.(C) Costs by cost type.(D) QALYs lost by patients and caregivers.2020 and 2050, QALYs lost by caregivers will almost double (+72%), similar to the doubling of QALYs lost by patients (+122%; Table costs.Incorporation of established temporal trends in incidence rates results in more realistic projections, which substantially attenuate the projected increase of dementia costs and QALYs lost.Transporting the results of this study to regions with fairly similar disease incidence and F I G U R E 3 Tornado diagram of sensitivity and scenario analyses for 2020, 2030, and 2050.(A) Costs (in billions of euros).(B) QALYs lost (in thousands).life expectancy, compared to the Netherlands, would imply costs of USD 2 trillion by 2050 in the United States (instead of USD 2 trillion already by 2030 3 ) and £75 billion by 2050 in the UK (instead of £94 14

Table 1
To estimate the QALYs lost due to dementia, we multiplied the time spent in the different dementia stages, as predicted by MISCAN- Dementia, by the corresponding difference in utility compared to the age-specific utility norm (ie, QALYs lost = age-specific utility norm − stage-specific utility estimate).We assumed that the QALYs lost had to be non-negative.Hence, if the stage-specific utility estimate was higher than the utility norm of the age group, we assumed that no QALYs were lost due to dementia.This applied mostly to the oldest age group and caregivers (Supplement A).Similarly, we multiplied the total life years lost due to dementia by the corresponding age-specific utility norm to compute QALYs lost due to premature dementia death.In addition to estimating the total QALYs lost, we also estimated QALYs lost by severity stage (MCI, mild, moderate, and severe) and separately for patient and caregiver.
Finally, when assuming different sets of caregiver disutilities, life years lost and QALYs lost by individuals with dementia remained the greatest source of QALYs lost.TA B L E 2