The costs of dementia in England

Objectives This study measures the average per person and annual total costs of dementia in England in 2015. Methods/Design Up‐to‐date data for England were drawn from multiple sources to identify prevalence of dementia by severity, patterns of health and social care service utilisation and their unit costs, levels of unpaid care and its economic impacts, and other costs of dementia. These data were used in a refined macrosimulation model to estimate annual per‐person and aggregate costs of dementia. Results There are around 690 000 people with dementia in England, of whom 565 000 receive unpaid care or community care or live in a care home. Total annual cost of dementia in England is estimated to be £24.2 billion in 2015, of which 42% (£10.1 billion) is attributable to unpaid care. Social care costs (£10.2 billion) are three times larger than health care costs (£3.8 billion). £6.2 billion of the total social care costs are met by users themselves and their families, with £4.0 billion (39.4%) funded by government. Total annual costs of mild, moderate, and severe dementia are £3.2 billion, £6.9 billion, and £14.1 billion, respectively. Average costs of mild, moderate, and severe dementia are £24 400, £27 450, and £46 050, respectively, per person per year. Conclusions Dementia has huge economic impacts on people living with the illness, their carers, and society as a whole. Better support for people with dementia and their carers, as well as fair and efficient financing of social care services, are essential to address the current and future challenges of dementia.

research, particularly to find disease-modifying as well as symptomatic interventions.
Cost-of-illness (COI) studies aim to identify and measure all costs of a disease or condition to estimate its total impact on society in monetary terms. 4 In the case of dementia, this involves, but is not confined to, estimating total costs of health, social, and unpaid care for all people with dementia. COI studies can raise awareness of the substantial and rising financial impact of dementia and shed light on the adequacy or otherwise of responses to it, thereby acting as a lever for potential reprioritisation of resources.
Studies worldwide have estimated the costs associated with dementia. Some studies focussing on particular population subgroups. [5][6][7][8][9] Two closely related COI studies have been conducted in the United Kingdom, 3,10 with total costs of dementia estimated at £26 billion in 2013, 24% higher than previously estimated in 2007 (adjusting for inflation and additional coverage, largely due to an increase in the number of people with dementia). Two-thirds of the cost arose from unpaid care and payments for privately funded social care borne by people with dementia and their families.
These two UK estimates of the costs of dementia relied on data, now over a decade old, derived from a number of small studies each with criteria that excluded certain groups, such as people with severe dementia. Using up-to-date prevalence estimates, service utilisation and unpaid care data from multiple sources, and a more refined modelling approach than previously employed, we report the per-person and total societal costs of dementia in England in 2015.

| Overall estimation approach
We sought to estimate the societal costs of dementia in England for 2015, encompassing costs of health, social, and unpaid care. Our modelling refines approaches previously used, 3 which in turn built upon related studies by research team members. The availability of data for older people (age 65 and over) and younger people (age 35 to 64) varied; hence, different models were used to estimate costs for these two age groups. Although fewer data are available for younger people with dementia, we have included estimates for younger adults to aid comparability with earlier estimates and to be inclusive of all age groups.
Our model for older people has three parts. First, we divide the older population into subgroups according to relevant characteristics.
Second, we estimate the number of older people with dementia using different types of community care and care home services in each subgroup. Third, we calculate average per-person cost and aggregate costs for older people with dementia at national level ( Figure 1A).
We used the best available secondary data sources to derive reliable estimates for (a) the number of older people with dementia in England in 2015, (b) their receipt of health, social, and unpaid care, and (c) costs associated with that care and other related activities. Our model for younger adults follows the same approach as for older people but has a simpler structure and provides less detailed cost estimates due to data limitations for this group ( Figure 1B).
All three cost categories (health, social, and unpaid care) were estimated separately by severity of dementia (mild, moderate, and severe) and by year since onset of dementia (first year and subsequent years).
National Health Service (NHS) costs were split by primary and secondary care. Social care costs were split between publicly and privately funded care. All reported estimates are annual costs for England for 2015, in pounds sterling (£) at 2015 price level. Cost estimates represent a snapshot for 2015, not lifetime costs. Discounting was unnecessary since all costs refer to a 1-year period.

| Data sources
Estimates were derived from multiple sources: (1) Numbers of older people and younger adults in England, disaggregated by age and gender, come from the 2015 population estimates published by the Office for National Statistics. 11 (2) Estimates from the Population Ageing and Care Simulation (PACSim) model 2,12,13 were used to generate prevalence rates of cognitive impairment by severity and care needs by "interval need" in the older population. 14 Rates were estimated by age, gender, and education. Full details of validation of the PACSim model have previously been described. 13 (3) The proportion of older people with dementia, by age, gender, and education, and the proportion of those with dementia who receive unpaid care, formal community care, and care home services according to their characteristics (age, gender, education, and severity of cognitive impairment) were estimated using

Key points
• On the basis of the newly available data and refined modelling, we estimate the total annual cost of dementia in England to be £24.2 billion in 2015, £2 billion higher than the previous estimate for 2013.
• Family and other unpaid carers make substantial contributions to the support of people with dementia.
Given rapid population ageing, the already substantial demand for, and costs of, unpaid care is expected to increase enormously in the future, which calls for better support for carers.   Table S1.

| Measurement of dementia
To estimate the numbers of older people with dementia in 2015, we applied prevalence rates by age, gender, and education from the PACSim model, which drew on CFASII data, to ONS population estimates for 2015. We took account of years of education because the prevalence of cognitive impairment varies with years of education.
Identification of dementia in CFASII was based on the well-established AGECAT algorithm. 15,22,23 Incidence and prevalence rates derived from CFASII relate to this definition. Overall numbers of people with dementia were divided into three severity levels (mild, moderate, and severe) using a breakdown that maps to the conventional Mini-Mental State Examination (MMSE). As previously, 3 the following cut-off points were used: 21 to 26 for mild, 10 to 20 for moderate, and less than 10 for severe dementia. Numbers by severity were further divided by extent of care needs (independent, requiring help less often than daily, requiring help at regular times of the day, and requiring 24-hour care) based on Isaac and Neville's interval needs classification 14 to ensure greater accuracy in cost calculation. We conducted multinomial logit regression analyses of CFASII to examine proportions of older people in each subgroup by age, gender, education, and severity of dementia who received no care, unpaid care only, formal community care only, both unpaid and formal community care, or care home services. We included education as an explanatory variable because the receipt of unpaid care and formal care varies with socio-economic group with which education is closely associated. We used the fitted values from the regression model as the estimated proportions of each subgroup of the older population who received no care, unpaid care only, etc. We then applied these estimated proportions to the numbers of older people in each subgroup to estimate total numbers of older people nationally receiving unpaid care, formal community care, and care home services. We scaled the resulting national estimates for formal community care and care home services (but not unpaid care) to externally derived total numbers of older service users in England. 21,24 On the basis of CFASII, we assumed that 70% of older care home residents and 25% of older users of community care services in England have dementia.

| Costing health and social care
We applied annual costs calculated from the MODEM cohort to our estimates of numbers of people with dementia using health, social, and unpaid care to calculate annual total costs in the older population with dementia. All health care costs are assumed to be met entirely by the NHS. Social care costs are divided between costs met by local authorities and those met by service users on an assumption that service users with dementia are divided between publicly and privately funded users in line with the breakdown for all older care service users in England. 24 The MODEM cohort found considerable differences in service receipt between first and second interviews, except for those living in care homes. Since most members of the community sample were recruited from memory services, they are likely to have received a dementia diagnosis not long before their first interview. This may explain why they received more secondary health care but less formal social care at first interview than at second interview (12 months later). Therefore, we used service use data from the first interviews for the first year of care (incidence numbers) and data from the second interviews for second and subsequent years of care.

| Costing unpaid care
Consistent with previous studies, 3 At baseline, unpaid carers spent 10% of their caring time on assisting with ADLs, 19% assisting with IADLs, and 72% on supervision.
Total time spent on ADL tasks was calculated by multiplying the total time spent caring by the unpaid carer (from CSRI) by the proportion of time spent on ADL tasks (from BADLS). This was valued at £18 per hour, the replacement cost of an hour of formal home care. 20 The remaining proportion of time spent caring was valued using an opportunity cost approach, assuming that carers who were not retired were forgoing employment to provide care. For each carer, this approach applied a value for unpaid care equal to the average wage for an individual with the same age, gender, and occupation (or the same age and gender if they did not report a specific occupation). For retired carers, the opportunity cost applied was the 2016 National Living Wage (£7.20 per hour) deflated to 2015 prices. 26 The survey additionally asked about time provided by other carers.
This time was also costed to reflect total unpaid care time received by the care recipient. Time contributed by other carers was allocated between ADL tasks, IADL tasks, and supervision based on the   spending, but they also relate to a general sample of people receiving services rather than a more specific (and almost certainly less representative) group consenting to participate in trials. We have thus addressed some of the limitations of the two previous UK estimates by using data from surveys, which are more recent, have larger samples, and are more representative of people with dementia than data sources used in the earlier studies.
Consistent with similar studies in other countries, 31 we find that unpaid care accounts for a substantial proportion of the total cost of dementia. It is projected that both the number and proportion of the older population, especially those aged 85 and over (among whom prevalence rates of dementia are highest), will continue to increase rapidly in England. 11 The demand for and costs of unpaid care for people with dementia can be expected to grow substantially in the coming decades.

| Strength and limitations
Drawing on newly available data from multiple sources including 2015 ONS population data, NHS Digital data on receipt of publicly funded social care, the National Living Wage, CFASII, and MODEM, our study presents comprehensive, up-to-date evidence on a range of costs including those associated with end-of-life care and young-onset

| Policy implications
People with dementia receive combined support from health care and social care professionals and unpaid carers. Hence, coordination, synergy, and mutual support between sectors should be encouraged to better serve the needs of people with dementia and tackle the associated social and economic challenges. Given the substantial contribution of unpaid carers and the scale of unpaid care costs, support is essential for those carers, in order to promote their health and wellbeing and enable them (if they wish) to combine caring with employment or other activities. This support could include increased information and advice services, increased resources for respite care, or increased cash payments to carers.
The economic impact of dementia is not evenly shared between the health and social care systems, but weighs heavily on the already underfunded social care sector. The reduction in central government funding for local authorities in recent years has impacted on resources for social care leading to a decline in the numbers of older people receiving publicly funded community-based and residential care.
Unlike health care that is free of charge at the point of use in England, the entitlement to publicly funded social care depends upon service users' income and assets, and eligibility criteria for publicly funded care vary considerably across England due to the discretionary power of local authorities. 33 Around 70% of care home residents have dementia, 23 and a substantial proportion of the social care costs is met by people with dementia or their families. These findings further highlight the importance of addressing the challenges of social care financing. In particular, social care should be financed fairly and efficiently to make sure that high-quality care services can be delivered in a timely fashion to those people who need them.

DATA AVAILABILITY
The modelling uses a range of data sets, as explained above and in Figure 1. The ONS data are available at www.ons.gov.uk. The NHS Digital data are available at www.digital.nhs.uk/data-and-information/areas-of-interest/social-care. The CFASII data are available by application to the CFAS research team as explained at www.cfas.
ac.uk. The MODEM study data will be deposited with the UK Data Archive.