Formal and Informal Costs of Care for People With Dementia Who Experience a Transition to Hospital at the End of Life: A Secondary Data Analysis

to evaluate

Although reducing multiple transitions to hospital is likely to result in lower hospital costs, 11 it is not clear how this would affect community or residential care, or the cost for families.
Current estimated costs of dementia care per person are significant. In England, these are projected to increase 68% by 2040, with social care costs projected to increase by 92%. 12 Informal care costs are defined as the cost of the time spent by family or friends caring for a person without payment and they can include the costs of stopping or reducing work. 13 Informal care constitutes an important component of dementia care. 14,15 It has been estimated that informal care costs account for around 40% of all total costs in this population 12 and can result in a significant and multidimensional caregiver burden. 16,17 Although end-of-life care costs have been investigated in other populations, 18e20 and dementia care costs have been previously explored, 12,14,21,22 fewer studies have focused specifically on the endof-life period for people with dementia, in particular considering informal care costs. 23 The aim of this study is to explore formal and informal care costs in the last 3 months of life for people with dementia and to evaluate the association between a transition to hospital and usual place of care, with costs, to understand the effect of transitions to hospital on formal and informal care costs.

Study Design
Cross-sectional study using pooled data from 3 mortality followback postal surveys: the International Access Rights and Empowerment studies I (IARE-I) 17 and II (IARE-II), 1 and OPTCare Elderly study. 24 Adults 65 years or older with and without access to palliative care services were identified from hospital records in the IARE-I and IARE-II studies, respectively. 1,17 In the OPTCare Elderly study, decedents 75 years or older with causes of death common in advanced age and suitable for palliative care were identified from death registration records by the Office of National Statistics. 24 The 3 surveys were sent by post 3 to 10 months after the participant died to their next of kin in IARE-I and IARE-II, or the person who registered the death in the OPTCare Elderly study (Supplementary Table 1).

Population
We included people with dementia as underlying or contributing cause of death.

Outcome
The primary outcome was total costs of care in the last 3 months of life. Costs were calculated using a bottom-up approach. 25 The Client Service Receipt Inventory was used to collect information on health and social care services use and informal time spent caring for the participant. 26 We calculated the costs of services by multiplying the number of specific services used and 2020 unit costs according to the Client Service Receipt Inventory (Supplementary Table 2). When unit costs were only available for previous years, values were inflated to year 2020. When a service was reported as used but not the quantity, we imputed the median costs of the sample.
Secondary outcomes were cost components, calculated by aggregating the costs into 5 categories based on Gardiner et al 13 :

Explanatory Variables
Transition to hospital was defined as 1 or more days in hospital during the last 3 months of life for people whose usual place of care was home or care home, or people whose usual place of care was reported as hospital but spent less than 90 days in hospital.
Usual place of care was derived from respondents' answers to the question "Where did the decedent spend most of his/her last 3 months?" and were grouped by home (including own home or a relative/friend home), care home (including nursing home or residential home), and hospital.

Covariables
We included sociodemographic information collected in the surveys on age at death, gender, ethnicity, marital and financial status, having a key contact in the last 3 months of life, and whether proximity to death was discussed with the decedent or their family. Categories of ethnicity were grouped as White or non-White. Marital status was recorded as married or with partner, widowed, divorced, or single. Financial status was collected using the question "How well would you say she/he was managing financially during the last 3 months of life?" and recoded as living comfortably or doing alright, just about getting by, or finding it quite or very difficult.
We calculated the EQ5D index based on the value set for England in Devlin et al. 27 The EQ5D is a standardized measure of health status developed for clinical and economic appraisal. 28 The instrument measures self-perceived health status in 5 dimensions: mobility, selfcare, usual activities, pain or discomfort, and anxiety or depression. The EQ5D-5L version measures each dimension using a 5-level Likerttype scale. The EQ5D index can be calculated from the answers to these 5 dimensions, and it is used to represent people's values of different health states. The index ranges from less than 0 to 1, where 0 is the value of a health state equivalent to dead, negative values represent values worse than dead, and 1 the value of full health. 27

Analysis
We reported the proportion of people receiving services in the last 3 months of life and the mean number of days (for hospital and care home care), contacts (for other services), or hours per week (for informal care) for the whole sample, and people with and without a transition to hospital. We calculated costs by cost components (hospital, care home, community, palliative care and informal care costs) and total costs in the last 3 months of life. We calculated the proportion of the total costs that are hospital, community, care home, palliative care, and informal care costs by dividing the mean costs for each cost component by the mean total costs.
We used a generalized linear model with a gamma family and log link to explore the association between a transition to hospital, usual place of care, total, formal (hospital, community, care home and palliative care) and informal care costs, adjusted by covariables. We did not include ethnicity in the models as most people were White. We reported the adjusted estimated mean of total, formal and indirect costs and 95% CIs for transition of care and usual place of care, with covariables at mean.
We used generalized linear models with a gamma family and log link, adjusted by covariables to explore the effect of transition to hospital on the different cost components. We reported the adjusted estimated means and 95% CIs for transition to hospital as main explanatory variable, for the main model and by usual place of care (home or care home), with covariables at mean. Palliative care costs were included in total costs but were not analyzed as a secondary outcome, as costs were small.
Missing values in covariables were assumed missing at random and were imputed using chain multiple imputations to minimize bias and maximize available information. A sensitivity analysis was performed for the total costs and subgroup analysis including covariables without imputation.
The analysis was performed using Stata MP 16.1.

Characteristics of the Study Sample
We identified 146 decedents (IARE-I n ¼ 27; IARE-II n ¼ 22; OPT-Care Elderly n ¼ 97) with a diagnosis of dementia. The mean age was 88.1 years (SD 6.0), 137 (93.8%) were White, 98 (67.1%) female, 110 (75.3%) reported to live comfortably, and 93 (63.7%) were widows or divorced. The place of death was hospital for 57 (39.0%) people and care home for 65 (44.5%) ( Table 1). The mean EQ5D index was 0.68 (SD 0.13), with a response of slight or moderate to most of the 5 EQ5D dimensions. Twenty-three respondents (15.8%) reported the decedent experienced moderate pain or discomfort, and 31 (21.2%) reported anxiety or depression in the last 3 months of life.
Respondents had a mean age of 64.9 years (SD 9.8), 64.4% were the decedent's son or daughter, 67.1% female, and 89.0% had secondary school or higher level of education (Table 1).

Transition to Hospital and Usual Place of Care
Of 146 people, 85 (58.3%) experienced a transition to hospital in the last 3 months of life. The usual place of care was reported as care home for 85 (58.2%), or their own home or a relative's or friend's home for 58 (39.7%). In addition, 74.1% of people whose usual place of care was home experienced a transition to hospital vs 45.9% among those in care homes.

Service Use in the Last 3 Months of Life
The average number of days in hospital for the whole cohort was 6.9 (SD 14.0). People had on average 0.8 (SD 1.2) visits to the emergency department (ED), 0.8 (SD 1.2) ambulance calls, and 0.3 (SD 1.1) outpatient appointments. Ninety people (61.6%) spent at least 1 day in a care home [average number of days 39.9 (SD 42.8) for total cohort]. Among the 85 people reporting care home as their usual place of care, the average number of days spent in a care home was 67.4 (SD 35.8).
Overall, 128 (87.7%) and 95 (65.1%) had at least 1 face-to-face contact with GPs and community nurses, and the average number of contacts with GP was 0.6 (SD 1.7) and 2.7 (12.7) for community nurses. Contacts with other health care professionals in the community were less frequent. Forty-five people (30.8%) had a paid carer for an average of 8.0 (SD 48.0) hours per week. Only 3 people in this cohort spent a day in a hospice, and 22 (15.1%) and 10 (6.8 %) had at least 1 contact with a palliative care physician or nurse. Ninety-two people (63.0%) received some sort of informal care related to personal care, support with medical procedures, appointments, or household tasks, and 125 people (85.6%) had a relative spending time with them ( Table 2).
People with a transition to hospital had a higher average number of ED visits, ambulance calls, outpatient appointments, contacts with community nurses, GPs and palliative care specialists. A higher proportion of people with a transition to hospital had a paid carer (35.3% vs 24.6%), but they had a lower average number of hours per week of paid care (3.5 vs 14.3) than people without a transition to hospital. Both the proportion of people with informal care and number of hours per week of informal care received was higher among people with a transition to hospital than people without transition ( Table 2).

Costs of Care
The average total cost of care in the last 3 months of life was £31,224.7 (SD 23,536.6). Unadjusted total costs were higher for people with a transition to hospital than without transition (£39,605.5 vs £19,546.6) ( Table 2). In people with a transition to hospital, 39.2% of the total cost was explained by informal care costs, followed by hospital costs (29.6%) and care home costs (22.1%). In people without transition to hospital, most of the total cost was explained by care home costs (45.6%), followed by informal care costs (37.1%) ( Figure 1).
After adjusting for confounders, total costs were significantly higher for people with a transition to hospital (£33,239.2, 95% CI 28,301.8-39,037.8) than without transition (£21,522.0, 95% CI 17,784.0-26,045.8). People whose usual place of care was a care home had lower total costs (£23,801.3, 95% CI 20,172.0-28,083.6) compared with people whose usual place of care was home (£34,331.4, 95% CI 27,824.7-42,359.5). This was mainly explained by lower informal care costs for care home residents (Table 3 and Supplementary Table 3).
Transition to hospital was associated with higher hospital costs but not with the other cost components. The analysis by subgroup showed that people with a transition to hospital had higher total costs for people whose usual place of care was home, but not for those in care homes. Informal care costs were higher for people whose usual place of care was home than care home but did not differ according to whether patients had a hospital transition (Figure 2 and Supplementary Table 4).
The sensitivity analysis with complete-case analysis (without imputation) shows similar results (Supplementary Table 5).

Discussion
In a sample of people who died with dementia in England, we described total costs of care in the last 3 months of life from a societal perspective, including formal and informal care costs, and explored the effect of transition to hospital on total costs and cost components. We found that the estimated average total cost per person was £31,224.7 and informal care costs represent 35% of the total costs of care. After adjusting for confounders, people with a transition to hospital had higher costs than people without transition, mainly explained by hospital costs. People whose usual place of care was care homes had lower total costs compared to home, mainly explained by lower informal care costs.
Informal care costs formed a major component of total costs in our study (35% of all costs). A similar pattern is observed in the USA, with    informal care costs representing approximately 30% of the total costs of dementia care in the last 5 years of life. 29 In England, an audit of 9 people who had died of dementia reported an average cost of formal care in the last 6 months before death of £25,000 per patient, with high estimated informal care costs for people living at home. 30 We found that people with a transition to hospital in the last 3 months have higher total costs than people without transitions. This was mainly explained by hospital costs. In fact, people with a transition to hospital had similar community care, care home, and informal care costs than people without transition after adjusting for confounders. This suggests reducing admissions to hospital might lead to lower total costs and has important implications for policy. Further research exploring the cost-effectiveness of interventions aiming at reducing end-of-life transitions to hospital in this population is needed.
We found that people whose usual place of care was care homes had significantly lower total costs than people at home, even considering the cost of the care home. This is explained by people living in care homes being less likely to transition to hospital and also having lower levels of informal care. Our findings support evidence suggesting care home expenses act as substitute for hospital and informal care costs in older adults. 31e33 There is some evidence suggesting that people dying in care homes might have better access to health care support, 34 advance care planning, 35,36 and symptom control 37 than people living at home, which might contribute to reducing transitions to hospital in this population. There is evidence suggesting the number of older people living at home needing informal care will rise by 63% in the next decade. 38 It is therefore critical to improve the level of support for people living at home with dementia and their informal carers. Interventions such as multidisciplinary home care, 39 communitybased integrated palliative care services, 40 and advance care planning 31,41 have been associated with lower risk of transitions of care and lower costs in people dying with dementia.
This study only considered costs in the last 3 months of life, and the number of days spent in care homes was low. Care home costs might increase significantly and exceed hospital costs if a longer end-of-life period is considered 42 and, therefore, future research with a longer time frame might be important. However, when exploring the effect of transitions to hospital, the last 3 months of life is particularly relevant as hospital admissions and ED visits increase exponentially during this period. 43,44 In this cohort, people without a transition to hospital had similar informal care costs to people with transition. This is likely to be explained because care provided by family does not stop when  someone is admitted to hospital. Hospital admissions are burdensome not only for the patient but also for their carers who still need to be involved. 45 Admissions to hospital do not provide "respite" for carers, and traveling and adaptations to their routine can be stressful. 46

Strengths and Limitations
This study has some limitations. We used information from 3 mortality follow-back surveys that had different recruitment strategies. However, we adjusted by study in the analysis. Surveys were completed 3-10 months after death by proxies, who were generally the next of kin. This could lead to recall bias, under-or overreporting, and missing information due to memory issues. However, there is evidence suggesting the Client Service Receipt Inventory questionnaire has a good level of agreement with administrative sources in terms of health care utilization 47 and that proxies close to the person can reliably report on the use of services and observable symptoms. 48 Our small sample size (146 people) might have led to a low power to detect differences in sociodemographic variables related to total costs. However, the study had enough power to detect a statistically significant difference in total costs for our main explanatory variables. Our method might overestimate hospital costs, as we only had information on the number of days in hospital but not the length of stay per admission. In addition, 93.8% of the sample was White and 75.3% referred to living comfortably; hence, more research is needed to understand if these results apply to more ethnically diverse and socioeconomically deprived populations.
Nevertheless, this study has important strengths, such as reporting informal care as well as formal care costs, which has been highlighted in the literature as an important component of total costs among people with dementia at the end of life.

Conclusions and Implications
Total and informal care costs are high among people with dementia at the end of life. Our results highlight the importance of informal care costs in this population. Considering informal care costs in policy evaluations is key, as the pattern of informal care costs differs depending on the place of care. Preventing transitions to hospital in people with dementia might reduce overall costs from the health care perspective, but not from patients' and families' perspectives. Transitions to hospital have an important impact on total costs, mainly explained by higher hospital costs. Access to care homes could help reduce transitions to hospital transitions, as well as reduce formal and informal care costs.