Socio‐economic predictors of time to care home admission in people living with dementia in Wales: A routine data linkage study

Abstract Objectives Limited research has shown that people with dementia (PwD) from lower socio‐economic backgrounds can face difficulties in accessing the right care at the right time. This study examined whether socio‐economic status (SES) and rural versus urban living location are associated with the time between diagnosis and care home admission in PwD living in Wales, UK. Methods/Design This study linked routine health data and an e‐cohort of PwD who have been admitted into a care home between 2000 and 2018 living in Wales. Survival analysis explored the effects of SES, living location, living situation, and frailty on the time between diagnosis and care home admission. Results In 34,514 PwD, the average time between diagnosis and care home admission was 1.5 (±1.4) years. Cox regression analysis showed that increased age, living alone, frailty, and living in less disadvantaged neighbourhoods were associated with faster rate to care home admission. Living in rural regions predicted a slower rate until care home admission. Conclusions This is one of the first studies to show a link between socio‐economic factors on time to care home admission in dementia. Future research needs to address variations in care needs between PwD from different socio‐economic and geographical backgrounds.

people with dementia (PwD) are encouraged to stay in their own home for as long as possible. However, at some point, the care needs of a person cannot be supported sufficiently within the community any longer, for example, due to increased support needs with daily activities, 5 so that entering a care home is the best solution to care for PwD appropriately. Other people might feel isolated at home however by living alone, and the family carer might live far away, so that a faster care home entry might be more desirable and suitable to their lifestyle. However, decisions to enter a care home depend on a variety of circumstances, including unpaid carers' and PwD's wishes as well as which country people reside in. 6 Several studies have explored the predictors of accessing a care home in dementia, and found that difficulties with daily activities, behavioural problems, cognitive deficits, depression, reduced carer quality of life, and carer burden all predict care home admission, whereas problems with daily activities was often the most significant predictor. [7][8][9][10][11] Considering the large costs associated with staying in a care home, 3 accessing a care home might be difficult for people from more disadvantaged backgrounds. For Wales specifically, people with savings or financial assets worth £50k or more have to fully fund their care home stay, 12 which can be a barrier. Across the 22 local authorities in Wales, care home provision varies with many providers owning a single care home, whilst other providers own multiple care homes and others are run by local authorities. 13 Limited previous research indicates that people from more disadvantaged backgrounds experience health inequalities in accessing dementia care, 14,15 with a recent data linkage study reporting that PwD in general are more likely to live in a deprived area 16 and developing dementia was higher for those people living in the most deprived areas. 17 Van de Vorst and colleagues 15 for example reported socioeconomic disparities in mortality after a dementia diagnosis, with people from a low socio-economic status (SES) having higher risks of mortality. Moreover, Cooper and colleagues 14 showed that PwD from more affluent backgrounds were 25% more likely to access anti-dementia medication than those from more disadvantaged backgrounds. Specifically, to date, little research has looked into socio-economic factors and individual background characteristics that might predict care home admission in dementia. It appears that only one study 10 has found that being from a White ethnic background (and living alone) predicted increased likelihood of care home admission for people with Alzheimer's disease (AD). In their analysis of 3000þ PwD, Knapp et al. 10 only included people with AD however and those living in one urban area, therefore limiting the generalisability of the findings. Thus, their findings provide no insight into the time to care home admission, but instead the general likelihood of admission. Yaffe and colleagues 18 (2002) explored care home placement in PwD utilising medicare data from the United States, reporting placement likelihoods of 22%, 40%, and 52% in Year 1, 2, and 3 since diagnosis. Whilst Yaffe et al. 18 (2002) also explored the factors contributing to placement, no focus was placed on socioeconomic background. Therefore, there is a gap in the evidence base on how SES and rural living location contribute to care home placement in dementia over a substantial period of time, both globally but also specifically in Wales. Moreover, this leaves out many other important factors of someone's socio-economic background, such as education, income, and the level of deprivation of the neighbourhood they live in, clearly highlighting an important gap in the evidence base. The latter can be measured by the Index of Multiple Deprivation (IMD), and has been shown in other studies to be linked to differences in healthcare utilisation. 19,20 With limited evidence on socio-economic predictors of care home admission in dementia, the aim of this study was to use linked routine electronic health record (EHR) data to explore the effects of SES and other factors on the time between dementia diagnosis and care home admission in PwD living in Wales. We hypothesised that people from disadvantaged backgrounds and those living in rural areas access a care home later. Findings from this EHR data linkage study can have implications for addressing some of the priorities of the Dementia Roadmap 2025. 21 By understanding potential health inequalities in accessing care homes timely, we can develop possible solutions to address these barriers with findings informing policy guidance on enabling PwD from any background to access care homes, and thus the right care, more timely.

| Study design
We used longitudinal anonymised EHR and administrative data from the Secure Anonymised Information Linkage (SAIL) Databank (9)(10)(11) to conduct a retrospective cohort study.

| Data sets
Our cohort was created using data held within the SAIL Databank.

| Participants and sample selection
Data from people with any diagnosis of dementia having been admitted into a care home in Wales were included. Figure

| Variables
Demographic characteristics of age and gender were obtained from the WDSD. Date of diagnosis was measured as the first clinical record of dementia in the SDEC data set, derived from the SAIL primary and secondary care data, from which we also collected data on the dementia subtype (AD dementia, vascular dementia, frontotemporal dementias, and Lewy body dementia). Each person could have more than one subtype diagnosis, so that no one diagnosis is mutually exclusive. Data on mortality was also obtained from the SDEC, derived from the SAIL ADDE mortality data.
Date of care home admission was generated by combining the WDSD with the CARE data set to create an individual-level admission date. Living situation at time of diagnosis and living situation at time of care home admission was recorded as 'living alone' if the PwD had lived alone in the address preceding the diagnosis/care home admission.
Frailty was measured using the eFI, which inquires 36 deficits to identify older adults with no (fit), mild, moderate, or severe frailty at the time of care home admission, with data obtained from the WLGP.

Comorbidities were measured using the Charlson Comorbidity
Index, 25 which collects information on 22 comorbidities at the time of care home admission. A score of 'À 1' indicates diabetes with longterm effects, '0' indicates no chronic conditions, and each comorbidity receives a score that is added up. For the purpose of this analysis, we have categorised the Charlson Index into 'À 1' and '0', 1-10, and >10.

Index of Multiple Deprivation (WIMD) quintiles version WIMD 2011
of the last residence prior to CH admission, with '1' indicating the least disadvantaged neighbourhoods, and '5' the most disadvantaged neighbourhoods. The WIMD measures income, employment, health, education, access to services, housing, community safety, and physical environment in declining levels of importance to produce overall rankings of neighbourhood deprivation, and has been utilised in previous explorations of how SES is linked to health outcomes. 26,27 Rural and urban location of the living situation a day prior to care home admission was derived from the Office of National Statistics rural urban classification and linked to the Lower layer Super Output Area (LSOA) of residence version LSOA 2001, with areas of 10k population size or more classed as 'urban'.

| Data analysis
Data were analysed using SPSS 25, with significance level set at p <  To explore whether IMD quintile and geographical living location were associated with the time to mortality after care home admission, ANOVA with Bonferroni correction and independent t-tests, respectively, were used, with time to mortality as outcome variable.

| Variations in predictors of care home admission by SES
Kaplan-Meier survival curves showed that PwD from the most disadvantaged background had a longer duration from dementia diagnosis to care home admission compared to those living in the most advantaged neighbourhoods. PwD from rural backgrounds were delayed in entering a care home compared to those living in urban environments (see Figure 4).  -515 dementia diagnosis can be a limitation in the present study, as it is well known that there are severe delays in getting a diagnosis in the first place. 30 This means that some people might receive a diagnosis whilst already being in the more advanced stages of the condition, whilst others might have received a diagnosis more quickly, which will inevitably have an implication on the time to care home placement. Unfortunately, no data were available on first symptom recognition, as this data linkage used routinely collected data.

| Association between SESs and living location on mortality rate
Further research needs to link up routine data with additional primary data collection on first symptom recognition, or link up with data on the severity of dementia. Nevertheless, the present findings not only contribute novel evidence to a growing research field, but can also have important implications for policy guidance on care homes and their potential (self-) funding. which equates to an estimated £13.9 billion each year in the UK.
However, once family carers experience high levels of burden, 35 their relative with dementia is often admitted into a care home, 36 suggesting that family members should be supported to cope and have a good quality of life to enable their relative to stay well in the community for longer. For those PwD who live alone, better access to post-diagnostic community support services needs to be put in place so that PwD can access the support they need.

| Limitations
This study was based on routinely collected EHR data linked with a specifically designed care home data. The care home data (CARE) was created using data of current care homes in the Care Inspectorate Wales database in 2018. By using routine data, no data are included on severity of the dementia. In addition, the date of diagnosis is based on the first clinical record of dementia. However, people may have been diagnosed before this date, and people may have been delayed in going to their doctor to get a diagnosis, possibly due to their SES. The sensitivity of using UK routinely collected primary care, hospital admissions and mortality data in combination to identify PwD is not known. 37,38 It is likely that a proportion of 'true' dementia cases would have been misclassified as non-cases. Under-recording of dementia in EHR data may itself be related to SES. Data linkage also resulted in 20,000þ missing cases on dementia diagnosis in this study, which however did not affect the power of the sample, as we were still able to include 34,514 PwD. Considering the analysis, we acknowledge the simplicity of the Kaplan-Meier analysis and the limitation of only including a single explanatory variable, it was our aim to give a representation of the differences between levels of SES over time. The Cox regression models extended this analysis to incorporate additional variables. Lastly, we used the WIMD as a neighbourhood deprivation index, which does not provide individual-level data on for example income and education. The mechanisms between neighbourhood and individual level of deprivation might vary however, which needs to be considered as a limitation. It is important to highlight though that many studies employ a deprivation index as opposed to individual-level data, so that our study is not an exception.

| CONCLUSIONS AND IMPLICATIONS
This is the first study based on population scale (Wales) linked routine HER data to show how SES and geographical living location are associated with time to care home admission in PwD. Future research needs to explore the underlying reasons for these relationships, and variations in care needs at care home admission.
These findings clearly address the Dementia Roadmap 2025, 21 and provide novel insights of existing health inequalities in dementia care, by addressing one of the five essential conditions for more equal health for all as outlined in the recently published WHO European Health Equity Status report. 39