The relative impact of socioeconomic position and frailty varies by population setting

Abstract Introduction Frailty and socioeconomic position (SEP) are well‐established determinants of health. However, we know less about the contributions of frailty and SEP in older adults, especially in acute settings. We set out to answer how frailty and SEP might influence health outcomes in older people, comparing a population sample and patients managed by a speciality acute frailty service. Methods We used the Delirium and Population Health Informatics Cohort, a population sample of 1510 individuals aged ≥70 years from the London Borough of Camden and 1750 acute frailty patients. SEP was determined using the Index of Multiple Deprivation. Linear and Cox proportional hazard regression models were conducted to assess SEP on frailty, readmission, and mortality outcomes. Results In the population sample, SEP was significantly associated with frailty and mortality with successive increases in rate of death for each IMD quintile (HR = 1.28, 95% CI 1.11 to 1.49, P < 0.005). Increasing SEP, age, and admission status among hospitalized individuals were associated with greater frailty. For individuals seen by the speciality frailty service, SEP was not associated with frailty, mortality, or readmission. Discussion When older people experience acute illness severe enough to require secondary care, particularly specialist services, this overcomes any prior advantages conferred by a higher SEP.

socioeconomic position and frailty. 10 In the Survey of Health, Ageing and Retirement in Europe, a participant with the lowest levels of education, occupation, income, and wealth was as frail as a participant 7 years older with the highest levels of these measures. 11 In older adults, lower socioeconomic position is associated with more frequent episodes of acute health problems leading to deterioration and instability in baseline frailty and increased mortality. 12 However, the extent to which these relationships hold true across different settings within health and social care systems is less clear. For example, how much does SEP continue to affect clinical outcomes once an older person is admitted to hospital? Quantifying these effects might have implications for assessing older people with acute illness at the individual level, as well as the service design at the population level.
To address the question of how frailty and SEP might influence outcomes in older people, we used overlapping prospective clinical and population data. We hypothesized that lower socioeconomic position might be associated with more frailty and mortality. We addressed these by focusing on four specific questions ( Figure 1): What is the relationship between SEP and frailty in: 1. a population sample? 2. those acutely admitted to hospital? 3. those seen by a specialist frailty service? 4. What is the relationship between SEP and mortality in a population, compared with a specialist frailty service? While there are different priorities for different settingspublic health approach for addressing health inequality, and a direct clinical one in the acute context-our aim in these analyses was to describe the points at which these two factors transition within a defined geographic health service.

| Data sources Population sample
DELPHIC is an ongoing population-representative study, following individuals aged ≥70 years from the London Borough of Camden.
Its primary purpose is to undertake longitudinal assessments across community and acute hospital settings, focusing on measures often unreliably coded in electronic health records (e.g. cognition, physical function, frailty). University College Hospital, London, UK, is one of two acute hospitals in Camden (see below). Full details of the study have previously been reported. 13,14 In brief, the sample was mainly enrolled from primary care lists and is representative of the borough in terms of age distribution and income deprivation indices. We assessed participants through telephone interviews and on each hospital admission. In addition, researchers had access to all health and social care data to corroborate clinical information. The data pre-

| Outcome measures
The outcomes of interest were current frailty (Questions 1, 2, and 3), and subsequent readmission and mortality (Question 4).
In the population sample, we derived a 35-item Frailty Index representing the proportion of accumulated health deficits (0 to 1) in the population sample. Items used included: self-rated health, comorbidities, sensory difficulties (including vision and hearing), incontinence, falls, mobility, personal and instrumental activities of daily living, polypharmacy, cognitive function, and quality of life. The Frailty Index was calculated using standard procedures. 16 In the acute frailty service cohort, frailty was measured using the Clinical Frailty Scale (CFS), referring to a period of health 2 weeks before acute presentation. The CFS is a 9-point score where 1 represents robust and active people and 9 are those approaching the end of life.
For Question 4, we considered any hospitalization (population sample) or readmissions (acute frailty service). We determined mortality through notifications to the NHS Spine, a statutory register of

F I G U R E 1 Schema showing inter-relationship of datasets.
Questions: What is the relationship between SEP and frailty in: 1. a population sample? 2. those acutely admitted to hospital? 3. those seen by a specialist frailty service? 4. What is the relationship between SEP and mortality in a population, compared with specialist frailty service? all deaths in England. We considered all-cause mortality from the date of admission to hospital to December 2018 (44 months).

| Socioeconomic position and frailty
Question 1 and 2: The frailty index was the primary outcome, with educational attainment (three levels), occupational class (four levels), IMD or income deprivation affecting older people index (centile rank), and hospitalization status (yes/no) estimated in linear regression models adjusted for age and sex.
Question 3: We used linear regression to estimate the relationship between CFS score (as a continuous measure) and IMD decile, adjusted by age (in years), sex, and presence of dementia.

| Population sample
The average age was 78 (SD 16.7) years and 57% were women.
Frailer participants were older and more likely to have been admitted to hospital at least once. Educational attainment was high in this cohort, with 1092 (72%) having at least a bachelor's degree. This group showed the lowest levels of frailty (lowest education n = 521, highest education n = 220, P < 0.01). Occupational class followed a similar distribution (Table 1).   Table S1). The inverse association between educational attainment and frailty persisted, but this was not the case for occupational class (Figure 1). We did not demonstrate any interaction between IDO and admission status (P > 0.05); the degree of the relationship was the same regardless of all-cause hospitalization ( Figure 3).

F I G U R E 2
The relationship between socioeconomic position and frailty in a population sample (Question 1)

F I G U R E 3
The relationship between SEP and frailty in a population sample, stratified by admission status (Question 2)

| Acute frailty service
Among 2259 admissions in 1750 individuals, mean age was 85 (SD 6.8) years and 57% were women ( Table 2). The lowest two IMD quintiles accounted for 63% of admissions; <1% were in the most advantaged quintile.  (Figures 3 and 4).

Question 4:
In the population cohort, SEP was associated with mortality, with successive increases in rate of death for each IMD quintile (HR = 1.28, 95% CI 1.11 to 1.49, P < 0.005), even after adjustment for age, sex, frailty, delirium, and falls history (Table 3).

| DISCUSS ION
We showed different relationships between SEP and frailty in overlapping population and specialist clinical samples. In a population cohort, there were clear associations between SEP and frailty. Frailty scores were higher in those hospitalized, but the gradient of the SEPfrailty relationship was consistent. Though patients with high SEP were much less likely to present to the Acute Frailty Service, those subsequently being managed by this service did not vary by degree of SEP. Taken together, our findings suggest that by the time patients were selected into a specialist service, SEP was no longer a driving part of ill health, and frailty-specific factors may predominate.
The relationship between socioeconomic position and frailty presenting to an acute frailty service (Question 3) A scoping review investigating frailty in the acute setting demonstrated associations with mortality, increased length of stay, and institutionalization post-discharge. 18 Most other studies of SEP are in community-based populations [19][20][21] ; socioeconomic position is rarely assessed in acutely hospitalized patients, and even less so in older adults. For example, while an association between social deprivation and in-hospital mortality has been reported in critical care, this did not extend to complications (acute kidney injury or ICU admission), length of stay or readmissions. 22 Increased social vulnerability is predictive of long-term care outcomes in hospitalized older adults, but only for the oldest-old with lower levels of frailty. 23 What accounts for the lack of association between IMD and frailty in the specialist cohort? It would seem that the social determinants of health operate at each level (population, general admission) until patients are selected into a specialist frailty service. Once managed in this setting, SEP no longer appears to drive mortality or readmission. This may be specific to a publicly funded health system able to counter the socioeconomic position health gradient, and international comparisons would be needed to confirm this. On the other hand, it is possible that the specialist cohort are so frail that any proximal effects from SEP no longer come into play. To examine this, future studies may look at conducting analysis in a group of patients where SEP is less accounted for in care plans, for example in a surgical setting.
Our study demonstrates that illness severe enough to require secondary care, particularly specialist services, overcomes prior advantages conferred by a higher SEP. Though policies that aim to reduce socioeconomic inequalities may be of benefit at the population level, acute frailty services in health systems comparable to the UK are likely to provide the same benefit to individuals across the spectrum of socioeconomic advantage.

F I G U R E 5
The relationship between socioeconomic position and mortality (Question 4)