Frailty and socioeconomic position: A systematic review of observational studies

Background: Frailty, an age-related state of reduced physiological reserve, is often associated with lower socio-economic position (SEP). This systematic review synthesised observational studies assessing (i) the association between SEP and frailty prevalence; (ii) how changes in frailty status over time vary by SEP; and (iii) whether the association between frailty and clinical outcomes is modified by SEP. Methods: We searched three electronic databases from 2001 to 2023. We included observational studies measuring early, mid-, and late-life indicators of SEP (education, income, wealth, housing, occupation, and area-based measures of multiple deprivation) and frailty (assessed using any validated measure). Screening and extraction were performed in duplicate. Findings were synthesised using narrative synthesis. Results: We included 383 studies reporting findings from 265 independent samples/cohorts across 64 countries. Lower SEP was associated with higher frailty prevalence across all indicators (childhood deprivation 7/8 studies, education 227/248, occupation 28/32, housing 8/9, income 98/


Background
Frailty is a major challenge facing healthcare systems worldwide, particularly in the context of ageing population demographics (United Nations, 2022;Victora et al., 2018).Frailty is often observed to be more prevalent in contexts of socioeconomic deprivation (Hale et al., 2019).Responses to these challenges are constrained by the inequitable distribution of healthcare resource, characterised by the pervasive inverse care law; the availability of good medical care tends to vary inversely with the need for it in the population served (Hart, 1971).At the interface of ageing populations and rising inequalities (World Health Organization, 2023), it is important to understand how the development of frailty relates to socioeconomic position, and to consider implications for healthcare systems and the distribution of healthcare resources.
Frailty is a state of increased vulnerability to physiological decompensation in response to a stressor event (Clegg et al., 2013).There are a large number of diverse measures used to assess frailty (Box 1) (Dent et al., 2016), however each shares some core features (Hoogendijk et al., 2019).First, frailty is associated with age.However, frailty is not universally found even at the extremes of age and may also manifest in relatively younger people, particularly in the context of socioeconomic deprivation (Hanlon et al., 2018).Second, frailty is dynamic: the degree of frailty within an individual changes over time (Welstead et al., 2021).Third, people living with frailty are at greater risk of a range of adverse health outcomes, including falls, cardiovascular events, functional decline, loss of independence, hospital admission and mortality (Hoogendijk et al., 2019).
Socioeconomic position refers to "the socially derived economic factors that influence what positions individuals or groups hold within the multiple-stratified structure of a society" (Galobardes et al., 2007).Socioeconomic position has been linked to a wide range of adverse health outcomes and to healthy ageing (Kivimäki et al., 2020;Wagg et al., 2021).Socioeconomic position may be quantified, in an epidemiological context, using a number of different individual-level indicators.These include, but are not limited to, childhood circumstances including parental occupation or household income, education, occupation-based indicators, income, wealth, and housing.The latter three may be operationalised to reflect socioeconomic position at different stages of life (e.g.working-age income versus retirement income, or housing status reflecting early to mid-life circumstances, or housing conditions in older age/retirement).Area-based measures of socioeconomic position are also used, either as proxies of individual-level socioeconomic position or as indicators of the environment in which a person lives.Each of these indicators measures an aspect of socioeconomic position, and may be more or less relevant to a given health outcome.
Various studies have explored associations between socioeconomic position and frailty, employing a range of different analytical approaches, using diverse frailty measures and assessing different socioeconomic indicators at play in different periods of the lifecourse (Wang and Hulme, 2021).These include early-life socioeconomic indicators such as education or parents' occupation, midlife factors such as occupation, as well as factors such as income, wealth and housing in older age.Therefore, this systematic review seeks to provide a comprehensive overview of the relationship between socioeconomic position throughout the life-course and frailty.We aim to assess (i) the cross-sectional association between socioeconomic position and the prevalence of frailty, (ii) how changes in frailty status over time vary by socioeconomic position, and (iii) whether the relationship between frailty and adverse health outcomes varies by socioeconomic characteristics.

Methods
This systematic review was conducted according to a pre-specified protocol (PROSPERO registration CRD42023425882) and is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines (Page et al., 2021).

Search strategy and eligibility
We searched three electronic databases (Medline, Embase and Web of Science Core Collection) from 2001 to April 2023 (as the original descriptions of the frailty phenotype and frailty index, the two most commonly used measures of frailty, were published in 2001) without language restriction.The search combined frailty and socioeconomic terms, using a combination of Medical Subject Headings and keyword searches.Full search terms are shown in the Supplementary appendix.
• Exposure: Studies had to assess at least one measure of socioeconomic position (childhood socioeconomic position, education, working-age occupation, income, wealth, housing status, area-based measures, or composite measures combining two or more indicators).• Comparison: Studies had to compare participants at different levels of a given socioeconomic indicator.• Outcomes: Studies had to assess either (i) frailty prevalence, (ii) changes in frailty status over time (including incidence of frailty, transitions between frailty states, or trajectories of frailty extent or severity), or (iii) relationships between frailty and adverse health outcomes (where socioeconomic position was assessed as an effect modifier).
• Setting: We only included studies in community settings (defined as any setting outside of a hospital, and including nursing homes and other supported accommodation).• Study design: Observational studies including cross-sectional analysis or cohort studies.
Studies using any measure of frailty were eligible for inclusion, providing the measure was fully described or referenced.We excluded single-measure proxies of frailty status (such as grip-strength or walking speed alone).We excluded conference abstracts and grey literature.
We anticipated that most studies of frailty would include only older people (e.g.> 65 years), however we included wider age eligibility criteria as frailty has also been described in younger populations and socioeconomic position has been proposed as one driver of youngeronset of frailty (Brunner et al., 2018).We therefore included studies assessing adult populations of any age range to allow exploration of differences in the relationship between socioeconomic position and frailty between age strata (within a given study) or between studies with different age profiles.
Two independent reviewers screened all titles and abstracts identified from database searches, and subsequently screened full texts of all potentially eligible studies.This process was supplemented by hand searching reference lists of all eligible studies and relevant systematic reviews, as well as forward citation searching of all eligible studies using Web of Science.

Data extraction
We extracted data from all included studies using a piloted data extraction template.One reviewer extracted data, which was then checked for accuracy by a second, independent reviewer.Data extracted included bibliographic details, country, study cohort, setting, sample size, age-criteria, age and sex demographics, details of frailty measure and measure(s) of socioeconomic position, numbers of participants at each level of frailty and socioeconomic position, analytical strategy, summary statistics of the association between frailty and socioeconomic position, and details of covariate adjustment.

Quality assessment
We assessed the quality of the included studies using the Joanna Briggs Institute risk of bias assessment checklists appropriate to the study design of the included studies (cross sectional or cohort studies) (Joanna Briggs Institute, 2023).

Data synthesis
We performed a narrative synthesis of the review findings.Summary data of each included study are presented in tables.We summarised findings related to frailty prevalence and changes in frailty over time using Harvest Plots (which allow synthesis of data across diverse exposures and outcomes, and allows synthesis of heterogeneous data when meta-analysis is not appropriate).Harvest plots are a matrix on which each included study is represented by a bar.The position of the bar on the matrix indicates the overall finding of the study relating to the outcome in question (e.g. the association between socioeconomic position and frailty prevalence).We used height to indicate the study size, and colour to denote the measure used for frailty or socioeconomic position.We used 95 % confidence intervals or a p-value of < 0.05 (whichever was reported in the included studies) to assess whether studies showed a significant association (positive or negative) or a neutral association between lower socioeconomic position and frailty.Where papers reported more than one statistical model, we used results from the fully adjusted models in the Harvest plots.Where no statistical test was reported but the study provided sufficient data (e.g.mean and standard deviation of a frailty index at each level of socioeconomic position) we calculated the p-value.Where two or more studies used data from the same cohort or sample, and used similar measures of frailty and socioeconomic status, we assessed for consistency of the findings and presented all relevant studies as a single bar on the study (using the largest reported sample size).

Results
The results of the search and screening are shown in Fig. 1.Overall, 383 articles were eligible for inclusion, comprising data from 265 separate cohorts or samples.Studies were from 64 different countries (38 high income, 19 upper-middle income, 3 lower-middle income and 4 low-income countries), as shown in Fig. 2. Aggregate data summarising the included studies is shown in Supplementary Appendix page 3. Frailty was measured using a range of measures including the frailty phenotype (214 studies), frailty index (95 studies), FRAIL scale (23 studies), Tilburg frailty indicator (18 studies), Edmonton frailty indicator (14 studies) and Groningen frailty indicator (4 studies).Early-life socioeconomic indicators included childhood socioeconomic status (assessed by either a composite measure or by parental occupation: 8 studies) and education (313 studies); midlife socioeconomic status was assessed using workingage occupation (51 studies); and current socioeconomic status among older people was assessed using income (personal or household: 145 studies), wealth (66 studies), housing status (10 studies) or composite area-based measures (44 studies).
Quality assessment showed that while eligibility criteria and frailty qualification were reported to a high standard, assessment of socioeconomic indicators was more variably described and quantified.This included limited detail on how these data were collected, and variable approaches to defining constructs such as education (e.g. years of education or completion of various types), income (household versus individual, specifically quantified versus dichotomisation as 'high' or 'low').This latter variability, along with the sensitivity of frailty prevalence estimates to the specification of different measures, led to the decision to perform narrative synthesis and not a meta-analysis.Identification of confounders was often lacking in cross-sectional studies, with 171 not adjusting for covariates such as age and sex.Detail on quality assessment is shown in the Supplementary appendix.

SEP and frailty prevalence
Two-hundred and eighty-six studies assessed the cross-sectional relationship between socioeconomic position and frailty.Of these, 256 assessed early-life factors (248 education, 8 childhood socioeconomic position, typically using parent's occupation as a proxy), 32 assessed working-age occupation as a mid-life determinant, while the others assessed socioeconomic position at the time of data collection, typically later-life (108 income, 44 wealth, 34 composite area-based measures and 9 housing status).Of the cross-sectional analyses, 276 studies (using data from 227 different datasets) assessed differences in crude (unadjusted) frailty prevalence between participants of different socioeconomic position and are summarised in Fig. 3.For each indicator, lower socioeconomic position was associated with higher frailty prevalence in most studies.This pattern was consistent across geographical region.
One-hundred and fifteen studies quantified the association between socioeconomic position and frailty prevalence (or degree of frailty) after adjustment for additional covariates.All included studies adjusted for age and sex, with most also including other covariates including marital status, behavioural risk factors (such as smoking, alcohol consumption or physical activity), and long-term conditions (although the covariates included in adjusted models were highly variable between studies, see Supplementary appendix).Fig. 4 summarises their findings, and demonstrates that lower educational status, lower occupational class, lower income, lower wealth, poorer housing status, and lower area-based socioeconomic position were each associated with higher prevalence of frailty.Lower childhood socioeconomic position was associated with frailty in 3/5 studies.

Longitudinal changes in frailty
Seventy-five studies (using 49 different datasets, mostly from higher or upper-middle income settings) assessed changes in frailty status using a variety of methodological approaches, frailty measures, and socioeconomic indicators.Sample size ranged from 150 to 94,550 and followup ranged from 6 months to 17 years.Their findings are summarised in Fig. 5 and presented in detail below.

Changes in the frailty phenotype
Twenty-four studies assessed incidence of frailty in people who were robust at baseline.Lower educational status (12/14 studies), lower income (6/6 studies), lower occupational class (3/3 studies) and higher area-based socioeconomic deprivation (1/1 study) were all associated with a higher incidence of frailty in people who were robust or pre-frail at baseline.
Twenty studies assessed the relationship between a range of factors and the likelihood of transitions between frailty states (robust, pre-   showed a positive association with frailty (higher frailty prevalence associated with lower socioeconomic position at a threshold of p < 0.05), neutral association (no statistically significant difference in prevalence), or negative association with frailty (higher frailty prevalence associated with lower socioeconomic position at a threshold of p < 0.05).Y-axis is truncated at 100,000 participants, three studies had > 100,000 participants (493,737, 968,885, and 1,470,000, respectively).Abbreviations: SEP = socioeconomic position.FRAIL = Fatigue, Resistance, Ambulation, Illness and weight Loss scale.Tilburg = Tilburg Frailty Indicator, Edmonton = Edmonton Frailty Indicator, Groningen = Groningen Frailty Indicator.
frailty and frailty).These found that lower educational status (11/13 studies), lower occupational status (2/3 studies), lower income (7/9 studies) and lower wealth (3/3 studies) were associated with greater likelihood of transitioning from a robust or pre-frail state to a frail state.
One study showed that area-based socioeconomic deprivation was associated with frailty worsening, with two others showing no significant difference.Lower education (4/5 studies), lower income (2/2 studies) and lower wealth (1/2 studies, the other showing the inverse relationship) were associated with lower likelihood of improvement from frailty to a pre-frail or robust state, however two studies showed no association between area-based socioeconomic position and frailty improvement.The colour shows the measure of frailty used in the analysis.The position of the bar on the matrix indicates whether the socioeconomic indicator showed a positive association with frailty (higher frailty prevalence associated with lower socioeconomic position at a threshold of p < 0.05), neutral association (no statistically significant difference in prevalence), or negative association with frailty (higher frailty prevalence associated with lower socioeconomic position at a threshold of p < 0.05).Y-axis is truncated at 100,000 participants, one studies had > 100,000 participants (493,737).Abbreviations: SEP = socioeconomic position.FRAIL = Fatigue, Resistance, Ambulation, Illness and weight Loss scale.Tilburg = Tilburg Frailty Indicator, Edmonton = Edmonton Frailty Indicator, Groningen = Groningen Frailty Indicator.

Changes in the frailty index
Studies using the frailty index modelled change in frailty in different ways.Ten studies used either growth curves or generalised estimating equations to assess the relationship between socioeconomic characteristics and the rate of accumulation of age-related deficits.All showed that lower socioeconomic position was associated with a higher intercept (indicating that the starting frailty index at a given age was higher in people with lower socioeconomic position); however, analyses of the rate of change in the frailty index showed variable results.Five analyses (using data from the English Longitudinal Study of Ageing, Osteoporotic Fractures in Men study and the Lothian Birth Cohort) showed that lower education, lower occupational class, and lower area-based socioeconomic position were associated with more rapid accumulation of deficits.Conversely, five studies (using Survey of Health, Ageing and Retirement in Europe, Longitudinal Aging Study Amsterdam and the Helsinki Birth Cohort) found that the rate of accumulation of deficits did not vary by education, income or wealth.In either case, regardless of whether the rate of increase differed significantly, intercepts were higher in people with lower socioeconomic position (indicating that people enter older age with socioeconomic inequalities in frailty) and the average trajectory of the most socioeconomically deprived participants at a given age was comparable to that of the most affluent participants who were approximately 10 years older.
Other studies assessed incident frailty among participants who were robust at baseline, with 4/5 analyses showing lower socioeconomic position was associated with the development of frailty.Finally, eight studies identified groups of participants with different trajectories of frailty, all of which demonstrated that lower socioeconomic position was associated with more rapidly progressive trajectories of the frailty index.

Interplay between lower SEP and frailty
We identified 6 studies exploring the relationship between frailty, socioeconomic position, and all-cause mortality, which analysed whether the association between frailty and mortality was modified by socioeconomic position.Five of these found no evidence of statistical interaction, suggesting that the relationship between frailty and mortality was similar across levels of socioeconomic position.A sixth study did report a significant interaction, suggesting that the relative association between frailty and mortality was greater for people with high socioeconomic position compared to lower.However, this finding was largely attenuated after additionally accounting for interactions with age.
Fig. 6.Summary of review findings.

Discussion
This systematic review, synthesising 383 observational studies across 64 countries, demonstrated that lower socioeconomic position, defined by a range of indicators, was associated with higher frailty prevalence.This finding was replicated across geographical areas, regardless of the frailty measure used, and after adjustment for age, sex, and additional covariates.The social gradient in frailty was most evident in relatively younger populations, and both early-and mid-life socioeconomic disadvantage was associated with earlier onset of frailty.The relationship between socioeconomic position and changes in frailty after its development were less consistent but did show a clear association between lower socioeconomic position and higher incidence of frailty as well as less-favourable frailty trajectories in most studies.Finally, the association between frailty and mortality was not modified by socioeconomic position.These findings are summarised in Fig. 6.These findings have implications for healthcare services seeking to support people living with frailty (addressing the inverse care law), for frailty prevention, and for future research to understand the development of frailty.
Higher prevalence and incidence of frailty associated with lower socioeconomic position indicate that healthcare providers serving areas of higher socioeconomic deprivation are likely to see higher levels of frailty, and encounter frailty developing at younger ages.People living with frailty require a comprehensive and individualised approach to healthcare, typically managing multiple long-term conditions, experiencing more frequent and severe episodes of illness, and often requiring rehabilitation or support with activities of daily living (Dent et al., 2019).This requires robust and accessible primary healthcare with availability of appropriate multidisciplinary services and geriatrician-led secondary healthcare to manage and support rehabilitation from episodes of acute illness and resultant decompensation.In this context, the inverse care lawwhereby people living with the highest need often experience poorer and more limited access to healthcarepresents a profound challenge to the management and support of people living with frailty.Primary healthcare coverage is limited in many settings (World Health Organization, 2023), and people living with high socioeconomic deprivation frequently experience poorer access (Mercer and Watt, 2007;Ellis et al., 2017).Number of clinicians per patient and time spend per consultation are also typically lower in more deprived areas, and are lower in lower-income countries (Mercer and Watt, 2007;McConnachie et al., 2023;Greg et al., 2017).In some systems, the financial cost of healthcare is prohibitive for many.As such, ensuring equitable access to appropriate healthcare for the growing proportion of older people is a challenge for healthcare systems and policymakers across the world (World Health Organization, 2023), as people living with frailty (and thus having some of the greatest health needs) disproportionately experience the barriers associated with socioeconomic deprivation.Therefore, to avoid widening existing inequalities, efforts to reduce or manage frailty should be based on proportionate universalism with greater efforts and resources required in the context of higher deprivation.
The social gradient of frailty reflects known inequalities in health that have been observed across multiple contexts (Cookson et al., 2021).The association between frailty and a wide range of socioeconomic indicators (reflecting early-life factors such as education, mid-life factors such as working-age occupation, and later-life factors such as income or wealth) also suggest that a life-course approach to frailty prevention is required to delay the onset of frailty and reduce its overall prevalence.Frailty is a "multiply determined" state, in that there are multiple potential causal pathways to the development of frailty, often reflecting accumulated dysfunction across multiple domains and systems (Howlett et al., 2021).The importance of early-life factors demonstrates that frailty prevention is a population-level challenge and needs to start before the onset of demonstrable indicators of frailty.This could span early life enrichment (comprising the benefits gained from stimulation across settings such as home, school and community) (Frank et al., 2023), preventing or reducing modifiable behavioural risk factors, and enabling protective factors such as social networks (Hanlon et al., 2024).This would require intervention at multiple levels, addressing both 'upstream' structural sources of inequality and 'downstream' individual risk factors.Attempts to address behavioural risk factors also need to consider the context in which these occur, social practices and cultural norms that perpetuate these behaviours (Blue et al., 2016), the resources and support available to those most affected, and the capacity of individuals to undertake the work involved in addressing risk factors, particularly in the context of socioeconomic deprivation (Hanlon et al., 2021).
Frailty is increasingly recognised to be a modifiable state, with potential to prevent or delay frailty onset and (in some circumstances) to reduce and reverse frailty at an individual level (Welstead et al., 2021;Sun et al., 2023).Interventions such as nutritional support and structured exercise have shown promise in delaying or reversing frailty (Sun et al., 2023).However, if such interventions are to be refined and scaled up, it will be necessary to address the potential barriers to their implementation in contexts of high socio-economic deprivation (such as high costs of nutritious food, 'food deserts', high energy costs for storage and cooking, or access to facilities supporting structured exercise) (Karpyn et al., 2019;Bai et al., 2021).Modifiable risk factors such as smoking, obesity and excess alcohol consumption are also linked to the development of frailty (van Assen et al., 2022;Niederstrasser et al., 2019).These risk factors are more common, and often co-exist, among people from lower socioeconomic position (Shankar et al., 2010).People in this context also face additional structural barriers to addressing these risk factors, which further exacerbates inequalities (Blue et al., 2016).Our findings that a high proportion of people living with frailty have lower socioeconomic position, particularly at relatively younger ages when frailty may be more amenable to intervention or prevention, demonstrate the scale and importance of this challenge.
Our findings suggest that the association between socioeconomic position and frailty may be of greater magnitude at relatively younger ages (e.g.under 65 years).This could be for several reasons.Frailty is rare in younger populations but, when it occurs, is strongly associated with multiple long-term health conditions and behavioural risk factors (Hanlon et al., 2018).These factors show a strong socioeconomic gradient and may lead to earlier accumulation of deficits or functional impairments in people living with higher socioeconomic deprivation.Second, individual level social vulnerability (such as loneliness or social isolation) is linked to lower socioeconomic position and also with earlier onset of frailty (Hanlon et al., 2024;Politis et al., 2024).Thirdly, socioeconomic position is strongly associated with mortality.This could lead to survivor bias when assessing the association between socioeconomic position and frailty, with people with lower socioeconomic position having a lower probability of living to older age (when frailty is more likely to develop), reducing the association between socioeconomic position and frailty at older ages.While this review cannot differentiate between these (and other) explanations, the findings suggest that frailty manifesting earlier in life is a phenomenon that is predominantly encountered in contexts of high socioeconomic deprivation.Interventions to prevent frailty development should ideally target a broad age range (including people under 65) and consider the specific challenges faced by people living with high socioeconomic deprivation.There is also a need for greater understanding of the clinical implications of frailty at younger ages, which may differ from frailty presenting in later life, and for evidence to support the management of frailty in services catering for populations experiencing high socioeconomic deprivation.
Strengths of this systematic review include a comprehensive search strategy without language restriction and robust screening and data extraction processes.The broad inclusion criteria also allowed synthesis of a wide range of literature including multiple measures of frailty and socioeconomic indicators.However, this also resulted in high heterogeneity which limited quantitative synthesis of the results.The large number of studies allowed a broad overview of the topic, however limited more detailed exploration of specific relationships.The review included studies from a wide range of geographical regions, however studies from high income countries were over-represented and there is a need for greater exploration of the impact of frailty and health inequalities in middle-and lower-income countries.There is also a risk of publication bias and given the high methodological heterogeneity of the included studies it was not possible to formally assess this.Finally, the studies included are observational in nature and the associations between socioeconomic position and frailty are prone to bias (e.g., by confounding, reverse causation, or selection bias).Our findings therefore reflect the social gradient observed in frailty, but are not sufficient to demonstrate causal relationships.However this does suggest, given the apparent links between socioeconomic disadvantage at various life stages and greater frailty in later life, that future research into the development of frailty requires a life-course approach.
Frailty disproportionately affects people with lower socioeconomic position, in whom it occurs earlier and is more likely to worsen or progress.This pattern is seen across socioeconomic indicators throughout the life-course.As frailty on a global scale continues to increase, there is a pressing need to address structural inequalities in both the determinants of health and in the availability and delivery of healthcare.

Fig. 2 .
Fig. 2. Location of included studies: This figure shows the countries from which participants were recruited in each of the included studies.In addition to those shown on the map, 11 studies used pooled data from the Survey of Health, Ageing and Retirement in Europe.

Fig. 3 .
Fig. 3. Unadjusted cross-sectional association between frailty and socioeconomic indicators: This Harvest plot shows the findings for unadjusted crosssectional analyses of the relationship between socioeconomic indicators and frailty.Each study is represented by a single bar.The height of the bar shows the study sample size.The colour shows the measure of frailty used in the analysis.The position of the bar on the matrix indicates whether the socioeconomic indicator showed a positive association with frailty (higher frailty prevalence associated with lower socioeconomic position at a threshold of p < 0.05), neutral association (no statistically significant difference in prevalence), or negative association with frailty (higher frailty prevalence associated with lower socioeconomic position at a threshold of p < 0.05).Y-axis is truncated at 100,000 participants, three studies had > 100,000 participants(493,737, 968,885, and 1,470,000, respectively).Abbreviations: SEP = socioeconomic position.FRAIL = Fatigue, Resistance, Ambulation, Illness and weight Loss scale.Tilburg = Tilburg Frailty Indicator, Edmonton = Edmonton Frailty Indicator, Groningen = Groningen Frailty Indicator.

Fig. 4 .
Fig. 4. Age-and sex-adjusted association between frailty and socioeconomic indicators: This Harvest plot shows the findings for cross-sectional analyses of the relationship between socioeconomic indicators and frailty after covariate adjustment.All analyses are adjusted for age and sex, plus additional covariates which varied between studies and are shown in the Supplementary appendix.Each study is represented by a single bar.The height of the bar shows the study sample size.The colour shows the measure of frailty used in the analysis.The position of the bar on the matrix indicates whether the socioeconomic indicator showed a positive association with frailty (higher frailty prevalence associated with lower socioeconomic position at a threshold of p < 0.05), neutral association (no statistically significant difference in prevalence), or negative association with frailty (higher frailty prevalence associated with lower socioeconomic position at a threshold of p < 0.05).Y-axis is truncated at 100,000 participants, one studies had > 100,000 participants (493,737).Abbreviations: SEP = socioeconomic position.FRAIL = Fatigue, Resistance, Ambulation, Illness and weight Loss scale.Tilburg = Tilburg Frailty Indicator, Edmonton = Edmonton Frailty Indicator, Groningen = Groningen Frailty Indicator.

Fig. 5 .
Fig. 5. Association between socioeconomic indicators and changes in frailty status: This Harvest plot shows the findings for longitudinal analyses of the relationship between socioeconomic indicators and frailty.Each row represents a different methodological approach to modelling changes in frailty over time.Each study is represented by a single bar.The height of the bar shows the study sample size.The colour shows the socioeconomic indicator used in the analysis.The position of the bar on the matrix indicates whether the socioeconomic indicator showed a positive association with the marker of frailty in question (at a threshold of p < 0.05), neutral association (no statistically significant difference in prevalence), or negative association.