Introduction

In 2008/9, a financial crisis in the US spread quickly across the world, forcing governments in many countries to borrow significant sums of money to prevent collapse in the banking sector and broader economy. Subsequently, policies of austerity were implemented to varying degrees in many countries as governments sought to reduce budget deficits through combinations of increased taxes and/or reduced public spending. In the UK, the government focused particularly on reducing public spending, making cuts across government spending from 2010 including welfare [1], local government budgets [2], and social care [3]. NHS spending increased from less than 5% of GDP in the early 2000s to a highpoint of 7.6% in 2009–10 but austerity was a factor here too and spending began to fall slightly as a percentage of GDP during the 2010s while the population continued to age [4]. Research has suggested that the impact of austerity, while most focused on the poorest people [5], also affected more affluent groups [1]. Beyond the obvious impacts such as waiting lists for elective surgery growing from 2.5 million people in April 2012 to 4.1 million people in 2017 [6], researchers have connected the broad and significant extent of austerity in the UK to a range of concerning trends including rising use of food banks [7], deterioration in community infrastructure, services and environment quality [5, 8], increasing barriers to receipt of social care services [9], deterioration in indicators of mental health and rising suicide, and elevated levels of homelessness and housing insecurity [10]. Cuts to local government spending have been directly associated with increasing levels of multimorbidity [11] and, perhaps the starkest potential impact of austerity, is the levelling off in the rate of increase in life expectancy from 2014 (with decline for some groups) [12], following 200 years of increasing life expectancy [13, 14].

Frailty is where an individual has a limited homeostatic reserve and is most commonly a consequence of multi–system decline over a lifetime [15]. Frailty is an important outcome in later life since it leads to a lack of resilience and reduced capacity to recover from shocks, meaning that relatively minor events can trigger major changes in health status. Frailty is associated with multiple adverse consequences including increased risk of falls [16], hospitalisation [17], increased length of hospital stay [18], institutionalisation [19], poorer quality of life [20], and death [21–23]. A body of literature has demonstrated increasing frailty with age, higher frailty among women compared to men, and stark and widening inequalities in frailty across indicators of socio–economic position (for example, Harttgen and colleagues [24] and Marshall and colleagues [25]). Research on changes in frailty over time have found little evidence for improvement in levels of frailty, despite general trends of improved longevity. For example, results comparing Swedish cohorts born 30 years apart found similar levels of frailty [26] and, more pessimistically, frailty was observed to be higher in later–born US [27, 28], Netherlands [29] and Hong Kong Chinese [30] population cohorts compared to earlier born. These findings are in line with a broader body of research across a wide range of health indicators that shows growing inequalities in health, with improvements in healthy life expectancy lagging improvements in life expectancy [31, 32]. The bulk of evidence appears in line with the expansion of morbidity theory, whereby life expectancy increases faster than healthy life expectancy [33].

Understanding change in frailty before and during austerity is important as it provides a marker of the health of the older population and their longer–term capacity for independent living. In the context of austerity and population ageing, there is little health and social care capacity to cope with increasing frailty. This paper uses nationally representative data from the English Longitudinal Study of Ageing (2002 to 2018), to examine whether frailty increased at different rates during austerity than before. Our work was written according to the guidelines developed by the STROBE initiative [34].

Methods

We hypothesised that the introduction of austerity policies in 2010 would be associated with steeper increases in frailty with age. We used multilevel interrupted time series analyses to explore whether the change in frailty with age was different after the implementation of austerity policies compared to before, with a sensitivity analysis using an accelerated longitudinal design to compare changes in frailty for different birth cohorts, at the same age, from 2002 and 2012.

Results

A total of 16,410 individuals were included in the analysis (8,977 women), with a mean age of 67 across all years (see Table 1). On average, each individual contributed 4.5 frailty index scores to the modelling (range 1–9, median 4). These scores ranged from 0–0.76 with a mean of 0.15 and a median of 0.12. Men had lower mean and median frailty index scores (0.14 and 0.11), compared to women (0.16 and 0.13). Raw mean frailty index scores over time are shown below (Fig 1), stratified by age in 2002 which is the strongest single determinant of frailty although with large variation between individuals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Mean frailty at each wave of data collection, stratified by age in 2002.

https://doi.org/10.1371/journal.pone.0296014.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the population.

https://doi.org/10.1371/journal.pone.0296014.t001

Interrupted time series model

Model coefficients are in Table 2, but interpreting the individual coefficients is complicated by the square–root transformation, and the interaction terms. Fig 2 therefore shows the model predicted frailty index score (with model estimates squared to give mean frailty). On average, frailty index score increased strongly with age, was routinely higher for women than men of the same age, and was lower for richer people. The wealth effect was nonlinear with the poorest third having distinctly higher levels of frailty than the richest and middle thirds. For the youngest age groups, the difference was equivalent to the poorest third being as frail as people in the richest third of the population who were ten years older, but this association attenuated with age. The association between frailty index score and sex was weaker than that with wealth, with women having similar frailty in middle–age as men who were approximately five years older.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Modelled mean frailty over time for different age groups, stratified by wealth tertile and sex using an interrupted time series approach.

https://doi.org/10.1371/journal.pone.0296014.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Interrupted time series model output predicting the square root of the mean frailty index score.

https://doi.org/10.1371/journal.pone.0296014.t002

The rate of increase in frailty scores accelerated for all age groups following the best–fit modelled interruption of 2012 (see Fig 2). Over the period of four waves of data collection (2012–2018), the increase in frailty index score is approximately equivalent to an additional five years of ageing, similar in size to the difference between men and women. The estimated impact of austerity over the six years of follow–up is smaller than the estimated difference between the poorest and richest tertiles, but changes in frailty are seen across all wealth groups. As can be seen in Fig 2, the absolute increase was greater in the older age groups. For example, the mean frailty index score for the richest men aged 55–59 in 2002 was increasing by 0.0014 per year from 2002–2010, rising to 0.0038 per year from 2012–2018. By contrast, the mean frailty index score for the poorest women aged 80–84 in 2002 was increasing by 0.0098 per year from 2002–2010, rising to 0.015 per year from 2012–2018.

Accelerated longitudinal modelling (ALM)

Modelling the periods separately facilitates direct comparison of frailty trajectories between, for example, those who were 65–70 in 2002 and those who were 65–70 in 2012. The models are detailed in S1 File and their outputs are given in S5 Table and shown in Fig 3. Here people in 2012 start with lower frailty than their age-matched counterparts in 2002, suggesting that frailty had broadly improved during the 2000s, but frailty index scores then accrue more quickly for all age groups in the later time period. This is true for men and women, richer and poorer, but is particularly evident in the oldest ages groups. The increased rate of accrual results in average frailty scores in 2012–2018 going above the 2002–2010 levels for women aged 80+ and men aged 85+, meaning the oldest groups were frailer at the end of the 2010s compared to the equivalent age groups in the late 2000s.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Modelled mean frailty over time for different age groups, stratified by wealth tertile and sex using an accelerated longitudinal design approach (2002–2010 and 2012-–2018 modelled separately).

https://doi.org/10.1371/journal.pone.0296014.g003

Discussion

This study compared changes in frailty with age before and after implementation of austerity policies reducing public spending in many areas including welfare, social care and local government. The study finds increasing frailty with age, and higher frailty for women compared to men and for the poorest compared to the richest groups. In both the ITSA and ALM analyses, mean frailty score increases faster during the period of austerity (2012–2018) than in the pre–austerity period (2002–2010).

The key strength of this research is that we utilise a nationally representative social survey that includes a rich set of information from people born before the first world war to those born in the 1960s. We were able to construct a robust frailty index score and individual wealth measures, with sufficient data before and after austerity to estimate trends but there are several limitations to our research.

Firstly, the frailty index score is built using self-reported measures. Whilst much of the self-reporting comes from well-validated questionnaires, we cannot rule out changing trends in how people respond to such questions. Secondly, it might be argued that the increased rate of change in frailty from 2012 compared to before is simply a consequence of a non–linear increase in frailty as cohorts age. We explored this possibility by modelling trajectories before and during austerity separately, with age groups assigned by age in 2002 and 2012 respectively. Under this approach the 2012 model predicts faster accrual of deficits than the 2002 model, in line with the ITSA. Thirdly, ELSA is subject to non–random attrition as respondents choose not to take part in waves of the survey, they move into residential care, or they die. Those who drop out are more likely to be male, in poorer health and in lower socio–economic groups bringing potential bias to results. Since people with higher levels of frailty are more likely to drop out, we might expect our results to understate the extent of steepening in frailty trajectories after austerity compared to before, however it may also be the case that people have been supported to live in their own homes for longer, effectively increasing the frailty in the ELSA population but not the wider UK population. We were not able to control for this and it should not be dismissed as an issue, particularly in the oldest age groups.

Finally, as with all observational studies, this was an analysis of associations and causality cannot be strongly inferred. As was seen in previous work by Mousa and colleagues in England [38], changes in frailty may be driven by approaches to medical care that affect both ascertainment and prevalence of conditions. They compared frailty 1991 and 2011 and found higher levels of frailty in the 2011 group, driven by hypertension and diabetes, the former being better diagnosed and the later likely being both better diagnosed and increasing in prevalence. Whilst diabetes was not a key driver in our findings, we did not examine broader impacts of increases in sedentary behaviour and obesity which are also likely to be playing a role in changing health over time. We did however see frailty index score increase in all wealth groups (the richest typically having low levels of obesity) and our findings are consistent with other analyses finding associations between austerity and food insecurity [7], housing precarity [10], multi-morbidity [11] and mortality [12].

The UK experienced the highest (pre-pandemic) annual spike in mortality in 2015, which occurred predominantly among frail older people and was sustained in subsequent years [12]. Darlington–Pollock and colleagues [14] estimate that the levelling off in mortality rates from 2010–2018 cost around 230,000 lives compared to a continuation of prior trends, and increasing frailty is a logical precursor to these increased deaths. Mortality attributed to austerity in older groups occurred in all geographic areas (affluent and deprived) [12]. This has been echoed by our finding that frailty index score increases faster in all population groups (rich and poor), which accords with previous studies showing that austerity impacted across all population groups [1], although there is variation in the impact of austerity on excess mortality. For example, the extent of local cuts to social care and to financial support aimed at older people are associated with area level rises in mortality among those aged over 85 [12], which is the age group where we observe the largest change in frailty in the accelerated longitudinal design models.

A policy implication of our research is to recognise that public spending reductions have impacts on health and mortality across all population groups. Most governments internationally face challenging fiscal circumstances as societies emerge from the acute stages of the pandemic compounded by rising energy costs and general inflation. Proposals to cut public funding should therefore balance perceived benefits with likely consequences for health and other social outcomes. This study also finds that from the early 2000s to the early 2010s, when health spending was increasing, mean population frailty was falling, but this improvement was largely lost due to accelerated accrual of frailty deficits during austerity. Population resilience to further austerity in the 2020s may therefore be lower than it was in the 2010s.

There is a need for further research to better understand what caused the deterioration in health outcomes, such as frailty, during austerity compared to before. Although we did not find any one domain of frailty deficits to be driving results, future research might focus on specific deficits that one might expected to be particularly linked to austerity in the short term. Additionally, research exploiting the harmonised longitudinal data on older cohorts including the English Longitudinal Study of Ageing, the Health and Retirement Study in the USA, and the Survey of Health and Ageing and Retirement in Europe has the potential to explore differences in frailty trajectories in national contexts where different policy responses to the financial crisis were implemented. This has the potential to help evidence whether austerity policies were causally related to changes in frailty, as countries each approached the financial crisis differently. Such differences may also identify elements of 2010s austerity that were potentially the most harmful, facilitating mitigation should further austerity be pursued. If the political response to the pandemic and subsequent economic instability is further austerity, then further longitudinal research to examine its effects across a range of outcomes including frailty will be needed.

Supporting information