Cutting through the intersections to care for caregivers: Secondary data analysis of a carers support service in Glasgow, Scotland

As the number of people living longer with life-limiting conditions grows, so too does the number of caregivers and the complexity of the caring role. To understand more about the role and how caregivers can be supported, local and national registers have been created that collect data on caregivers. Our objective was to undertake comparative analysis of female and male adult caregiver assessments from a caregiver database created from a carers support service running in Glasgow, Scotland. Assessments were carried out over a 12-month period (01/04/17– 29/03/18). We aimed to identify the prevalence of negative consequences of caring through descriptive statistical, comparative analysis. Seven hundred and eighty-three assessments were eligible for inclusion. In our dataset

about the role and how caregivers can be supported, local and national registers have been created that collect data on caregivers.Our objective was to undertake comparative analysis of female and male adult caregiver assessments from a caregiver database created from a carers support service running in Glasgow, Scotland.
Assessments were carried out over a 12-month period (01/04/17-29/03/18).We aimed to identify the prevalence of negative consequences of caring through descriptive statistical, comparative analysis.Seven hundred and eighty-three assessments were eligible for inclusion.In our dataset, 69% were female (n = 552), and 29% were male (n = 231).Female caregivers were more likely to be of working age but unemployed (p = 0.03) and experiencing mental ill-health (p = 0.011).Male caregivers were more likely to be retired (p < 0.001), caring for a parent (p = 0.017) and living with heart disease (p = 0.0004), addiction issues (p = 0.013) or diabetes (p = 0.042) than female caregivers.For caregivers using this support service, female and male caregivers experienced, recognised or reported negative impacts from caring on their personal identity, social life, ability to self-care and relationships similarly.Furthermore, a caregiver whose relationships had been negatively impacted was 13.8 times more likely (p > 0.00) to report a reduction in psychological well-being.Sex disaggregated data are an important consideration for caregiver research due to socio-political influences that impact caring roles and expectations.Disaggregating data by sex allow researchers to understand how the caring role differs between subsets and allow for the development of more targeted, sensitive support.

K E Y W O R D S
caregiving, carer, carers' services, family caregiving, gender, statistical analysis

| INTRODUC TI ON
As a direct result of people living longer with life-limiting conditions, the number of caregivers continues to grow.A caregiver is usually, although not always a family member who supports at least one other person to undertake activities of daily living to varying extends depending on need and capability.The intensity and complexity of the role often results in a shared illness experience with the person for whom they care (Gardiner et al., 2020).There are an estimated 43.5 million caregivers in the United States of American (Family Caregiver Alliance, 2019), 8.8 million caregivers in the United Kingdom (Carers UK, 2019), and between 20% and 44% of the population of Europe consider themselves caregivers (Verbakel et al., 2017).However, many countries such as India do not formally recognise this role, so worldwide caregiver population statistics are unknown.
Although not all countries recognise the role, the societal benefits created by people who care cannot be overlooked.Oxfam's Time to Care (Lawson et al., 2020) report calculated that the unpaid labour of caregivers is valued at, at least $10.8 trillion annually, which is three times the tech industry.In the United Kingdom, their support is valued at £132 billion (Buckner & Yeandle, 2015).
To put this in perspective, the annual spending on health and social care in the United Kingdom is currently valued at £140 billion (Kings Fund, 2020).Furthermore, Fraser (2016) described unpaid caring labour as the foundation, not only for economic productivity but also for culture and political organisation.The contribution made by caregivers is, therefore, essential to maintaining functional communities and our health and social care systems.
Although the caring role is of significant collective gain, it can come at personal loss, with a diverse and multidimensional range of negative consequences related to it.Research suggests that caregivers are exposed to an increased risk of physical injury, psychological trauma and social isolation (Gusdal et al., 2016;Irfan et al., 2017;Vasileiou et al., 2017).The array of negative consequences coupled with increasing demand for people to undertake the role has created a growing body of research related to caregiver needs assessment and supportive services and interventions.However, delays in timely assessments (Carers UK, 2019), barriers to implementing evidencebased assessments (Diffin et al., 2018) and variability in outcome measures (Drummond et al., 2019) have been identified.Moreover, not all caregivers experience negative consequences, nor do all require support to continue in their caring role (Hazell et al., 2020).
Therefore, due to finite resources, researchers have sought ways to identify or predict those most likely to require support.
It has been suggested that negative consequences can be somewhat predicted by considering caregiver characteristics such as age (Chow & Ho, 2015), income (Hu et al., 2016) and duration of the caring role (Litzelman et al., 2015).However, the influence of caregiver sex on negative consequences and outcomes is an area that is overlooked within caregiver research (Gilmore-Bykovskyi et al., 2018).And this oversight is not unique to caregiver research.
Criado Perez (2019) identified the extend of the 'gender data bias'.A lack of sex disaggregated data has resulted in dangerous oversights that disproportionately affect the lives of women and girls all over the world (Criado Perez, 2019).However, as caregivers are predominantly a female population and existing findings are therefore, predominantly related to women, this oversight could potentially negatively impact male caregivers.Because of these issues, we were interested in describing a contemporary, UK-based caregiver dataset to look at the potential effect of sex as a caregiver outcome.
Secondary data analysis of routinely collected health and social care data is an area of increasing research interest.This is because data collected from some research methods, such as randomised controlled trials, do not necessarily account for real-world influences and variances.However, there are limitations to this type of data.
Issues related to quality control and data validity have been identified because researchers are not typically involved in designing the data collection and management methods (Polit & Beck, 2004;Trenner et al., 2019).Furthermore, as digitised health and social care data grows exponentially, so too do the variances within it, which can make it difficult to create new knowledge (Feldman et al., 2018).
Nevertheless, secondary analysis of existing health and social care data is a cost-effective way for nurses to participate in service evaluation and population health research (Garmon Bibb, 2007).
Furthermore, as there is potential to identify patterns and trends that could potentially save lives and improve care, researchers should seek to overcome the challenges (Raghupathi & Raghupathi, 2014).
In sharing experiences of secondary data analysis of electronic, routinely collected data, researchers can contribute to the improvement

What is known about this topic?
• Caring responsibilities increase a person's risk of experiencing physical injuries, mental ill-health and financial hardship.
• The risk of caregivers experiencing negative consequences of caring can be somewhat predicted by considering the caregivers age, income and diagnosis of the cared-for person.
• The demand for caregivers continues to grow due to people living longer with life-limiting conditions.

What this paper adds?
• Disaggregating caregiver data by sex is a necessary consideration in caregiver research because female and male caregivers experience different socio-political influences that impact their caring role.
• Caregivers of working age require flexible support that works around employment and other caring responsibilities to reduce the risk of financial hardship and poverty.
• When designing support for caregivers, health and social care services should consider that older and retired caregivers will likely have their own set of health and social care needs that may impact their caring role. of quality, procedure and analysis in this field of research (Weiskopf & Weng, 2013).

| Objective
• To undertake comparative analysis of female and male adult caregivers in relation to the prevalence of negative consequences of caring.
There was limited data provided relating to the care recipient because the focus of the support service is the caregiver.There is no set of criteria related to diagnoses, time spent caring or living arrangements that caregivers must meet to be entitled to assessment and support.Caregivers only need to be identified as or self-identify as caregivers.Therefore, we do not report on health data related to the cared-for person.This ensures the health and well-being of the caregiver remains the focus and does not become side-lined by the needs or perceived needs of the person for whom they care.

| Sample
Our analysis used a dataset held by the Glasgow Carers Team, which is curated by the Glasgow City Health and Social Care Partnership.
Glasgow is an urban city in Scotland with a population of 626,410.
The Carers Team undertake individualised assessments of caregivers' needs.Caregivers are not considered to be in paid employment for their role, although some might receive state benefits related to it.Referrals to the service can be from hospital and community services, concerned friends and family, as well as the caregivers themselves.Assessments are carried out by social workers, usually in the caregiver's own home.Consultation style for data collection varied between practitioners, and there was no set list of questions used for assessments.The dataset is collected for service evaluation and routine documentation of practice; we had no control over what was included in it.Our dataset forms part of a larger database that records all caregiver assessments carried out by the service.The database is reviewed and updated annually.The dataset includes socio-demographic and healthcare-related information of the caregiver and cared-for person.There were also several free-text fields that included other relevant information from the assessments.
Nine-hundred and ninety-eight caregiver assessments were submitted for analysis.Although the service also assesses young caregivers, we included caregivers who were 16 years and over in our analysis.

| Ethics
We obtained ethical approval to undertake secondary analysis of anonymised data from the University of Glasgow ethics committee (reference number 200170185).

| Data cleaning and categorising
Data cleaning and categorising was an arduous process due to the variations in documentation styles and nature of the Excel spreadsheets, as each assessment was recorded manually.We initially removed any duplicate assessments and cleaned the demographics fields.The free-text fields contained insightful data detailing common areas of the caregivers' lives that were detrimentally affected by their caring role(s).Therefore, we decided it was important to explore the possibility of including data from the free-text fields in our analysis.Because the disparities in the way the free-text fields were recorded, we scoped the first 25% of assessments to establish if there were recurring concerns identified by caregivers.We did not limit ourselves to any specific categories during this process.Then, we reviewed the categories and identified the following aspects of their lives that caregivers stated had been negatively impacted due to their caring role: • Social life/recreation.

| Statistical analysis
We used IBM SPSS Statistics for Windows, Version 27 to undertake descriptive statistical analysis.We used the chi-squared test to compare univariable categorical variables (using a threshold of p < 0.05) comparing female and male caregivers in age, demographics, health diagnosis and the coded free-text fields.When there was found to be a statistically significant difference between female and male caregiver variables, we carried out binary logistic regression analysis.We also performed binary logistic regression of issues identified from the coded text and caregiver sex that might contribute to poorer psychological well-being because mental illness was the most reported health issue for female and male caregivers.Due to the significant difference between female and male caregivers by age, we controlled for age in both regression models.

| Socio-demographic and health data
Due to variation in consultation style and documentation of data, the cleaning and categorisation, particularly of the free-text fields, was a lengthy process.In this dataset, of the 783 caregivers included (Figure 1), 552 (69%) were female, and 231 (29%) were male (Table 1).Fifteen caregivers (2%) had no sex listed.Table 1 illustrates the significant difference in age between female and male, as female caregivers tended to be younger than male caregivers.For example, 71% (n = 394) of female caregivers were aged under 65 years of age, compared to 51% (n = 120) of males.A Mann-Whitney U test indicated that overall, male caregivers (min = 21, max = 96, mean = 60.5) were significantly older than female caregivers (min = 17, max = 92, mean = 52.1,p = 0.0001).
Table 2 outlines the socio-demographic findings from our dataset.The only caregiver/cared-for relationships that differed significantly between sexes was when a parent was caring for a child (female caregivers, p < 0.001) and when the caregiver was caring for a parent (male caregivers, p = 0.017).Unemployed caregivers were significantly more likely to be female (p = 0.03), and retired caregivers were significantly more likely to be male (p < 0.001).There was no significant difference among caregivers who were willing or able to continue in their caring role.When considering the types of support requested by the caregivers, male caregivers were significantly more likely to ask for respite than female caregivers (p = 0.046).
Two hundred and fifty-nine female caregivers (46%) and 100 male caregivers (43%) had a health issue listed.Mental health was the most recorded health complaint for female and male caregivers (Table 3).Specific mental health conditions were not listed; instead, mental health was used as an umbrella term.We know this from reading the free text, where specific conditions such as anxiety, depression and schizophrenia were mentioned.In this dataset, male caregivers were significantly more likely to have heart disease, addiction issues, diabetes or health issues related to advancing age than female caregivers (Table 3).
Where p had suggested a significant difference between female and male caregivers in Table 2, a binary regression was performed to establish how the caregiver/cared-for relationships, employment status, support requests and caregiver diagnoses (Table 4) were influenced by sex.Overall, this model did not find a significance.
However, female caregivers were 13.7 times more likely to be caring for a child (p = 0.001).
Table 5 outlines the findings from the categorisation of free-text fields that relate to various aspects of daily living and personhood that caregivers expressed were negatively affected by their caring role.The reporting of the psychological impact of caring (p = 0.038) and personal/family finances (p = 0.044) differed significantly between female and male caregivers.Table 5 suggests that for caregivers using this support service, male and female caregivers experienced, recognised or reported most of the common negative consequences of caring equally.
Because mental ill-health was the most common health complaint listed by female and male caregivers, a binary regression was performed to establish the effects of caregiver sex, social life, independent identity, personal/family finances, self-care, and relationships on the likelihood of the caregiver experiencing a negative psychological impact from their caring role (Table 6).The binary regression model was statistically significant, with all but personal identity influencing psychological impact.The most significant influence on psychological impact was from caregiver relationships.
A caregiver whose relationships had been negatively impacted was 13.8 times more likely to report a negative psychological impact from their caring role.

| Overview of findings
Previous research related to caregivers of people at end of life and living with life-limiting conditions and mental illness identified that the influence of caregiver sex on the risk of negative outcomes and consequences is often overlooked and misunderstood (Gardiner et al., 2020;Ghosh et al., 2020;Pinquart & Sorensen, 2006;Sharma et al., 2016).This study demonstrates that there is meaning and knowledge to be found by disaggregating caregiver data by sex.In our dataset, the most prominent difference between female and male caregivers related to age.Male caregivers were significantly more likely to be retired, caring for a parent, and living with heart disease, diabetes or addictions, whereas female caregivers were more likely to be of working age but economically inactive due to sickness and living with mental ill-health.
Female caregivers were also more likely to report that their caring role was detrimental to their mental health.Greenwood and Smith's (2016) review suggested that the importance and influence of age is often overlooked in caregiver research.This led us to consider how the age and sex of caregivers interact and influence the use of support services, such as the one the dataset was obtained from.

F I G U R E 1 Dataset summary of results from data cleaning and reviewing inclusion criteria
Crenshaw's seminal work on intersectionality was a move away from single-axis analysis towards a more nuanced understanding of how characteristics such as race, sex and class intersect to create complex subordination (Crenshaw, 1989).However, in recent years, there has been much heterogenity around the concept.
Nevertheless, it is generally understood to be a way of understanding and analysing complex human experiences and the organisation of power within society (Hill Collins & Bilge, 2018).This framework has already been successfully adopted within caregiver research to explain the effect of the relationship between caregiver and the person for whom they care on well-being (Chappell et al., 2015).Akkan (2019) describes the intersections of caring and its multiple social relations as a complex 'inequality-creating phenomenon'.The intersection of age and sex was first identified by Krekula (2007).Krekula (2007) suggested that societies 'othering' of older women was so endemic that the lives of older women were even overlooked in feminist research and theory.Instead, our understanding of female experiences primarily focused on young, fertile Note: Abbreviation: COPD, chronic obstructive pulmonary disease.
p is 0.05 and was calculated from chi-squared testing.
women (Krekula, 2007).Indeed, this ageism also extends to caring roles and responsibilities because of prejudices that exist in society related to who should do unpaid, caring labour (Akkan, 2019).This is because within the family, the age and stage of women and girls predicts how likely she is to have caring responsibilities expected of her (Akkan, 2019;Young & Grundy, 2008).

| The 'crisis of care'
Most female caregivers in our dataset were under 65 and therefore, within the working age population.However, a significant number of the female caregivers in this dataset were unemployed or economically inactive (46% of females).This compares to 31% of the general population and 44.5% of caregivers in Scotland from the 2011 census (Scottish Government, 2015).The Scottish census also found that 32% of female caregivers felt their caring responsibilities had impacted their employment, with 8.1% working fewer hours and 5.5% had left their work altogether.Indeed, the challenges experienced by Scottish women with caring responsibilities entering or maintaining employment are echoed by women all over the world.
Caring is the main barrier to women entering, progressing and as likely to take urgent time off due to caring responsibilities (Smith et al., 2019).The risk of financial hardship or poverty because of caring responsibilities, therefore, disproportionately affects women and girls.
Caregivers in our dataset were 3.3 times more likely to report a negative psychological impact if they were experiencing financial hardship from their caring role.Poverty levels are highest among caregivers who care for over 20 hr a week and those caring for someone living in the same household (Aldridge & Hughes, 2016).
In the United Kingdom, caregivers report delays in needs assessments, regularly using their own money to buy necessary equipment, and most (69%) do not receive financial support from their local authority or health body (Carers UK, 2019).Furthermore, Gardiner et al. (2016Gardiner et al. ( , 2020) ) identified that these common financial strains increase significantly when caring for someone receiving palliative and end of life care.End-of-life caregiving often causes further stressful life events, such as moving to a new house or early retirement to cope with the increased care requirements (Gardiner et al., 2020).However, the financial risks associated with the role for working age women extend beyond the caregiver, because women with children are more likely to take on caring responsibilities (Young & Grundy, 2008), and women are far more likely to be single parents (OECD, 2018b).Therefore, any financial hardship experienced by working age female caregivers has the potential to impact children.
As most of the caregivers in our dataset were women of working age, this suggests that this demographic is the most likely to be identified or identify as requiring help.Fraser (2016) argued that capitalism in its current form reinforces this 'crisis of care', whereby women balance limited, feminised employment options with an increasing number of caring responsibilities.Indeed, the influence of capitalism on the division of both paid and unpaid labour compared to men persists to varying degrees all over the world (Barrett, 1988;Ferrant et al., 2014;Smith et al., 2019;Stanfors et al., 2019).
Therefore, it may not be caring alone, but the pressure to balance employment with care heightens the need for support.Greenwood and Smith's (2016) review identified some evidence that indicates that older caregivers might view their caring role more positively and cope better than younger, working age caregivers.Therefore, to reduce the risk of financial hardship and poverty, caregivers require flexible support that works around their caring responsibilities and employment.This, in turn, will enable caregivers to be more stable, dependable and productive workers.Furthermore, Gardiner et al. (2016Gardiner et al. ( , 2020) ) recommended that services, professionals or researchers who work with caregivers, monitor and assess caregivers using evidence-based tools for financial burden and signpost to appropriate support when necessary.Female caregivers in our dataset were also significantly more likely to have mental ill-health listed and to report that caring was psychologically detrimental to them.This supports previous research that found female caregivers articulate more strain (Fromme et al., 2005) and experience higher levels of burden (Chiao et al., 2015) than male caregivers.However, why women and men report different levels of common caregiver stressors has been debated in the literature (Pinquart & Sorensen, 2006).Fromme et al. (2005) suggested that women report more strain because women are more likely to share their emotions.Indeed, females do tend to have better social support (Wong & Hsieh, 2019), and poor social networks are linked to an increased risk of caregiver burden (Ghosh et al., 2020).Nevertheless, research suggests that not only are women more likely to care for more hours than men (Pinquart & Sorensen, 2006), caring for people with more complex needs.
However, the magnitude of these differences has been found to be influenced by how gendered national labour and workforce policy is (Stanfors et al., 2019).
So, if we consider the increased risk of financial burden and likelihood that female caregivers will have multiple caring responsibilities, women may indeed be more likely to express their struggles, but they will likely have more stressful demands.Fortunately, addressing the psychological impact of caring is a priority in caregiver literature.
A review of caregiver outcomes identified outcomes that aimed to measure psychological symptoms (such as anxiety and depression) were the most popular types used in caregiver intervention research (Drummond et al., 2019).

| Physical health, sex and care
The 'othering' of older women as outlined by Krekula (2007) has some considerations for the male caregivers in our dataset.
Older caregivers were more likely to be male, and this seems to be an increasing trend across the United Kingdom, as caregivers over the age of 85 are more likely to be men (Carers UK, 2019).
'Othering' has many negative consequences; it frames people who do not fit the preconceived expectations of a group as inferior and contributes to deeper marginalisation.Therefore, as caregivers are a predominantly female population, there is a risk of 'othering' male caregivers.Furthermore, male caregivers are exposed to harmful social expectations based on sexist stereotypes.For example, Williams et al. (2017) examined how caregivers construct their views of the appropriate roles for women and men in providing palliative and end of life care for family members.They found that female participants described men as strong, independent and self-sufficient (Williams et al., 2017).These masculine stereotypes are at odds with the reality that many men will have less experience of caring and household labour throughout their life course (Ferrant et al., 2014).Therefore, the help and support they need are likely very different to women.However, Houde's (2002) literature review detected methodological issues in caregiver research in relation to men.Therefore, it was recommended that researchers proactively seek to recruit men into caregiver research, especially intervention trials (Houde, 2002).
As well as being older, male caregivers in our dataset were found to have significantly higher rates of heart disease and diabetes mellitus than female caregivers (Table 3).There is limited research available related to the prevalence of heart disease and diabetes among male caregivers.However, it is established that long-term exposure to stressful situations can be detrimental to cardiovascular health (Dimsdale, 2008).Further, Lebrec et al. (2016) found that caregivers with diabetes were more likely to be males and that caregivers with diabetes had a significantly higher rate of outpatient visits.Men in our dataset were also more likely than women to be caring for their parents.Morgan et al.'s (2020) literature review highlighted the complexity of caring as an older adult and outlines specific considerations for health and social care services.They will likely have their own set of health and social care needs, may have their own caregiver(s) or be in a mutually caring relationship with the person for whom they care (Morgan et al., 2020).
Regarding caregiver's social life, independent identity, self-care and relationships, there was no significant difference experienced between female and male caregivers in our dataset.This echoes a meta-analysis on differences in caregiver stressors and social resources, which found only small to very small differences between female and male caregivers (Pinquart & Sorensen, 2006).However, this is inconsistent with findings from Wong and Hsieh (2019).They identified gendered differences between female and male caregivers' ability to maintain and manage their relationships and social networks when caring for a spouse or partner.This study analysed data from heterosexual caring relationships aged between 57 and 85.These varying results between studies demonstrate the need for further analysis of how age and sex influence one another and the caring experience.

TA B L E 5
The presented data report aspects of caregivers' lives and personhood that they expressed during their assessments had been negatively impacted due to their caring responsibilities It has been suggested that caregiver researchers should consider incorporating a more nuanced understanding of the caring role through the use of conceptual and theoretical frameworks (Dilworth-Anderson et al., 2020).One suggested framework is intersectionality.Intersectionality is a concept that was developed by Black feminists in the 1980s and 1990s to describe the multidimensional oppression experienced by Black women.Kimberle remaining in paid employment and why women tend to work in part-time, lower paid and less secure jobs (International Labour Organization, 2018).Women spend more hours than men undertaking unpaid work and care (OECD, 2018a) and more minutes per day in their caring role than they do in paid work (International Labour Organization, 2018).Smith et al. (2019) carried out secondary analysis of data from Statistics Canada's Labour Force Survey collected over 19 years on the impact of caring responsibilities on labour market involvement.They found that women were 73% more likely to leave the labour market, over 5 times more likely to work part-time, and twice Note:We have also included comparison data from the 2011 Scottish National Census.Data presented is the socio-demographic information recorded caregiver assessments included in our analysis TA B L E 1 Data presented are related to caregiver age and categorised by sex TA B L E 2 Presented data are a binary regression model related to data from Table2Where p suggested significance in Table2, we explored each (the dependent variables) with its association with the sex of the caregiver (the independent variable) and calculated the odds ratio and confidence interval (CI).With this model, we controlled for age due to the significant difference in age between female and male caregivers.
TA B L E 4 These data were collected from the categorisation of free-text fields included in caregiver assessments and presented disaggregated by sex, with p of 0.05.