STRENGTHS AND LIMITATIONS OF THIS STUDY

A strength of our study was the use of the Australian population representative data sets, the Survey of Disability, Ageing and Carers.

A further strength is the use of established microsimulation models (Australian Population and Policy Simulation Model and Static Incomes Model) to project economic impacts to 2030.

Our study only includes carers caring for a person whose main chronic condition is arthritis.

The sample size of our analysis was relatively small as the study has primarily focused on those carers who were not in the labour force due to caring for a person with arthritis.

Underemployment of informal carers is not included in our analysis.

Introduction

Arthritis, the most frequent type of musculoskeletal disease (MSK), is an umbrella term encompassing a large range of chronic conditions where inflammation of the joint, muscle or bone causes pain, swelling, stiffness and redness.1 The condition can be debilitating, with cycles of exacerbations and remissions often impacting a person’s ability to perform tasks, their psychological health and quality of life.2 3 Arthritis can affect anyone at any age but it is more common in older people and women.2 4 Risk factors, including modifiable factors (excessive weight, manual and repetitive tasks, physical injuries, smoking and dietary) and non-modifiable factors (age, gender and genetics) are associated with all forms of arthritis.5 There were roughly 3.6 million (15%) Australians living with arthritis in 2017–2018, with osteoarthritis (OA) and rheumatoid arthritis (RA), being the two most common types.1 This prevalence is expected to increase, with one study estimating 3.1 million and 580 000 Australians would be affected by OA and RA, respectively, by 2030.6 Other studies have estimated that the future prevalence of arthritis would increase by almost 50% in the USA and Canada,7 8 with the Global Burden of Disease (GBD) study reporting this as one of the fastest-growing health concerns and a leading cause of disability.9 The increasing growth and ageing of the population are the main causes for the increasing prevalence, including future prevalence, of arthritis worldwide.6–8 10 11 Other causes include the increasing rates of obesity and sports injuries.6 11

Arthritis is costly to treat due to its high prevalence and the lifelong nature of the disease. Several studies estimate the increases in healthcare expenditure from the estimated increase in the prevalence of arthritis.6 12–14 In one Australian study, the estimated total healthcare cost of OA and RA are expected to increase from $A2.6 billion in 2015 ($A109 per capita)15 to $A3.6 billion by 2030 based on the estimated increase in the prevalence of OA and RA from 2015 (2.2 million and 422 309 Australians, respectively) to 2030 (3.1 million and 579 915 Australians, respectively).6 However, this is likely an underestimate as the Australian study noted some important health costs including costs associated with medical specialists and allied healthcare were excluded. The findings from one US study on medical expenditures and earning losses associated with arthritis and other rheumatic conditions showed that the increase in the prevalence of arthritis from the year 1997–2003 (36.8 million and 46.1 million US adults, respectively) led to an increase in total medical expenditure of about US$88 billion (from US$233.5 billion in 1997 (US$876 per capita)16 to US$321.8 billion in 2003 (US$1105 per capita).12 17 One Canadian study estimated the total (direct and indirect) expenditure of OA and RA at $C33.2 billion in 2010 ($C975 per capita)18 and has predicted that this would increase to about $C1712.1 billion in 30 years if nothing is implemented to prevent the rising prevalence.13

Several studies also report lower levels of workforce participation and early retirement among those with arthritis, with some reporting indirect costs outweighing direct healthcare costs.8 14 19–23 An Australian study reported arthritis as the second most commonly reported condition for Australians aged 45–64 years old who had left the workforce prematurely, with almost 50% of the respondents reporting arthritis as the main health condition causing this.19 Those who had left the workforce prematurely due to arthritis had an estimated income that was approximately one-fifth of their employed counterparts ($A257 and $A1226 per week, respectively).20 They were also more likely to fall into poverty, especially 1–3 years after the diagnosis of arthritis, compared with those who did not have arthritis.21 Lacaille and Hogg8 reported that the working life expectancy of those with arthritis or rheumatism is 3–4 years (9–10%) less than those without the disorder. This is equivalent to an average lifetime cost of $C177 435 for men and $C86 296 for women, and $C208 billion to the government.8 A 2013 US study showed that the total expenditure on arthritis was US$303.5 billion per year, with over half of the cost comprising indirect costs (US$163.7 billion).22 From this study, those with arthritis were more likely to be out of the labour force (23.1%) compared with their counterparts without the disease (8.8%).22 A review of the literature on eight conditions requiring rehabilitation in the USA found back pain or arthritis (OA or RA) as the most common and expensive conditions, where the cumulative results across the cohort can amount to tens of billions of dollars per year in lost wages and reduced productivity.23 In the UK, the prevalence of MSK conditions is projected to increase within the labour force population due to the average age of the working population estimated to increase from 39 years old in 2016 to 43 years old by 2030.14 In this study, it is estimated the indirect costs from days lost from work for those affected with OA and RA is expected to increase from £2.58 billion in 2017 to £3.43 billion by 2030.14

Although expenditure on arthritis is widely reported, very few studies have reported the economic impact associated with informal caregivers of those with arthritis. A carer is a person who provides care to another person who is unable to perform daily activities themselves because of limited personal autonomy, and either receives a monetary payment (formal carers) or does not receive any monetary payment (informal carers) in exchange for the care provided.5 Like those with arthritis, informal carers are less likely to participate in the labour force due to the demands of caregiving duties.24 In a recent systematic review, Oliva-Moreno et al 25 identified studies on the cost of informal caregivers providing care to people with one of the six conditions with high limitations in personal autonomy, thus, more likely to require a carer.25 Out of the 91 studies identified in this review, only 4 reported the cost of informal care to a person with arthritis or OA, the lowest number of studies identified out of all six chronic conditions.25 However, this review only included cost-of-illness studies valuating the cost of informal care from the hours of caregiving duties provided by the carers estimated. Similarly, in a literature review by Salmon et al,26 out of the 32 studies identified that have an indirect cost component, only one estimated the cost of informal care. However, this was limited to cost-of-illness studies on lower limb OA.26 The paucity of data on the cost of informal caregiving to those with arthritis is observed in one New Zealand study, where the cost of informal care of those with arthritis in 2018 (NZ$1.58 billion) was estimated based on data retrieved from the published studies that were mostly on RA.5

Given the importance of arthritis and its future impact, we aim to address this gap in the literature on the cost of informal care to a person with arthritis by using primary data and a microsimulation model to project the economic costs of leaving the workforce early to provide informal care for persons with arthritis in 5-year intervals from 2015 to 2030.

MethodsData sources

We used data from the Care and Work Microsimulation Model (Care&WorkMOD), a microsimulation model built to analyse the economic impact of lost productive life years of informal carers who care for at least one person with a chronic condition. Care&WorkMOD comprises four Australian population representative data sets (figure 1). The base data set is the unit record data of individuals aged 15–64 years from the Australian Bureau of Statistics (ABS) Survey of Disability, Ageing and Carers (SDAC) 2003, 2009 and 2012.27–29 The SDAC occurs every 3–4 years, surveying individuals from three target populations throughout Australia: (1) the elderly (aged 65 years and older); (2) Australians with a long-term disability or chronic illness; (3) informal carers of either (1) or (2).30 Australians are selected randomly by the ABS and those who meet the eligibility criteria are invited to participate, with data weighed based on the probability of the households being selected in the sample and benchmarked.31 The full responses of all consenting participants are used to generate a snapshot for Australia.30 Other uses of this nationally representative data set include research and government policies aimed at allocations of services and needs for the three target populations.30 The reason for not being in the labour force for carers and non-carers (including caring for someone else due to their illness, retired and so on) was included in the SDACs, as well as the main condition of the caree, with arthritis being one of the identified conditions. This paper focuses on persons who participated in the SDACs who reported that the main reason for them being out of the labour force was because of caring for ‘someone else’s ill health or disability’ and reported arthritis as the main chronic condition of their main care recipient. We excluded informal carers caring for people with other chronic conditions or if arthritis was not their main chronic condition.

Figure 1

Care and Work Microsimulation Model. ABS, Australian Bureau of Statistics; STINMOD, Static Incomes Model.

The other three components in Care&WorkMOD are: (1) the population and labour force projections for 2015–2030 from the 2015 Intergenerational Report32; (2) projected distributions of sociodemographic variables, including the proportion of informal caregivers from the Australian Population and Policy Simulation Model (APPSIM)33; and (3) output data sets from the Static Incomes Model (STINMOD), a microsimulation model of Australia’s income, tax and transfer system.34 A detailed description of the development of the model is given elsewhere.35–37 Briefly, Care&WorkMOD used static ageing techniques, such as ‘reweighting’ to project the sociodemographic and economic profile of the Australian population at 5 yearly intervals from 2015 to 2030. Weights were derived based on the projected number of carers and projected labour force participation rates.

Carer status and their labour force participation (including full-time and part-time employment) were captured in the SDAC data sets. We reweighted the SDACs at the individual record level to reflect the projected profile of the population including labour force participation based on the 2015 Australian Intergenerational Report,32 over the modelled years (2020, 2025, 2030). The projected growth in primary carers was based on APPSIM age-specific and sex-specific rates of primary carers for 2015–2030 by 10-year age group and gender every 5 years from 2015 to 2030. Reweighting was undertaken to these benchmarks using an algorithm based on generalised regression reweighting techniques, coded in an SAS macro called GREGWT, developed by the ABS.38 Growth in the number of patients with arthritis is driven by ageing and the growth of the population, and is captured by reweighting.

Data from STINMOD, such as income tax paid and welfare payments, were imputed onto Care&WorkMOD by statistically matching the unit records of STINMOD with the Care&WorkMOD base population39 based on nine matching variables: labour force status, income unit type, income quintile, gender, age group, hours worked per week, highest education level attained, whether or not the individual is a homeowner and whether they receive the Carer Payment (a social security benefit in Australia population). Income and other economic data from STINMOD were indexed to produce projections for 2030. Earnings, income taxes and welfare payments: Age Pension, Disability Support Pension and Carers Payments were increased at a rate of 1% per annum in real terms (ie, 1% above inflation). All other welfare payments were indexed at 0% in real terms based on the Australian government policy of increasing welfare payments other than pensions in line with the Consumer Price Index (CPI).40

Statistical analysis

We estimated the mean, SD and median for weekly income, weekly welfare payments and weekly taxes paid for informal carers not in the labour force (NILF) to provide care for someone with arthritis; and for non-carers employed full-time and non-carers employed part-time. Counterfactual values for economic outcomes (weekly income, welfare payments or taxes) that carers NILF might have achieved on average, had they not been engaged in the provision of care, were estimated using Monte Carlo methods. A counterfactual record was selected at random with replacement from the pool of non‐carers who were in the labour force (full-time or part-time employed or unemployed) for each record of those NILF due to their caregiving roles, with records matched for age group, sex and the highest level of education. Losses in economic outcomes such as weekly income, welfare payments and taxes for carers NILF due to their caregiving roles were estimated as the difference between the carer-reported economic outcomes and their counterfactuals. A total of 5000 simulations were run, generating 5000 counterfactual data sets, which is sufficient for the estimates to converge. The average of the 5000 simulations and the 95% CIs, using the percentile method, were estimated to provide the degree of uncertainty of the estimates reported. Similar simulations were undertaken with the counterfactual pool of full-time employed non-carers and the counterfactual pool of part-time employed non-carers to estimate the differences between informal carers NILF owing to care provision and full-time employed non-carers and part-time employed non-carers, respectively. To estimate national costs, counterfactual records were selected from the pool of non-carers who were in the labour force (employed full-time and part-time and unemployed). All costs were expressed in 2020 Australian dollars ($A).

Analysis was undertaken in SAS, V.9.4 for Windows (SAS Institute, Cary, North Carolina, USA). The use of ABS data was approved by the Microdata Review Panel at the Australian Bureau of Statistics.

Patient and public involvement

The SDACs are nationally representative surveys undertaken by the Australian Bureau of Statistics. Patients and the public were not involved in the development of the microsimulation model.

Results

Within the base population of Care&WorkMOD, which uses the three SDAC data sets (SDAC 2003, 2009 and 2012), there were 110 119 records (unweighted) of those aged between 15 and 64 years old. Of whom, 295 were informal carers who cared for a person with arthritis and 67 of them were NILF due to providing care. Once weighted, these records represented about 11 000 informal carers NILF to provide care for someone with arthritis in 2015. By 2030, the number of informal caregivers NILF to provide care to those with arthritis is expected to increase by 21% to about 13 300 (online supplemental table 1).

Weekly income of carers and non-carers

In 2015, informal carers NILF to care for a person with arthritis received an estimated median weekly income of $A360, about 26% of the median weekly income of full-time employed non-carers ($A1397) and about 68% of the median weekly income of part-time employed non-carers ($A533). By 2030, the estimated median weekly incomes were projected to increase to $A395 for informal carers NILF to care for a person with arthritis compared with $A1640 and $A680 for full-time employed non-carers and part-time employed non-carers, respectively (table 1).

Table 1

Weekly total income, welfare payments and taxes paid by non-carers and informal carers who were not in the labour force (NILF) due to caring for someone with arthritis, Australian population aged 15–64 years, in 2020 $A

Carers of a person with arthritis who were employed full-time had median weekly incomes of $A1299 per week compared with $A1397 for those employed full-time but who were not carers (table 1).

The estimated median total weekly welfare payments received by informal carers NILF to provide care to those with arthritis were estimated to be relatively constant at $A299 from 2015 to 2030 (table 1).

Difference in weekly income of carers and non-carers

After adjusting for age, gender and the highest level of education attained, the estimated difference in the average weekly total income between employed full-time non-carers and informal caregivers NILF to provide care for someone with arthritis was $A1051 (95% CI: $A927 to $A1204) in 2015 and projected to increase by 22% to $A1283 (95% CI: $A1139 to $A1454) by 2030 (table 2). When compared with employed part-time non-carers, the difference in the average weekly total income in 2015 was estimated at $A340 (95% CI: $A257 to $A441), which was projected to increase to $A415 (95% CI: $A324 to $A521) by 2030. Informal carers NILF to provide care for those with arthritis received $A269 (95% CI: $A254 to $A281) more in weekly welfare payments compared with employed full-time non-carers in 2015, which was projected to rise to $A310 (95% CI: $A296 to $A321) in 2030, a 15% increase.

Table 2

Differences in average weekly income, weekly welfare payments and weekly tax payments between non-carer employed full-time or part-time and informal carers who were not in the labour force (NILF) due to caring for someone with arthritis, Australian population aged 15–64 years, in 2020 $A

Aggregated national annual loss of income to informal carers

The aggregated national annual loss of income to informal carers NILF to care for someone with arthritis was estimated at $A388.2 million in 2015 (95% CI: $A324.3 to $A461.9 million), rising to $A576.9 million (95% CI: $A489.2 to $A681.8 million) by 2030, an increase of about 48% (table 3; figure 2). The national annual loss in tax revenue that the Australian government would have received if those informal caregivers were in the labour force was estimated at $A99 million in 2015 (95% CI: $A77.9 to $A126.4 million), increasing to $A147.5 million (95% CI: $A118.1 to $A185.9 million) by 2030, a 49% increase. The total annual cost to the government of providing extra welfare payments to informal carers NILF to care for those with arthritis was estimated at $A119.9 million (95% CI: $A110.6 to $A128.5 million) in 2015 and is projected to increase by about 39% to $A167.1 million (95% CI: $A156.8 to $A176.4 million) by 2030. Note that costs from the carer and government perspective cannot be summed as this would result in double counting for welfare payments and taxes associated with carer income.

Figure 2

National annual costs due to primary carers of people with arthritis not being in the labour force, Australians 15–64 years old, in 2020 $A millions.

Table 3

National annual costs due to primary carers of people with arthritis not being in the labour force, Australians 15–64 years old, in 2020 $A millions

Discussion

Our study shows that the financial impacts on informal caregivers who leave the labour force early to provide care for someone with arthritis were substantial. Informal carers NILF to provide care for those with arthritis received a weekly income that was about 74% lower than that of those working full-time non-carers. When adjusted for age, gender and highest level of education achieved, the disparity in the average weekly income of informal carers NILF compared with full-time employed non-carers is substantial and expected to widen by 68% in 2030. In one study, the median weekly income of a person with arthritis who left the workforce early in Australia was estimated at $A260 in 2009, roughly five times less than the median weekly income of their healthy counterparts.9 When compared with patients who left the workforce because of arthritis, carers who left the workforce to care for a patient with arthritis were found to have a similar income loss.

To our knowledge, this is the first study projecting the economic impact on informal carers providing care for someone with arthritis. This is an important gap in the literature given its significant effects on household finances.41 42 In the few studies available, informal care was included as a small component of the indirect costs.5 43 44 An Australian study estimated the cost of providing informal care to a person with arthritis and other MSK conditions, such as dorsopathies (back pain) and osteoporosis at approximately $A1.21 billion in 2012.44 There is an increase in the interest of including the cost of informal care as part of the non-healthcare costs in health economic studies.25 45 These inclusions are particularly important in cost-of-illness studies as these are frequently used by policymakers and researchers to investigate the economic impact of a disease and to populate cost-effectiveness studies.45

Our study shows the number of informal carers leaving the labour force to provide care for people with arthritis is expected to rise from 2015 to 2030 (table 1). This finding is supported by other studies.25 46 In one Australian study, the number of informal carers in 2020 (674 000) is projected to increase by 16% by 2030.46 However, this does not meet the demand for informal carers in Australia, where it is projected to increase from 1.25 million in 2010 to 1.54 million by 2030.46 In another study, the prevalence of informal carers in European countries in 2010 is expected to increase up to 140% by 2030.25 In both these studies, the informal carers are not exclusive to those caring for a person with arthritis.

A strength of our study is the use of the Australian population representative data sets of informal carers and one of the best available to analyse the main reason for not being in the labour force. We used SDACs as one of the four data sets in the model. SDAC is a nationally representative data set of informal carers, individuals with a chronic condition or long-term disability and the elderly in Australia, with surveys held regularly every 3–4 years.30 The data on arthritis from SDAC is one of the best available and is used in other studies to estimate the cost of informal caregiving to a person with arthritis.5 43 Another strength is the use of the results from the well-established Australian microsimulation models such as APPSIM and STINMOD to project the sociodemographic distribution including the proportion of informal carers in the Australian population. We generated counterfactual simulated data sets using Monte Carlo methods and used them to estimate the degree of uncertainty in the results. One of the benefits of using this counterfactual analysis with Monte Carlo methods is that it does not require to have any assumption about the statistical distributions of the outcomes.

One limitation of our study is the lack of a breakdown by arthritis type in Care&WorkMOD. The SDACs, which comprise the main base data set of Care&WorkMOD, have one broad group for arthritis which includes OA and RA. Further, the relatively small sample size limits the capacity for subgroup analysis. This study included a small sub-subsample (n=67) of primary informal carers not in the labour force, as we focused on primary carers caring for a person with arthritis as their main condition only. Weights and projections were based on socioeconomic trends and numbers of carers but not arthritis specifically, due to the unavailability of such data. We were not able to compare the prior income or employment (including whether employment was full-time or part-time) of informal carers with their current employment or income as the SDACs are cross-sectional surveys and thus do not include this information. Further, our estimates of the economic costs of informal caring for arthritis do not include the costs associated with underemployment.

There is some evidence that within families where there are multiple potential carers, care more often falls to those on the lower income.47 To take account of this, we estimated carer counterfactual income based on people of the same age, education and sex. Thus, we capture, for example, that carers with a lower level of education are likely to have had a lower income prior to caring than those with a higher level of education. However, this methodology would not capture where carer income had been relatively lower than those within the same age, sex and education.

Informal carers of people with arthritis are economically worse off when compared with non-carers, with the difference in the average income received by these two groups expected to widen by 2030 and the losses to the government to grow substantially. Implementing effective interventions is important in reducing the burden of arthritis.48 The reduced burden of arthritis would also reduce informal caring demand for people with arthritis, enabling informal carers to continue working, thus reducing the associated economic costs, including increased government tax revenue and decreased welfare payments.

Data availability statement

Data are available upon reasonable request. Data from the Surveys of Disability, Ageing and Carers are available from the Australian Bureau of Statistics. Other data, such as Care&WorkMOD, STINMOD and APPSIM, are available from the corresponding author on reasonable request.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Not applicable.