Article Text
PDF
Abstract
Purpose To assess sickness absence and transitions from employment for employees with arthritis compared with employees without arthritis over time.
Methods We use 10 waves of the UK Household Longitudinal Survey (2009–2019). The sample (n=38 928) comprises employees aged 50 years to state retirement age. Arthritis was self-reported and could refer to people with conditions under the umbrella term ‘inflammatory arthritis’ or osteoarthritis (hereafter ‘arthritis’). Weighted random-effects multivariable linear probability models were estimated for two employment-related measures (1) sickness absence and (2) transitions from employment to: (a) unemployment; (b) long-term sick; (c) early retirement. These were regressed against a variable for arthritis and confounding factors (age, socioeconomic job classification, employing sector, year and additional health conditions). Additional analyses examined an interaction between the variable arthritis and these factors to test whether the effect of arthritis differs between these groups.
Results Employees reporting having arthritis were more likely to have sickness absence (1.35 percentage points greater rate (95% CI (0.92, 1.78)) and to transition to long-term sick (0.79 percentage points (0.46, 1.13)) and early retirement (0.58 percentage points (0.05, 1.11)). No effect was found for transitions to unemployment. There was limited evidence that the effects of arthritis vary for employees in different socioeconomic classifications.
Conclusions Employees living with arthritis have higher rates of sickness absence and greater rates of transitions from employment to long-term sick and early retirement. Further work could look at ways to quantify the implications for individuals, employers and the state and ways to alleviate the effects of living with arthritis on work participation.
- Arthritis
- Osteoarthritis
- Arthritis, Psoriatic
- Economics
Data availability statement
Data may be obtained from a third party and are not publicly available. Data are accessible via the UK Data Service under license (https://ukdataservice.ac.uk/deposit-data/sharing-experiences/understanding-society/).
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
The evidence base on the effects of living with different types of inflammatory arthritis, such as rheumatoid arthritis, psoriatic arthritis and osteoarthritis (hereafter ‘arthritis’) compared with people without arthritis, on work outcomes has been limited to cross-sectional associations that can bias the measurement of effect. Appropriate analysis of longitudinal (panel) data can help reduce bias to better understand the effects of arthritis on work outcomes. There are few studies comparing the effects of arthritis on employees compared with employees without arthritis.
WHAT THIS STUDY ADDS
The study provides evidence that employees living with arthritis are more likely to experience sickness absence and are at greater risk of exiting employment due to long-term sick and early retirement. The observed effect of living with arthritis does not differ depending on sex, age, employment type or socioeconomic job classification.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
The study highlights the detrimental effect living with arthritis can have on sickness absence and participation in the labour market. This indicates that employees living with arthritis may need more support to enable them to continue to participate in the labour market. Future research could explore the financial impacts to employees, employers and the state and how interventions can reduce the socioeconomic consequences for employees with arthritis, employers and society.
Introduction
Ageing populations and rises in the age of pension eligibility in many economies mean there are increasing rates of people in the labour market living with the effects of long-term health conditions. Inflammatory arthritis (including rheumatoid arthritis, ankylosing spondylitis and psoriatic arthritis) and osteoarthritis are part of a broader group of rheumatic musculoskeletal diseases (RMDs). In general, the prevalence of RMDs rises with age and therefore presents as increasing numbers among older age employees, currently 7% of UK employees are reported to have musculoskeletal diseases.1 2 RMDs are the first or second most common cause of work disability and the second most common cause of sickness absence in the UK.3 As the working age population becomes increasingly susceptible to RMDs, there is a need to understand whether support may be needed for employees living with arthritis and the economic arguments for state or employer intervention.
Focusing on the effect of adult-onset inflammatory and osteoarthritis (hereafter ‘arthritis’), studies suggest people living with one of those conditions have a high rate of sickness absence4–14 and are less likely to be in employment.7 12 13 15–23 The current evidence base understanding the effects of living with arthritis on employment is largely based on cross-sectional data,7 12 13 16 20–23 limiting the ability to identify causal relationships. Using these data, and associated analytical methods, may introduce bias and limit the inference on the effect of living with arthritis on employment. Few studies have been longitudinal15 17–19 24 and fewer assess the effects of living with arthritis on transitions from employment.15 17 18 24 Of these most are Scandinavian studies,17 24 one was done in Canada18 and one in England that focused on early retirement and any transition from employment.15 For sickness absence, all studies explore cross-sectional associations with the majority being focused in Scandinavian or USA and Canada.
The study aimed to identify the extent of the association between arthritis and work outcomes in people aged 50 years and over in employment. Work outcomes were defined as: transitions from employment to early retirement, unemployment or long-term sick, and the rate of sickness absence.
Methods
Study design
This was a comparative observational study using longitudinal (panel) data from a publicly available UK household survey.25
Ethical approval for the use of the data is covered by the University of Essex Ethics Committee. Participants gave informed consent to particpate in the survey before taking part.
Data
We use 10 (of the available eleven) waves of the UK Household Longitudinal Survey (data collection points spanning January 2009–May 2020).25 Analyses did not include data from wave 11 of the survey due to the impacts of the COVID-19 pandemic which altered the approach for survey data collection. Additionally, changes were made to the reporting of arthritis in wave 10.
The survey provides longitudinal data on the health, employment, income, education, family and social life of the UK population. The survey was constructed to provide a representative sample of households in the UK with an initial sample covering approximately 40 000 households, comprising data from 77 309 individuals.26 All individuals in households in the first wave were followed across subsequent waves, with the sample expanding when new households were formed from these individuals. Individuals aged 16 years and over completed an individual questionnaire generally via face-to-face interviews or a self-complete online survey.
The initial sample contained 444 181 individual-wave observations which included respondents from the main Understanding Society general population sample, an ethnicity minority boost sample, an immigrant and ethnic minority boost sample, the British Household Panel Survey (BHPS) sample and its regional booster samples. From the sample, we restrict to those that form part of the UKHLS in wave 1 (including the ethnic minority boost sample). We exclude BHPS samples. This reduced the sample to 289 768 individual-wave observations. Weights, contained in the UKHLS dataset, were applied to ensure representativeness of the sample as at wave 1.27 Restricting the sample to those who were present in wave 1 resulted in a sample of 273 612 individual-wave observations. With arthritis increasing in prevalence with age, and low prevalence in ages under 50 years, the available sample was further restricted by dropping individual-wave observations where the respondent is aged under 50 years of age, giving a sample of 139 871 individual-wave observations. Individuals could therefore be observed in later waves once they turn 50 years old and would need to be observed over at least two consecutive waves (in order to model changes in employment, detailed below).
Definition of arthritis
The UKHLS contains several measures of self-reported health conditions. Arthritis was defined based on responses to the question “Has a doctor or other health professional ever told you that you have any of these conditions?” where ‘arthritis’ is listed as one of the conditions. The type of arthritis is not identified, hence in the data the term ‘arthritis’ could refer to people with inflammatory arthritis (including rheumatoid arthritis, ankylosing spondylitis and psoriatic arthritis) or osteoarthritis. Respondents were asked this question in their first wave of observation, and subsequent waves prompted the respondent to report any new diagnoses and whether the respondent still has the condition. We assumed individuals reporting arthritis had arthritis in subsequent waves (treating the label of ‘arthritis’ as an absorbing state).
Transitions from employment measures
Those employed were identified based on self-reported labour market status. The UKHLS ‘jbstat’ variable contains responses to the question “Which of these best describes your current employment situation?” Employment was defined for those reporting ‘Paid employment (pt/ft)’ or ‘Self-employed’. The measures of labour market status used to model transitions include ‘Unemployed’, ‘Retired’ and ‘long-term (LT) sick or disabled’. The data enabled a longitudinal picture of employment status and any changes to unemployment, retirement or long-term sick (long-term sick or disabled) over time. Individuals can have multiple spells of employment in the data, for example, they can transition to unemployment and re-enter the employment sample should they return to employment.
From the sample of 139 871 individual-wave observations, there are individuals who, for some waves, were not employed in the labour market and for which the impacts of arthritis were not of focus in this study. These included: those reporting ‘On maternity leave’, ‘Family care or home’, ‘Full-time student’, ‘Government training scheme’, ‘Unpaid, family business’, ‘On apprenticeship’, ‘On furlough’, ‘Temporarily laid off/short term working’ and ‘Doing something else’. Restricting the sample to exclude these individual-wave observations reduced the available sample to 134 860 individual-wave observations.
Labour market status of ‘Retirement’ includes those eligible for state pension and the early retired. The study focused on ‘early retirement’ rather than retirement per se given transitions to retirement will be largely due to state pension age eligibility rather than arthritis. Individuals were eligible for a state pension at age 60 or 65 for women and men respectively; however, changes in legislation have meant this now varies depending on the year of birth.28 Changes have resulted in harmonisation in the retirement age for men and women to an equivalent age for those born after 6 December 1953 and an increase in retirement age from 65 to 66 for those born between 6 December 1953 and 5 April 1960. We identified early retirement in the sample by removing respondents once they are eligible for state pension in England. This approach meant that the ‘Retirement’ status comprised only those individuals reporting retirement prior to being eligible for state pension. Restricting the sample to those aged below the state pension age resulted in a sample of 61 542 individual-wave observations. The remaining respondents reporting ‘Retirement’ were assumed to reflect early retirement. Early retirement is possible under private pensions and depends on the pension available, this can sometimes be from age 55.29 It could also be possible that respondents self-identify as being in retirement should they self-fund their retirement prior to state pension. It was not possible to distinguish between those able to call on their pension early and those self-funding their early retirement.
Modelling transitions
Transitions (or ‘changes’) were defined for those in employment by comparing labour market status in the proceeding wave of the survey. Three types of transitions (changes) were modelled representing movement from employment to: unemployment; long-term sick; early retirement. Having constructed indicators for changes in labour market status in the proceeding wave, wave 10 was omitted resulting in a sample of 56 220 individual-wave observations containing those employed, early retired, unemployed or long-term sick. Reducing the sample to exclude those not in employment, having generated the indicator for a transition in the proceeding wave, resulted in a sample of 39 806 individuals in employment.
Sickness absence
Sickness absence was measured for those in employment, where respondents reported ‘Yes’ to ‘Did not do paid work last week but does have a job’. For these respondents, the survey includes a subsequent question, “What was the main reason you were away from work last week?” The reasons for absence include: ‘Maternity/Paternity leave’, ‘Other leave/holiday’, ‘Sick/injured’, ‘Attending training course’, ‘Laid off/on short time’, ‘Other personal/family reasons’, ‘Furloughed’ and ‘Other reasons’. We restricted sickness absence to absences reported as being due to ‘sick/injured’.
Confounding factors
To model the effect of living with arthritis on employment it was important to account for potential confounding related to factors associated both with arthritis and the (work) outcomes of interest. To mitigate potential bias here, we adjusted for three contextual factors: sex (male, female); age (50–54 years; 55–59 years; over 60 years); ethnicity (white; other). Please note respondents are asked their sex and not gender in the questionnaire. In addition, we adjusted for the year of the interview (2009–2019) and additional health conditions. Respondents were asked if they had any health conditions based on a predefined list of health conditions at the first interview and for any changes in subsequent interviews. From the range of conditions noted from responses to the question “Has a doctor or other health professional ever told you that you have any of these conditions?” (see online supplemental table 1), a count measure was generated which was then grouped into a number of additional health conditions (0; 1; 2, or more).
Supplemental material
Further controls included employment-specific measures. The UKHLS asked those in employment “Do you work for a private firm or business or other limited company or do you work for some other type of organisation?”, we generated a binary indicator for the public sector, where ‘other type of organisation’ was reported versus ‘Private firm or business, a limited company’. Socioeconomic status was inferred from the National Statistics Socio-economic Classification provided by the UKHLS for those employed, the five-category version included: ‘Management and professional’; ‘Intermediate’; ‘Small owners and own account’; ‘Lower supervisory and technical’ and ‘Semiroutine and routine’.30
Analysis
Separate multivariable weighted linear probability regression models were estimated for sickness absence and transitions (changes) from employment to: unemployment; long-term sick; early retirement. The main explanatory variable of interest was the binary indicator for the self-reported presence of ‘arthritis’ (yes, no). The ability to identify proceeding changes in labour market activity helps mitigate the potential endogeneity concerns of labour market status being a determinant of arthritis. Additional explanatory variables that may be associated with arthritis were included in the regressions to reduce potential confounding on the estimated arthritis effect. Additional variables were as follows: age; gender; ethnicity; presence of additional health conditions; job-specific measures (job sector and employing sector). The panel nature of the data permits accounting for unobserved heterogeneity (random effects), this reduces bias caused by unobserved factors related to individuals in the sample that correlate with sickness absence or transitions.
Sensitivity analyses
Three sensitivity analyses were performed to understand the potential impact of varying the methodological approach on the observed results from the main analyses. The first sensitivity analysis compared linear probability models with the non-linear models to explore whether modelling sickness absence and employment transitions in a non-linear approach impact quantitatively and qualitatively the findings. The second sensitivity analysis added an interaction term to account for the joint effect of arthritis with the confounding variables in the model to test whether the impact of arthritis differs between these. The third sensitivity analysis explored the effect of attrition from completing the survey. Since the analysis conducted is longitudinal, the sample may not remain representative beyond wave 1, meaning the cross-sectional weights applied may not be sufficient to ensure representativeness. To test how sensitive the results are to this, we restricted the analyses to those respondents who responded in each wave of the survey and applied respective longitudinal weights contained in the UKHLS to ensure the representativeness of the sample.27
Results
Of the 39 806 individual-wave observations, missing data in any variable amounted to 878 individual-wave observations (2.21%). Missing data were treated as missing at random. Removing those individuals with any missing responses from survey questions reduced the sample to 38 928 (table 1).
Characteristics of the sample population overall and stratified by arthritis
In the final estimation sample (n=38 928), 16.73% reported arthritis (6514, table 1). The sample illustrates how those employees living with arthritis were more likely to be female, in older age groups, have additional health conditions and be in lower socioeconomic classification groups. These differences highlight the need to account for these factors in the regression analyses, as omission could bias the estimated effect of arthritis.
In the sample, transition from employment was seen in: 3.10% transition to early retirement; 1.79% to unemployment; 0.65% to long-term sick over the period. A total of 1.33% of individuals reported sickness absence due to health.
Characteristics of those reporting sickness absence and those transitioning from employment are provided in table 2. Employees with arthritis appear to have greater rates of transitions to early retirement (3.84% vs 2.96%), unemployment (2.06% vs 1.73%), long-term sick (1.38% vs 0.50%) and sickness absence (2.64% vs 1.07%).
Characteristics of the sample population that are off sick or transition from employment
The unadjusted relationship between arthritis and sickness absence and transitions are provided in table 3. The estimated coefficients are percentage point differences and are interpreted in relation to the base category. For example, the 0.0155 estimate in the sickness absence model for those with arthritis is interpreted as those with arthritis having 1.55 percentage points higher rate of sickness absence compared with those without arthritis (this is similar to the 1.57 percentage point difference seen in table 1 where the sickness absence rate was 2.64% for those with arthritis and 1.07%, a difference of 1.57 percentage points, for those without). The 95% CIs give the range of values where we can be 95% confident the true value lies between, giving a 5% possibility of error. Returning to the example, this means we can be 95% confident that the true effect of arthritis lies between 1.13 and 1.96 percentage points. If the 95% CIs include 0, then this is interpreted as an insignificant effect, as we are not confident there is a difference between the two groups. In the unadjusted analyses living with arthritis was associated with a higher proportion of: reported sickness absence (1.55 percentage points; 95% CI (1.13, 1.96)); transition to early retirement (0.87 percentage points; 95% CI (0.34, 1.41)); transition to long-term sick (0.77 percentage points; 95% CI (0.49, 1.06)).
Weighted linear regression model estimates for the effects of arthritis on (1) labour market sickness, (2) transitions to early retirement, (3) transitions to early long-term sick and (4) transitions to unemployment.
Table 4 presents the estimated effects for the adjusted analyses where weighted random-effects linear regression models were estimated. Living with arthritis was associated with a higher proportion of sickness absence (1.35 percentage points; 95% CI (0.92, 1.78)) and transition to: early retirement (0.58 percentage points; 95% CI (0.05, 1.11)); long-term sick (0.79 percentage points; 95% CI (0.46, 1.13)). There were signals that arthritis was positively associated with transitions to unemployment though this was statistically insignificant at a 5% significance level (95% CI (−0.06, 0.81)).
Weighted random-effects linear regression model estimates for the effects of arthritis and confounding measures on (1) labour market sickness, (2) transitions to early retirement, (3) transitions to long-term sick and (4) transitions to unemployment.
Sickness absence was greater for those with additional health conditions, those employed in the public sector and for lower socioeconomic classification groups (table 4). There were differences in the factors associated with transitions out of employment (table 4). Females were more likely to transition to long-term sick than males and less likely to transition to unemployment, but no statistically significant difference was found for early retirement transitions. Older people were more likely to transition to early retirement and long-term sick, but no effect of age was found for unemployment transitions. Minority ethnic groups were more likely to transition to unemployment and less likely to transition to early retirement (no significant effect was found for long-term sick). While those employed were more likely to transition to early retirement than those who reported being self-employed, no significant effect was found for unemployment or long-term sick transitions. Those with additional health conditions were more likely to transition to unemployment and long-term sick, but no significant effect was found for early retirement transitions. In terms of employer type, public sector employees were more likely to retire early and less likely to move into unemployment or long-term sick. All socioeconomic classification groups were less likely to transition to early retirement than employees in management and professional classifications. Those in the semiroutine and routine classifications were more likely to transition to unemployment and long-term sick.
The estimated effects of the calendar year for each are presented in online supplemental figure 1. The sharp decrease in all measures in 2019 is a reflection of the small volume of respondents completing wave 9 of the survey in that year, with the majority of wave 9 interviews conducted in 2018.
Sensitivity analyses
Estimates from non-linear (logistic) and linear regressions were compared to assess whether estimating linear models for binary measures of sickness absence and transitions could impact the findings (table 5, full set of results in online supplemental table 2). The samples differed here due to the omission of individuals in 2019 (where failure is predicted perfectly). The results were qualitatively and quantitatively similar. Where positive associations were found in linear regressions, odds greater than 1 (positive associations) were found in logistic regressions. Most notably greater sickness absence and transitions to long-term sick and early retirement for those with arthritis were found in either linear or non-linear estimated effects.
Random-effects weighted linear regression and weighted non-linear (logistic) regression model estimates for the effects of arthritis and confounding measures (not reported) on (1) labour market sickness, (2) transitions to early retirement, (3) transitions to long-term sick and (4) transitions to unemployment
Additional analyses tested whether the effects of arthritis differed between the observed personal characteristics and job- and employment-related groups. Here, arthritis was interacted with the additional confounding variables, and the interaction term provides an estimate of any difference in the effect of arthritis. The interaction models are provided in online supplemental tables 3–6. There was no evidence of any statistically significant difference in the effects of arthritis by gender, age, ethnicity, employment type, employer type, additional health conditions or socioeconomic classification groups for all transitions. For sickness absence, a significant greater risk of sickness absence was found for semiroutine and routine classification employees compared with employees in other classifications with arthritis (online supplemental table 3).
Third, as the analyses conducted is longitudinal, we restricted the analyses to those respondents who responded in each wave of the survey and applied respective longitudinal weights contained in the UKHLS to ensure the representativeness of the sample.27 The sample was reduced to 27 483 individual-wave observations by incorporating this restriction. Table 6 provides estimates from regressions where the longitudinal weights were applied and the cross-sectional weight applied on the same sample (full sets of estimates are provided in online supplemental table 7). There were no substantial differences between the weighting approaches in the estimated effects of arthritis on sickness absence and either transition model, under either weighting approach the inference of the effects of arthritis is qualitatively the same—the weighting approaches do not impact on inference with regards to statistical significance or the direction of the relationship (whether arthritis has greater or less rates for the measure being modelled).
Cross-sectional weighted and longitudinal weighted random-effects linear regression model estimates for the effects of arthritis and confounding measures on (1) labour market sickness, (2) transitions to early retirement, (3) transitions to long-term sick and (4) transitions to unemployment.
Discussion
The results from this study indicate that employees aged 50 and over with arthritis were more likely to have sickness absence and more likely to transition from employment to long-term sick and transition to early retirement compared with those without arthritis. No effect of arthritis was found across models for transitions to unemployment. The results were robust to whether linear or non-linear models were estimated, and to whether longitudinal rather than initial cross-sectional weights were applied.
The effect of arthritis did not vary between personal characteristics or job- or employment-related groups. An outlier was found for those in semiroutine or routine occupations where arthritis was associated with a relatively greater sickness absence rate than those with arthritis in other occupations. These are occupations which tend to be more physical and where accommodations are not as common as in other occupational areas, from which the capability to mitigate arthritis symptoms is less likely.
This study focused on arthritis, but the analyses also highlight the differential impact that employer type and socioeconomic classification have on sickness absence and employment transitions more generally. Of note, public sector employees and those in less skilled classifications were specific subgroups in the sample. Public sector employees were found to have higher rates of sickness absence, higher rates of transitions to early retirement and lower rates of transitions to long-term sick or unemployment. Those in less skilled sectors had higher rates of sickness absence, lower rates of transitions to early retirement and higher rates of transitions to long-term sick and unemployment. Minority ethnic groups were found to have higher rates of transitions to unemployment and lower rates of transitions to early retirement. These effects suggest there could be inequalities in the labour market with public sector working seemingly having a protective factor against transitions to unemployment and long-term sick and early retirement being a relatively more accessible option for employees who are higher skilled, white employees and employees who are not self-employed. Research is needed to understand the drivers for these differences which may be due to a range of factors including cultural factors, family issues and financial circumstances.
Although not a focus of this study, the estimated changes in sickness absence and transitions over calendar years are broadly in line with national trends over the period 2009–2018 in sickness absence,31 workers aged 50–69 years old moving from employment into retirement,32 working-age disability benefit claiming33 and unemployment rates.34
Like previous Scandinavian and US studies,4–14 35 we find people living with arthritis have double the rate of sickness absence. The findings also concur with previous studies that those reporting arthritis are more likely to transition to long-term sick,17 18 24 more likely to transition to early retirement,15 and that transitions to unemployment are indifferent.17 Niederstrasser et al15 assessed transitions from employment in England for those with musculoskeletal pain (a broader group than arthritis as studied here), finding 1.25 greater odds of early retirement and slightly smaller odds for movements out of employment. We find slightly smaller odds for early retirement. We build on the study by Nielderstrasser et al by exploring a wider range of transitions finding greater transitions to long-term sick but no difference in the likelihood of transitions to unemployment. We also explored whether the effects of arthritis may differ by job sector and employment sector finding limited evidence that those with arthritis are impacted differentially across the job sector and employment sector.
The study has a number of strengths and weaknesses. The study used data from a representative sample of the population, exploited the panel nature of the data to model the effect of arthritis in the proceeding wave of the survey and found results that were largely robust to model specification and approach to weighting. The study used a range of personal and employment-specific characteristics to control for factors that may bias the relationship between arthritis and employment measures. In the data it was possible to identify early retirement and sickness absence for health reasons, limiting potential bias caused by state pension eligibility and absence from work for non-health reasons.
The study, while from a large sample of the UK population, focused on adults who were employed and aged 50 to retirement age. This was due to low arthritis prevalence in earlier age groups and eligibility for state pension once an individual reaches state pension age. This is an important population with current plans to extend working lives to mitigate the impact of increasing life expectancies. The measure of sickness absence was limited to absence within the past week, which may be a poor measure (underestimate) of sickness absence prevalence over the full wave. Though the analyses find little evidence of differential effects of arthritis by a range of different socioeconomic classification groups, the approach was limited by small samples, which may result in a lack of statistically significant effects when there was one. Comorbidity was captured using only a count of additional health conditions, and interactions between arthritis and specific health conditions may have highlighted particular at-risk groups of employees though sample size becomes a limiting factor here. There may be other confounding factors that have not been controlled for in the analyses. Missing data were assumed to be missing at random, this may not have been the case; however, the percentage of the sample omitted due to missing data amounted to only a small proportion (2.21%). Finally, the measure of arthritis is self-reported which may introduce bias due to under-reporting or over-reporting and it was not possible to differentiate different forms of arthritis in the dataset.
The findings provide several suggestions for future policy. Those aged 50 and in employment with arthritis have greater rates of sickness absence and transitions to long-term sick and early retirement. Supporting a workforce resilient to the impact of arthritis requires initiatives to support employees with arthritis in the workplace. These could include healthy workplaces and supportive attitudes from line managers and employers, early workplace interventions (eg, adaptations, flexible working), access to physiotherapy and/or psychosocial and rehabilitation programmes.36 37 The findings suggest initiatives may reduce sickness absence (at the benefit to employees and employers) and reduce dependency on state transfers related to long-term sick benefit claiming (at the benefit to the state), staff turnover (at the benefit to employers) and support employees to realise the benefit of employment.38
Conclusion
Living with arthritis results in higher rates of sickness absence and greater rates of transitions from employment to long-term sick and early retirement. Further work could look at ways to quantify the implications for individuals, employers and the state and to alleviate the effects of living with arthritis on work participation.
Data availability statement
Data may be obtained from a third party and are not publicly available. Data are accessible via the UK Data Service under license (https://ukdataservice.ac.uk/deposit-data/sharing-experiences/understanding-society/).
Ethics statements
Patient consent for publication
Ethics approval
The Understanding Society study protocols and research programme are scrutinised by a number of research ethics committees to assure that ethical and legal obligations are respected at all times. Ethics Committee approval for the use of the data is covered by the University of Essex Ethics Committee. Details can be found online (https://www.understandingsociety.ac.uk/documentation/mainstage/user-guides/main-survey-user-guide/ethics/). Participants gave informed consent to participate in the survey before taking part.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
X @HealthEcon_Will
Presented at Part of this data has been published as an abstract by the Annals of the Rheumatic Diseases and presented as a poster at the EULAR conference 2023. Reference: Higgerson J, Whittaker W, Eden M, Payne K, Verstappen S. POS0569 Identifying the impacts of arthritis on labour market participation: an observational study in Britain [abstract]. Ann Rheum Dis. 2023; 82 (Suppl 1). DOI: 10.1136/annrheumdis-2023-eular.4178
Contributors WW designed the study, analysed the data, interpreted the results and wrote the manuscript. WW is the guarantor for the manuscript. JH designed elements of the study, analysed the data, interpreted the results and revised the manuscript. KP and SMMV designed elements of the study, interpreted the results and revised the manuscript. ME and RW interpreted the results and revised the manuscript.
Funding The research was funded by the Centre for Musculoskeletal Health and Work, Versus Arthritis (Award Number: N/A) and supported by the NIHR Manchester Biomedical Research Centre (Award Number: NIHR203308).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.