Article Text

Consequences of inflammatory arthritis for workplace productivity loss and sick leave: a systematic review
  1. Marie-Louise B Lenssinck1,2,
  2. Alex Burdorf1,
  3. Annelies Boonen3,
  4. Monique A Gignac4,
  5. Johanna M W Hazes2,
  6. Jolanda J Luime2
  1. 1Department of Public Health, Erasmus MC, University Medical Centre Rotterdam, Rotterdam, The Netherlands
  2. 2Department of Rheumatology, Erasmus MC, University Medical Centre Rotterdam, Rotterdam, The Netherlands
  3. 3Department of Internal Medicine, Division of Rheumatology, University Hospital Maastricht and Caphri Research Institute, Maastricht, The Netherlands
  4. 4Arthritis Community Research & Evaluation Unit, Toronto Western Research Institute and Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
  1. Correspondence to Dr Jolanda J Luime, Department of Rheumatology, Erasmus MC, PO Box 2040, 3000 CA, Rotterdam The Netherlands; j.luime{at}erasmusmc.nl

Abstract

Objective To review the occurrence and magnitude of workplace productivity loss and sick leave in inflammatory arthritis (IA) patients and to identify determinants.

Methods PubMed, EMbase, PsycINFO and CINAHL articles to July 2012 on IA and workplace productivity loss or sick leave were reviewed. Methodological quality was assessed by a criteria list developed by the authors.

Results 47 original studies were identified. The occurrence of sick leave in IA patients varied from 3.7% in the past 4 days to 84% in the past 2.5 years. Total duration of sick leave ranged from 0.1 to 11 days over 1 month. Pain and functional disability were associated with sick leave and workplace productivity loss. About 17%–88% of IA patients experienced workplace productivity loss, four studies investigated determinants. Tumour necrosis factor inhibitors were associated with reduced workplace productivity loss and sick leave.

Conclusions IA impacts worker productivity, but its estimated magnitude varies. Higher levels of sick leave and workplace productivity loss were reported for increased levels of pain and decreased levels of functional ability.

  • Arthritis
  • Social work
  • Rheumatoid Arthritis

Statistics from Altmetric.com

Introduction

Inflammatory arthritis (IA) is characterised by pain and stiffness of the joints that limits normal movement and reduces participation in work and leisure time activities. Limitations in activities and restrictions in participation have important implications for an individual's life. As a consequence, IA leads to substantial direct costs in terms of healthcare utilisation and to large indirect costs due to work disability, sick leave and early retirement.1 ,2

From the onset of IA onwards, the consequences for participation are substantial. Around a third of rheumatoid arthritis (RA) patients already leave the workforce within 2–3 years after disease onset.3 ,4 Most studies, however, have focused on permanent work disability among patients with established RA (ie, arthritis-related unemployment, early retirement or disability pension) and its underlying risk factors.5–8 Two reviews showed increased risks of work disability with higher physical job demands, higher disability using the Health Assessment Questionnaire (HAQ), older age and lower education.6 ,7 There is some evidence that these factors are also related to sick leave, defined by temporary absence from work due to illness, and workplace productivity loss, characterised by a reduced performance while at work due to illness.5 ,9 The evidence on determinants of workplace productivity loss and sick leave is limited in contrast to work disability.5

In order to prevent patients with IA becoming work disabled, it is important to identify factors associated with sick leave and workplace productivity loss early in the disease course. Insight into modifiable causes of workplace productivity loss and sick leave could contribute to the development of effective interventions aimed at reducing sick leave and job loss. In this systematic review of the literature we aimed to describe the occurrence and magnitude of workplace productivity loss and sick leave in IA patients and to identify individual, disease and work-related determinants of workplace productivity loss and sick leave.

Materials and methods

Literature search

Studies on workplace productivity loss and sick leave among persons with IA were identified by a comprehensive search in PubMed, EMbase, PsycINFO and CINAHL databases from database inception to July 2012 (see search strategy in online supplementary file S1). Articles were limited to English, French, Dutch or German language studies. The reference section in review articles and original studies were searched for additional studies.

Study selection

We retained case control studies, cross-sectional studies, cohorts and clinical trials in which subjects were workers who had IA, that is, RA, psoriatic arthritis, ankylosing spondylitis, undifferentiated arthritis or early arthritis. If studies included individuals with other types of arthritis (eg, osteoarthritis), they were included only if data on IA could be extracted separately. We limited inclusion to studies in which workplace productivity loss or sick leave was quantified. Qualifying definition for sick leave was any paid or unpaid temporary absence from work because of illness or injury in those with a paid job. The definition for workplace productivity loss was being at work, but performing below par due to illness. Experimental studies (ie, clinical trials) were specifically included in this review to understand the potential effect of drug treatment on workplace productivity loss and sick leave.

One reviewer (ML) screened titles and abstracts of identified articles for eligibility. The reviewer was not blinded with regard to authors, journal and institution. In case of doubt regarding a study's eligibility, a second reviewer made the final decision (JL or ABu). All potentially relevant studies were retrieved as complete manuscripts. Second, the selection criteria were applied on full text of all articles, which had passed the first eligibility screening.

Data abstraction and quality assessment

One reviewer (ML) systematically abstracted data from each study on: patient demographics, diagnosis, disease duration, study design, occurrence and duration of workplace productivity loss and sick leave and its determinants. Data abstraction was independently checked for all reports by a second reviewer (ABu).

Studies were evaluated for susceptibility to bias using a quality assessment list developed by the authors, based on recommendations of Sanderson et al.10 The constructed checklist covered 10 items related to selection methods, measurement of study variables, design-specific sources of bias, control for confounding and appropriate use of statistics (see online supplementary file S2). Each item could be scored ‘yes’, ‘no’ or ‘unclear’. A quality score was derived for each study by assigning one point for each quality measure. Each study was independently rated by two reviewers (JL and MG or JL and AB). In cases of a disagreement between the two reviewers, consensus was used to resolve disagreement. The German articles were assessed by one author (ABu).

Analysis

Data were abstracted from original studies on sick leave days and recalculated to mean days of sick leave per month over follow-up to ease comparison between studies. The descriptive tables highlighted substantial heterogeneity in definitions of sick leave and productivity loss at work as well as its determinants, so this precluded combining data in a meta-analysis. Instead, findings were summarised descriptively across studies. Determinants were classified into individual, disease-related, work-related and treatment factors.

Best evidence synthesis

The information was synthesised into: (a) limited evidence: only one study available; (b) consistent evidence: two or more studies available of which at least 75% of the studies find a significant association of a factor with workplace productivity loss or sick leave in the same direction; and (c) inconsistent evidence: remaining situations. The evidence synthesis was based on multivariate associations when available or univariate associations when multivariate associations were unavailable.

Results

Selection

In all, 56 publications reporting on 47 original studies out of 986 titles and abstracts met our inclusion criteria.3 ,9 ,11–64 Initially 170 out of 986 articles were retrieved for further consideration (see flow diagram of the selection process in the online supplementary file S3). A total of 36 articles did not assess workplace productivity loss or sick leave; in 35 articles data could not be extracted separately on diagnosis or employed versus non-employed persons and another 45 articles were not full original publications.

Five of the original studies were published in more than one paper. Puolakka et al reported data from the same study population in five published articles of which two were used here.41 ,45 Geuskens et al9 presented cross-sectional and longitudinal data of the same patients,19 while the articles of Mau et al33 ,34 and Merkesdal et al 36 presented data on work incapacity and work disability in separate papers. Also Zhang et al59 ,61 and Bansback et al12 reported data from the same study population in three different articles. One article by Grazio was written in Croatian and therefore excluded. Three additional articles were retrieved via screening of references.18 ,32 ,38

Studies on occurrence and duration of sick leave are presented in tables 1 and 2 and on occurrence and level of workplace productivity loss in tables 3 and 4. Findings with respect to determinants of sick leave and workplace productivity loss are given in tables 5 and 6.

Table 1

Reported occurrence and duration of sick leave among persons with arthritis in paid employment in observational studies

Table 2

Reported occurrence and duration of sick leave among persons with arthritis in paid employment in clinical trials

Table 3

Reported occurrence and magnitude of workplace productivity loss among persons with arthritis in paid employment in observational studies

Table 4

Reported occurrence and magnitude of workplace productivity loss among persons with arthritis in paid employment in clinical trials

Table 5

Determinants of sick leave among persons with arthritis in paid employment multivariate regression analysis; OR (95% CI) unless otherwise indicated by %

Table 6

Determinants of workplace productivity loss among persons with arthritis in paid employment

Study populations

Of 47 original studies, 42 included RA patients,3 ,11–19 21–24 ,26 ,29–40 ,45 ,46 ,48–53 55–64 four included psoriatic arthritis patients,20 ,25 ,27 ,30 two included early arthritis patients9 ,19 ,64 and three included patients with sponyloarthropathies.28 ,30 ,47 In 21 studies, RA patients satisfied the 1987 American College of Rheumatology criteria,11 ,13 ,15–17 ,22 ,23 ,26 ,32–36 ,41 ,45 ,46 ,48 ,51 ,52 ,54 ,56 ,60 ,62 in 15 studies patients were clinically diagnosed12 ,14 ,18 ,21 ,24 ,29 ,30 ,37–39 ,48 ,53 ,57–59 ,61 ,63 and in four studies patients satisfied the ICD-9 code.40 ,49 ,50 ,55 A total of 11 studies reported on early RA (<2 years).3 ,11–15 ,23 ,33 ,34 ,36 ,41 ,45 ,54 ,56 ,59 ,61 ,62 There were 11 clinical trials,11 ,13 ,25–27 ,32 ,45 ,51 ,54 ,56 ,60 19 cross-sectional articles,3 ,15–17 ,19 ,24 ,31 ,33–35 ,47 ,48 ,50 ,52 ,55 ,57–59 ,63 15 prospective cohort articles,9 ,18 ,20 ,22 ,23 ,29 ,30 ,36–38 ,46 ,53 ,61 ,62 ,64 two case control studies14 ,40 and four retrospective cohort studies.21 ,28 ,39 ,49

Occurrence of sick leave

Overall, 44 of 47 studies (94%) reported information on sick leave. A total of 34 of these were cross-sectional studies or longitudinal cohorts and 10 were clinical trials (see tables 1 and 2). Definitions of sick leave ranged from ‘any absence from work’ to ‘absences of 2 weeks or more’.

Self-constructed questionnaires to obtain sick leave data were used in 24 studies (51%),9 ,11 ,13 ,16 ,18 ,19 ,22 ,23 ,25–27 ,29 ,30 ,32 ,37 ,46–48 ,50 ,51 ,55 ,56 ,62–64 while nine studies used standardised questionnaires,3 ,12 ,15 ,20 ,24 ,52 ,53 ,59–61 seven studies (15%) used registries,14 ,28 ,38–41 ,45 ,49 one study reported physician-certified data33 ,34 ,36 and two studies conducted interviews to obtain sick leave data.17 ,21

The proportion of patients with sick leave was reported in 28 studies (65%),3 ,9 ,14–17 ,19 ,21–23 ,25 ,28 ,30 ,33 ,34 ,36–40 ,46 ,48 ,49 ,51–53 59–64 while the number of sick leave days was determined in 30 studies.3 ,11 ,13 ,16 ,18 ,20–24 ,26 ,27 ,29 ,32–34 36–40 ,45 ,47 ,49 ,50 ,52 ,55 ,56 ,60–63 The number of sick leave days per month varied from 0.121 to 11 days per patient33 ,34 in the observational studies. Authors tend to present on sick leave due to any cause,9 ,12 ,14–16 18–20 ,23 ,28–30 38–40 ,49 ,50 ,53 ,55 ,59 ,61 ,62 sick leave due to arthritis3 ,17 ,21 ,22 ,24 ,33 ,34 ,36 ,37 ,46 ,47 ,52 ,63 ,64 or both.48 Any sick leave ranged from 0.7 to 8.7 days per month and sick leave due to arthritis ranged from 0.1 to 11 days per month. In all 17 of 23 articles reported less than three sick leave days per month. In the clinical trials the number of sick leave days per month varied from 0.332 to 4.526 per patient. Any sick leave ranged from 1.0 to 4.5 days per month and sick leave due to arthritis ranged to 0.3–2.8 days per month. The proportion of persons who were on sick leave ranged from 3.7% to 84%,36 largely because recall periods ranged from 4 days to two-and-a-half years. There was no difference between early RA and established IA in proportion of persons who were on sick leave.

Occurrence of workplace productivity loss

In all, 17%–88% of IA patients experienced workplace productivity loss, reported in 109 ,11 ,15 ,25 ,37 ,54 ,56–58 ,60 of 20 studies.9 ,11 ,15 ,21 ,24–26 ,31 ,32 ,37 ,47 ,52 ,54 ,56–61 Of these 20, seven were clinical trials and 13 were cross-sectional or cohort studies (see tables 3 and 4). Definitions of workplace productivity loss varied and included ‘reduced productivity on the last working day’ to ‘productivity lost in the past year’. Productivity loss was measured by self-constructed questionnaire,26 Visual Analogue Scale questions21 ,25 ,32 ,37 ,56 and standardised questionnaires, such as the Health Limitation Questionnaire15 ,52 ,60 and Work Limitation Questionnaire.11 ,47 ,54 ,57 ,58

Determinants of sick leave

Of 43 studies examining sick leave, eight studies (21%)9 ,12 ,14 ,19 ,22 ,33 ,35 ,41 ,56 reported determinants of sick leave (see table 5). Multivariate logistic regression analysis9 ,12 ,19 ,22 ,33 ,35 ,41 was used in seven articles, while univariate analysis56 and variable influence on projection (VIP)14 were both used in one article to identify variables contributing most to the model. The VIP statistics represents the predictive power of (sets of) variables and a variable with a VIP > 0.8 was considered a relevant predictor for sick leave.

Regarding individual factors, inconsistent evidence was found for the association among older age, lower levels of education, female sex and sick leave. Four14 ,22 ,33 ,41 of nine articles showed that older age was related to more sick leave. Lower educational level was associated with sick leave in two studies14 ,41 and was not significantly associated in three other articles.9 ,19 ,35 In seven out of nine studies female sex was not related to greater sick leave, while one paper showed an increased risk35 and another article a decreased risk.19

The most common disease factors measured were erythrocyte sedimentation rate, disease duration, pain, number of swollen joints and functional disability. Consistent evidence was found for higher values of pain and greater functional disability, measured using the HAQ, lower physical functioning in the Short Form 36 Health Survey9 ,19 and disease severity (ao DAS28)35 on sick leave. Consistently no relationship was found for higher erythrocyte sedimentation rate and swollen joints. Inconsistent evidence was found for the following factors associated with sick leave: disease duration,9 ,12 ,19 ,22 ,33 ,35 ,56 erosive disease, fatigue,9 ,19 ,56 any comorbidities,9 ,19 passive coping by decreasing activity to cope with pain,19 and limited evidence was found for lower mental health56 and patient's and physician's global assessment.41

Work-related factors were assessed in eight9 ,12 ,14 ,19 ,22 ,33 ,35 ,41 of nine articles. Inconsistent evidence was found for the association between heavier type of work,12 ,14 ,22 ,33 ,35 ,41 frequent manual materials handling9 ,19 and sick leave. Regarding psychosocial aspects, limited evidence was found for the association among frequent time pressure at work,33 low job control19 and sick leave. There was also limited evidence for the relation between sick leave with being a supervisor9 ,19 and frequent overhead work.33

Pharmaceutical treatment was measured in two determinant studies and showed inconsistent evidence for the association with a shorter duration of sick leave.41 The role of antitumour necrosis factor (anti-TNF) agents on sick leave was evaluated in eight clinical trials11 ,13 ,25–27 ,51 ,56 ,60 and one cohort study.37 All showed a reduction in sick leave due to anti-TNF agents. However, only one of these studies56 included other potential determinants such as individual and disease-related factors.

Determinants of workplace productivity loss

Four studies measured determinants of workplace productivity loss (see table 6),9 ,12 ,47 ,56 while six trials evaluated the effect of treatment on workplace productivity loss.

Regarding individual factors, limited evidence was found for the association between older age and magnitude of workplace productivity loss.12 ,47 ,56 There was no significant association between female sex and workplace productivity loss.9 ,12

Disease factors were measured in all four studies. Consistent evidence was found for correlations among higher functional disability (HAQ), higher levels of pain,9 ,47 poor physical functioning and poor mental health (Short Form 36 Health Survey),9 ,12 fatigue47 ,56 and workplace productivity loss. Limited evidence was found for associations among a higher DAS28 score,56 different specific ankylosing spondylitis measures and increased workplace productivity loss.

Work-related factors were measured in two studies.9 ,12 Limited evidence was found that low support from colleagues9 and high physical demands12 will decrease workplace productivity and that jobs with an influence on scheduling one's own activities decreased productivity loss.12

Five clinical trials11 ,25 ,26 ,56 ,60 and one cohort study37 evaluated pharmaceutical treatment by means of the role of anti-TNF agents. Reduced workplace productivity loss was shown in all of these studies, although only one study56 also looked at individual and disease-related factors.

Methodological quality assessment

Details on the methodological quality are presented in the online supplementary file S4. Overall, 43 of 50 articles (86%) provided a clear sampling frame.9 ,11–17 19–23 25–27 29–36 38–41 45–58 In all, 19 articles (38%) had a patient participation of more than 80% or between 60% and 80% and reported the reasons for dropout.3 ,9 ,11 ,14 ,15 ,17 ,19 ,27 ,28 ,30 ,36 ,40 ,41 ,45 ,46 ,49 ,51 ,54 ,60 ,62 A total of 36 articles (72%) reported a reproducible method for assessing workplace productivity loss or sick leave.3 ,9 ,11–17 19–21 ,23 ,24 ,26–28 ,31 ,32 ,38–41 ,45 ,46 ,49 ,51–53 56–61 ,64 A short recall period (sick leave <6 months, workplace productivity loss <3 months) was used to limit recall bias in 35 articles (70%).3 ,11 ,13–15 ,19 ,23 ,24 ,26–28 30–33 ,35 ,37–41 ,45 ,46 ,48 ,49 ,51–54 56–58 60–62 In 32 articles (64%) serious non-response bias was reduced.3 ,9 ,11–14 ,16 ,17 ,19–21 24–30 ,36 ,38–41 ,45 ,49 ,51 ,54–56 ,60 ,62 Seven9 ,12 ,14 ,19 ,33 ,35 ,41 ,56 of nine determinant studies (78%) controlled for confounding in the analysis and six9 ,12 ,14 ,19 ,33 ,35 ,41 of nine studies also quantified the effect of confounding. All nine studies scored positive on reproducible data collected and presented about determinants. In seven determinant studies (78%) quantitative measures of association were presented.9 ,12 ,14 ,19 ,33 ,41 ,47 ,56 And in four of these studies the number of cases in the multivariate analysis was at least 10 times the number of independent variables in the analysis.9 ,12 ,19 ,33 ,41

Discussion

This systematic review showed that the occurrence sick leave and workplace productivity loss varied widely among studies. The large differences in frequency and duration of sick leave and workplace productivity loss can be partly attributed to large heterogeneity in study populations, studied periods, and definitions and measurements of workplace productivity loss and sick leave. It is of interest to note that even in studies with similar definitions heterogeneity was large, most likely also reflecting national differences in social security and legislation. Most studies included in this review focused on RA, indicating gaps in research into other forms of IA.

The heterogeneity in findings hampered clear summary statements on the consequences of RA for sick leave and productivity loss at work. This implies also that results from one study cannot easily be applied to RA patients in other settings and countries. This will restrict a more quantitative underpinning of indirect costs of RA in guidelines for treatment across the world. Reducing the heterogeneity between studies using similar definitions for sick leave and productivity loss, however, may help in comparing indirect costs between countries. This goes beyond the use of validated questionnaires as even validated questionnaires do not always measure the same and interpretation is not always comparable.65 A recent study with four scales to measure productivity loss in the same patients resulted in substantial differences in lost-hour estimates and corresponding costs, making interpretation of workplace productivity loss and economic evaluations across studies difficult.65 A factor to consider in the definition of sick leave is the recall period. Validation studies suggest using 6 months or less.66 ,67 Another factor that may affect sick leave rates is whether participants are asked to attribute their absence to their arthritis. Although results of our review seem not to suggest there is a difference, there is no evidence of direct comparison of disease attribution. Additional research is needed to examine the attributions and reasons for sick leave time.

Another issue might be the methodological quality of the primary studies. Although on average most studies scored well on the items assessed only 38% of the studies had a response rate over 80% or ‘60%–80% response with no selection in the non-response’. This is a common issue in observational studies, where the response is often lower compared with clinical trials; however, addressing the reasons for non-response would help in understanding whether the lower response rate would be a serious threat for the validity of the study. The quality assessment of studies in this review was performed with a quality assessment tool that enabled us to assess cross-sectional, case-control and longitudinal study designs. Although a wide variety of quality assessment lists exist, there is no universal recommended tool and no quality assessment list that is applicable to all study designs.10 We, therefore, composed a list according to recommendations of Sanderson et al,10 addressing the most important domains of bias for these studies: patient selection, recall period, response bias, reproducibility of outcome measure and if studies reported on determinants also reproducibility of the determinants, controlling for confounding and sufficiently sized sample. An item not addressed specifically was selection of patient towards the more severe spectrum of disease activity in randomised controlled trials. Although these studies provided valuable data on effect of treatment, reported influence on sick leave and workplace productivity loss will likely not to reflect an average patient in daily clinical practice.

Overall, few studies (about a 20%) examined either determinants of workplace productivity losses or sick leave. The results for individual factors are mixed, but tend not to support the views that among RA patients older workers or women are absent more often or are less productive. There is clear support that functional limitations and higher levels of pain are important in understanding sick leave and productivity losses. Pain could have a direct effect on sick leave and productivity loss, but could also function as a modifier via functional limitations or fatigue.4 A limited number of work-related factors were evaluated, suggesting that physically demanding work, lack of job control and lack of support may contribute to productivity losses at work and sick leave. These results are similar to studies outside the field of IA,68 but more research is needed in the field of arthritis.

Treatment is an important predictor for disease course in RA,69 and therefore it is likely to be an important determinant of sick leave and workplace productivity loss. However, it is not often studied as such in cohort studies. Effects of treatment are mainly available from anti-TNF clinical trials. Unfortunately, in general these studies collected little or no information on work-related factors (eg, job type, physically demanding work) or detailed individual factors (eg, education). Given that these work-related factors have been shown to be important in other studies, they need to be included in future randomised clinical trials to get a better appreciation of whether treatment is beneficial for all groups equally, or whether it may have differential effects across work-related factors. Vice versa treatment needs to be included in longitudinal studies on workplace productivity loss and sick leave as it is an important prognostic factor for disease course.

In conclusion, IA leads to problems in participation at work as characterised by workplace productivity loss and sick leave, even during early phases of disease. There are large differences in frequency and duration of sick leave and workplace productivity loss that may be attributed to large heterogeneity in study populations, definitions and measurements of workplace productivity loss and sick leave, but also to treatment modalities. Although higher levels of pain and functional disability were consistently associated, only few studies provided data on determinants of sick leave and workplace productivity loss.

Application of validated and agreed upon outcome measures for workplace productivity loss and sick leave in IA would enhance comparability and uniformity of studies. The OMERACT worker productivity group has suggested that workplace productivity loss and sick leave are important endpoints and they have started efforts to advance on this issue.70 Since treatment is an important determinant for prognosis of RA and possibly also for IA, it is advised to include (i) (drug) treatment in determinant studies of (reduced) workplace productivity loss and sick leave and (ii) work-related, psychosocial characteristics and detailed individual factors (eg, education) in clinical trials for treatment effect.

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.

    Files in this Data Supplement:

Footnotes

  • Contributors ML substantially contributed to study conception and design, acquisition of data and analysis and interpretation of data and drafted the article. ABu and JL substantially contributed to study conception and design and substantially contributed to analysis and interpretation of data and critically revised the article for important intellectual content. AB, MG and JH substantially contributed to analysis and interpretation of data and critically revised the article for important intellectual content.

  • Funding This study was conducted with a grant (300020004) within the Disease Management Chronic Illnesses Program of the Netherlands Organisation for Health Research and Development (ZonMw).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.