The Effect of Osteoarthritis of the Hip or Knee on Work Participation

Literature search

In June 2009, we searched Medline, Embase, CINAHL, and PsycInfo with the following terms and combination of terms: ((knee OR hip) AND (artrosis OR arthrosis OR osteoarthritis)) OR coxarthrosis OR gonarthrosis AND (‘work participation’ OR ‘paid work’ OR occupation* OR employment OR ‘sick leave’ OR burden OR impact OR ‘work transitions’ OR ‘work adaptations’ OR ‘work changes’).

First, titles and abstracts obtained by the search were screened for relevance to our study questions by 2 of the authors independently. Second, after this preselection, full-text articles of relevant titles and abstracts were also screened by 2 authors independently for final inclusion. Reference lists of these articles were analyzed for additional titles. In case of disagreement on the selection, a consensus meeting was held between the 2 authors. If disagreement was still present, a third author acted as referee.

Selection criteria

Studies were included if they met the following criteria: (1) a study population of working age (18–65 years) people with OA was used, or a part of the study population was working-age people with OA and there were separate reports on these people, or having OA was analyzed as a determinant; (2) data were included on work participation and a quantification of the effect of OA on work participation (decrease in productive hours, sick leave, work disability, work adaptations, early retirement); and (3) the study was published in English, German, French, or Dutch and was available as a fulltext article. Articles were excluded if they presented only estimates in terms of money lost, without data on the factors upon which those costs were based.

Assessment of risk of bias

Two authors independently assessed the methodological quality of the articles in the final selection. A specific set of assessment criteria were formulated, based on existing criteria lists (Appendix). The internal validity was the main aspect judged, to assess the risk of bias and to inform the reader about the quality of the studies regarding our research questions. The validity of studies assessing the effect of OA on work may be threatened in different ways: by selection bias, in the case of disproportionate inclusion of either relatively healthy patients or patients with severe complaints; by confounding, if other patient characteristics (age, education level) are related both to the OA and to effects on work participation; or by information bias, in the case of unreliable or invalid measurements. The criteria were therefore grouped into 4 categories: the study population (selection bias), the validity of assessing determinants (OA and possibly confounding determinants of work outcomes), the validity of reported work measures (information bias), and the quality of data analysis (to correct for all factors). The possible judgments were “yes” (coded +), “no”, and “unclear”, (both coded –; Table 1). Cohen’s κ were calculated to assess agreement between the reviewers (before consensus was reached).

Study selection

The searches in Medline, EMBASE, CINAHL, and PsycInfo resulted in 1476, 261, 108, and 16 titles, respectively. Screening of these 1861 titles and abstracts resulted in a selection of 53 fulltext articles that were studied thoroughly. From the reference lists, 1 additional title was added. Finally, 14 articles were included in the review, from which the data were extracted and analyzed.

Quality assessment: risk of bias

Results of the quality assessment are presented in Table 1. Two reviewers independently scored 154 items and agreed on 120 (78%; Cohen’s κ = 0.53). Disagreement was mostly caused by differences in interpretation of the criteria list or unclear reporting in the article and considered mainly the items of standardized and valid measurements of outcome measures, presentation of outcome measures, and multivariable estimates. Agreement was reached by consensus after a discussion in which the referee participated.

Study characteristics

Of the selected articles, 4 studies concerned large population surveys or database surveys^26,27,28,29; 2 prospective cohort studies and 1 cross-sectional study concerned workers, all with OA or including patients with OA^30,31,32; and 7 cross-sectional studies concerned patients with OA^{33,34,35,36,37,38,39}. The characteristics of the included studies are presented in Table 2.

Two studies were prospective: 1 was an OA cohort with 4.5 years followup³⁰ and 1 was a cohort of workers, among them a group with self-reported OA, with a 6-month followup³¹. One study had included a population of 10,412 patients diagnosed by a physician, of which 1750 had paid work³³. Several studies reported on older populations in which only a small minority of subjects were still working^34,35,36,38. Seven studies were performed in Europe^{26,27,31,33,35,28,39}, 6 in North America^{29,30,32,34,36,38}, and 1 in Asia³⁷. Eight studies were published in the last 5 years, two 5-10 years ago, and the other 4 more than 10 years ago.

Outcome measures

The results of the included studies are presented in Table 3.

The only prospective study with a substantial followup period (4.5 years) demonstrated that 37% of 490 working patients with arthritis [57% OA, 10% both OA and rheumatoid arthritis (RA), 33% RA] left the labor force in this period³⁰. Leaving the labor force was related to higher age, lower education, having less control over one’s work schedule, working as a health or education professional, and reporting previous job disruptions and reductions to work hours. A weakness of this study was that it used patient reports of a physician’s diagnosis and that the body sites of the arthritis were not specified.

OA was independently related to being out of work, having work limitations, and being on sick leave in 2 large population surveys^26,28. Work participation rates, matched for age and sex, were equal for patients with OA and healthy controls in 2 cross-sectional studies^33,39. Another study²⁹ showed that the work participation in subjects with OA (age 18–64) was lower than in controls without arthritis, in both men and women, but additional analysis demonstrated that age, education level, and comorbidity explained a large part of this difference. The effect on work participation of OA was smaller than that of RA in 3 comparative studies^29,36,38. The other 6 cross-sectional studies did not report work participation rates, or no comparisons with controls were made.

The overall conclusion regarding the effect of OA on work participation is that it varied. Some studies found similar rates as in controls, 1 found that more than one-third of the patients dropped out of work in 4.5 years. Many of the results were confounded by age, comorbidity, and education level. On average there seems to have been a mild decrease of work participation at a population level.

Productivity, work disability, sick leave, and early retirement

The occurrence of occupational limitations leading to reduced productivity during work was reported in 3 studies^28,32,33 and found to be 3–5 times higher than in controls. Reports on lost working days because of sick leave showed varied results, but seemed to be not very high³⁰, similar to controls³⁹, or slightly higher^31,37. Sick-leave data do not show a normal distribution; many subjects had no (or only a few) days of absence, and the small proportion of subjects with a long absence has a strong influence on the mean duration. One study²⁷ showed that OA caused a substantial part of all temporary work disability periods and early retirement in Germany. Regarding early retirement, 2 studies^34,35 reported exactly the same proportion (2.5%) of patients who indicated they were not working because of OA, and 1 study³⁸ reported a higher proportion (13.7%).

The pattern arising from these studies is that many workers with OA do not reach their optimal productivity during work. On the other hand, OA is responsible for long periods of sick leave or for early retirement only in a small proportion of workers.

Work adaptations

Three-quarters of the working subjects with OA in a followup study of 4.5 years reported any kind of change to their work situation³⁰. This was the only study specifically designed to monitor changes in work; work transitions were related to subsequently leaving employment. In most cross-sectional studies, work changes were either not reported at all, or small proportions of patients (1–10%) reported changes in their work^35,37,38.

Work participation and disease progress

Both the mean age, the mean disease duration, and the disease progress of subjects in the studies varied, from a disease duration of < 1 year³⁹ to about 9 years^30,33,35. From the cross-sectional studies, no information on the effect of disease progression can be drawn, but the studies do demonstrate that many subjects with longer duration of OA are (still) working. The only longitudinal study showed that in 4.5 years, 63% of the subjects (mean age 50.9 years, mean disease duration at baseline 9.2 years) remained employed. The longitudinal course of work participation in OA has not yet been completely described.

Main findings

Many individuals with OA had paid work, and OA could not be proven to be a strong reason for leaving the work force through sick leave or early retirement. Occupational limitations and reduced work capacity or job effectiveness were reported more frequently by patients with OA than by controls. Sick leave mostly did not differ from healthy populations. Work adaptations were measured only occasionally; however, they were revealed as important factors that may precede changes such as leaving the work force³⁰. Because of its high prevalence, OA was a significant factor in longterm disability statistics²⁷. As a result of the differences among the studies, the magnitude could not be expressed quantitatively. Overall, it appeared that many patients with OA were faced with problems in their work, but only a relatively small proportion left the workforce because of these problems. However, the course of OA in relation to work participation has not yet been described completely; neither regarding changes in time, nor influencing factors.

Search strategy

Despite a broad search strategy that resulted in 1861 titles, only 14 studies were included. Many of the included studies reported on the effect on work only as a secondary or even lower outcome measure. The majority were designed for an overall assessment of the burden and the costs of OA for patients in a wide, but mainly higher, age range. Consequently, current effects on work were only relevant for subjects in their working years, which were often a small minority; errors of recall when answering retrospective questions on work in the past may have introduced bias into the estimation of relevant outcome measures. This resulted in a limited amount of relevant information or data that could not distinguish between workers and nonworkers. We confirmed the conclusion that studies on the effect of OA on work are still scarce²².

Quality assessment

At first a systematic difference was noticed between the 2 reviewers concerning applying the criteria specifically to the questions of our review. These were different from the primary questions that were frequently formulated in included studies. This dilemma was reflected in the κ scores for agreement between the reviewers. An example is that OA was associated with older age and comorbidity and that patients frequently had limited education^26,29. These factors are well-known determinants of a lower work participation rate. Therefore the results of studies that included mainly older individuals were probably confounded and the effects on work were not independently determined by OA. Discussions in which the referee participated clarified this matter and thereafter consensus was easily reached.

The diagnostic methods to determine OA varied from self-report or patient report of a physician’s diagnosis, to physician diagnosis and/or radiological assessment. Besides that, all studies included patients with complaints of knee and/or hip, but sometimes also of other body regions. OA in the hands and the back may obviously have an added or different effect on work participation than knee or hip OA only. Four studies^26,29,30,31 used self-reported diagnosis only, which harmed the validity. The differences in diagnostic methods have probably led to the inclusion of different patient categories, which also hampered valid comparisons. On the other hand the association between OA-related impairments (radiological status, pain, stiffness) and limitations in activities is moderate⁴⁰, while participation in work is a result of even more factors and interactions among these factors. From this perspective, in future studies on this issue the aspects of body structures and functions, activities, and participation should be validly measured⁴¹ to enable appropriate analyses of the relations between them.

Several outcome measures were reported, for example, work status, sick leave, work disability, reduced productivity, lost work days, and work transitions; the amount of information was often very limited. Definitions or standardization of these variables were not always presented and different methods were used for measurement. This has limited interpretation and made comparison of the results of these studies more difficult. Differences in conceptualization and outcome measurements have been addressed⁴². Standardized instruments for work-related outcomes need to be studied better and applied in research more often^43,44,45.

Implications

Occupational and ergonomic interventions may be applied more often to help people to stay in their jobs and to prevent progression of work-related complaints and limitations⁴⁶. Patients who cannot meet job demands should be supported in attempts to switch to another job that matches their physical capacity^7,30,47,48. Our review demonstrated that limited research has been performed on the time course of work participation in OA. The critical periods that precede an employee’s decision to leave the workforce because of OA have hardly been analyzed. Extrapolations and projections based on the current literature may have overestimated the effect of OA on work, because studies appear to have included mainly patients with relatively severe complaints or long disease duration, while patients who are functioning well in their work were outside the scope of research.

Limitations of the review

The question of the effect on work of disease progress and duration could not be answered adequately, since the continuum from early complaints through more progressed stages until joint replacement was not covered in the review literature. A number of studies have been published^{49,50,51,52,53} of work participation by subjects awaiting or having had surgical interventions such as total hip arthroplasty (THA) or total knee arthroplasty (TKA). These operations are carried out almost exclusively for OA, but the manuscripts did not meet the inclusion criteria of our review. Considering the increase in THA and TKA, their application at younger ages and the progress in surgical techniques, evaluating their effect on work participation and return to work is relevant.

We realize that most of the included studies were not primarily designed to answer our research questions regarding work outcomes. As a consequence, bias and confounding with regards to this outcome measure may have been introduced in some studies and precautions should have been taken in extrapolations to conclusions on the effect of OA on work. Valid information on the influence of job type and workload was not reported either. We believe this is the first systematic review that revealed these methodological shortcomings, and its value is that we gathered basic insight into the issue of OA and work.

Recommendations

Studies on the effect of OA on work participation should preferably include both working and non-working individuals with OA and compare these to both working and nonworking controls, over a longer period of time. Cohort studies with a followup of at least 5 years could yield valuable information. Different stages of disease progress should be studied and body sites of OA specified, as well as specific aspects of work participation. Multivariable regression analysis should be applied to control for confounding factors such as age, comorbidity, and education level.

Work is an important aspect of people’s social participation, irrespective of their health condition. Staying at work depends on several critical factors, and specific interventions may be needed to reinforce the work ability of patients with OA. To support their work participation, this issue should be addressed regularly in contacts with healthcare professionals.

In this review, a mild negative effect of OA on work participation was found. Many patients with OA may experience difficulties in their work, but they seem to cope with it. However, the longitudinal course of work participation in OA has not been described completely. Considering the need for increasing numbers of people to continue working at an older age, this issue needs attention in well-designed studies and in clinical practice.