Common prognostic factors of work disability among employees with a chronic somatic disease: a systematic review of cohort studies

Common prognostic factors of work disability among employees with a chronic somatic disease: a systematic review of cohort studies. Scand J Work Environ Health. 2009;35(4):261–281 Objective Based on prospective and retrospective disease cohort studies, the aim of this review was to determine common prognostic factors for work disability among employees with rheumatoid arthritis, asthma, chronic obstructive pulmonary disease, diabetes mellitus, and ischemic heart disease (IHD). Methods A systematic literature search in Medline (1990–2008) and Embase (1990–2008) was carried out to identify relevant cohort studies using a well-defined list of inclusion and quality criteria. Results We identified 43 relevant cohort studies with sufficient methodological quality (20 for rheumatoid arthritis, 3 for asthma and 20 for IHD). The common prognostic factors for work disability found in all the diseases were: perceived health complaints, limitation in daily physical activities caused by the disease (high versus low), heavy manual work, and female gender. The common positive prognostic factors for rheumatoid arthritis and IHD were age (high versus low) and sickness absence. The common negative factors for rheumatoid arthritis and IHD were education (high versus low) and ethnic origin (white versus non-white). Conclusions As many prognostic factors for work disability are similar for employees with various chronic diseases, it is possible to detect high risk groups. This information supports the development and implementation of a general disability management intervention for employees suffering from a chronic disease to overcome health-related limitations at work.

The percentage of the working-age population with a long-standing health problem or disability (including non-specific low back pain and mental disorders) varies widely among countries in Europe, with the highest percentage (32.2%) found in Finland and the lowest in Romania (5.8%) (1). Over the next 20 years, prognostic studies predict an increase in the incidence of chronic diseases in the working population, such as rheumatoid arthritis, chronic obstructive pulmonary disease (COPD), asthma, diabetes mellitus, and ischemic heart disease (IHD), merely due to the aging of the working population (2)(3). Despite improvements in facilities and medical care for people with a chronic medical condition, the presence of chronic disease is still a major cause of long-term sickness absence and job loss in Europe.
Labor force participation in Europe is much lower for disabled people. For individuals with a chronic medical condition between 16-64 years of age, the unemployment rate is nearly twice as high as nondisabled individuals. Only one in six employees with a long-standing health problem or disability, who face work restrictions in Europe, is provided some assistance to work (1).
In the Netherlands, only one third of the people between 16-64 years of age with a chronic disease have a paid job in comparison to two thirds of the general population. Of those employees who have a chronic disease, 30% experience problems at the workplace related to the disease (4). The unemployment rate among the disabled is higher than that of the non-disabled for various reasons. One reason is probably the "discouraged worker" effect. Many disabled persons may believe that they are very unlikely to get a job, so they do not even attempt to find one and are, therefore, classified as inactive (5).
In studies of work disability due to rheumatoid arthritis, a relation has been found between becoming work disabled and variables such as: demanding physical work, low education, increasing age (>50 years), pain and self-reported functional status measured with the (modified) Health Assessment Questionnaire (HAQ), disease duration, and motivation to work (6)(7)(8). In studies on work disability due to asthma, a relation was found between work factors, education, general health, and motivation to work (9)(10)(11). Until now, no reviews have systematically evaluated the correspondence between prognostic factors of work disability among employees with a variety of chronic somatic diseases.
In a qualitative (concept mapping) study to explore n a qualitative (concept mapping) study to explore the prerequisites for employees with a chronic somatic disease to function at work, working with employees working with employees and health professionals, we found relatively more we found relatively more disease-generic (ie, common) than disease-specific factors to enable employees to cope. These included: self-acceptance, self-efficacy, support from management and colleagues, and professional advice (12). It is important to search for evidence of common factors as this knowledge can be used to train health professionals in occupational and general health to become aware of employees who have an increased risk of becoming work disabled. Consequently, knowledge about these factors is needed to develop and implement evidencebased interventions to support patients/workers not only in retaining their jobs, but also for the purposes of rehabilitation and return-to-work programs. It might be postulated that many employees with a chronic disease (with the exception of specific work-related chronic diseases caused by occupational exposure) could continue working if they received support in coping with their disease at work. Such employees should be facilitated to continue any form of work participation. This includes the option to change jobs if the employee is not able to continue in his or her current job due to a diseaserelated limitation or continued occupational exposure. Occupational health services, primary health care, medical specialists, nurses, and allied health personnel can improve the quality of care by adjusting their actions to the needs of specific risk groups.
The objective of this study was to search for prognostic factors of work disability which are common to various chronic somatic diseases. A prognostic factor is a risk factor for becoming work disabled, which -when identified -can be used in training health professionals to recognize and support employees who are at risk. The identification of such prognostic factors also enables the development and implementation of disability management interventions.
In this systematic review, we focused on rheumatoid arthritis, COPD, asthma, diabetes mellitus, and IHD as these diseases have a high prevalence in the Netherlands and in other countries (13). Our hypothesis was that several common personal, work-and disease-related factors predict work disability for a variety of chronic diseases. For the purpose of this study, we selected a number of frequently occurring chronic somatic diseases representing a wide variety of diseases in an effort to look for common prognostic factors.

Search strategy
In January 2007, with a follow-up in October 2008, we we conducted a search in Medline (1990Medline ( -2008 and Embase in Medline (1990Medline ( -2008 and Embase (1990Embase ( -2008 using the following search terms both as MeSH term and text word in the title and abstract: (i) rheumatoid arthritis or diabetes mellitus or COPD or asthma or myocardial ischemia; and (ii) occupation or participation or sustainability or employment or unemployment or workplace or career mobility or rehabilitation, vocational or vocation or work or working or labor or job or jobs; and (iii) ability or abilities or disability or disabilities or return to work or continuing to work or functioning or performance or participation or work capacity or sick leave or absenteeism or vocation/disability or disability evaluation; and (iv) prognostic factors or predictive factors. The search strategy was formulated in NCBI (Medline) and was adapted for use in Webspirs (Embase). A combination of the words from (i), (ii), (iii), and (iv) above was used and, as limitation, only English publications were selected.

Selection and methodological �uality assessment methodological �uality assessment
Two reviewers independently selected relevant abstracts from the articles retrieved from the search strategy. If abstracts provided insufficient information, the full text of each article was used. We applied two sets of criteria: the first set to select the study population and type of study, and the second set to screen the articles on quality. In the first screening round, studies were selected if they met all of the following criteria: (i) the study population consisted of subjects with rheumatoid arthritis, COPD, asthma, diabetes mellitus or IHD; (ii) the chronic diseases were not work-related (caused by work); (iii) the study examined work disability, return to work, or continuing to work as the outcome; (iv) the subjects were 18-65 years old (working population); (v) the subjects were employed at the start of the study; Detaille et al (vi) the subjects were not fully work disabled and had not been absent from work for more than two years at the start of the study; (vii) the study must have been a cohort study (prospective/retrospective); and (viii) the study must have applied an appropriate statistical model (univariate or if available multivariate). Following a reading of the abstract, the authors decided either to include or exclude each study; in some cases further discussion was needed to address whether a study should be included (indecisive result). Disagreements regarding inclusion status were resolved by consensus. When no consensus was reached, the abstract or full text was screened by a third reviewer who then decided.The methodological quality of the selected studies was assessed using the criteria list of the Dutch Cochrane Centre for cohort studies (14) which is set out in table 1. Based on the aforementioned eight criteria, the studies were classified as being of "high  quality" when meeting ≥7 criteria, "medium quality" when meeting 5-6 criteria, and "low quality" when meeting <5 criteria. Low quality studies were excluded from the review.

Data extraction
We extracted data from the selected studies regarding study population, outcome measures, and prognostic factors. To facilitate the interpretation and comparison of the results, we categorized the studies by a specific prognostic factor. We distinguished five groups of prognostic factors for work disability based on the core concepts used in the International Classification of Functioning, Disability and Health (15): (i) disease-related factors (eg, severity of the disease, duration of the disease); (ii) body function or structural impairment factors (eg, breathlessness, pain, and body mass index); (iii) activity limitation and participation restriction factors (eg, walking, self-care, communication, and participation at work); (iv) environmental factors (eg, all factors related to work content, work environment, work conditions, care received [including medication], and situation at home); and (v) personal factors (eg, age, gender, education, coping style, and co-morbidity).

Best evidence synthesis
We synthesized the data using the "best-evidence synthesis" procedure (16). The levels of evidence were determined using a rating system similar to those used by de Croon et al (17) and Steenstra et al (18). The information was synthesized into different evidence levels as shown in figure 1. For the purpose of synthesizing the results, we clustered similar work-related outcomes -such as work disability, work disability pension, continuing to work, and return to work (not returning to work) -as the same outcome of "work disability" as there is no single or "gold standard" to define work disability. We did not cluster the outcome "sickness absence" with work disability as we believe that different prognostic factors may play a role in sickness absence (ie, "temporal" inability to work) than in work disability (ie, "permanent" inability to work). The prognostic factors predicting sickness absence were not used in the final results; the direction of the evidence level (ie, positive or negative) was based on the outcome "work disability".
Univariate or, if available, multivariate data are presented in tables. Univariate data were only shown in tables 2, 3, and 4 if there were only univariate results for the factor or if a factor was not significant in univariate analysis and, therefore, not included in the multivariate analysis.
Most articles found with the MeSH term "myocardial ischemia" dealt with a variety of coronary artery diseases including acute coronary syndrome, coronary disease, myocardial infarction, and angina pectoris. Some articles focused on return to work after a coronary artery bypass graft or a percutaneous transluminal coronary angioplasty (PTCA). In the best evidence synthesis, we considered using the categories "IHD with operation" and "IHD without operation", but we found that the diagnosis of the patients included at the    Table 2. Results of the best evidence synthesis for employees with rheumatoid arthritis. [Positive or (+) = a significant positive relation between the prognostic factor and the outcome; Negative or () = a significant negative relation between the prognostic factor and the outcome]

Factor
Result Evidence Outcome a

Disease-related factors
Duration of illness Inconsistent Inconsistent Mau et al (35) No effect Work disability Straaton et al (37) No effect Work disability Reisine et al (7/22) (25) No effect (Nashville/Jyväskylä) b Work disability Sokka et al (40) No effect Work disability Puolakka et al (24) No effect Work disability Straaton et al (37) Negative Return to work Albers et al (38) Negative    (25) No effect (Nashville/Jyväskylä) b Work disability Ever/never methotrexate Weak (+) Chung et al (25) Positive (Jyväskylä) b Work disability Chung et al (25) No effect (Nashville) b Work disability Ever/never used prednisone Weak (+) Chung et al (25) Positive (Jyväskylä) b Work disability Work disability Chung et al (25) No effect (Nashville) b Work disability (continued)  (57) Positive Work disability Barrett et al (24) No effect Work disability Albers et al (38) Positive Work disability Fex et al (42) No effect Work disability Wolfe & Hawley (36) Positive Work disability Young (26) Positive Work disability Borg et al (43) No effect b Work disability Eberhardt & Fex (44) No effect Work disability Puolakka et al (24) No effect b Work disability Sokka et al (40) No effect Work disability Chung et al (25) No effect b Work disability Straaton et al (37) No effect Return to work  (38) No effect Work disability Borg et al (43) No effect Work disability Fex et al (42) No effect Work disability Sokka et al (40) No effect Work disability Holte et al (57) No effect Work disability Married (only in women) Straaton et al (37) No effect Return to work Reisine et al (41) Positive Work disability (continued)

Impairments in body function or body structure
Forced Expiratory Volume (%) Insufficient Blanc et al (10) Positive Work disability Activity limitations and participation restrictions SF36 (physical) Insufficient Yelin et al (11) No effect Cont. employ  (11) No effect Cont. employ

Age
No evidence Yelin et al (11) No effect Cont. employ Alexpoulos & Burdorf (21) No effect Return to work Blanc et al (10) No effect Work disability Education Insufficient Yelin et al (11) No effect Cont. employ Ethnic origin Insufficient Yelin et al (11) No effect Cont. employ Cont. employ

Smoking status
No evidence Yelin et al (11) No effect Cont. employ Blanc et al (10) No effect Work disability Perceived severity score Insufficient Yelin et al (11) No effect Cont. employ a Based on the best evidence synthesis level b Results based on univariate analysis.   (19) No effect Return to work Mark et al (32) No  (32) No effect Return to work Caine et al (53) No effect Return to work New York Heart Association classification Angina class III or IV No evidence Engblom et al (49) No effect Return to work Mark et al (32) No effect Return to work Exercise test f Insufficient Sellier et al (33) No effect Return to work (continued)    (32) No effect Return to work Length of time with an employer Insufficient Caine et al (53) No effect b Return to work Mental stress of job No evidence Hlatky et al (31) No effect Return to work Mark et al (32) No effect Return to work Flexible work schedule Insufficient Earle et al (58) No effect Return to work Able to work fewer hours Insufficient Earle et al (58) No effect Return to work Allowed to change work tasks Insufficient Earle et al (58) No effect

Return to work
Job held open after operation Insufficient Caine et al (53) No effect Return to work   Gender (male versus female) Weak (+) Sellier (28) No effect Return to work Return to work Hlatky et al (31) Positive Return to work· Bhattacharyya et al (30) No effect Return to work Drory et al (34) No effect Return to work Samkange et al (47) No effect Return to work Mark et al (32) No effect Return to work   (19) No effect No effect Return to work Health concern Insufficient Soejima et al (45) No start of the study (inception cohort) are not always well defined and may change in the course of the follow-up study. On the basis of this finding, we decided to use one category (ie, IHD). The only exception was the article of Boudrez (2007) (19), in which a difference was made in the analyses between patients with and without an operation. These two groups are described as two different cohorts.

Selection and methodological �uality assessment
The literature search resulted in 620 hits, of which 89 articles were marked "probable inclusion" or "indecisive result". The third reviewer decided on 11 articles, on which no consensus had been reached. Of the 89 articles, 48 were excluded. Reference checking yielded ten additional articles, three of which were included in this review, bringing the total to 44 studies: 20 for rheumatoid arthritis, two for asthma, one for asthma and COPD, 21 for IHD. All 44 articles were screened for quality; one low quality study on IHD was found (20) and excluded. In the final results, 43 studies were included, 32 of which were of high quality and 11 were of medium quality.
No cohort studies were found for diabetes mellitus. COPD was excluded from the results as a minimum of two articles is needed for inclusion and only one (21) was found. In this article, asthma and COPD were studied; consequently, only the results of the asthma patients were included in this review. Background information and the quality criteria for the selected cohort studies are presented in table 1. Data from the same cohort studies mentioned in more than one article have only been presented once in the final results. This applies to the articles of Reisine (7,22) and Puolakka (23,24). Chung et al (25) described two cohort studies (Nashville/Jyväskylä) in one article, but we presented the results as two studies. Also the results of the two cohorts of Boudrez et al (19) are described separately (IHD with and without (IHD with and without surgery). .

Study characteristics
The studies concerned different populations and countries, and did not use the same data sources nor define work disability or return to work in the same way. Puolakka defined work disability in two different ways. In one article (24), work disability is defined in terms of work disability days; in another (23), it is defined in terms of becoming a disability pensioner. Other authors have defined work disability in terms of losing of a full-time job or early retirement because of a chronic Table 5. Prognostic factors with sufficient evidence [weak (x), moderate (xx) or strong (xxx)] for work disability in rheumatoid arthritis (RA), asthma, and ischemic heart disease (IHD). Factors Factors which increase (+) or decrease () the risk of work disability are indicated. The direction of the prognostic factors is presented as, for example, pain (high/low) which means a high pain condition versus a low one. (PTCA = purcutaneous transluminal coronary angioplasty, AMI = acute myocardial infarction).

RA
Asthma IHD  (26). There were also variations in the definition of return-to-work employees with IHD. Some authors (27,28) used the definition of return to work as working (full-time) after coronary bypass surgery, while others (29,30) defined it as the resumption of a former job or starting a new job, on a full-or part-time basis.

Prognostic factors for work disability
The prognostic factors found in the literature for rheumatoid arthritis, asthma, and IHD are presented in tables 2-4 including the original outcomes mentioned in the articles (ie, work disability, return to work, etc).
To facilitate the interpretation and comparison of the results, in table 5 and the following sections, we have synthesized the results for the outcome work disability only. Table 5 presents all the prognostic factors supported by weak, moderate, or strong evidence for each disease for the outcome work disability.

Disease-related factors.
For employees with rheumatoid arthritis, there is currently not enough evidence on disease-related factors for any conclusion to be drawn. Among employees with asthma, the asthma severity score has been found to have a weak positive relation with work disability. One study found a positive relation between the asthma severity score and work disability (10) while another study found no such relation at all (11).
In the case of employees with IHD, a weak negative relation has been found between the presence of congestive heart failure and return to work: one article found no relation (31); one article found a negative relation (32). The presence of angina pectoris has been found to be a weak positive factor of work disability: one article found a positive relation (33); another article found no relation (32). Two factors, a history of PTCA and a complete revascularization, have been found to have a weak negative relation for work disability. For PTCA, one article found a negative relation (34); one found no effect (33). For a complete revascularization, one study found a positive relation (19); one study found no effect (33).

Impairments in body function or body structure.
In employees with rheumatoid arthritis, various factors related to impaired body function or body structures have been found to be predictive of work disability. A moderate positive relation has been found between the presence of a rheumatoid factor (RF+) and work disability: four studies found a positive relation (35)(36)(37)(38) and four studies found no relation (24,25,(39)(40). The amount of deformed and swollen joints has been found to be a moderate positive prognostic factor for work disability: three studies found a positive relation between the amount of deformed joints and work disability (22,35,41), and four studies found no effect (7,24,26,42).
For employees with asthma, there is currently not enough evidence on factors which are predictive of work disability in this category.
The presence of depression and the level of somatic complaints (high versus low) have been found to be strong positive prognostic factors of work disability for employees with IHD. Six studies found a positive relation (30,(45)(46)(47)(48)(49) and one article found no relation (49) between depression and work disability. The level of somatic complaints was found to have a positive relation with work disability in three articles (19,28,50), while two articles found no effect (19,46). For employees with IHD, Q wave in cases with acute myocardial infarction (yes versus no) (34,51) and chest pain (yes versus no) (32,52,53) were found to be moderate positive prognostic factors for work disability. The presence of dyspnea (no versus yes) (33,53) and left ventricular function (normal versus abnormal) (19,31) have been shown to be moderate negative prognostic factors for work disability. Weak negative factors that have been found include peak/final workflow (33,49) and a sum score of psychological problems (19,32).
To conclude, we did not find any studies which identified common prognostic factors valid for more than one disease in this category.

Activity limitations and participation restrictions.
For employees with rheumatoid arthritis, two factors have been found to be predictive of work disability: a high score on the health assessment questionnaire (HAQ) (strong positive) and previous sickness absence (weak positive). Eight studies found a positive relation between a high score on the HAQ and work disability (23/24, 26, 39, 42-44, 54, 55). Two studies found no relation (25,36).
In employees with asthma, there was not enough evidence on any factors in this category for any conclusion to be drawn.
For employees with IHD, sickness absence before operation has been found to be a strong positive prognostic factor of work disability: five articles found a positive relation with work disability (19,(27)(28)(29)53) and one article found no relation (49). The patient's view on the limitations caused by the disease at work (strong versus light) has been found to be another strong positive prognostic factor for work disability in employees with IHD: two articles found a positive relation with work disability (47,56), and one article found no relation (49). Functional activity (the patient's view) has been found to be a weak positive factor of work disability: one article found a positive relation (32) and the other no relation at all (31).
Based on the evidence, we can conclude that the limitation in daily activities caused by the disease (HAQ and Duke activity scale) and sickness absence (for IHD before operation) are common prognostic factors for rheumatoid arthritis and IHD.
Environmental factors. The type of employment (heavy manual work/blue-collar work has been shown to be a strong positive prognostic factor for work disability in rheumatoid arthritis. Six studies reported a positive relation with work disability (7,26,36,40,44,57) and four studies found no effect (24,(40)(41)(42). Another moderate positive prognostic factor found in the literature was precision versus non-precision work (35,43). Studies have shown weak positive factors to include the following: non-sedentary versus sedentary work (25), complex work (high versus low) (22,36), working hours (part-time versus full-time) (7,22,57), and the use of methotrexate and prednisone (25). Another weak negative prognostic factor found in the literature was the commencement of medication use (<4 months versus >4 months) (24,25).
For employees with asthma, heavy manual work/ blue-collar work was reported to be a moderate positive factor of work disability (11,21).
Studies have also shown heavy manual work/bluecollar work to be a moderate positive prognostic factor of work disability for employees with IHD. Four articles have shown a positive relation between work disability and heavy manual work/blue-collar work (19,33,34,47), while six articles found no relation between the type of occupation and work disability (27,31,45,49,50,53). Living in a rural versus urban area was identified in the literature as another moderate positive factor for work disability (27,33,52); the attendance of a cardiac rehabilitation program was found to be a moderate negative factor (50,58). Support from friends (yes versus no) and health insurance were shown to be weak negative prognostic factors for work disability (19, 45, 50, and 31, 58, respectively).
In the category "environmental factors", we can conclude that the type of employment (manual/bluecollar work versus non-manual/white-collar work) is a prognostic factor that results in a higher risk of work disability in all diseases.

Personal factors.
Studies have shown that age (high versus low) has been found to be a strong positive prognostic factor of work disability for employees with rheumatoid arthritis. Ten studies reported a positive relation (7, 22-25, 35, 37, 40, 43, 57), while four studies found no relation (25,38,44,54). A moderate positive relation has been shown between gender (female versus male) and work disability. Five studies found a positive relation (26,36,38,54,57) and ten studies found no relation at all (7,22,25,37,39,40,(42)(43)(44). The literature has also shown the patient global severity score (high versus low) to be a moderate positive factor (24,25,36). Education (high versus low) has (24,25,36). Education (high versus low) has . Education (high versus low) has been shown to be a moderate negative prognostic factor of work disability in rheumatoid arthritis. Six studies reported a negative relation (25,36,37,42,54,57) and seven studies found no relation (7,22,24,39,40,43,44). Ethnic origin (white versus non-white) has also been identified as a moderate negative factor for work disability (25,36,41). The literature has identified age at the onset of the disease to be a weak positive prognostic factor (26,39); weak negative prognostic factors include the desire to remain employed (yes versus no) and satisfaction with working conditions (7, 22, and 41, 42, 44 respectively).
For employees with asthma, gender (female versus male) was found to have a weak negative relation with continuous employment and thus a weak positive relation with work disability. One study reported a negative relation (11) while another found no effect (10).
Age (high versus low) was shown to be a strong positive factor of work disability for employees with IHD. A positive relation between an older age and work disability was reported in 14 studies (27-29, 31-34, 46-48, 50-52, 56), while three studies found no relation (30,45,53). The literature has identified education (high versus low) as being a strong negative factor of work disability. Three studies report a negative relation (32,48,58); one study found no relation (45). The patient's expectation (yes versus no) regarding return to work after a coronary bypass operation was reported to be a strong negative prognostic factor of work disability. Two articles showed a negative relation (19,31) while a third found no relation (49). An internal locus of health control in patients with IHD has also been found to have a moderate negative relation with work disability. One study reported a positive relation with return to work (46), while three studies found no relation (19,45,58). The perceived importance of work (high versus low) was identified as a moderate negative factor (47,50) (47, 50) and the presence of diabetes (yes versus no) a moderate positive factor for work disability (31,32,34,51). A (31,32,34,51). A . A weak positive relation has been shown between gender (female versus male) and work disability (30)(31)(32)(33)(34)47), (30)(31)(32)(33)(34)47), , while a weak negative relation was found between ethnic origin (white versus non-white), socioeconomic status (high versus low) and work disability (32, 58 and 46, 49, respectively). .
To summarize, in the category "personal factors", gender (female) has been found to be a common positive determinant of work disability for asthma, rheumatoid arthritis and IHD. Being of older age has been identified as a strong positive prognostic factor of work disability in rheumatoid arthritis and IHD. Furthermore, education (high versus low) and ethnic origin (white versus non-white) have been shown to be negative prognostic factors for work disability in employees with rheumatoid arthritis and IHD.

Relevant factors
Using prospective and retrospective disease cohort studies, the aim of this study was to identify common prognostic or risk factors for work disability in employees with rheumatoid arthritis, asthma, COPD, IHD, and diabetes mellitus. We found only one cohort study for COPD and none for diabetes mellitus. We did, however, identify a number of studies which point to common prognostic factors for rheumatoid arthritis, asthma and IHD.
In the category "impaired body function or body structure", studies have identified many prognostic factors. For rheumatoid arthritis and IHD, the perceived health complaints were mostly related to somatic complaints. In IHD, depression was also found to be a positive prognostic factor of work disability. In the category "activity limitations", studies examining rheumatoid arthritis and IHD have found that disease-related restrictions in daily physical activities (high versus low) is a positive prognostic factor for work disability. Studies have also identified sickness absence as a common prognostic factor between rheumatoid arthritis and IHD. In the category "environmental factors", studies have pointed to the type of employment (ie, manual work/ blue-collar work versus non-manual work/white-collar work) as a prognostic factor that results in a higher risk of work disability in all diseases. While studies have found social support at work to be a prognostic factor of work disability in IHD, no evidence has been found for rheumatoid arthritis or asthma.
In the category "personal factors", studies have identified several common factors: being of an older age is a strong prognostic factor for work disability in all three diseases. Female gender is a strong positive prognostic factor in employees with rheumatoid arthritis and asthma. High education is a negative prognostic factor for work disability in employees with rheumatoid arthritis and IHD. In the latter, an internal health locus of control and perceived importance of wanting to work have been found to be a negative factor for work disability. For employees with rheumatoid arthritis and employees with rheumatoid arthritis and asthma, we did not find any cohort studies identify-we did not find any cohort studies identifying the prognostic factors of health locus of control or coping with the illness. In rheumatoid arthritis, studies have found the desire to remain employed to be a negative prognostic factor of work disability. Factors such as coping with disease symptoms and health locus of control are possibly important for continuing to work or return-to-work intervention programs but they have not yet been analyzed in observational cohort studies for all the diseases. This review confirms the prognostic factors found in rheumatoid arthritis for work disability in other reviews (17,59). We did not find any other reviews pointing to prognostic factors for work disability in asthma, IHD, and diabetes mellitus.
Most risk factors with weak, moderate or strong evidence found in this systematic review were disease/impairment-related, socio-demographic-related, or work-related. Common prognostic factors that we found in the literature were: (i) various severity of disease/ impairment factors, (ii) gender (female versus male), (iii) age (high versus low), (iv) education (low versus high), (v) heavy manual work (blue-versus white-collar work), (vi) sick leave, and (vii) perceived health complaints (high versus low). Factors such as low functional status, age, gender, education and blue-collar work are already known to be risk factors for work disability for employees without a chronic disease (60,61).
A serious weakness of most of the cohort studies reviewed was that they had not analyzed the role of psychosocial personal factors (eg, health locus of control or coping). Cross-sectional studies have suggested that there is a causal relationship between low psychological well-being and high health-related distress, and an increase in work limitations among employees with various chronic diseases (62)(63)(64). A cross-sectional study (65) has also identified an association between psychological distress, an external health locus of control, and reported disabilities in patients with lower back pain. Working women, non-whites, and low-educated employees suffer more psychological distress because of a chronic medical condition (66,67). This finding may partly explain the excessive risk of work disability in women and low-educated, blue-collar workers.
A well-known pathway is the relationship between coping style and the health locus of control on one side, and gender, education, age, and blue-collar work on the other. A meta-analysis of Tamres et al (68) showed that, compared to men, women are more likely to use an external rather than internal health locus of control (ie, have a passive versus active coping style). A study of van der Linden et al (69) showed that there is an association between low education, a higher age, blue-collar work and an external health locus of control.
More research is needed on the influence of workrelated and psychosocial personal factors on work disability as such factors can be influenced by work-oriented behavioral and environmental interventions. Supporting employees at work to cope with the limitations of the disease (in terms of self management behavior at work) has been found to be associated with lower psychological distress and a greater control over the disease. These employees are also more likely to be proactive in seeking support at work (70)(71). Active coping has also been shown to reduce the chance of work disability (72). One article on IHD (28) found a positive relation between return to work and coping with or accepting the illness (ie, active versus passive sick role).

Methodological problems
We encountered five methodological problems in this systematic review. Firstly, there were methodological differences between the studies: work-related outcomes and definitions of work disability were different, measurements were taken at different times, and definitions of prognostic variables (eg, various age categories and cut-off points) were not standardized. Of paramount importance is the fact that national differences in the organization of the social security system and unemployment rates are major determinants of work disability risks. These overarching differences can also have an impact on the influence of prognostic factors, which has to be taken into account in the interpretation of findings and the implementation of measures.
Secondly, it was difficult to find cohort studies on asthma, COPD, and diabetes mellitus. For diabetes mellitus, we found only cross-sectional studies providing "determinants" of the relation between the disease and work disability. The determinants found were similar to the prognostic factors identified in the cohort studies in the present study, for example, age (high versus low) and gender (female versus male), both of which were positively associated with work disability, and education (high versus low) which was found to be negatively associated with work disability (73)(74)(75).
Thirdly, the results of the review were based on the evidence provided in the cohort studies. This may imply that some variables, which in occupational practice are supposed to have an effect on work disability, may not have been analyzed in the studies. An example of this is the lack of data regarding the prognostic factors for work disability in asthma. More research is needed on the effects of environmental and occupational exposure as these factors are known to be related to work disability in employees with asthma.
Fourthly, we decided in this review to give priority to data from multivariate analysis over data from univariate analysis as the former controls for confounding variables like gender and age (76) -particularly as controlling for confounding is included in international guidelines as an important criterion to screen the quality of an observational study (77,78). Nevertheless, a disadvantage of multivariate analysis is the difficulty faced by or the subjectivity of the researchers when constructing adequate models. In articles, information is not always presented on the intercorrelations between the independent factors involved and whether one factor overrides the other in the final model. An example is the forced expiratory volume (FEV%) and the asthma severity score in employees with asthma -two prognostic variables that highly correlate with each other (10). Multivariate analysis can select one variable to the exclusion of another when there may be only minor differences in the strength of their prediction. In the final In the final model, the researcher may decide to exclude or to keep the researcher may decide to exclude or to keep one of the variables. In univariate analysis, more factors are visible as having a relation with work disability but an important disadvantage is that confounding and effect modifying is invisible.
Finally, due to the inclusion of a variety of chronic diseases and the wide range of prognostic variables found, it was not possible to effectuate a meta-analysis (79). In order to overcome the limitations of the methods of a traditional narrative review and a meta-analysis, we used the method of "best-evidence synthesis" (16). This method combines the strengths of the methods of both traditional narrative review and meta-analysis, incorporating the statistical rigor of the latter to synthesize quantitative findings as well as the flexibility of the former. Subjectivity in this method is, to a certain extent, excluded by the use of well-justified and well-described inclusion criteria for the studies (16).

Recommendations for further research
From the findings in this study, we derived some recommendations for further research.
Firstly, more observational cohort studies with objective data on the effect of coping style and health locus of control on work disability are needed, as these variables can be influenced by work-oriented behavioral interventions. Secondly, more observational cohort studies are needed on the effect of work factors on work disability, as these factors can also be influenced by interventions at the workplace. Thirdly, more observational cohort studies in general are needed for COPD, asthma, and diabetes mellitus. Fourthly, more research and discussion is needed on the advantages and disadvantages of different prognostic models as data may be lost by using a final multivariate model of analysis. A combination of strategies may provide more transparency, better insights of data, and more efficient control of confounding.

Concluding remarks
This review shows that several prognostic factors for work disability have been found to be common amongst employees with a chronic physical disease. This implies that these prognostic or risk factors have a more generic character beside the more specific problems for working life that may be related to every chronic disease. The common factors are presumably related to the impact of a long-lasting or recurrent health problem on working life as such. As many prognostic factors for work disability are common, we are now better equipped to distinguish groups of employees at high risk of becoming work disabled, regardless of the employee's specific disease. In short, we have to pay special attention to (i) employees with a more severe chronic disease, including a high level of perceived health complaints, (ii) employees with disease-specific impaired body functions, such as pain and swollen/deformed joints in rheumatoid arthritis and depression in IHD, and (iii) employees with more daily physical limitations caused by the disease. Special consideration of older workers, women, bluecollar workers and low-educated employees is also needed. We may be able to use this information in the development and implementation of general disability management interventions for employees with a chronic disease to overcome health-related limitations at work.