Work disability figures in most western European countries have more than doubled since the 1970s and nowadays more than 5% of the working population receives a disability pension.1 In most cases, before a pension is granted work ability is assessed by a medical professional in order to predict fitness for work. A scientific basis for these assessments is lacking, however.2 3

A number of medical professionals may be involved in the work ability assessment process, including general practitioners, occupational physicians, medical specialists and insurance physicians. Communication between these parties is advised4 but may be limited in practice.5 The different medical professionals concerned may have diverse points of view, interests and concerns6 and it is not clear which items they assess for work ability. In this respect, universally accepted lists of items for consideration in an evaluation of work ability may help identify aspects that are relevant to patient–professional communication, may be useful in helping professionals to prevent long-term work disability and useful for encouraging work ability.

The assessment of work ability concerns a prediction of future fitness for work in the case of a certain disease. Because, as stated by the WHO’s International Classification of Functioning (ICF) model,7 work ability is multi-causal and not only dependent on the disease, the list of items for consideration can be expected to contain disease-specific and non-disease-specific prognostic factors.

To address this issue, a study was set up to research prognostic factors for return to work for the three diseases for which disability pensions are most frequently granted in the Netherlands: myocardial infarction (MI), chronic low back pain (cLBP) and major depressive disorder (MDD).8 The research question was formulated as follows: What are prognostic factors for work ability in sicklisted employees with MI, cLBP and MDD?

METHODS

Systematic search strategy

A systematic search of the PubMed electronic database was carried out to identify relevant studies using Yale University’s methodological research filter ‘Prognosis and Natural History’, in which the keywords were connected with ‘OR’ (Table 1). The different keywords relating to the concept of work were connected with ‘OR’ and the different keywords relating to the concept of ability were also connected with ‘OR’ (Table 1). Different keywords for MI, cLBP and MDD were connected with ‘OR’. MI or cLBP or MDD (Table 1) were combined by ‘AND’ with the methodological research filter, work and ability. Limits were set on age (19–65 years), publication date (1 January1990 to 1 July 2006), English and human.

Selection of papers

The following inclusion criteria were applied to the identified studies:

(a) MI: diagnosed by a cardiologist and requiring hospital admission; cLBP: at least 12 weeks’ lower back pain and not having a specific cause; MDD: according to DSM diagnostic criteria

(b) studies with a prospective or retrospective cohort or case–control design

(c) at the start of the study all participants should be disabled for work

(d) outcome of return to work or long-term financial compensation for work disability

The first author (F.S.) applied the inclusion criteria. In the event of uncertainty, the other authors (J.S., P.K., M.F.) were consulted as a group. For each included study a data extraction form was used to note the following: patient sample; duration of work disability at the start of the study; moment of measurement of prognostic factor in the study; follow-up; loss to follow-up; outcome measure of return to work or compensation status; adjustment for other possible prognostic factors; and the rationale of the studied prognostic factors. Each data extraction form was discussed by the authors (F.S., J.S., P.K., M.F.). The included studies were then assessed to ensure that they met at least four of the six formulated quality criteria according to Straus et al.9 as follows: (1) all participants should be employees; (2) all participants should be work disabled at the start of the study; (3) the follow-up should be at least 1 year; (4) loss to follow-up should be less than 20%; (5) there should be adjustment for important prognostic factors; and (6) the used set of prognostic factors should be justified. When the discussion regarding inclusion was inconclusive, J.S., P.K. and M.F. studied the original paper, and a further discussion about inclusion took place. Upon reaching a consensus the article was included or excluded.

Further selection of papers

When the discussion regarding inclusion yielded no papers at all for disease-specific prognostic factors, studies from the initial identified papers with a cross-sectional design were also considered.

RESULTS

The search strategy identified 961 studies. A total of 955 studies failed to meet the inclusion criteria. The six remaining studies met at least five of the six formulated quality criteria according to Straus et al.9 (Table 2).

Prognostic factors for work ability in MI patients

Study characteristics

The search strategy identified 164 articles on MI. After applying the inclusion criteria four articles on MI remained. The sample sizes of the MI studies ranged from 9011 to 507415 and the follow-up range was one,11 two10 15 and four years.14 Loss to follow-up was not mentioned in the study by Hamalainen et al.15 and was less than 5% in the other studies. Three of the four studies concerned employees who were admitted to hospital because of MI.11 14 15 The study by Froom et al. concerned employees consulting an occupational health clinic after 1 to 14 months.10 The studies concerned different countries and did not use the same data sources. Return to work was not defined in the same way in the included studies. Froom et al. defined return to work as an 8-h working day,10 while Nielsen et al. defined return to work as the resumption of a former job or starting a new job, on a full-time or part-time basis.14 All studies were adjusted for other relevant prognostic factors.

Prognostic factors

As shown in Table 2, younger age10 14 15 and having lower physical demands at work10 11 14 are mentioned as predictive factors for return to work in three out of the four studies. Prognostic factors were determined shortly after hospital admittance in three out of the four studies11 14 15 and after an average of 3 months in the fourth study.10 Some factors, such as Q waves, angina before MI and age, cannot be expected to change in the course of the disease. Others, such as anxiety, diabetes and workload, may reasonably be expected to change.

Prognostic factors for work ability in cLBP patients

Study characteristics

The search strategy identified 353 articles on cLBP. After applying the inclusion criteria two articles on cLBP remained. The sample sizes of the cLBP studies ranged from 32813 to 275212 and the follow-up was 1 year in both studies. Loss to follow-up ranged from 10%13 to 15%.12 The study by Van der Giezen et al.13 concerned Dutch employees sicklisted for 3 to 4 months. The study by Hansson and Hansson12 concerned employees from different countries who were sicklisted for 3 months. The exact duration and profile of cLBP was not mentioned in the studies. It was assumed that because employees were sicklisted for 3 months because of LBP that it concerned cLBP. Both studies defined return to work as the resumption of work. Both studies were adjusted for other relevant prognostic factors.

Prognostic factors

As shown in Table 3, younger age is a predictive factor for return to work in both studies. The prognostic factors found in the studies are determined after 3 to 4 months’ work disablement by Van der Giezen et al.,13 and after at least 3 months’ work disablement by Hansson and Hansson12 Some factors, such as age and gender cannot be expected to change in the course of the disease. Others, such as pain, general health and physical job demands, may reasonably be expected to change.

Prognostic factors for work ability in MDD patients

MDD study characteristics

The search identified 444 studies on MDD. After applying the inclusion criteria no studies on MDD remained.

DISCUSSION

Four prognostic studies on MI, in which participants were recently work disabled at the start of the study, and two prognostic studies on cLBP, in which the participants had been work disabled for 3 to 4 months at the start of the study, were found. The studies found met five or more of the six quality criteria formulated according to Straus et al.9 For MDD, no studies that dealt with prognostic factors for work ability were found. No studies in which, at the start of the study, the participants had been work disabled for more than a year, i.e. the period after which long-term disability pensions were granted in the Netherlands in 2004,8 were found.

Although we performed a sensitive literature search, our search yielded only six studies. The studies that were found did not use the same sets of potential prognostic factors. A sound theoretical background for which prognostic factors should be investigated is missing. As a consequence, studies identified prognostic factors that were not investigated in other studies. Finding only a few studies that did not investigate the same prognostic factors limits the generalisability of the results.

Although determined in different phases of work disablement, the studies on MI and cLBP identified common prognostic factors. LVEF > 35%, light or sedentary job, no financial basis on which to retire and no anxiety attacks in the MI studies seem comparable with pain intensity, physical demands at work, being a breadwinner and general health in the cLBP studies. Generally speaking, disease-specific and non-disease-specific prognostic factors appear for work ability. Therefore, in addressing work ability, treating physicians should, in general, on the one hand treat the disease and on the other hand focus on non-disease-specific factors that are amenable to change. However, it cannot be ruled out that some of the prognostic factors are significant by chance. There is as yet no evidence that just because a prognostic factor is modifiable, it will change the prognosis for work ability. At present, the prognostic factors found should be used with caution and only as flags for work ability and as indicators for its prognosis.

The MI studies described prognostic factors determined among recently hospitalised MI patients. Because prognostic factors for return to work may change,16 17 it is not clear whether described factors are also relevant in the prediction of work ability in later phases of disablement. Both the course of predictive factors and the relation of this course to work ability in work-disabled MI employees are relevant in this context and no such studies have been carried out to date on this topic.

Two studies on cLBP in which the participants were 3 to 4 months work disabled at start of the studies were identified. Checking for the prognostic factors may indicate recommendations for adequate pain management, for the improvement of the patient’s general health, for the reduction of obstacles at work that aggravate symptoms and, for return to work.

MDD is the fourth leading cause of disease burden on society18 and is, at least in the Netherlands, the most common diagnosis in long-term work-disabled employees. No studies for prognostic factors were found, however. It has been demonstrated that in many cases MDD has a chronic relapsing course and that work ability fluctuates with the severity of MDD.19 20 Therefore, until such time as more evidence becomes available, the course and the severity of MDD could be considered when giving advice on work ability.

The prognostic factors identified in the present study do not belong to the same domains of health as defined by the WHO’s International Classification of Functioning (ICF) model.7 Our findings are in accordance with the ICF model because the model states that work participation is multi-causal7 and not only dependent on the disease. Supporting disabled patients in returning to work may therefore exceed the expertise of the individual doctor who operates in a certain health domain. Therefore cooperation between different professionals may be necessary. Categorising the prognostic factors according to the ICF domains may be beneficial in this respect. Disease-specific factors such as pain intensity, LVEF and atrial fibrillation point to possible disease-specific MI or cLBP interventions. Personal factors like age, gender, disease history and comorbidity point to interventions that can empower the employee as an individual. Environmental factors like physical demands at work, psychological demands at work and decision latitude are directed at workplace interventions from which not only the work-disabled employee but other employees could benefit. Tools for handling work disability should therefore encompass all domains of the ICF model and also address the cooperation of different professionals.

Since work disability figures are rising, every doctor will encounter short-, medium- or long-term disabled patients. Patients and/or stakeholders in the disability determination process will enquire as to the prognosis for work ability. Although relevant studies were found, this study demonstrates the strong need for more evidence on prognostic factors for work ability. Because the study concentrates on three common diseases it is reasonable to assume that this lack of knowledge applies for other diseases as well.

In the present study many studies were not included because they did not concern (only) work disabled employees; they concerned depressive disorders other than MDD; heart disease but not MI per se; acute or sub-acute LBP instead of cLBP; a cross-sectional instead of a longitudinal design; a short-term follow-up; or they did not concern return to work or its equivalent as outcome.

Future studies on prognostic factors for work ability in chronic diseases should be planned and can learn from the present study. The outcome of future studies should be return to work with long-term follow-up. In each particular study participants should all have the same disease and should all be in the same (short-, medium- or long-term) phase of the disablement process. Because functioning in work is multi-dimensional, the factors to be explored in these longitudinal future studies should at least encompass all components from the ICF model. In this respect qualitative research to elucidate possible barriers and facilitators for return to work known by employees, employers and other stakeholders in the work disablement process may be helpful.

CONCLUSION

In the earlier phases of work disablement in MI and cLBP patients, only a few studies describe disease-specific, environmental and personal prognostic factors for return to work. No studies describe prognostic factors for MDD. More evidence is needed on the topic of prognostic factors for return to work for chronic diseases.