A mismatch of the work capacity of an employee with a health problem on the one hand and the task demands in his/her job on the other may lead to sickness absence. Such a mismatch is a consequence of impairments and disabilities caused by the health problem and a lack of possibilities to adjust the job demands to the diminished work capacity of the employee. The term 'impairment' is used to indicate a medical condition that leads to disability, and 'disability' refers to a restriction in functioning or an activity limitation. Mostly, the afore-mentioned mismatch will be temporary, and the employee will resume work after recovery from a sickness spell. In some cases, an employee does not (fully) recover over time, and will remain disabled in his/her work to some extent. If this disability is only minor, the employee may be able to resume his/her own work with increased effort, eventually with minor adaptations of the task demands, tailored to his/her disability. In case the disability is more severe, major interventions in the person domain and/or in the job will be required to facilitate return to work. Examples of such interventions are: enhancement of abilities by education and by training of skills; adaptation of the task demands, and so on [4-6]. Another way to deal with the consequences of the medical situation of these employees may be to reappoint him/her to another job, more suited to his or her reduced abilities.

If the work ability of an employee is affected by health problems in such severe way that return to his/her former job is impossible, then the reduced work ability of the employee will impose limitations on the search for job alternatives. The accessibility of a job for an employee with a specific type of disability will partly be dependent of the (maximum) level of the corresponding task demands in that job. As those demands in the job increase, jobs will be accessible to a lower percentage of sick-listed employees with that disability. Besides, as the severity of the disability increases, the dependency on low-demand jobs will increase, and the number of accessible regular job alternatives will decrease (in this manuscript, the expression 'low-demand jobs' is used to indicate jobs low in task demands). The plausibility of such a negative relationship can be illustrated with the results from a literature review by Turner et al. [7], who reported positive relations between (self-reported) pain and functional disability on the one hand and subsequent duration of work disability on the other.

We can illustrate the relation between the functional abilities of an employee and the task demands in a job with a simple, one-dimensional model. In this simple model, the work load of a job is a one-dimensional work demand, measured by a single parameter, and the corresponding one-dimensional functional ability determines whether an employee is able to cope with the work demand (Figure ​(Figure1a).1a). Our model excludes jobs with major adaptations - for example in sheltered workshops - and employees too severely disabled to work in any regular job. The functional ability of employees increases from low to high, and employees can be sorted accordingly. Likewise, jobs can be sorted on work demand from low to high. Employees high on functional ability have basically access to all jobs (Figure ​(Figure1b).1b). Conversely, jobs very low in work demand are in principle accessible to all employees who can work in a regular job (Figure ​(Figure1c).1c). For employees low on functional abilities (i.e. relatively severely disabled, but still able to work), the number of accessible jobs is restricted: they can only work in jobs with the lowest level of work demand (Figure ​(Figure1d).1d). By selecting employees low on functional abilities and studying the regular jobs accessible to them, one can also identify jobs lowest in task demand, which are the few accessible jobs for these employees. In addition, these low-demand jobs can be used as a classification instrument, to distinguish two groups. The first group consists of all employees whose functional abilities minimally equal the work demands in these low-demand jobs: they have a remaining capacity to work in normal jobs. This group includes all employees who still are able to work in regular jobs (Figure ​(Figure1c).1c). The second group consists of the employees for whom the work demands in the afore-mentioned low-demand jobs exceed their functional abilities. In our simple model, the work demands in all jobs exceed those in the low-demand jobs, and therefore the employees in the second group are not able to work in any regular job.

In reality, the functional disability of long-term sick-listed employees varies not only in severity, but also in nature. Moreover, combinations of health problems within one employee will occur (co-morbidity), mostly resulting in combinations of functional disabilities. Within one specific type of disability, the severity of the disability of an employee will determine the number of accessible jobs, and the most severely disabled are not able to work at all in regular jobs, i.e. in jobs without major adaptations, just as in the afore-mentioned simple model. Employees whose work disability is a little bit less severe will be on the borderline between being able to work and work disability. We expect that only a restricted number of jobs will be suitable/accessible for them, just as in the simple model (Figure ​(Figure1d).1d). For employees having a specific type of disability of a 'medium' severity, jobs low on the corresponding demands are the rare opportunities for them to work in a regular job. The composition of a set of low-demand jobs should be comprehensive to contain accessible jobs for as many disabled employees as possible on the one hand, while the size of the set of jobs has to be manageable on the other hand.

The size of a set of jobs can be reduced by classifying those jobs according to their occupational title into occupations, for example according to the Standard Classification of Occupations (abbr. SCO-'92) of Statistics Netherlands CBS [8]. The SCO-92 distinguishes 1211 occupations to classify all jobs in the Netherlands. Therefore, there is a considerable diversity in task demands and working conditions between occupations. Given a specific type of disability, occupations very low in the corresponding demands can be identified. In rehabilitation efforts for disabled employees with that type of disablement, the low-demand occupations have to be considered for job opportunities. Moreover, if an employee with this specific type of disability cannot work in any of the corresponding low-demand occupations, he/she will not be able to work in any regular occupation, and major efforts will be needed in job adjustments; or work in supported employment may be necessary; or a work disability pension remains as a last resort. This way, the presence or absence of work capacity for regular jobs for employees with the specific type of disability can be assessed by using those low-demand occupations: the so-called reference occupations.

Regular occupations still accessible for severely disabled employees are the most eligible to be used as reference occupations. To identify those reference occupations, firstly functional abilities of sick-listed employees have to be summarised in a limited number of scale variables, each of which represents a common type of disability of the claimants. Secondly, subgroups of claimants with the same main type of disability need to be identified, characterised by high scores on one of the scale variables, and relatively low scores on the other scale variables. Thirdly, the jobs judged to be accessible to such a subgroup by a labour expert in work disability assessments need to be classified into occupations according to the SCO-'92 [8]. Fourthly, the occupations most frequently used need to be presented as potential reference occupations.

The indirect method of selecting low-demand jobs, using the data of severely disabled employees and the jobs accessible to them, seems a cumbersome manner to identify those jobs. It may seem more effective to measure specific task demands in a variety of jobs, and to select the jobs lowest in those demands [9,10]. However, the low level of some of the task demands may be combined by high levels of other demands, or with other thresholds, for example the required education. A labour expert takes all these factors into account in selecting suitable jobs for claimants in the work disability assessment procedure, which justifies this indirect method of selecting low-demand jobs.

In Norway, data about functional capacities of sick-listed employees were summarised in seven scales, based on analyses resembling the analyses for the first step in our identification process of reference occupations, as proposed above. The functional capacities of sick-listed employees in Norway have been registered using the Norwegian Function Assessment Scale (NFAS) [11,12], consisting of 40 items. Four scales about physical dimensions of functioning were identified, i.e. walking/standing; holding/handling; lifting/carrying; and sitting. The other three scales measured mental dimensions: coping; communication; and the senses. The seven scales of the NFAS were developed to be used by rehabilitation professionals to assess the work capacity of sick-listed employees.

The LFA and some other employee characteristics, and the jobs selected in the assessment process as accessible to him/her, were registered/monitored by the computer system CAMS. The cumulative files containing the LFA data of the claimants were previously used to investigate the dimensional structure within the items of the LFA [17], and this resulted in 15 sum variables, to be used in various future applications. However, the number of 15 variables was too large for our present purpose to identify a limited number of dimensions within functional disabilities, and to reduce that number we used a second order factor analysis. The Cronbach's Alpha of one of the 15 sum variables ('Communication') was only 0.54 [17], and therefore we excluded this variable from the present analysis. The second order factors, based on the LFA data of 50,931 claimants, were used to aggregate the items into new scales.

The interrelations between the first order factors were analyzed using factor analysis of the SPSS package, with principal components analysis and varimax rotation. Furthermore, reliability analyses (Cronbach's Alpha) of the scales were conducted.

Within each of the new scale variables, the relation between the value of the sum variable, irrespective of the value of the other sum variables, and the probability of entitlement to a full disability pension was used to classify the values as low, medium or high. For this purpose, two threshold values were set, arbitrarily, at 50% and 66% probability. For each of the sum variables, the monotonic relation between the scale variable and the probability of total disability entitlement were used to identify the two threshold values for 50% and 66% probability. For each sum variable, a group of claimants was selected with high scores on that sum variable (> 66% probability) and low scores on the other sum variables (< 50% probability). Therefore, each group had one main disablement. To be used in the selection of reference occupations, the claimants had to be assessed as able to work in regular jobs. This outcome of the assessment was based on the identification of a minimum of three suitable occupations from the job file of the CAMS, each occupation with at least three observations of actually existing jobs [13]. The occupation most frequently selected for a subgroup of severely disabled claimants was the most eligible candidate to be selected as reference occupation.

Four new sum variables were based on these analyses, and the relation between the sum score and the risk of an assessment result of total work disablement was established. Four groups of claimants with one main type of disability were selected, each with a score above the high threshold value on one sum variable and scores below the low threshold value on the three other sum variables. According to the LE, some claimants within each of those groups were able to work in normal jobs, notwithstanding their high sum score on one of the LFA-scales. Those normal jobs were classified into low-demand occupations for that type of disability. Table ​Table33 lists some of the characteristics of those four subgroups of claimants, and of the group of all other claimants not belonging to the former four subgroups.

The most accessible occupations of the four subgroups of severely disabled of claimants are presented in Table ​Table4.4. The educational level of all of these jobs was 'elementary' or 'low'. Within each of the four subgroups of employees, four to six occupations were sufficient to cover about 90% of the assessments. For the claimants in subgroups 1 and 4, i.e. disabled in general physical ability and manual skills, respectively, the occupational level of most of the suitable jobs was 'low'. Five out of the six most suitable jobs for subgroup 3 with psychological disabilities were on the elementary level, and for the claimants of subgroup 2 (autonomy) two occupations of each level were most frequently used.

For many of the assessments of the four subgroups of claimants, more than one of the selected four or six occupations were used. Moreover, some occupations appeared in the list of the most frequently selected occupations of more than one subgroup of claimants. Five occupations were frequently used for two subgroups of claimants, and one occupation for three subgroups (Table ​(Table4).4). By selecting a subset of six out of the 13 occupations of Table ​Table4,4, we took advantage of the overlap in frequently selected occupations between subgroups of claimants, and so we reduced the redundancy. In this way, we were able to explore the coverage of such a limited number of occupations over all subgroups of claimants. The choice of occupations was based on both the accessibility within each subgroup of claimants and the overlap in accessible occupations between subgroups of claimants. The coverage of the four subgroups of claimants with the six selected occupations are presented in Table ​Table5.5. The coverage per subgroup of claimants varied from 84% to 89%. The weighted mean coverage for all four selected subgroups of claimants was almost 87% vs. almost 91% for the four separate sets of Table ​Table4.4. The coverage of the six general occupations of Table ​Table55 roughly equalled the coverage of the four specific occupations of Table ​Table44 for the second and fourth subgroup of claimants. The reduction in coverage, compared to the four specific sets of occupations per subgroup of claimants, was relatively large in the third subgroup. However, the overall mean coverage per occupation increased considerable because of the decrease in the number of occupations: from 13 to six occupations. Depending on the kind of application, one can choose between the coverage of a large majority of claimants by a restricted set of six occupations of Table ​Table55 and a more comprehensive coverage by the broader set of 13 occupations of Table ​Table44.

Four scale variables were computed to measure the four disability types. The Cronbach's Alpha of one of those scales was only 0.46, probably due to the low number of two variables on which the scale was based. For the application of the scale scores in the identification of subgroups of severely disabled claimants in this study, the low value of the Cronbach's Alpha of this scale was considered acceptable.

To identify severely disabled claimants with one of the four types of disabilities (but still assessed as being able to work), two threshold values were arbitrarily set at 50% and 66% probability of a full disability pension. Additional sensitivity analyses in the vicinity of the 66% threshold did not show a sudden change in accessible occupations (results not presented here), indicating that the choice of the threshold level did not dominate the selection of reference occupations.

The four subgroups of severely disabled claimants differed in some characteristics from each other and from the other claimants (Table ​(Table3).3). The subgroups disabled in psychological abilities and autonomy were more represented in the two younger age categories, especially compared to the subgroup disabled in general physical abilities. In the two subgroups disabled in general physical abilities and in psychological abilities included a higher percentage of female claimants. The four disabled subgroups were better represented in two higher educational categories taken together than the group of all other claimants. However, the results for the four educational categories separately differed between the four subgroups severely disabled claimants. Although some of the differences between the groups in the three characteristics of Table ​Table33 were substantial, none of these were indications that the four subgroups of severely disabled claimants were an extremely deviating selection from the total group of claimants.

The second order factor analysis on 14 of the 15 first order factors resulted in four factors, which was less than the seven factors that were identified within the Norwegian NFAS [11,12]. The Norwegian scale "holding/handling" showed a great resemblance to the scale "manual skill and grip strength". Within the NFAS, three other scales within the physical domain were distinguished, whereas in our second order factor analysis the remaining factors were combined to produce the scale "general physical abilities". Of the three factors in the mental domain of the NFAS, the scale "senses" resembled the scale "communication" of the first order factor analysis of the LFA, which we excluded from our analysis because of the lack of reliability. The remaining two scales in the mental domain of the NFAS differed from the two psycho-social scales of the LFA. This may be due to differences in the subject matter and the application between the two functional ability lists, although both lists were partly based on the ICF [14]. The main differences were: an emphasis on everyday activities vs. work-related activities; self-reported disability vs. assessment by an insurance physician; and duration of the sickness spell preceding the disability assessment: six weeks vs. two years.

The similarities and differences between the factors found in our study and in the Norwegian study [11,12] suggest limited possibilities to apply our results in other countries. It seems plausible that the relative prevalence of some types of disabilities in the work force will show similarities between countries, especially under similar conditions with respect to, for example, the level of prosperity; the nature of the jobs; the working population; the social security system; and so on. Rehabilitation professionals in other countries, consulted by disabled clients of such a similar type, may be inspired by our reference occupations in their advising about accessible occupations, although the task composition of specific jobs may differ between countries. In addition, the method for the selection of reference occupations may be applied in other countries, although the availability of a similar data set may be problem.

In the work disability assessment procedure, a job file with a total set of about 7,000 jobs was used to select suitable jobs from. Although this number of jobs was substantial, these jobs constituted only a limited part of the total of millions of regular jobs in the Netherlands. The size of the job file of the CAMS was limited because of the costs of the observations on location of the task demands of jobs within companies and of the scheduled updates of those observations. The coverage of the total of 1,211 occupational titles of the SCO-'92 by the occupations in the CAMS was relatively high in the three lower level occupations (coverage 47%; 39%; and 23%, respectively), whereas the occupations on the high level (14%) and the academic level (3%) were relatively sparsely covered (the second row of Table ​Table1).1). In the development and maintenance of the CAMS job file, this unbalanced composition of the job file was preferred because the lower level jobs were accessible to almost all claimants as far as the required educational level is concerned, while higher level jobs were only accessible to the higher educated claimants with the right specialization. However, the selection of only elementary and low level occupations as reference occupations was not only caused by the composition of the CAMS job file. Although the education requirements of high and academic level occupations that were included in the job file were all of a more or less general nature to avoid additional barriers caused by the required specialization, those educational requirements would only be met by a small part of the work force (and of the claimants). Some task demands in a number of higher-level occupations may be low, especially the physical demands. However, the requirements of those occupations are often very specific, and the higher educated claimants are not able to return to their former higher level jobs because of health problems, and probably similar problems will occur if they try to work in other higher level jobs of a comparable nature. Moreover, they do not have the right specialisation to function in completely different occupations, and therefore they miss the opportunity to apply their specific work ability, as developed in their education and their former job [18]. To exaggerate: the long-term sick-listed employee, formerly working in a high-level job, is apparently not able to function in that job, and probably neither in similar jobs for which he/she is qualified, and for other high-level jobs he/she lacks (some of) the specific qualifications needed in those jobs. Therefore, it is probable that the (lower level) reference occupations will be the only opportunities for the high-educated claimants to work in a regular occupation. Hence, it is improbable that including more higher-level occupations in the job file would have resulted in the selection of a higher-level reference occupation. It is plausible that the composition of the sets of reference occupations, with only elementary and low level occupations, was caused by the additional barriers in the higher-level occupation for many claimants, leading to a low a priori maximum coverage. The reference occupations at the elementary and low level are useful to explore the general work ability [18].

The SCO-'92 occupational classification comprises 1,211 distinct occupational titles. Each occupational title covers jobs from a variety of companies and institutions, and therefore with a certain variation in specific working situations. A limited selection of these jobs within each occupation was included in the job file of the CAMS. This subgroup of jobs may not be a representative sample of all the jobs within the occupation in the Dutch work force, for example relatively low in the level of demands or required skills and education. However, each occupational title covers jobs from a variety of companies and institutions, and therefore with a certain variation in specific working situations. Not all of the jobs within the job file had been used by the LE in disability assessments of moderate severely disabled claimants. In these disability assessments, relatively demanding jobs had a lower probability of being used. Therefore, the choice of a reference occupation will probably be based on the use of a relatively accessible subgroup of jobs within the occupation in the job file. The description of the reference occupations, including the task demands, to be used in the newly proposed disability assessment procedure should be based on this accessible selection.