The present study goal was to compare the assessment of work ability based on the use of the Work Ability Index (WAI) to the one based on the use of the first item of WAI, this single question being termed “Work Ability score” (WAS), in a population of workers occupying a wide variety of jobs or functions. The non-participation rate in WAI assessment was relatively high (44%) among the workers concerned but does not seem to have induced any recruitment bias insofar as the demographic characteristics of the respondents overlap with those of the non-respondents. This high non-participation rate could most probably be ascribed to difficulties in administrative management experienced within the collaborating medical centres.
Research question 1 - Validity of WAS compared to WAI and usefulness of WAS
The relative merits of using either a single-item measure or a multiple-item (or scale) measure have been discussed at great length in the occupational psychology literature for assessing job satisfaction [24, 25]. Although job satisfaction and work ability are constructs of a different nature, they are both complex constructs with multifaceted determinants. It is thus tempting to hypothesize, as these authors did for job satisfaction, that simply combining 7 pre-selected items or dimensions of work ability to obtain an overall index of work ability may in some cases exclude other significant aspects of the man–machine interaction that may be very influential in determining the worker’s own perception of his/her work ability. One could however argue that for assessing work ability, taking into account the number of diseases the worker is currently encountering (WAI item 3) and the importance of sick leave in the last 12 months (WAI item 5) should substantially increase the content validity of WAI in comparison to the WAS single-item measure.
In the present study originating in occupational health practice, the comparison between WAI and WAS was first guided by cost-effectiveness considerations. As stated by Wanous et al. for job satisfaction measures [24], a single-item measure is shorter in length, requires less time to complete and is more likely to be completed by the employee. Since the introduction of WAI in the medical surveillance routine, STM has experienced difficulties in the use of this tool, whether for the occupational health service (need for external expertise when defining work function categories and for a high degree of rigor when doing the data capture and calculating the index) or for the workers themselves. When the worker does not have a good understanding of the WAI aim, he/she can only with difficulty answer correctly all questions pertaining to the 7 items, which could account for the high proportion of questionnaires not completely filled in (28.3%) in the present study. Missing values were particularly frequent (> 15%) for WAI item 4 (estimation of work impairment due to the diseases), item 5 (sickness absence), item 6 (prognosis of ability over 2 years) and item 7 (psychological resources).
The results obtained in the study seem to indicate that using the single-item approach instead would not deteriorate the validity of the work ability information collected. The level of convergent validity observed between WAS and WAI was quite satisfactory (rs = 0.63) and of the same order of magnitude as the correlation obtained for job satisfaction measures [24]. In addition, the present analysis shows that the assessment based on WAS (1 item) highlighted the same factors of increase or reduction in work ability as did the 7-item WAI, with the exception of the effect of firm size (not observed with WAS in the multivariable model). WAS therefore appears as a tool to be used in priority in the future as its user-friendliness brings in a clear advantage for a systematic application during medical examinations performed within the field of occupational health care.
Research question 2 – Determinants of work ability
Relation with individual factors
The relations observed between work ability and the individual variables generally corroborate those reported in the literature. The results show a strong association between ageing and the decline in work ability, whether be it assessed by WAI or by WAS; numerous studies have indeed demonstrated that young workers estimate their work ability at a higher level than older ones [26–31]. In the results obtained, the relation observed is nevertheless not fully linear as work ability assessed in 60-65-year olds was better than in 55-59-year olds (Tables 2 and 3). The limited size of the older worker group when compared to the other age groups supports the hypothesis of a “healthy worker” effect, healthier individuals being able to stay longer on the labour market, a well-described phenomenon in several industries [3].
The data analysed did not highlight any relation between gender and work ability, an observation in line with the systematic review done by van den Berg et al. (2008) [2]. It is worth mentioning however that this relation varied according to the measurement tool used, either WAI or WAS. On the basis of WAI measurement, the probability of low or poor work ability was higher in women (significantly in univariable and not reported in multivariable). Yet, this association disappeared when ability was measured using WAS. Such a discrepancy between the two methods has also been reported in the Finnish health survey [9]. The authors suggested that the decrease in WAI but not in WAS level in women could be accounted for by a higher number of sick leaves and days of absence but also by lower psychological resources [9].
The population studied in Luxembourg included a high proportion of overweight workers and, in line with the observations reported in other studies [2, 26, 32–34], the present results suggest that those workers are more at risk of presenting moderate or poor WAI and that this risk increases as a function of the excess weight; but it must be noted that this association was either not significant (p=0.08) or not included in the multivariable model. This could be due to the inclusion in the model of the ‘number of diseases’ variable, the link between overweight status and several pathologies, especially cardiovascular ones, being well established. Another unhealthy behaviour significantly increased the risk of presenting low or poor work ability: a dose-effect relationship was observed in univariable analysis with the number of smoked cigarettes. This association was however not significant in the multivariable model, and this could possibly be ascribed to the importance of the diseases variable in the model. In the literature, an association between work ability and workers’ smoking habits was reported as significant in a single study only [26].
In the present study, the relation between health status and work ability has been explored on the basis of the number of major diseases recorded by the occupational physician. Mental diseases and musculoskeletal disorders exerted the strongest negative influence on WAI; the association with cardiovascular diseases was not so clear unlike the observations made in the Finnish health survey [9].
The prevalence of musculoskeletal disorders in the studied population (38.3%) seems in line with the results of the European survey in which 24.7% complained about backache and 22.8% about muscle pain [35]. Regarding mental health problems (psychosis, anxiety, addictive behaviour,…), the rate observed in the sample (10.3%) appears relatively low when compared to the rates reported in the European survey for stress (22.3%), irritability (10.5%), or anxiety (7.8%). Nevertheless, given the possible impact of such self-reported mental health problems on the “fit for work” decision to be issued by the occupational physician, some under-reporting bias could be hypothesized.
Relation with occupational factors
This study highlighted significant differences in self-estimated work ability according to the type of work function held. Workers assigned to a predominantly mental function presented higher work ability levels than those assigned to a mostly physical function. This trend was observed both for WAI and WAS levels, and this is in line with literature data [9, 27, 36, 37]. This association could not only reflect the detrimental effects of chronic exposure to biomechanical and postural stress in physical jobs but also the impact of low work control and poor job content [27].
Firm size, or in other words the number of workers employed, also influenced work ability as estimated by WAI. The probability of presenting a WAI level defined as moderate to poor was indeed lower in firms employing less than 50 workers. This observation could be linked to a more favourable relational environment in small and very small (<10 workers) firms. Literature data however does not provide any information on this issue.
Strengths and limitations of the study
A strength of the study lies not so much in the size of the population sample studied (over half of the workers in Luxembourg) but in the wide variety of professional sectors represented. Not all sectors were represented however: the STM service is not empowered to monitor occupational health in such sectors as banks and insurance companies or even hospitals. These sectors have their own occupational health service.
From a methodological point of view, the large sample of WAI data available made the exclusive selection of fully complete questionnaires a better option than the use of substitution algorithms for missing values.
Another strength of the study lies in its inscription in the real practice conditions of workers’ health surveillance as performed by an occupational health service; the study showed the difficulty of applying a standardized and systematic process for work ability assessment with more than 25 different nurses and physicians being involved.
The study design also has some limitations: the completion of the questionnaire being made on a voluntary basis, without direct supervision, one cannot rule out the idea that less educated workers decided not to fill in the questionnaire or when filling it in, failed to provide information for all the items. Another potential limitation has to do with the quality of medical variables. The CMR used in the STM service was not primarily intended for epidemiological studies and the lack of standardized anamnesis implies that the data-capture and exploitation of the CMR might have been influenced by features specific to each medical examiner.
Another limitation arises from the asymmetric gender distribution in the studied sample: any extrapolation of the observations of this study to other populations of workers with balanced gender distribution would require utmost caution.
