Work ability and percentage of hours worked related to limitations in patients with upper extremity musculoskeletal disorders: a cross-sectional cohort study

In this cross-sectional study, a questionnaire was sent out in April 2005 to all 3250 members of the Dutch National UEMSD patient association, irrespective of the severity of complaints. Participants that were unemployed or on disability pension were excluded. Each questionnaire (‘RSI: Complaints, health and possibilities’) was accompanied by a participant information form. Filled out forms were returned anonymously and directly to the research institute by an enclosed return envelope. After 1 week, the members received a reminder to fill out and return the questionnaire. All patients participated on a voluntary basis and data was stored anonymously. The study was conducted in accordance with the declaration of Helsinki [26]. The research proposal for this secondary analyses of the data was submitted to and approved by the Medical Ethical Committee of the Academic Medical Center, which decreed that a comprehensive evaluation was not required as this study was not subject to the Medical Research Involving Human Subjects Act (W19_211#19.256).

Respondent characteristics and basic information were gathered by a questionnaire. Besides the basic information about work, several instruments were used to measure the work hours, complaints and the work ability of UEMSD-patients [27]. The questionnaire further contained questions on (health) complaints, health and health perceptions, work and limitations [28].

The primary outcome, self-reported work ability, was measured with the Work Ability Score (WAS), the first question of the Work Ability Index (WAI). This single item WAI was included to assess the appraised work ability at the current time (according to the patient) on a scale of 0 to 10. This single item question showed sufficient convergent validity to the complete WAI [29]. A score of 0 meant not being able to work and a score closer to 10 meant increasing work ability compared to their lifetime best (score 10). Hours worked, was operationalized as the percentage of hours worked compared to fulltime employment (36 h). To calculate this variable, participants filled out the number of hours they actually worked in the past week. This value was chosen rather than the percentage of worked hours compared to the contract hours because multiple respondents were on a 0-h or flexible contract basis while working multiple hours. Many other respondents worked a numerous amount more than their contract hours, leading to abnormal values, which made the contract hours unsuitable for these analyses.

To answer our second research question, the limitations were operationalised in three ways: The degree of difficulties in performing daily activities in the last week was assessed by the complete Disability of the Arm, Shoulder and Hand (DASH) questionnaire, including the work and sport module [30, 31]. The DASH scores are scaled with 0 indicating no disability and 100 indicating most disability.

A second way to assess limitations was to ask whether patients have problems performing certain frequent grips of both hand and wrist. Illustrations of these common grasps and movements were included in the questionnaire, accompanied by a question on the limitations of performing these in daily work life. These illustrations are shown in the Appendix. These grasps and activities were, among others, the cylindrical grip, the lateral key grip, the precision grip or spherical grip, the hook grip or medium wrap, and common activities at work (i.e. using a computer mouse or keyboard, pushing with the hand or finger, and bending the wrist) [27, 32]. An example question next to an illustration was: When you have to perform activities during work using your hand as displayed in the illustration above, do you experience problems? The answer options to these questions were: ‘Yes, almost always’, ‘Yes, sometimes’, ‘No, no problems’, or ‘No, I don’t have to do this’. The outcome was divided into either ‘yes, almost always’ scored as 1, and all the other answer options grouped together scored as 0. This means that a score of 1 indicates problems with performing this movement compared to ‘no’ or ‘only sometimes’ problems, which was coded as 0.

The third way was to assess limitations in the health-related quality of life by using the Short-Form-36 (SF-36) [33‐35]. Of the nine subscales, only ‘physical functioning’, ‘pain’, and ‘physical role functioning’ were used in the analysis. The subscale of physical functioning measured ‘limitations in daily activities as a consequence of health issues’ during the last 4 weeks. The pain and physical role functioning subscales represent ‘pain and limitations due to pain’ and ‘limitations in work and other daily activities due to physical health issues’, respectively [35]. These scores are measured on a scale between 0 and 100, where a score closer to 100 indicates a better general health status. Averages for a Dutch population aged 35–44 were 90 (±14.4), 84 (±21.7) and 83 (±32.0) for physical functioning, pain and physical role functioning, respectively [35].

Additionally, demographic variables, among which age and gender were included, were used in the analyses to describe the study population and differentiate possible influencing characteristics. Males were coded as 1 where females were coded as 0.

Before the analyses were performed, cases with over 50% missing variables were deleted. Cases with one or both primary variables missing were also deleted from the analyses. Before the analyses, the MCAR test was executed to establish the randomness of missing values. If the MCAR test was significant, no imputation has taken place. First, the self-reported work ability was related to the percentage of hours worked using a Pearson correlation analysis. In this analysis, the coherence between both outcomes was analysed.

As a final check, we tested both models for heteroscedasticity (through a visual inspection of the residual plots) and collinearity (through collinearity statistics: tolerance and Variance Inflation Factor (VIF) score). A tolerance ≤0.4 and a corresponding VIF score over 2.5 was set as cause for concern of collinearity, with tolerance calculated as 1-R² and VIF as 1/tolerance. The statistical analyses were done performing multiple regression analyses in IBM SPSS Statistics version 25.

In total, 1129 UEMSD patients responded (35%), of which 794 were eligible for secondary analyses as they filled out both questions regarding work ability and hours worked and were currently employed [28]. Individuals who received disability pension, or did not have a job were excluded from the sample. The inclusion flowchart is shown in Fig. 1. The MCAR test showed that some missing values are not randomly missing. As a result, no imputation took place, and no data is listwise deleted.

The mean age at the time of participation was 40.0 (±8.5) years and one-third of all participants were male. Over 90% of included participants reported they were still working. The mean duration of complaints was 5.6 (±3.1) years, and the mean extensiveness of pain regions was 5.8 (±3.6) on a scale of 0 to 16 regions. Participant characteristics (of all respondents) are shown in Table 1. Population characteristics divided per job category according to the International Standard Classification of Occupations [36] are shown in the Appendix.

The percentage of hours worked correlates significantly (p < 0.001) with the self-reported work ability (n = 794). The correlation coefficient is r = 0.46; 95% CI [0.40, 0.53] p < 0.001, which is considered a modest or moderate correlation [37].

Results of the univariate analyses are shown in the Appendix. In the multivariate analyses, the prediction model for the WAS contained four variables. These variables were able to explain 52% of total variance in WAS. The DASH work module score, the SF-36 physical functioning and physical role functioning domains and the precision grip were predictors in this model. Multicollinearity was not of concern in the WAS model as there were no VIF-scores over 2.5.

The model that explained 21% of total variance in percentage of hours worked contained five variables, namely: SF-36 physical functioning domain, the DASH score, the DASH work module, limitations using a keyboard and gender. Multicollinearity was a concern in the worked hours model where the VIF-score of the DASH was 3.4. When this variable was removed from the model, there was no more concern for collinearity. However, the variables ‘ability to hold a precision grip’ and ‘ability to grasp a large object with one hand’ became significant contributors and were added to the model again in accordance with the methods. The explained variance of the new model, without the DASH score, but including the re-entered variables as predictors, is also 21%. A higher score of the B-coefficient means a more positive contribution to the work ability score. Both final models are displayed in Tables 2 and 3. A plot of the residuals was analysed for a check on heteroscedasticity, but there were no reasons to believe this was of concern in either model.

In this model, a higher SF-36 score, having problems with using a keyboard, having problems performing a precision grip and being male resulted in a higher percentage of worked hours, while a higher score on the DASH work module and limitations in grasping a large object with one hand predict a lower percentage of worked hours in individuals. In the univariate analyses, limitations using a keyboard and the ability to hold a precision grip are not significant. However, in the multiple regression, these variables become significant indicating suppression effects (or negative confounding) between variables. Limitation in using a keyboard is unsuppressed by the presence of the DASH work module.

Work ability has not been related to current work hours in populations with limitations in studies before. We could not identify other studies that studied this specific relationship. Leijten et al. (2014) did find that WAS and productivity at work (0–10), a concept closely related to working hours, were positively correlated in older employees with chronic diseases (r = 0.23, p < 0.01) [39]. Whereas Leijten studied the influence of chronic limitations on work ability 1 year later, our study provides knowledge on limitations at the current time in relation to work ability and worked hours at the same moment. Another study by Oakman et al. (2019) studied the relationship between the number of pain sites and work ability trajectories [40]. They found that the number of pain sites was associated with a decreasing work ability trajectory over time. A study by Kamaleri et al. (2009) showed the strong predictive power of the number of pain sites on work disability 14 years later [41]. In our study, pain was included as a potential predictor of workability, but was not retained in the final model. This means that not only pain predicts work ability, but that limitations in work might help better predict current work ability. However, these limitations are only moderately related to the hours worked. Although future work disability and current work ability represent different concepts, we would argue—based on our study—that not only the number of pain sites or number of limitations is important, but also the type of limitations in cohesion with the type of work.

A strength of this study is that we studied a population with a large variation in work ability due to the existence of limitations/health problems during work and daily life. This study population showed a larger variance in WAS in comparison to a general working population [42]. When studying the coherence between work ability and hours worked, a population with a large variation in work ability would be beneficial in order to study correlations between the two variables.

A limitation of this study is that the type of work that participants do was not taken into account. However, given the fact that participants have UEMSD complaints, it is plausible that the work of these participants involves working with arms and hands. All questions regarding movements and hours concerned the current job of the respondent, as is the assessment of self-reported work ability. If participants did not have to do certain activities during work, these were regarded as not being a limitation, and therefore not affecting work, and this should therefore not be a large source of bias. However, we expect some selection bias. Although the entire population of the UEMSD patient association received an invitation to participate in this study, it can be expected that the more severe cases were represented more strongly in the UEMSD patients association.

An aspect that might explain the difference in explained variance by limitations between work ability and the percentage of hours worked is the common method principle due to the cross-sectional nature of this study [43]. This principle suggests that two measurements are more strongly related to each other than a third variable due to the type of measurement. In this case, the limitations and WAS are both assessed by a questionnaire filled in by the workers themselves. The number of hours, however, is more factual and not only determined by the workers themselves, as the occupational physician and their employer play a role in determining the optimal number of work hours.

The outcome of hours worked was used to represent the impact of UEMSD-related limitations on the ability to perform work. However, since the variance in worked hours is only explained to a moderate extent by the factors in our model, we can conclude that hours worked is influenced by other factors than just these limitations. The family situation, social security system, psychological state, need for income generation or informal care can all influence the reason to work [44]. This can explain the moderate correlation between work ability and the hours worked at the current time.

It seems that the number of hours worked does not fully match with self-reported work ability. When determining the work hours of a UEMSD patient, it can be profitable to better align these work hours with work ability, as workers take their limitations at work into account when assessing their own work ability. In this way, the working hours of a UEMSD patient correspond better with their self-reported ability to meet the job demands. Limitations are individual and job-specific. Workers can estimate the influence of limitations on their work best because they know every details about their job demands in relation to the limitations they encounter. Therefore, the appraisal of their work ability can be used to better match work hours to their limitations in order to prevent workers from going on sick leave. In the Netherlands for instance, an occupational physician can be consulted when a worker is unable to meet the work demands. This occupational physician advises in consultation with the employee and employer how and if work can be continued. One potential outcome of the consultation is that the employer is advised to change the work demands or work hours. When this proves impossible, the occupational physician can advise other jobs within or outside the company, that can be met by the employee without compromising their health.

Occupational physicians can use the appraisal of UEMSD patients’ work ability to better match the ability to meet job demands with the working hours, keeping the limitations in mind. Occupational physicians can advise their patients to have their work hours and demands match their abilities. The results of this study are relevant for occupational professionals such as occupational physicians as they give an indication of which functional limitations affect the work ability most, for instance the ability to hold a cylindrical grip, precision grip and hook grip. The importance of the capability to perform certain actions can aid occupational professionals in determining the next steps in the treatment, to decrease the impact of limitations on work life. To confirm the additional value of taking work ability into account when determining work hours, a longitudinal study could be performed. Based on the univariate analyses, limitations in bending the wrist are quite important in some individuals, but not significant on a group level. For which individuals this limitation is of importance needs to be studied in future research. Future research can focus on differences between types of job demands. A division can be made between more physical demands using shoulders, or more desk work using fingers and hands. With this information, future work disability or sick leave might be prevented by using the right strategies to ensure work hours match work ability. Furthermore, the ability to use a keyboard, the ability to hold a precision grip and the ability to grasp a large object with one hand are only related to the amount of hours worked when the DASH work module was present in the final model, and the B-coefficient changed direction. Since this indicates paradoxical confounding, the relationship between these variables needs to be investigated further.