Objective To investigate the developmental pathways of multisite musculoskeletal pain (MSP) and the effect of physical and psychosocial working conditions on the development of MSP trajectories.
Methods The study was conducted among food industry workers (N=868) using a longitudinal design. Surveys were conducted every 2 years from 2003 to 2009. The questionnaire covered MSP, physical and psychosocial working conditions (physical strain, environmental factors, repetitive movements, awkward postures; mental strain, team support, leadership, possibility to influence) and work ability. MSP as an outcome was defined as the number of painful areas of the body on a scale of 0–4. Latent class growth modelling and multinomial logistic regression were used to analyse the impact of working conditions on MSP pathways.
Results Five MSP trajectories (no MSP 35.6%, persistent MSP 28.8%, developing MSP 8.8%, increasing MSP 15.3% and decreasing MSP 11.5%) were identified. In a multivariable model, the no MSP pathway was set as the reference group. High physical strain (OR 3.26, 95% CI 2.10 to 5.04), poor environmental factors (3.84, 2.48 to 5.94), high repetitive movements (3.68, 2.31 to 5.88) and high mental strain (3.87, 2.53 to 5.92) at baseline predicted the persistent MSP pathway, allowing for poor work ability (2.81, 1.84 to 4.28) and female gender (1.80, 1.14 to 2.83). High physical strain and female gender predicted the developing MSP pathway. High physical strain, poor environmental factors and high repetitive movements predicted the increasing and decreasing MSP pathways.
Conclusions A substantial proportion of individuals reported having persistent MSP, and one-third reported changing patterns of pain. Adverse physical working conditions and mental strain were strongly associated with having high but stable levels of MSP.
Statistics from Altmetric.com
What this paper adds
The development of multisite musculoskeletal pain (MSP) has received limited attention, despite the significance of the problem, especially in workplaces with heterogeneous work tasks.
Few studies have been undertaken that examine psychosocial, physical and environmental exposures in relation to MSP using a longitudinal design.
Five different developmental pathways of the number of MSP sites were identified: no MSP, persistent MSP, developing MSP, decreasing MSP and increasing MSP.
Nearly a third (29%) of the employees had persistent MSP, one-third (36%) had no pain and the remaining third reported changing patterns of pain.
Adverse physical working conditions and mental strain were strongly associated with having high but stable levels of MSP.
Musculoskeletal disorders (MSDs) are a major cause of work incapacity and sickness absence and early exit from paid employment1–3 and are a significant compensable burden across the world.4 MSDs are complex, multifactorial problems associated with a range of workplace physical and psychosocial hazards.5–7 Musculoskeletal pain (MSP) commonly occurs in multiple body sites8 ,9 with more severe consequences than single-site pain,8 ,10–12 resulting in reduced daily functioning,8 ,11 poor work ability,12 increased sickness absence13 and early exit from the workforce.14
Despite the significance of the MSP problem, its development and relationships with the work environment have received limited attention. Only a few earlier studies have reported the patterns of developmental pathways of multisite pain. A study of female kitchen workers identified four different patterns of MSP development over 2 years of follow-up,15 with more than 36% in the high MSP group. Finnish firefighters were followed up for 13 years, and three trajectories of multiple MSP and depressive symptoms were identified:16 high MSP but low depressive symptoms among 32% of the participants, high depressive but low pain symptoms (19%) and low symptoms overall (49%). The above studies15 ,16 were conducted in occupations with high physical demands and one only including female workers. Thus far, limited evidence exists to support the understanding of the development of MSP especially in workplaces with heterogeneous work tasks.
Some previous studies have examined MSP trajectory group membership to explore pain development and reported strong associations with physical and psychosocial work factors.15–17 Furthermore, several occupational exposures and lifestyle factors have emerged as determinants of MSP. Physical exposures such as heavy loads, bending or twisting, vibrations, awkward postures and occupational biomechanical demands9 ,18–20 and psychosocial exposures such as high job demands and low coworker support21 have been reported as strong predictors of MSP. Few studies have been undertaken, which examine the relationships between the psychosocial, physical and environmental exposures in relation to MSP using a longitudinal design.
Improved understanding of the influence of working conditions on the development of pathways of MSP over time will enable more focused interventions to be implemented in workplaces. In this study, our aim was to analyse the developmental pathways (trajectories) of MSP during 6 years of follow-up. The second aim was to explore whether the baseline physical or psychosocial working conditions predict the developmental pathways of MSP.
This study used data collected from a questionnaire conducted in a Finnish food company, which manufactured various products including meats, canned goods and convenience foods. The survey was conducted over 6 years, every 2 years from 2003 to 2009. Questionnaires were not addressed to individual employees, so personal reminders could not be sent. Respondents could respond anonymously or provide their name and consent for linking survey data with register data obtained from the company personnel registers. Ethics approval was obtained from the Pirkanmaa Hospital District, Tampere, Finland.
In 2003, a 63% (N=873) response rate was obtained. In 2005, 2007 and 2009, 1201, 1400 and 1398 people replied to the questionnaire, respectively. For inclusion in the current analysis, participants must have responded to the baseline survey and at least one of the follow-up surveys. A total of 560 people responded to the baseline and first follow-up survey, 522 people in the baseline and second round of follow-up and 402 people replied to the baseline and last follow-up. Respondents who responded to the baseline survey were aged between 18 and 64 years (mean age 40.5±11.1); almost 70% were women and blue-collar workers.
Measurements of variables
A modified version of the validated Nordic Musculoskeletal Questionnaire22 was used to assess MSP. Questions asked whether the employee had felt pain, aching or numbness in four anatomical areas (hands or upper extremities; neck or shoulders; lower back and feet or lower extremities) during the preceding week (from 0 (not at all) to 10 (very much)). The variables were dichotomised at the median score (less than median: 0=mild; more than median: 1=severe). The cut-off values for upper extremities, neck and shoulder, lower back and lower extremities were 4, 5, 2 and 2, respectively. The dichotomised variables were summed into a variable, expressing the number of areas with severe pain (from zero to four).12 ,19 In this study, we are modelling the number of painful sites, which are referred to as MSP.
Perceived physical strain was measured using a rating of perceived exertion with the question “how physically hard/exhausting do you feel your job is on a normal work day?” on a scale from 6 (not at all) to 20 (very much).23 Physical strain was dichotomised into ‘low’ (6–13) and ‘high’ (14–20) by using the median value as a cut-off point.
Environmental exposure was measured using questions adapted from a survey by Statistics Finland covering draught, noise, indoor climate, heat, cold, lighting and distractions impacting concentration in the work environment.24 The seven items were summed into a single variable ranging from 7 to 35. The variable was dichotomised into ‘low’ (7–18) and ‘high’ (19–35) using the median value as a cut-off point. Biomechanical factors at work were assessed through questions on ‘repetitive movements’ and ‘awkward postures’. A scale of 1 (not at all) to 5 (very much) was used. The variables were dichotomised into ‘low’ and ‘high’ at the median value (cut-off value 3 for both).
Perceived mental strain was assessed using a modified version of the occupational stress questionnaire25 (“stress means a situation in which a person feels excited, apprehensive/concerned, nervous or distressed or she/he cannot sleep because of the things on her/his mind. Do you feel this kind of stress nowadays?”) with a scale from 0 (not at all) to 10 (very much). The variable was dichotomised as ‘low’ (0–4) and ‘high’ (5–10) using the median value as the cut-off point.
Psychosocial factors used in this study have been described in detail elsewhere.13 Briefly, constructs on the following areas: ‘incentive and participative leadership’, ‘team support’ and ‘possibilities to exert influence at work’ were asked with a response scale from 1 (totally disagree/very probably not) to 5 (totally agree/very probably).26 Responses were summed and divided by the number of variables used in the index. Cronbach's αs of the measures were 0.71, 0.79 and 0.82, respectively. All psychosocial factors used in the analysis were dichotomised using the median value as the cut-off point.
Work ability was measured with the question “how is your current work ability compared with life time best”, with responses from 0 (absolutely incapable of work) to 10 (work ability at its best). In the current analysis, we dichotomised the variable as ‘poor’ (0–8) and ‘good’ (9–10) at the median. Age was categorised into four groups (<30, 30–39, 40–49 and 50+) to examine aged-related differences.
Latent class growth modelling (LCGM) was used to identify patterns of change (trajectories) in the number of painful sites (MSP). LCGM enables the simultaneous identification of different developmental patterns (also non-linear) over several measurement points. These patterns are empirically derived post hoc. LCGM is a special case of the growth mixture model, given the assumption of homogeneity of growth parameters within a latent subgroup.27 A four-step process was used to estimate LCGM: problem definition, model estimation, model selection and interpretation.28 A change function which best represents the patterns of change in the data was first identified, for example, straight line (linear function), one curve (quadratic function), two curves (cubic function) or two or more separate trajectories (piecewise model).29 In this study, quadratic function best represented the patterns of change in the data. The final model was chosen based on a range of fit criteria, which include Akaike information criterion (AIC), Bayesian information criterion (BIC), sample size-adjusted BIC, entropy and Lo-Mendell Rubin (LMR) p value. In the fit criteria, a lower BIC, AIC and sample size-adjusted BIC value and entropy close to one indicate a better model fit. Nylund et al 29 determined that BIC and sample size-adjusted BIC performed best after bootstrap likelihood ratio testing (BLRT). However, Nylund et al 30 also suggested that due to the increased computing time for BLRT, using BIC and the LMR p values is an appropriate alternative. On the basis of this, the model with a low BIC value and significant LMR p value (comparing the current model against the model with one less class than the current model of choice should give an LMR p value <0.05) was used in the current analysis. The BIC is commonly used within mixture modelling analysis31 and a decrease in at least 10 points shows a sufficient improvement in the model.32 On the basis of the above fit criteria, five trajectories were determined.
For inclusion in the final analysis, participants must have responded to the baseline and at least one follow-up survey. However, preliminary analysis was undertaken of those who responded to all four waves (n=325), and the trajectory shapes were unchanged. Therefore, a decision was made to include all respondents who replied to baseline (N=868) and at least one of the follow-up surveys. The trajectory groups were illustrated by plotting mean levels of MSP against year of the survey (figure 1).
Baseline characteristics of participants were examined by the trajectory group using the χ2 test. To investigate the relationship of the latent classes of MSP with baseline working conditions, we used multinomial logistic regression. ORs and their CIs were used to measure associations. Models were built in two steps; first, all studied variables were entered together in the model for simultaneous adjustment and, second, a backward stepwise method was applied to create the final model. In the final model, those predictors which remained statistically significant (p<0.05) in the backward stepwise analysis were included. The trajectory group with ‘no MSP’ was used as the reference category.
LCGM was analysed in Mplus V.7.2, and the regression analysis was performed in SPSS V.23.0.
Online supplementary table S1 shows the fit criteria for comparing the different numbers of latent classes of MSP in the follow-up. Trajectory models were tested from one to six classes, and several parameters for each class were calculated. According to fit criteria, five latent groups fitted the best with low AIC, BIC and sample size-adjusted BIC. The latent group also had higher entropy and significant LMB p value (<0.001).
Fit criteria for number of painful sites with different latent classes.
Figure 1 shows the results of the final trajectory solution. Five distinct trajectories with different mean numbers of painful sites were identified: no MSP (n=309, 35.6%), persistent MSP (n=251, 28.8%), developing MSP (n=76, 8.8%), increasing MSP (n=132, 15.3%) and decreasing MSP (n=100, 11.5%). The no MSP trajectory consisted of more than one-third of workers who reported pain at low levels throughout the follow-up. The persistent MSP group comprised almost 29% of the workers with three to four sites of pain throughout the follow-up. The other three groups were diverse in the changes reported. The developing MSP trajectory had the fewest workers (9%) with almost no pain at baseline but up to three pain sites at follow-up. The increasing MSP pathway consisted of 15% of the workers who had two pain sites at baseline that increased slowly to three sites during the follow-up period. Similarly, the decreasing MSP trajectory (12%) had either two and three sites of pain at baseline but only one pain site at follow-up.
Table 1 shows baseline characteristics and the MSP pathways. Of the 868 respondents at baseline, about one-third were aged between 40 and 49 years and almost 24% were in the oldest age (50+ years) group. Approximately 69% of the respondents were women, and about the same number employed in a blue-collar occupation. At baseline, 27.3% of the respondents were pain-free, 17.7% had one pain site and 55.1% had at least two MSP sites. Statistically significant differences between the different pathways were found in all studied variables except age. Briefly, those with persistent MSP often had poor work ability and high physical work strain, and the majority reported exposure to poor environmental conditions, high repetitive movements, awkward postures, mental work strain, poor team support and leadership and limited influence at work. The members of the other trajectories were more mixed in their reported exposures.
Table 2 describes the adjusted effect of baseline predictors on the development of MSP trajectories among all workers. Compared to men, women had a higher likelihood of having persistent MSP (OR 1.73, 95% CI 1.09 to 2.75) and increasing MSP (OR 2.10, 95% CI 1.13 to 3.90). Similarly, poor work ability strongly predicted persistent MSP (OR 2.60, 95% CI 1.68 to 4.30) and increasing MSP (OR 1.94, 95% CI 1.21 to 3.11).
High physical work strain, poor environmental factors, high repetitive movements and exposure to awkward postures significantly predicted persistent MSP with ORs 3.11, 3.39, 2.94 and 1.94, respectively. High physical work strain and poor environmental factors were associated with the pathway of decreasing MSP and increasing MSP. Similarly, among psychosocial factors, high mental work strain was a significant predictor of persistent MSP (OR 3.85, 95% CI 2.48 to 5.97) and decreasing MSP (OR 1.86, 95% CI 1.12 to 3.07). Poor team support was also found to be associated with decreasing MSP with borderline significance (OR 1.71, 95% 1.00 to 2.91). Age, occupational class, leadership and possibility to influence did not predict trajectory membership.
In the final model (table 3), women were more likely than men to have a pathway of persistent MSP (OR 1.80, 95% CI 1.14 to 2.83) and increasing MSP (OR 2.02, 95% CI 1.10 to 3.71). Poor work ability significantly predicted persistent MSP (OR 2.81, 95% CI 1.84 to 4.28) and increasing MSP (OR 2.06, 95% CI 1.37 to 3.53). High physical work strain was the only factor that significantly predicted membership of a trajectory group. However, the magnitude of the association was much stronger for the persistent MSP trajectory (OR 3.26, 95% CI 2.10 to 5.04) followed by increasing MSP (OR 2.20, 95% CI 1.37 to 3.53). Poor environmental factors and high repetitive movements were associated with the membership of MSP trajectories. Both factors predicted persistent MSP, decreasing and increasing MSP trajectories, but the association with developing MSP trajectory was not statistically significant. The magnitude of the association of poor environmental factors was highest with the persistent MSP trajectory (OR 3.84, 95% CI 2.48 to 5.94) followed by increasing MSP (OR 3.01, 95% CI 1.88 to 4.82). Similarly, the highest magnitude of the association of high repetitive movements was found for the persistent MSP trajectory group (OR 3.68, 95% CI 2.31 to 5.88) followed by increasing MSP (OR 2.18, 95% CI 1.31 to 3.65). High awkward postures were not retained as a predictor in the final model. Of the psychosocial factors, only high mental work strain remained in the final model as a predictor of the persistent MSP trajectory (OR 3.87, 95% CI 2.53 to 5.92). High mental strain was also associated with the decreasing MSP trajectory (OR 1.97, 95% CI 1.21 to 3.21) but not with the increasing MSP trajectory.
The principal aim of this study was to investigate the developmental pathways of multisite MSP and their relationship with work-related physical and psychosocial conditions. Five different developmental paths of the number of pain sites were identified during 6 years of follow-up: no MSP, persistent MSP, developing MSP, decreasing MSP and increasing MSP. Most people (almost two-thirds) remained in a particular pain pathway, with similar characteristics during 6 years of follow-up. Almost all baseline physical working conditions such as high physical strain, poor environmental factors and high repetitive movements predicted MSP pathways. High mental strain was the only psychosocial factor which predicted trajectory membership. Women were at a higher risk of having persistent MSP and increasing MSP. Individuals with poor work ability were more likely to have persistent or increasing MSP.
Nearly a third of the employees remained pain-free over the 6 years of follow-up, one-third (29%) of participants had persistent MSP and the remaining third reported changing patterns of pain. This is consistent with previous research in relation to the pain patterns, but with fewer trajectory groups examined.15 ,16
In our study, the prevalence of multisite pain (pain in two or more sites) was about 55% at baseline and remained the same in the last round of follow-up, suggesting that MSP is persistent.32 In the current study, 30% of the workers were engaged in white-collar work, in contrast to previous studies where participants were predominately involved in physically demanding work. Despite these occupational differences, reports of persistent MSP were similar, supporting the notion that factors beyond the tasks being performed are important.15 ,16
Predictors of MSP trajectories
High physical work strain, poor environmental factors and high repetitive movements at baseline were predictive of membership in the persistently high and the increasing MSP trajectories in the multivariable model. The magnitudes of associations with the persistent MSP trajectory were stronger than those with the increasing MSP trajectory. High-perceived physical workload was a strong predictor of a high MSP trajectory among female kitchen workers in Finland.15 Similar findings have also been reported in other studies exploring risk factors for MSP.9 ,11 ,19 ,33 Solidaki et al 33 reported that physical loading was associated with persistence of MSP among occupational groups in Crete. Notably, physical strain also predicted the pathway of developing MSP where the number of pain sites increased from less than one at baseline to more than three at the last follow-up in our study. This finding is in line with that among kitchen workers,15 where the statistical significance of physical load was lost when psychosocial factors at work and lifestyle were taken into account. Previous longitudinal studies have also suggested that occupational activities such as heavy lifting34 and working in bent positions35 are associated with new onset of widespread pain. In contrast, Solidaki et al 33 reported no significant association of number of strenuous physical activities with new onset of MSP in their study. Another large cross-sectional study using data from 18 countries in different parts of the world showed that exposure to multiple physically stressing occupational activities was associated with pain in 6–10 anatomical sites.9
High mental work strain was equally as important as physical factors in predicting persistent MSP consistent with previous studies.15 ,33 In a previous study, high mental work strain was found to have a stronger impact than physical factors in predicting persistent MSP in a 4-year follow-up of the current study material.36 Psychosocial factors were also as important as physical factors among French workers in predicting MSP.37 A cross-sectional study from the USA reported an association between workplace psychosocial strain and MSP.38 In another cross-sectional study, Coggon et al 9 reported that time pressure, lack of support at work, low job control and job insecurity were significant predictors of MSP. However, the current study did not find other psychosocial factors such as team support, leadership and possibility to influence to be associated with any trajectory group in the multivariate model.
An unexpected finding was that high physical strain, poor environmental factors and high mental strain at baseline were associated with the decreasing MSP trajectory. Speculatively, such relationships could be due to favourable changes during follow-up in work circumstances among participants with high MSP at baseline. Alternatively, participants with pain may report worse mechanical and psychosocial work exposures than participants without pain, and this could be related to the self-reported nature of the data. We further analysed the data to check if these unexpected findings can be explained by the regression to the mean;39 however, no substantive impact was found.
Earlier studies have reported MSP as a risk factor for poor work ability12 and work disability.10 However, the reverse associations have not been widely reported. The current results support that poor work ability was an independent predictor of stable high MSP and increasing MSP trajectories, suggesting a reciprocity of effects between work ability and MSP. Women were at greater risk for having stable high MSP and increasing MSP, consistent with a previous cross-sectional study where women were more likely to report pain in multiple locations.9 However, inconsistencies exist in the relationship between gender and MSP. A longitudinal study of MSP in postal clerks, nurses and office workers33 found no gender differences. Age was not a predictor of a particular MSP pathway. This is in contrast with previous studies.9 ,15 ,33 However, wide variations within the capacities of different age groups mean that this finding is not surprising.
The occupational group was not associated with any particular MSP trajectory, which was unexpected as the working conditions differ between blue-collar and white-collar workers. Analysis of trajectories among blue-collar and white-collar workers was undertaken to ascertain if they had different patterns of development of pain. Interestingly, both groups followed similar patterns of development of MSP during the follow-up (data not shown). This finding is consistent with a previous cross-sectional study where the variation of prevalence of MSP did not differ by occupational category.9
A strength of our study was the prospective design with several data collection waves. The 6-year period provided sufficient follow-up time to examine the influences of working conditions on MSP. A further strength was the inclusion of blue-collar and white-collar occupations and their different job demands. A potential limitation was non-response to one or two of the four surveys. To account for this, we analysed data for those participants who replied in all follow-ups (n=325) and found that the trajectory shapes and group proportions did not alter. The healthy worker effect may have influenced our results as those with significant problems are more likely to leave work,40 and the follow-up surveys addressed only the group who stayed with the same employer. We also cannot rule out the possibility that differential reporting of work exposures was dependent on the baseline pain status. The influence of this differential reporting on the associations of work exposures assessed at baseline with the development of pain during follow-up was presumably minor, however. Participants were asked to report pain from the past 7 days, which reduces recall bias but does not take into account episodic pain. Using a median cut-off point may result in some information loss but ensures sufficient cases in each category. No designated rule exists to support the choice of cut-off points to define MSP.9 In the current study, the number of pain sites was used as the outcome in line with previous studies.16 ,33 Information on lifestyle factors such as smoking, body mass index and physical exercise were not collected at baseline and not included in the current analyses although these may be associated with MSP.15 Moreover, the working conditions were measured only at baseline and therefore, we were not able to analyse whether the trajectories of MSP were affected by the changes in workplace exposure levels during the follow-up.
Findings from this study pose some interesting challenges for those charged with improving physical and psychosocial working conditions. Understanding the implications of changes to the work environment is important, in particular for the reduction of MSP. The important role of work ability for those with persistent pain suggests an imbalance between the work environment and an individual's capacity and an area in need of attention in prevention strategies. Modifications to the physical and psychosocial work environment would be an important first step in the development of interventions to change persistent pain pathways.
Contributors C-HN and PV designed and conducted the surveys. SN conceptualised the study, performed the statistical analysis and interpretation of the results and drafted the manuscript. PL-A, JO, PV and C-HN critically reviewed the manuscript. All authors read and approved the final version of the manuscript as submitted.
Funding SN was funded by Finnish Cultural Foundation (SKR) with a personal grant (grant number 00150637).
Competing interests None declared.
Patient consent Obtained.
Ethics approval The Pirkanmaa Hospital District, Tampere Finland.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional data are available.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.