Occupational and Environmental Medicine
Work-related Psychosocial and Mechanical Risk Factors for Low Back Pain
A 3-year Follow-up Study of the General Working Population in Norway
Tom Sterud, Tore TynesDisclosures
Occup Environ Med. 2013;70(5):296-302.
Abstract
Aims This study examines the impact of work-related psychosocial and mechanical exposure on the development of low back pain (LBP) in the general working population.
Methods A randomly drawn cohort from the general population in Norway aged 18–66 years was followed up for 3 years (n=12 550, response rate at baseline=67%). Eligible respondents were in paid work during a reference week in 2006 and 2009, or temporarily absent from such work (n=6745). Five work-related psychosocial factors and seven mechanical exposures were measured. Outcomes of interest were moderate or severe LBP at follow-up adjusted for baseline LBP.
Results In total, 12.8% (861 individuals) reported LBP during the last month at follow-up. Work-related psychosocial predictors of LBP were high job demands (OR 1.41, 95% CI 1.16 to 1.72) and low job control (OR 1.26, 95% CI 1.01 to 1.57). Mechanical factors were prolonged standing (OR 1.48, 95% CI 1.20 to 1.83), awkward lifting (OR 1.55, 95% CI 1.28 to 1.88) and squatting/kneeling (OR 1.29, 95% CI 1.04 to 1.61). The estimated population risk attributable to these factors was approximately 42%. The risk for LBP associated with psychosocial exposure was not influenced by adjustment for mechanical risk factors, and vice versa. There was no substantial confounding related to age, gender, education, occupation or psychological distress.
Conclusions Highly demanding jobs, prolonged standing and awkward lifting appear as the most consistent and important predictors of LBP.
Background
Low back pain (LBP) is a common health problem responsible for 11–13.5% of all sick days in the working population and is one of the largest single causes of absenteeism in western countries.[1 2] Globally, the annual prevalence of LBP has been estimated at 38%.[3] In general, LBP resolves within weeks, but may recur in 24–50% of cases within 1 year.[4] Thus, the identification of risk factors for LBP is important in the prevention of recurrent and possibly chronic LBP.
More than 100 risk factors for LBP have been identified.[5] In the majority of cases, a combination of individual and work-related as well as non-work-related factors is likely to contribute to the development of LBP.[4] Risk factors for occupational LBP have often been studied in relation to exposure to biomechanical processes in work situations. A wide range of work-related mechanical risk factors for LBP have therefore been reported in prospective studies. They include ‘bending or twisting’,[6 7] ‘kneeling or squatting’,[8] ‘prolonged standing’,[9] ‘heavy physical work’[10 11] and ‘nursing tasks’ (eg, manually moving patients).[12 13] Overall, however, the evidence showing work postures, manual handling and carrying to be risk factors for LBP remains inconclusive.[14]
In recent decades, there has been increased emphasis on work-related psychosocial factors in epidemiological studies of LBP. There is some evidence that psychological demands[15 16] and high job strain[17] are related to LBP and there is a possible but uncertain relationship between decision authority, skill discretion and LBP.[9 18] Lack of social support has been demonstrated to increase the risk of sick leave associated with LBP.[7 19] However, the level of evidence for most psychosocial factors is limited.[20 21]
Despite an increase in the number of studies addressing the relationships between work exposure to mechanical and psychosocial risks and LBP, the possibility of drawing conclusions is limited by the small number of prospective studies.[14 20] First, the effects of multiple factors are supported by evidence from few high-quality studies, and the consistency across studies is low.[14] Second, psychological demands may be confounded by physical demands and vice versa,[22] yet there have been few studies that take both of these factors into account.[20] Third, many studies have focused on specific occupations and conclusions from these studies may not be valid in the general working population due to specific sample characteristics. Moreover, when otherwise well-designed epidemiological studies are conducted in a specific homogenous population, risk factors may be interpreted falsely as non-significant due to low variance for a given risk factor within that population.[23]
The aim of the present study is to identify work-related psychosocial and mechanical factors that contribute to the risk of moderate to severe LBP. We simultaneously tested the impact of psychosocial and mechanical factors in work settings in a randomly selected 3-year prospective cohort in the general working population in Norway. Potential confounders such as age, sex, occupation, education and psychological distress[4 5] were taken into account.
Methods
Population
Data were provided from the nationwide study of living conditions–work environment, conducted by Statistics Norway (SSB). Data were collected during the two periods: from 18 September 2006 to 24 February 2007 and from 22 June 2009 to 09 January 2010. Data were collected by personal telephone interviews (0.5% of the completed interviews were face-to-face interviews). Prior to telephone contact, potential respondents were informed by mail about the study, the topic of the study and data/privacy protection.
Eligible respondents were community-living Norwegian residents aged 18–66 years. In 2006, this population was 2 941 281 persons. A gross sample of 18 679 was randomly drawn from this population. Of these, 12 550 (67%) persons were interviewed. Among those interviewed, 9961 were in paid work for at least 1 h during the reference week, or temporarily absent from such work. The baseline sample was compared with the gross sample according to the benchmarks of age, gender and region. No major differences were detected.[24]
The panel data, that is, those who participated in the survey in 2006 and 2009, consisted of 9371 persons (response rate=50.2% of the gross sample/74.7% of the baseline sample). Respondents who were in paid work for at least 1 h during the reference week, or were temporarily absent from such work, at both time points (n=6745) constituted the follow-up sample in the present paper.
Measurement
Psychosocial Measures. Job demands was measured by a single item, developed by SSB: ‘Do you sometimes have so much to do that you have to skip lunch, work longer than your normal working hours or take work home with you?’ (answer categories: every day, a couple of days a week, 1 day a week, a couple of days a month, less often or not at all). Monotonous work was measured by a single item: ‘Does your job consist of constantly repeated tasks, meaning that you do the same thing hour after hour?’ (answer categories: almost the whole time, around three-quarters of the working day, half of the working day, a quarter of the working day, seldom or never). Other dimensions were measured with scales from QPS Nordic.[25] Job control was measured with three items (α=0.69): (i) ‘To what extent can you decide the pace at which you work?’, (ii) ‘To what extent can you influence decisions that are important to your work?’ and (iii) ‘To what extent are you free to decide how to go about doing your work?’ Supportive leadership was measured with three items (α=0.71): (i) ‘If needed, how often can you get support and help from your immediate superior with your work?’, (ii) ‘Does your immediate superior appreciate your achievements at work?’ and (iii) ‘Does your immediate superior treat employees fair and impartially?’ Role conflict was measured with three items (α=0.64): (i) ‘How often do you have to do things that you think should be done in a different way?’, (ii) ‘How often are you given tasks without being given sufficient tools and resources to complete them?’ and (iii) ‘How often do you receive contradictory requests from two or more different people?’ The items were scored on a 5-point scale (very seldom or never, seldom, sometimes, rather often, quite often or always). Each factor was computed as the mean of valid responses. For all work-related psychosocial factors assessed, the 5-point-scale score was converted into three categories; low (1–2), medium (2.1–3) and high (3.1–5). In addition, all variables were analysed as continuous to test for linear trends. Change scores from baseline to follow-up were based on the categorical scores at baseline and follow-up and recoded into five exposure categories: constantly low, decreased (a change from medium or high exposure at baseline to low at follow-up), medium, increased (a change from low or medium exposure at baseline to high at follow-up) and constantly high.
Mechanical Exposure at Work. Perceived mechanical workload was measured with seven items. The items were developed by an expert group in a Nordic cooperation project,[26] and have been asked in a survey of living conditions since 1989: (i) ‘Are you, in your day-to-day work, exposed to vibrations that cause your whole body to shake, for example from a tractor, forklift truck or other piece of machinery?’ (Whole body vibration); (ii) ‘Do you work in positions where you are leaning forward without supporting yourself on your hands or arms?’ (Upper body forward bend); (iii) ‘Do you have to lift things in uncomfortable positions?’ (Awkward lifting); (iv) ‘Do you need to squat or kneel in the course of your work?’ (Squatting/kneeling) and (v) ‘Do you work standing up?’ (Standing) (Response categories: yes/no). ‘Yes’ respondents were asked to estimate the proportion of the working day during which they were exposed (response categories: almost the whole time, three-quarters of the working day, half of the working day, a quarter of the working day and very little of the working day). Scores were categorised into ‘none or very little of the working day’, ‘a quarter of the working day’, ‘half of the working day’ and ‘three-quarters of the working day or more’, and analysed as linear. In addition the scores were reported as pooled estimates (‘a quarter of the working day or more’). Heavy lifting was measured with a single item: ‘Do you have to lift anything that weighs more than 20 kg on a daily basis?’ (Response categories: 20 times a day or more, 5–19 times; 1–4 times, No). ‘Yes’ respondents were also asked: ‘Are you normally lifting people?’ (Response categories yes/no) (Lifting persons). Score changes from baseline to follow-up were based on the dichotomised scores (‘none or very little of the working day’ and ‘a quarter of the working day or more’) at baseline and follow-up recoded into four exposure categories: not exposed, decreased, increased, exposed at T1 and T2 (ie, exposed at baseline and at follow-up).
Other Variables. Occupation was based on an open questionnaire and coded by SSB into a professional title in accordance with the International Standard Classification of Occupations, and was recoded into 10 occupational groups. Educational level was based on administrative register data and was coded into five educational levels (see Table 1). Psychological distress was measured with two questions: ‘Have you during the past month been very bothered, quite bothered, slightly bothered or not bothered by dejection or depression?’ and ‘Have you during the past month been very bothered, quite bothered, slightly bothered or not bothered by nervousness, anxiety or restlessness?’ Psychological distress was dichotomised as being slightly or more bothered on either of the two questions.
Outcome. The outcome measure was the reported intensity of LBP during the 4 weeks prior to answering the questionnaire: ‘Have you, over the past month, been severely afflicted by, somewhat afflicted by, a little afflicted by or not afflicted at all by pain in the small or lower part of the back?’ Cases were defined as respondents who reported being severely or somewhat afflicted at follow-up. In total, 12.8% (861 individuals) were cases at follow-up, with 4.7% (319 individuals) being cases at both time points.
Statistics
In order to address the issue of possible selection bias related to dropouts between baseline and follow-up, we compared background characteristics and exposure levels in the baseline and follow-up samples. Moreover, the strength of the bivariate cross-sectional associations between exposure and LBP at baseline were calculated in the baseline sample and the follow-up sample to evaluate whether the strength of the association was likely to be influenced by selection related to dropout.
In the main analyses, LBP at 3-year follow-up was regressed on work-related mechanical and psychosocial factors at baseline (Table 2). Further, we regressed changes in work-related mechanical and psychosocial exposure from baseline to follow-up (Table 4). The associations were calculated as OR with 95% CI. Adjustments for potential confounders were made by logistic regression analyses. In model #1, we made adjustments for LBP at baseline, gender and age. In the fully adjusted model#2, further adjustments were made for educational level, occupation, psychological distress and work-related factors. To limit the potential for overadjustment each work-related predictor was adjusted only for other work-related predictors that were first estimated to exert an influence above a certain threshold level. This estimation was made a priori, based on the following procedure suggested by Rothman et al [27] applied to baseline data. In the first step, crude ORs were separately estimated for each work-related factor. In the second step, each of the other work-related variables was entered one at a time. If the inclusion of a potential confounder resulted in a change in the OR of 10% or more, that variable was treated as a real confounder in the multiple regression models. In additional analyses, fully adjusted ORs were estimated separately for respondents which at baseline were LBP cases and non-cases, respectively, in order to evaluate whether factors that may determine recurrent LBP are different from factors that may determine onset of LBP. We entered an interaction term between LBP at baseline and each of the work-related factors into model#2 to formally test if predictors of new and recurrent LBP were significantly different. All statistical analyses were conducted with PASW Statistics (formerly SPSS), V.19.0 (IBM, Armonk, New York, USA).
For statistically significant work-related factors in the regression analyses, we calculated the population attributable risk estimates (PAR) with 95% CI (Table 3). In contrast to OR estimates, the PAR estimate combines data on prevalence and a measure of association to provide a quantitative estimate of the proportion of cases in the population that is attributable to a particular exposure. Based on the method described by Natarajan et al [28] PAR was calculated using the formula (Pd*(OR−1/OR), where Pd is the proportion of cases (ie, respondents with moderate or severe LBP at follow-up in the present study) exposed to a risk factor and OR is the adjusted OR. The lower and upper limits of the 95% CI for PAR were calculated from the general PAR formula using the lower and upper limits of the 97.5% CI for Pd and OR. This constitutes the 95% Bonferroni CI for PAR. The 95% CI for PAR provides information regarding the significance of the result. PAR can only equal zero if the OR=1; thus, a test result of OR=1 is equivalent to a test result of PAR=0.
Results
Table 1 describes the distribution of sociodemographic variables and their association with LBP among respondents at baseline and at two time points. In the prospective sample, there is a slight under-representation of the youngest and oldest respondents. There are no sex differences in the two samples, but there is a slight under-representation of people with a low level of education. Correspondingly, there is a slight under-representation of ‘service workers and shop and market sales workers’, ‘plant and machine operators and assemblers’ and people working in ‘elementary occupations’.
The results in Table 1 show that the level of LBP was rather similar in the total baseline sample compared with the follow-up sample at baseline (13.0% and 11.7%, respectively). The associations between age, sex, education, occupation, psychological distress and LBP were largely similar across the two samples (Table 1). Correspondingly, the results show that the level of exposure was nearly identical across the two samples. The associations between different types of exposure and LBP in the two samples were also very similar across the two samples. The calculated OR for continuous variables varies by less than 10% (not shown).
Table 2 shows the results of multiple logistic analyses with baseline psychosocial and mechanical exposure as predictors and LBP intensity at follow-up as outcome. High job demands, low job control and monotonous work predicted LBP in model #1 (adjusted for LBP at baseline, age and sex). Estimates ranged from OR=1.22 (job demands) to OR=1.43 (job control). When evaluating the fully adjusted model, the effect of low levels of job control was reduced, but remained a borderline significant predictor (OR=1.26, 95% CI 1.01 to 1.57), whereas the effect of high job demands was increased (OR=1.41, 95% CI 1.16 to 1.72). A test for trend was statistically significant for job demands (p≤0.001). When we separately evaluated risk for LBP cases and non-cases at baseline, the effect of high job demands was lower among non-cases than among cases (OR=1.28, 95% CI 1.02 to 1.62 and OR=2.07, 95% CI 1.38 to 3.11, respectively). The difference was borderline non-significant when introducing the interaction term (p=0.052). No significant interactions were found for any of the other work-related psychosocial factors. No mechanical factors acted as confounders related to any of the psychosocial factors and there was no internal confounding between the psychosocial dimensions.
Awkward lifting, squatting/kneeling, prolonged standing, heavy lifting, lifting of persons and whole body vibration significantly predicted LBP at follow-up in model #1 adjusting for LBP at baseline, age and sex. Estimates ranged from OR=1.77 (awkward lifting) to OR=1.28 (lifting persons). When evaluating the fully adjusted model, the significant predictors were awkward lifting (OR, a quarter of the working day or more=1.55, 95% CI 1.28 to 1.88), squatting/kneeling (OR, a quarter of the working day or more=1.29, 95% CI 1.04 to 1.61) and standing (OR, three-quarters of the working day =1.48, 95% CI 1.20 to 1.83). A test for trend was statistically significant for standing at work (p≤0.001). No significant interactions were found between any of the work-related mechanical factors and LBP at baseline. There was some internal confounding between the mechanical dimensions. Lifting of persons was no longer significant when adjusted for standing, heavy lifting and awkward lifting.
The estimated PAR of LBP due to work-related psychosocial and mechanical factors is shown in Table 4. Based on the partially adjusted model #3 (Table 3), statistically significant PAR estimates were found for low job control, monotonous work, awkward lifting, squatting/kneeling, standing, heavy lifting and whole body vibration. Based on the fully adjusted estimates in model#2 (Table 3), the proportion of moderate to severe LBP attributable to work-related exposure was 41.9%. The highest PAR was estimated for high job demands (PAR=11.6, 95% CI 4.1 to 18.6) and prolonged standing (PAR=11.6, 95% CI 5.7 to 22.8). Other statistically significant mechanical factors were awkward lifting and squatting/kneeling.
Table 4 shows the results from the multiple logistic analyses with changes in psychosocial and mechanical exposure from baseline to follow-up as predictors and LBP at follow-up as outcome. In model #3 (adjusting for LBP at baseline, age and sex) significant associations with LBP were found for all psychosocial factors evaluated. In the fully adjusted model, increased and constant high job demands (OR=1.42, 95% CI 1.07 to 1.89 and 1.63, 95% CI 1.26 to 2.10, respectively) and decreased and constant high levels of role conflict (OR=1.83, 95% CI 1.27 to 2.62 and 2.16, 95% CI 1.39 to 3.36, respectively) were significant predictors.
In model #3, increased and/or constant high exposure across baseline and follow-up predicted LBP at follow-up for all mechanical exposures evaluated. When evaluating the fully adjusted model, the significant predictors were: increased and constant exposure to awkward lifting (OR=1.34, 95% CI 1.00 to 1.81 and 1.96, 95% CI 1.53 to 2.51, respectively), constant exposure to standing (OR=1.41, 95% CI 1.14 to 1.73), constant exposure to squatting/kneeling (OR=1.41, 95% CI 1.03 to 1.93) and increased exposure to whole body vibration (OR=1.98, 95% CI 1.33 to 2.96). There was some internal confounding between the mechanical dimensions. Lifting of persons and heavy lifting were no longer significant when adjusted for other mechanical factors.
Discussion
This study investigated the role of both psychosocial and mechanical factors at work in the development of moderate to severe LBP in the general working population. We estimated that about 42% of the cases with moderate to severe LBP were attributable to work-related exposure. The most important psychosocial factors were high job demands (PAR=11.6%) and low job control (PAR=4.9%). Mechanical factors were awkward lifting (PAR=8.3%), squatting/kneeling (5.5%) and standing most of the working day (PAR=11.6%). There were no indications of substantial confounding related to age, sex, education, occupation or psychological distress, and the risk for LBP associated with psychosocial exposure was not influenced by adjustment for mechanical risk factors, and vice versa. Moreover, our results indicate that predictors of new and recurrent or persistent LBP are not significantly different. Thus, work-related psychosocial and mechanical factors are independent risk factors that should be addressed when developing strategies to prevent the development or return of LBP. In accordance with what is generally shown in the literature, the relative risk estimates were small to moderate, but given that these types of exposures are prevalent, the total impact on LBP in the general working population is not negligible.
The strengths of this study are that it is a large nationwide study using random sampling, it uses a prospective design with a comprehensive set of exposures measured, it includes a thorough control of confounding and it has a high response rate. Despite a substantial proportion (33%) not responding at baseline, those who responded to the survey at baseline were not systematically different from those who did not respond across the benchmarks of age, gender and region. Moreover, the follow-up sample did not differ from the baseline sample with respect to the prevalence of LBP at baseline. The associations between different types of exposure and LBP in the two samples were also similar across the two samples. Thus, there is no reason to suspect any exposure or outcome bias due to dropout from baseline to follow-up, at least not to the extent that it impacts substantially on the observed results. On the other hand, we do not know whether people with poor health were less likely to respond at baseline. In addition, it is possible that the most vulnerable people had already left their jobs and were thus excluded from the cohort in this study. Both of these selection processes may lead to biased and most likely attenuated estimates, and thus threaten the internal validity. However, other studies have shown that health complaints do not differ between respondents and non-respondents[29] and that some differences do not produce biased risk estimates.[30]
All data were collected by self-report. Reporting bias (eg, due to negative affectivity) influencing exposure and outcome measures may have inflated the estimates. However, baseline pain adjustment eliminates reporting bias by biases that are stable across time or mediated by stable outcome measures. Moreover, adjusting for symptoms of psychological distress at baseline should further eliminate the risk of reporting bias due to negative affectivity or other personality dimensions related to psychological distress. Nevertheless, common method bias cannot be ruled out when interpreting the current results. On the other hand, baseline pain was related to all baseline exposures except job demands in the crude analyses. Thus, since baseline exposure may have caused baseline pain, overadjustment is plausible.
High job demand was a consistent predictor of LBP in the present study and represented a substantial population-attributable risk. Previous studies have reported increased, but statistically non-significant findings, with ORs ranging from 1.2 to 1.7.[8 9 31 32] The estimate provided in the present study is further supported by a recently published meta-analysis.[33] It has been suggested that high job demands may increase strain and subsequently, increase muscle tension or other physiological reactions that put individuals at a greater risk for developing LBP.[34 35] Our results indicate that high job demands may be more strongly related to recurrent or persistent LBP. This result needs to be confirmed in further population studies.
The present study supports the notion of a relationship between low level of job control and LBP. This result contradicts the conclusion of moderate evidence for no association between low job control and LBP made in a recently published review study.[21] However, our results are similar to recently published general working population studies from the Netherlands[36] and Denmark.[9] A possible interpretation may be that employees with higher job control may have more autonomy and thus take more breaks, providing relief to their muscles and reducing the risk of LBP.
In line with other general working population studies, we found low level of supportive leadership was not a significant predictor.[9 36] The results were less consistent for the other psychosocial factors assessed in the present study. Monotonous work has been indicated as an important risk factor in one general working population study[8] and was a statistically significant predictor in partially adjusted models in the present study. Role conflict was a statistically significant risk factor for LBP across time, for increased and constant high exposure levels. The assumption that exposure at baseline influences LBP at follow-up implies either stable exposure across time or a long-term effect after 3 years. Our results indicate that the effect of role conflict on LBP may be more acute. This result needs to be confirmed in further population studies.
Prolonged standing was the most important and consistent mechanical predictor and demonstrated a significant linear relationship with LBP in our analyses. Our finding is in keeping with a general population study from Denmark.[9] Other general working population studies with smaller sample sizes have reported increased risk estimates that were non-significant[8] or statistically significant among men[37] and women.[11] However, some studies have reported non-significant findings, but without reporting risk estimates.[38 39] Hence, prolonged standing should be considered as a risk factor in future population studies, preferably also taking into account exposure to prolonged standing or walking in non-occupational settings.
Lifting in an awkward position was a consistent risk factor for LBP in the present study. The effect of being exposed ‘a quarter of the working day or more’ was similar to that of being exposed ‘three-quarters of the working day or more’, indicating that even brief exposure to awkward lifting may influence LBP. Moreover, heavy lifting and lifting of persons were also significant predictors before adjusting for awkward lifting. This finding concurs with those of several prospective studies reporting an association between LBP and manually transferring patients,[12 13 38] whereas the association between lifting and LBP is less consistent.[8 9 32 37]
Squatting/kneeling was the third most important mechanical risk factor in the present study, which is in keeping with general working population studies from the UK[8] and Denmark.[9] For the other mechanical factors assessed in the present study, the results were less consistent. Working in a forward-bending position without support was not a statistically significant risk factor in the present study. Two studies have reported an association between occupational exposure to rotation and higher degrees of trunk flexion and disabling types of LBP,[32 40] whereas other studies have reported nil or non-significant associations for LBP symptoms.[8 32] Whole body vibration at baseline was associated with an increased risk in the partially adjusted model, but not in the fully adjusted model. The literature reports mixed findings regarding the importance of whole body vibration for LBP.[11 41]
Conclusion
This study provides evidence of a substantial relationship between work-related mechanical and psychosocial factors and the development of LBP. Highly demanding jobs, prolonged standing and awkward lifting appear as the most consistent and important predictors of LBP. Other work-related factors were low job control and squatting or kneeling. Interventions aimed at reducing the development or return of LBP in the general working population may benefit from focusing on a range of work-related mechanical and psychosocial factors.