This is the fourth publication on the results of the National Institute for Occupational Safety and Health (NIOSH) quadrennial Quality of Work Life Survey (QWL) related to prevalence of back pain and upper limb pain in the general working population. The QWL survey contains items that evaluate the individual, physical and psychosocial risk factors for musculoskeletal pain. With a 12-year window into the risk factors for musculoskeletal disorders (MSDs) this analysis of the most recent QWL data set provides information on changes (trends) in the American work force as well as the stability of some of the response indicators from the previous survey analyses. In the prior publication of this series³ it was concluded that physical exposure factors have been consistent indicators of musculoskeletal pain (back and arms) and that the association of psychosocial factors with musculokseletal outcomes may be more sensitive to economic conditions that have fluctuated over the 2002–2010 time period. This was more evident in the latter recessionary part of the 2000’s decade (2010 survey). The 2014 QWL survey analysis reported here is an opportunity to further examine the QWL survey results for work related musculoskeletal pain in the post-recessionary environment.

Previous publications in this series^1,2,3 have summarized literature on work related MSDs and there continues to be active research in this area with a focus on prevalence of MSDs by industry/occupation and intervention strategies to address risk factors and systematic reviews of their effectiveness. There have been additional relevant studies subsequent to the background described with the analysis of the 2010 QWL data³. The brief literature summary below is an update of key studies relevant to work related MSD prevalence published since the previous publication in this series.

Two studies by Yang et al4,5 report on neck and back pain using data from the National Health Interview Survey and 2010 Occupational Health Supplementary Survey with a sample size of 13,915 civilian workers ages 18–64. Results showed significant increases (ORs) in neck pain and low back pain with work-family imbalance, hostile work environment, and job insecurity. Significant organizational risk factors for neck pain were non-standard work arrangements, multiple jobs, and long work hours. Prevalence of neck pain was estimated at 14.3% and back pain at 25.7% for workers in the United States. Occupational classification showed no significant association with neck pain, however, respondents in two occupational categories, health care support and construction and extraction, reported greater prevalence of back pain.

Anderson and Oakman⁶ (2016) conducted a review of the prevalence and risk factors for work-related MSDs in allied health professions (physiotherapy, occupational therapy, sonography, medical imaging, and podiatry) using results for 22 studies representing 14 different countries. Lower back was the most commonly reported injury area and results showed that allied health professionals have a high risk of developing MSDs in their career with the highest risk in younger therapists.

Freimann et al7 conducted a cross-sectional study on 409 nurses at a University Hospital in Estonia to determine musculoskeletal pain prevalence and the relationship of psychosocial factors and mental health problems with pain reports. Seventy percent of participants reported at least one body part with musculoskeletal pain lasting longer than a day within the past year and 64% having pain in the last month. Lower back and neck were the sites most reported. Significant odds ratios for musculoskeletal pain were noted with several psychosocial factors (work demands, work pace, role conflicts, job dissatisfaction, work family conflicts, and with what the authors refer to as “somatic stress symptoms (stomach ache, headache, palpitations, tension in various muscles)”⁷ (p. 3).

A series of studies on food processing workers in Finland investigated the influence of physical (biomechanical), environmental (work climate), psychosocial⁸, age⁹, and sickness absence¹⁰ factors on musculoskeletal pain (MSDs). Two of these studies^8,9 surveyed Finnish food processing workers twice, at baseline in 2005 and at follow-up in 2009. The baseline group comprised 1201 respondents and at follow-up 734 workers. Results showed that multi-site pain was most prevalent at baseline (54%) and at follow-up with 69% reporting pain at baseline also reporting pain at follow-up¹⁰. Age and physical hazards and MSD risk were not consistent for each age group, which was stratified into younger (20–35), middle (36–49), and older (50 plus).⁹ At initial data collection, physical hazards were not associated with MSD risk for the younger group, but for the middle group, repetitive movements and awkward postures were associated with increased risk. In older workers, environmental factors were associated with increased risk. Authors concluded that the findings showed no clear pattern of prediction with age for MSD risk. The number of MSD absence days increased with pain sites.⁹ Single site pain did not predict absence, but multi-site pain did.

Gerr et al^11,12 reported the results from a prospective epidemiological study on the effects of physical risk factors, psychosocial stress and work organizational factors on MSDs in appliance manufacturing workers. The study ran from 2004 until 2007 with 318 workers continuing participation through the study. Clinical assessments, questionnaires, physical exposure measurements (electromyography, video-based posture analysis, ACGIH Hand Activity Level, Strain Index), and psychosocial measures (job content questionnaire, affectivity scale, job stress scale) were used to collect data for the outcome measures. The incidence rate for hand/arm symptoms was 58/100 person years (PY) and for neck/shoulder 54/100 PY. A statistically significant 7% increased risk of neck/shoulder symptoms was noted for each 1% of time with shoulder elevation >90°. Surprisingly, the separate physical exposure measurements did not show any significant associations, but the dichotomized Strain Index measure did for hand/arm symptoms and disorders. With the psychosocial measures, when compared to the low demand/high control category, significant hazard ratios (HRs) were reported for both the high demand/high control and high demand/low control categories for hand/arm disorders and symptoms. A significant HR was reported for low demand/low control with neck/shoulder disorders. Significant associations were between weekly stress level and neck/shoulder symptoms and disorders and weekly job change and neck/shoulder symptoms. No significant associations were found with co-worker support, supervisor support or negative affectivity.

Several systematic reviews have been performed that have provided evidential conclusions for the influence of physical and psychosocial risk factors on MSDs. Hoozemans et al¹³ reviewed observational studies to determine if workers that performed pushing/pulling activities at work have increased risk for upper extremity symptoms when compared to workers that perform less or no pushing/pulling activities at work. Results showed “strong evidence” of pushing/pulling significantly related to upper extremity and shoulder symptoms, “moderate evidence” for neck/shoulder symptoms and “insufficient evidence” or conflicting evidence for combinations of upper arm, elbow, forearm, wrist or hand symptoms.

An updated systematic review for preventing upper extremity MSDs reported that there was “strong evidence” that a resistance training exercise intervention can help prevent and manage upper extremity MSDs.¹⁴ There was a “moderate level of evidence” for the effectiveness of stretching programs, mouse use feedback and forearm supports, but there was also “moderate evidence” that EMG (electromyography) feedback, job stress management training, and office workstation adjustment had “no effect” on upper extremity MSD outcomes.

Lean production techniques, which the authors had previously suggested may be relevant to musculoskeletal symptoms³, have subsequently been the subject of systematic review.¹⁵ The potential intervention benefits of lean production techniques are autonomy, worker participation, empowerment, and job control and job enlargement. The author concluded that lean production promises did not appear to create challenging and fulfilling work, worker participation was limited and job autonomy depended more on social relations in a company. Implementation was uneven between countries and work sectors and in the automotive industry there was an increase in MSD symptoms with lean production because of increases in work pace and lack of recovery time. However, in services and other sectors, the effects of lean production were more positive if buffers such as job control and social support were present. Management style seems to be important with the implementation of lean production techniques.

Padula et al¹⁶ (2017) reviewed job rotation programs for the prevention of musculoskeletal disorders. The majority of studies featured automobile assembly functions and job rotation schedules with short-cycle work task times and repetitive movements. There was “weak evidence” supporting job rotation for the prevention of MSDs and “some evidence” (three studies) that job satisfaction was increased. A randomized control study of 497 workers conducted in Brazil evaluated job rotation in textile workers using sick leave as the primary outcome variable over a 12 month period, but also MSD symptoms, psychosocial factors, fatigue, general health, and productivity as additional secondary outcome measures.¹⁷ Results showed an increase in number of hours due to sick leave for both groups but there were no significant group differences and there were no significant differences in the secondary measures. In short, the job rotation program was not effective.

This paper extends the data analysis of the General Social Survey Quality of Work Life module to include 2014 data for investigating trends in the prevalence of back and upper extremity musculoskeletal symptoms and associated workplace factors. The analysis of the 2014 QWL data and its presentation is consistent with those previously reported^1,2,3 with three additional items added in 2014 to assess sleep problems and use of telework (working from home) and associations with musculoskeletal pain.

Consistent with the previous publications in this series^1,2,3 the 2014 data were collected as part of the biannual General Social Survey (GSS) conducted by the National Opinion Research Center (NORC) at the University of Chicago (http://gss.norc.org/). NORC describes the GSS as a nationally representative survey that has been conducted to monitor societal change and study American society. It is further described as the only full-probability, personal-interview survey designed to monitor changes in both social characteristics and attitudes currently being conducted in the United States. The Quality of Work Life (QWL) module is the module that was added to capture nationally representative opinions about work life. The survey is a face-to-face 90-minute survey administered to randomly selected, non-institutionalized U.S. adults 18 years and older. The National Institute for Occupational Safety and Health (NIIOSH) added the quadrennial QWL module to the GSS core survey (https://www.cdc.gov/niosh/topics/stress/qwlquest.html) administered to individuals who indicate they were employed or self-employed for pay in the week before the survey.³ Respondents were required to be working ≥ 20 hours per week or currently missing work because of vacation, illness, injury or on strike. The 2014 sample size was 1,124, an increase over 2010, but not as high as 2002 and 2006. These fluctuations likely reflect employment figures as the GSS samples approximately the same number of respondents each year.

As with the previous analyses in this series^1,2,3 primary outcomes were the prevalence of back pain and pain in arms through a yes/no response to “In the past 12 months, have you had back pain every day for a week or more?” and “In the past 12 months, have you had pain in the hands, wrists, arms, or shoulders every day for a week or more?” Individual, physical, and psychosocial/work organization factors that have been relevant to the occurrence of MSDs were analyzed. These variables chosen for analysis (listed in Table 1) were based on known relationships with back pain and upper extremity pain reported in the literature, or those that were tested in previous analyses of the 2002, 2006, and 2010 QWL data sets.^1,2,3 These variables were also described fully in Waters et al¹.

Changes in the questions over the time period include “How often work from home” responses being available for 2002–2014, but the “sleep problems” and “why work from home” questions are available only for 2010 and 2014. Another change has been the re-coding of occupations using the 2010 Census Occupational Classification System. This system provides a better definition of occupations with 11 classifications. Two classifications, “military related” and “farming, fishing, forestry” are only listed in Table 2. Both classifications are considered unreliable for data analysis because of the possibility that some respondents misread instructions and should not have answered civilian employment questions (military) and low respondent numbers (farming, fishing, forestry). With the change to the 2010 codes comparisons with the 8 codes used in previous publications are not available for backward comparison with the 1977 Quality of Employment Survey. All tables now include a trend analysis statistic (χ²) that shows any significant differences in trend for the outcome measures from the 2002 base year.

The data analysis procedures were similar to previous reports.^1,2,3 As a brief description: continuous distribution data were re-categorized into smaller ranges and some question items with highly correlated response choices were combined to create new variables (see Table 1 - supervisor support, safety climate). Odds ratios (ORs) were calculated using univariate logistic regression and multivariate, stepwise regression to identify statistically significant bivariate interactions. All univariate factors were forced into these models. Bivariate interaction ORs were calculated by combining the levels of two variables into one variable and univariate regression performed. Pearson Chi-squared tests were used to compare the outcome measurements by occupation for all survey years and Wald chi-squares from logistic regression models were used to assess trends for each risk factor and their response items by survey year. A statistical significance criterion of 0.05 is used in all tables and ORs with confidence intervals that do not include 1.0. All calculations were performed with SAS^® (Version 9.3, SAS Institute Inc., Cary, NC).

In the 2010 data analysis there was evidence that unemployment and overtime rates had some association with psychosocial risk factors. Thus, we report unemployment and overtime rates averaged for the months (March to September) in which the QWL survey interviews were conducted. Unemployment rates for 2002 were 5.77% for 2002; 4.64% for 2006;, 9.61% for 2010; and 6.22% for 2014. Overtime rates, as reported by the Bureau of Labor Statistics¹⁸ for 2002, 2006, 2010, and 2014, were 4.10 hrs, 4.44 hrs, 3.81 hrs., and 3.45 hrs, respectively.

Table 2 presents the frequency results and trend statistics for the outcome measures back pain and pain in arms by 11 occupational classifications for all four QWL survey administrations. There were, however, two significant changes in the percentage of respondents by occupation over the four surveys. Compared to the 2002 year, the percentage in “service” occupations increased from 16.6% in 2002 to 19.6% in 2014 (χ² = 10.8; df = 3; P = 0.013) and “production” occupations decreased from 8.2% in 2002 to 5.3% in 2014 (χ² = 14.0; df = 3; P = 0.003). In the 2014 survey reports of back pain and pain in arms did not differ significantly by occupational classification (back pain: χ² = 13.8; df = 10; P = 0.17; pain in arms: χ² = 16.4; df = 10; P = 0.09).

Data for 2014 showed a drop in reports of back pain at 22.8% (χ² = 12.1; df = 3; P = 0.007), pain in arms at 23.9% (χ² = 7.2; df = 3; P = 0.07), and “both pain” at 11.9% (χ² = 8.8; df = 3; P = 0.03), over the four survey reporting years. For the survey years 2004–2010 back pain reports averaged 27.3 %, “pain in arm” 27.8 % and “both pain” 15.4 %. Trend analysis results in Table 2 indicate few significant changes over the four survey years by “occupation”. The category “office/administrative” shows a significant decrease for back pain reports (χ² = 7.8; df = 3; P = 0.05) and “pain in arm” reports (χ² = 8.7; df = 3; P = 0.03). Professional and related also showed a significant decrease in “pain in arm” reports (χ² = 8.0; df = 3; P = 0.05), but there were no significant changes with “both pain” reports by occupation. Figures 1 and 2 show the reports of back pain and pain in arms by occupation over the four survey years. For easier illustration, the responses have been dichotomized into “Light” (very light, light) and hard (somewhat hard, hard and very hard) and averaged for the 2010 and 2014 years.

To examine potential trends in telework we examined the “never” telework responses from 2002–2014, assuming that a percent decrease in the “never” response indicated an increase in telework. In this analysis the Cochrane-Armitage trend test (Agresti, 2002) was used and the trend was not significant (Z = −1.35, df =3, P =0.18). Further examination of occupation data showed that for “management, financial, and business (Z = −2.31, df = 3, P = 0.02),” “office/administrative support (Z = −2.61, df = 3, P = 0.009,” and “construction and extraction (Z = −2.15, df = 3, P = 0.03)” the percent of “never” reports decreased significantly, but for “service (Z = 2.78, df = 3, P = 0.005)” and “sales and related (Z = 2.18, df = 3, P = 0.03)” the percent “never” responses increased significantly.

A key finding in the 2014 QWL analysis is the decrease in prevalence of back pain, pain in arms, and both back and arm pain over the four survey waves by an average of almost five percent from 2002 to 2014. Yang et al^4,5 reported a prevalence of low back pain (defined by that occurring “during the past three months”) from the National Health Interview Survey (25.7%) that was, and is still, very consistent with that found in the present series of QWL surveys, which dropped below 25% between the 2010 and 2014 survey administrations. Ferguson et al¹⁹ recently reported a comparable back pain prevalence of 25% across a consortium of pooled studies including 2000 workers participating from six US states with jobs in a variety of distribution centers and manufacturing facilities that involved manual materials handling.

Although the decline in back pain appears to be fairly consistent across all occupations there were some trends in terms of overall survey responses by occupation when related to physical effort ratings. The percentage of participants in Management, Financial, and Business, Professional and Related, Service Occupations, Sales and Related, and Office and Administrative Support - those occupations reporting less physical effort at work - increased from 76.2% to 79.2% between 2002 and 2014. Conversely, the number of respondents in Construction and Extraction, Installation, Maintenance, and Repair, Production, and Transportation and Material Moving - occupations with greater physical effort ratings - decreased from 23% to 18.8%. The only occupational groups which show significant participation rate changes from 2002 and 2014 were for Service which increased from 16.6 % to 19.6 % and Production which decreased from 8.2 % to 5.3 % of the respondents. This shift in distribution away from occupations with greater ratings of physical effort may explain an overall decline in pain reports. Other explanations for decreased reports of musculoskeletal pain have been described. For example, Wang et al²⁰ reported a significant drop (66%) in work-related musculoskeletal disorders (WMSDs) in the U.S. construction industry from 1992 to 2014 which was attributed to intervention efforts, changes in reporting requirements (Occupational safety and Health Administration), and possible under reporting in the industry. Interestingly, the four QWL survey analyses do not show a significant decline in back pain or pain in arms prevalence in the Construction and Extraction occupation category. Mustard et al²¹ reported a decreasing trend in the Ontario musculoskeletal disorder prevalence during a similar analysis time period (2004–2011). Their study integrated three independent population-based data sources to describe trends in the incidence of work-related MSDs over an 8-year period. The authors observed decreases in MSD prevalence through emergency department utilization and workers’ compensation claims and parallel decreases through the Ontario health interview surveys that are comparable to and consistent with the present QWL findings.

Further examination of the trend analysis results show that statistically significant changes in pain reports over the four survey waves, by occupation, were minimal. Only the categories Office/Administrative and Professional and Related showed a significant decrease in reports of pain in arms. These declines may reflect the use of strategies for prevention in office environments (e.g., ergonomically designed furniture, mini-breaks and other strategies to increase mobility in sedentary work).

Having sustained an injury at work (“hurt at work”) in the last year exhibits significant ORs with back pain and pain in arms for all reporting years. The trend statistic for the zero times hurt at work comparison choice was significant for back pain and showed a pattern of a decrease in reports. A similar pattern was also evident pain in arms, though not reaching statistical significance (p = 0.056). Three general indicators of health (physical health, mental health, health days) have consistently exhibited significant ORs with back pain and “pain in arm” for the same response choices for all reporting years. Significant trend statistics were noted for the comparison choices of 0–13 days of poor health versus greater than14 days (of the prior 30 days) of poor health. With the comparison choice there was a decrease in reports, but with the ≥14 day choice it was an increase. Sleep problems showed significant ORs on the “sometimes” and “often” choice for both back pain and pain in arms.

A second key finding is that the physical exposure risk variables “heavy lifting” and “hand movement” continue to show significant ORs for back pain and pain in arms in the 2014 survey. None of the other risk factors exhibit such consistent risk associations across all survey years. This result is in agreement with many reviews of evidence for the relationship between physical risk factors and MSDs.^22,23,24 The physical exposure risk factor “physical effort” has also shown significant ORs for back pain and pain in arms when dichotomized as “hard” versus “light” physical effort.

Most psychosocial and work organization factors are consistent with previous year’s results and there were some significant trend statistics that are consistent with the overall decrease in both back pain and pain in arm reports since the 2002 base year. Low “job satisfaction”, lack of ”supervisor support”, poor “safety climate”, and always having “work stress” all showed significantly reduced associations with back pain and pain in arms over the four survey years. Reduced job satisfaction level, however in 2014, was significant for pain in arms with only the “somewhat satisfied” level and not the higher levels of non-satisfaction as in previous years. For back pain all levels of job satisfaction less than “very satisfied” were significant. Psychosocial factors that have not consistently shown significant associations with the same response items were “work freedom,” “work hours” and “work schedule,” for back pain and “work hours” and “work schedule” for “pain in arms.” Work freedom (“freedom to decide how to do my own work”), while not showing any significant relationships with back pain since 2002, has shown significant protective associations (OR < 1) with pain in arms. Significant trends were noted for the comparison choices on back pain for “supervisor support”, “work hours”, “safety climate”, and “work schedule” for back pain and “work hours”, and “work schedule” for “pain in arms.” All reflected decreases in pain reports. The only increases in pain reports were on isolated response choices with “work schedule,” (rotating shift for pain in arms and night shift for back pain) and “not too true” with “work freedom” for “back pain.”

Strengths of this study are the national representativeness of the GSS and that the survey was administered to participants directly through face-to-face interviews. This approach is believed to lessen the likelihood that respondents would have misunderstood any questions or have been confused by response options. An additional strength is the consistency in the survey administration over the 2002, 2006, 2010, and 2014 surveys, which exhibit consistent statistical associations between musculoskeletal pain and physical exposure variables. Limitations of the study include that the QWL survey is based on recall of self-reported back and arm pain and does not confirm more specific criteria consistent with a clinical musculoskeletal diagnosis. Secondly, the workplace physical exposure factors were also obtained through self-report, rather than measurement, and inconsistencies in subjective interpretation between respondents is a threat to interpretation of those findings. Thirdly, the cross-sectional design of the survey in successive administration years does not establish causality, as the successive survey administrations are not from a consistent cohort.

From the 12-year period covering the QWL survey years 2002–2014, it can be concluded that physical exposure risk factors of heavy lifting (repeated lifting, pulling or pushing) and hand movement (repetitive or stressful hand movements or awkward postures) are associated with increased risk of back pain and pain in arms. These physical risk factors continue to have significant odds ratios for increased reports of back pain and pain in arms – consistent with prior survey reporting years. Over the analysis period psychosocial and work organizational factors associated with back and arm pain prevalence include mandatory working of extra hours (“must work”), reporting frequent work stress, not having enough time to get work done, low job satisfaction, low supervisor support, and reporting a poor safety climate in the work environment. Having high “work freedom” suggests a protective effect mostly with decreased prevalence of pain in arms. Individual factors associated with increased back pain and pain in arms are having sustained a previous injury at work and general indicators of overall health (e.g., physical health, mental health, health days, and sleep problems).