The NIOSH Lifting Equation and Low-Back Pain,Part 2

Abstract

Objective:

The aim of this study was to investigate the relationship between the revised NIOSH lifting equation (RNLE) and risk of seeking care for low-back pain (SC-LBP).

Background:

The RNLE is commonly used to quantify low-back physical stressors from lifting/lowering of loads in workplaces. There is no prospective study on relationship between RNLE and SC-LBP.

Method:

A cohort of 258 incident-eligible workers from 30 diverse facilities was followed for up to 4.5 years. Job physical exposures were individually measured. Worker demographics, medical history, psychosocial factors, hobbies, and current low-back pain were obtained at baseline. The cohort was followed monthly to ascertain SC-LBP and quarterly to determine changes in physical exposure. Associations between SC-LBP and both the peak lifting index (PLI) and peak composite lifting index (PCLI) were tested in multivariate models using proportional hazards regression.

Results:

SC-LBP lifetime prevalence at baseline was 31.9%, and there were 24 incident cases during follow-up. Factors predicting SC-LBP included job physical exposure (PLI and PCLI), history of low-back pain, age, female gender, and lower body mass index. In adjusted models, risk (hazard ratio [HR]) increased per unit increase in PLI and PCLI (p = .03 and .02, and maximum HR = 23.0 and 21.9, respectively). Whereas PCLI suggested a continuous increase in risk with an increase in PCLI, the PLI showed elevated, though somewhat reduced, risk at higher exposures (HR = 14.9 at PLI = 6).

Conclusion:

Job physical stressors are associated with increased risk of SC-LBP. Data suggest that both the PLI and PCLI are useful metrics for estimating exposure to job physical stressors.

Keywords

epidemiology ergonomics occupational cohort job analysis risk assessment

Introduction

Nonspecific low-back pain (LBP) is a common problem in workplaces (Bureau of Labor Statistics, 2011; Waddell & Burton, 2001). Several different definitions have been used to determine prevalence, incidence, and risk factors for work-related LBP, and there is no consensus definition of LBP. Some definitions of LBP include (a) discomfort, ache, or pain in lumbar area of any intensity and of any duration; (b) minimum duration LBP (such as LBP lasting at least 1, 3, 7, or 30 days); (c) LBP with certain minimum intensity (≥3/10, ≥6/10, etc., on pain scale); (d) seeking medical care for LBP (visit to a health care provider, often referred to as LBP with health care utilization); (e) seeking treatment for LBP (nonprescription and prescription medication, other medical treatment); (f) LBP resulting in sick leave or absenteeism; and (g) LBP radiating down one or both legs (Ozguler, Leclerc, Landre, Pietri-Taleb, & Niedhammer, 2000). Seeking medical care for LBP or a visit to health care provider is also referred as “care-based” LBP or episode (de Vet et al., 2002). LBP is one of the most common reasons for seeking care (Waddell & Burton, 2001). There is some evidence to suggest that the risk factors for LBP may vary with the LBP case definition used (Ozguler et al., 2000).

Wadell and Burton (2001) concluded that seeking care due to LBP (SC-LBP) depends more on complex individual and work-related psychosocial factors than on physical demands of work. However, most past studies on SC-LBP are cross-sectional, are community or clinic based, have involved patients with chronic LBP, and/or have relied on self-administered questionnaire to ascertain job physical demands (Adamson, Hunt, & Nazareth, 2011; Alexopoulos, Tanagra, Konstantinou, & Burdorf, 2006; Cote, Baldwin, & Johnson, 2005; Freburger et al., 2009; IJzelenberg & Burdorf, 2004, 2005; Kent & Keating, 2005; Mortimer, 2003; Ozguler et al., 2000; Vingard et al., 2000; Walker, Muller & Grant, 2004). Authors of only a few studies have investigated the association between biomechanical stressors and SC-LBP, and those results are inconsistent (Alexopoulos et al., 2006; IJelenberg and Burdorf, 2004; Plouvier, Leclerc, Chastang, Bonenfant, & Goldberg, 2009; Vingard et al., 2000). Plouvier et al. (2009) suggested that SC-LBP is a consequence that could be related to occupational exposure. A case-referent population-based study showed that current and past physical and psychosocial risk factors were gender specific and had a moderate impact on SC-LBP (Vingard et al., 2000). However, exposure assessments in the study were made through questionnaire and interviews rather than through direct observation and measurement. We were unable to find any prospective workplace studies in the literature that have reported on the association between measured job physical exposure and SC-LBP.

This study’s hypothesis is that there is a relationship between job physical factors, measured by the revised NIOSH lifting equation (RNLE), and risk of SC-LBP in a cohort free of LBP at baseline.

Method

This study was approved by the University of Wisconsin–Milwaukee Institutional Review Board (#04-02-049). Workers’ demographics, psychosocial factors, hobbies and physical activities outside of work, LBP history, and LBP status were determined at baseline by a health care assessment team. Similarly, a job assessment team quantified job physical exposure at baseline. These teams were blinded to one another throughout data collection. Workers were followed monthly to determine changes in SC-LBP status and quarterly to measure changes in job physical exposure. Details of data collection and instruments used are provided in Part 1 (Garg et al., 2014).

Case Definition

Cases were defined as those workers seeking care from any caregiver for a new episode of LBP. Caregivers were defined as physicians (MD/DO), chiropractors, physical therapists, occupational therapists, and/or massage therapists.

RNLE Analyses

Lifting Index (LI) and Composite Lifting Index (CLI) were computed for each subtask and task that a worker performed at baseline as well as for those tasks that changed during the follow-up period. Using Waters, Putz-Anderson, and Garg’s (1994) methodology, we calculated LI for each unique combination of weight, horizontal and vertical location of hands, travel distance, grasp, asymmetric angle, and frequency of lifting/lowering both at origination and destination (subtask). We computed CLI for each task from LIs for subtasks using the procedure recommended by Waters et al. The CLI included all lifts and lowers performed in a task. For example, for grocery warehouse order selectors, the task was one order, and that task included all lifts and lowers performed in that grocery order. Within a task, an LI was calculated for each unique combination of object weight, horizontal location, vertical location, travel distance, and type of grasp. In those situations in which lifts/lowers exceeded Waters et al.’s recommended values, the corresponding multipliers were capped at their maximum penalties (e.g., horizontal multiplier was capped at 0.4 for horizontal distances greater than 63 cm). We treated LIs and CLIs as continuous variables. Then, using the Waters et al. prescribed cut points, we classified LI score into one of the three categories: LI ≤ 1.0, 1.0 < LI ≤ 3.0, and LI > 3.0. The same procedure was used to classify CLI.

Statistical Analyses

Baseline prevalence of LBP was determined, and those workers with LBP at, or during the 90 days prior to, baseline were subsequently excluded from analyses (i.e., LBP-free cohort). We modeled time from enrollment into the study to first event of SC-LBP using Cox proportional hazards (PH) regression (Cox, 1972). To account for changes in job physical exposure during follow-up, we determined unadjusted univariate hazard ratios (HR) for incident cases of LBP and 95% confidence intervals for PLI and PCLI using time-varying covariates. All other covariates, such as age, gender, body mass index (BMI), hobbies, psychosocial factors, and past medical history, were treated as time-independent covariates (see Part 1). We performed analyses using the coxph() function in R-64 Version 2.13.1 for Macintosh (R Development Core Team, 2011). The statistical procedure used in the current paper was similar that used in Part 1.

BMI appeared to have a linear relationship with SC-LBP; however, age, PLI, and PCLI all exhibited a strong linear trend in the mid-regions followed by either a smaller upward trend, a leveling off, or a downward trend for the remaining values (i.e., nonlinear relationship with SC-LBP). To model the nonlinearity in the upper regions of these variables, we included a single linear spline term for each variable (see Part 1). We used the closest quantile of cases to this point as the knot for the linear spline (i.e., the 65th percentile for age, 50th percentile for PLI, and 50th percentile for PCLI). In simple terms, these linear splines are two line segments connected by a single point (called a knot). The slopes of the line segments were allowed to be different, thus allowing for a nonlinear relationship between PLI or PCLI and risk of SC-LBP (measured as log[HR]).

We built multivariate models to test whether PLI and PCLI were associated with increased risk of SC-LBP after controlling for potential confounders and/or effect modifiers (covariates). First, we built a covariate model using potential covariates that (a) were not collinear, (b) were biologically plausible, and (c) had a univariate p value ≤.20 (Table 1). All covariates were treated as time independent and formed a pool of candidate variables for inclusion in the final covariate model. Physical exposure variables (PLI and PCLI) were withheld during the covariate model building process. We determined the final covariate model using a best subsets variable selection procedure with the corrected Akaike information criterion (AICc score; Collett, 2003; Hurvich & Tsai, 1989). Age, gender, and BMI were forced into every covariate model. Then, PLI and PCLI were separately forced into the final covariate model to determine their association with SC-LBP.

Table 1:

Univariate Hazard Ratios for Peak Subtask Lifting Index (PLI) and Peak Task Composite Lifting Index (PCLI) Covariates (n = 258)

Variable (Overall p Value)	Categories	n (Cases)	HR^a [95% CI]	p Value
Demographic
Age (years) (p = .09)	Linear spline terms
	Per year increase for age ≤ 42.7	176 (15)	0.93^b [0.86, 1.00]	.04
	Per year increase for age > 42.7	82 (9)	1.19^b [1.01, 1.39]	.03^c
Gender	Male	173 (13)	1.0	—
	Female	85 (11)	1.8 [0.82, 4.09]	.14
Anthropometric
BMI (kg/m²)	Continuous	258 (24)	0.93 [0.86, 1.00]	.04
Medical history
High blood pressure	No	221 (22)	1.0	—
	Yes	37 (2)	0.5 [0.13, 2.32]	.37
High cholesterol	No	221 (17)	1.0	—
	Yes	37 (7)	2.2 [0.91, 5.30]	.10
History of LBP^d	No	193 (10)	1.0	—
	Yes	65 (14)	4.5 [1.98, 10.06]	<.001
LBP radiating to leg(s)	No	234 (20)	1.0	—
	Yes	24 (4)	2.1 [0.72, 6.13]	.21
Sought care for LBP	No	208 (15)	1.0	—
	Yes	49 (9)	2.3 [0.99, 5.12]	.06
Modified duty for LBP	No	244 (20)	1.0	—
	Yes	14 (4)	3.4 [1.16, 9.92]	.05
Lost time for LBP	No	244 (22)	1.0	—
	Yes	14 (4)	1.9 [0.44, 1.94]	.44
Received workers’ compensation for LBP	NoYes	251 (22)7 (2)	1.03.6 [0.83, 15.18]	—.15
Psychosocial
APGAR composite	Continuous	258 (24)	1.08 [0.95, 1.23]	.24
Zung composite^e	Continuous	258 (24)	1.02 [0.91, 1.14]	.78
Tense-Edge-Nervous composite	Continuous	258 (24)	1.16 [0.81, 1.66]	.41
Hobbies/activities^f
Swimming	No	236 (19)	1.0	—
	Yes	22 (5)	3.6 [1.35, 9.71]	.03
Biomechanical stressors
PLI (p = .08)	Linear spline terms
	Per unit increase for PLI ≤ 2.2	138 (12)^g	3.68^h [1.00, 13.52]	.05
	Per unit increase for PLI > 2.2	120 (12)^g	0.70^h [0.37, 1.35]	.05^c
PLI	PLI ≤ 1.8PLI > 1.8	88 (6)^g170 (18)^g	1.01.8 [0.72, 4.58]	—.21
PCLI (p = .07)	Linear spline terms
	Per unit increase for PCLI ≤ 2.6	141 (12)^g	2.80ⁱ [1.04, 7.57]	.04
	Per unit increase for PCLI > 2.6	117 (12)^g	0.85ⁱ [0.55, 1.31]	.06^c
PCLI	PCLI ≤ 2.2	112 (6)^g	1.0	—
	PCLI > 2.2	146 (18)^g	2.8 [1.11, 7.04]	.03

Note. HR = hazard ratio; CI = confidence interval; BMI = body mass index; LBP = low-back pain.

HR of 1.0 with no CI indicates reference category for the variable.

For age, ${\hat{b}}_{1} = - 0.0782$ , and ${\hat{b}}_{2} = 0.1701 . HR = e^{b_{1} \cdot age}$ and $HR = e^{b_{1} \cdot age + b_{2} (age - 42.7)}$ for age ≤ 42.7 and > 42.7, respectively. HR per unit increase in age is $e^{b_{1}}$ for age ≤ 42.7 and $e^{b_{1} + b_{2}}$ for age > 42.7.

This p value is for the second spline term and represents a test for change in slope at the knot (i.e., b₂ = 0). Thus, this p value does not correspond to the given confidence interval, which is for the HR (i.e., $e^{b_{1} + b_{2}}$ ) beyond the knot.

History of LBP was defined as LBP resulting in workers’ compensation, lost time, light duty, or visit to a care provider or LBP radiating down the leg(s).

Modified Zung Depression Scale; please see the Appendix of Part 1 for details (Garg et al., 2014).

Aerobics, baseball, basketball, bicycling, football, motorcycling, remodeling, running, soccer, tennis, and woodworking showed no statistically significant association with increased risk of any LBP (p > .20).

For cases, n is based on exposure at the time a worker became a case. For noncases, n is based on exposure at baseline.

For PLI, ${\hat{b}}_{1} = 1.3039$ , and ${\hat{b}}_{2} = - 1.6564$ . $HR = e^{b_{1} \cdot PLI}$ and $HR = e^{b_{1} \cdot PLI + b_{2} (PLI - 2.2)}$ for PLI ≤ 2.2 and > 2.2, respectively. HR per unit increase in PLI is $e^{b_{1}}$ for PLI ≤ 2.2 and $e^{b_{1} + b_{2}}$ for PLI > 2.2.

For PCLI, ${\hat{b}}_{1} = 1.0310$ , and ${\hat{b}}_{2} = - 1.1963$ . $HR = e^{b_{1} \cdot PCLI}$ and $HR = e^{b_{1} \cdot PCLI + b_{2} (CPLI - 2.6)}$ for PCLI ≤ 2.6 and > 2.6, respectively. HR per unit increase in PCLI is $e^{b_{1}}$ for PCLI ≤ 2.6 and $e^{b_{1} + b_{2}}$ for PCLI > 2.6.

Results

The cohort for this paper included 258 workers who were LBP free for ≥90 days at baseline and were eligible to develop an incident case of SC-LBP. The median observation time for this study’s report was 19.3 months, with an average of 20.4 ± 11.6 (range = 0.7 to 53.7) months. Demographics and the descriptive statistics of the incident eligible cohort are provided in Part 1.

Lifetime prevalence of SC-LBP at baseline was 31.9% (164 workers out of 514 workers). During follow-up (average 20.4 months), there were 24 new SC-LBP cases (n = 11, 12.9% of females, and n = 13, 7.5% of males) resulting in an incident rate of 5.5 cases per 100 person-years. Females had a higher incident rate than did the males, 7.8 versus 4.4 per 100 person-years, respectively. A majority of cases (21) believed their LBP was either work related (14 cases) or of an “unsure cause” (n = 7). Three of the 24 cases believed their LBP was caused by something outside of work. The mean pain rating at the time these 24 cases sought care was 6.1 ± 2.1 (range = 3 to 10, on a 10-point scale). Lumbar paraspinal region was the most common body region for LBP, with 13 cases experiencing pain only in this region and 20 experiencing pain in this and other regions. Pain lasted a median of 135 days (mean = 235 ± 267 days, range = 11 to 1,238), ending due to either resolution of the pain (n = 19, resolution of pain defined as LBP free for ≥90 consecutive days), departure from the study (n = 3), or the study’s conclusion (n = 2). One out of three cases sought care from multiple specialties. Fifty percent of cases sought care from physicians, 36% from physical or occupational therapists, 25% from chiropractors, and 16.6% from massage therapists (total percentage is greater than 100% due to multiple caregivers involved).

During follow-up, there were 14 incident cases among the 65 workers with history of LBP (21.5%). There were 7 SC-LBP cases among the 37 workers with high cholesterol (18.9%) (Table 1).

Descriptive statistics for covariates and baseline job physical exposure variables are provided in Part 1.

Univariate Results

Age, PLI, and PCLI were fit with linear spline functions with a single knot (K), resulting in two spline terms for each variable. Estimated HR for first spline terms reported in Table 1 as well as in multivariate tables is $HR = e^{(\hat{b} i)}$ for each unit increase in x up to x = K. Estimated HR for each unit increase in x for x > K is $HR = e^{({\hat{b}}_{1} + {\hat{b}}_{2})}$ (see Garg et al., 2012, for details). For the first spline term, the reported HR represents per each unit increase in risk below the knot. For the second spline term, the reported HR represents per each unit increase in risk above the knot (with a reference point at or above the knot). Equations are provided in Tables 2 and 4 to calculate HR (relative to unexposed) for PLI and PCLI above the knot. The reported p values for the second spline terms in Table 1 as well as multivariate tables are for a test of change in slope at the knot (i.e., test for b₂ = 0) in the linear spline transformed PH model. Thus, this p value does not correspond to the reported confidence intervals (CIs) for second spline terms (i.e., 95% CI for HR = $e^{b_{1} + b_{2}}$ ) in Tables 1 through 5.

Table 2:

Multivariate Model for Risk of SC-LBP With the Peak Subtask Lifting Index (PLI) as Continuous Variable

Variable (Overall p Value)^a	Category/Function	n (Cases)	HR	95% CI	p Value
PLI (p = .03)^a	Spline terms
	Per unit PLI ≤ 2.2	138 (12)^b	4.16^c	[1.12, 15.52]	.03
	Per unit PLI >2.2	120 (12)^b	0.89^c	[0.44, 1.82]	.08^d
	HR for PLI = 2.2		4.16^2.2 = 23.01
	HR for PLI > 2.2		23.01*0.89^(PLI–2.2)
Covariates
Age (years) (p = .07)^a	Spline terms
	Per unit age ≤ 42.7	176 (15)	0.92^e	[0.85, 1.00]	.04
	Per unit age > 42.7	82 (9)	1.11^e	[1.00, 1.22]	.03^d
Gender (p = .03)^a	Male	173 (13)	1.0	—	—
	Female	85 (11)	2.80	[1.11, 7.03]	.03
BMI (p = .19)^a	Continuous
	Per unit score	258 (24)	0.95	[0.87, 1.03]	.21
History of LBP (p < .001)^a	No	193 (10)	1.0	—	—
	Yes	65 (14)	4.92	[2.14, 11.28]	<.001

Note. SC-LBP = seeking care for low-back pain; HR = hazard ratio; CI = confidence interval; BMI = body mass index.

Overall significance associated with including each variable in the model using the likelihood ratio test.

For cases, n is based on exposure at the time a worker became a case. For noncases, n is based on exposure at baseline.

For PLI, ${\hat{b}}_{1} = 1.4262$ , and ${\hat{b}}_{2} = - 1.5411$ . $HR = e^{b_{1} \cdot PLI}$ and $HR = e^{b_{1} \cdot PLI + b_{2} (PLI - 2.2)}$ for score ≤ 2.2 and > 2.2, respectively. HR per unit increase in PLI is $e^{b_{1}}$ for score ≤ 2.2 and $e^{b_{1} + b_{2}}$ for PLI > 2.2.

This p value is for the second spline term and represents a test for change in slope at the knot (i.e., b₂ = 0). Thus, this p value does not correspond to the given CI, which is for the HR (i.e., $e^{b_{1} + b_{2}}$ ) beyond the knot.

For age, ${\hat{b}}_{1} = - 0.0814$ , and ${\hat{b}}_{2} = 0.1824$ . $HR = e^{b_{1} \cdot age}$ and $HR = e^{b_{1} \cdot age + b_{2} (age - 42.7)}$ for age ≤ 42.7 and > 42.7, respectively. HR per unit increase in age is $e^{b_{1}}$ for age ≤ 42.7 and $e^{b_{1} + b_{2}}$ for age > 42.7.

Table 3:

Multivariate Model for Risk of SC-LBP With the Peak Subtask Lifting Index (PLI) as Categorical Variable With Cut Point Equal to 25th Percentile of Cases (PLI ≤ 1.8)

Variable (Overall p Value)^a	Category/Function	n (Cases)	HR	95% CI	p Value
PLI (p = .06)^a	PLI ≤ 1.8	88 (6)^b	1.0	—	—
	PLI > 1.8	170 (18)^b	2.40	[0.90, 6.21]	.08
Covariates
Age (years) (p = .09)^a	Spline terms
	Per unit age ≤ 42.7	176 (15)	0.93^c	[0.86, 1.00]	.06
	Per unit age > 42.7	82 (9)	1.19^c	[1.01, 1.40]	.04^d
Gender (p = .02)^a	Male	173 (13)	1.0	—	—
	Female	85 (11)	2.68	[1.16, 6.12]	.02
BMI (p = .17)^a	Continuous
	Per unit score	258 (24)	0.95	[0.88, 1.03]	.20
History of LBP (p < .001)^a	No	193 (10)	1.0	—	—
	Yes	65 (14)	4.83	[2.10, 11.10]	<.001

Note. SC-LBP = seeking care for low-back pain; HR = hazard ratio; CI = confidence interval; BMI = body mass index.

Overall significance associated with including each variable in the model using the likelihood ratio test.

For cases, n is based on exposure at the time a worker became a case. For noncases, n is based on exposure at baseline.

For age, ${\hat{b}}_{1} = - 0.0733$ , and ${\hat{b}}_{2} = 0.1747$ . $HR = e^{b_{1} \cdot age}$ and $HR = e^{b_{1} \cdot age + b_{2} (age - 42.7)}$ for age ≤ 42.7 and > 42.7, respectively. HR per unit increase in age is $e^{b_{1}}$ for age ≤ 42.7 and $e^{b_{1} + b_{2}}$ for age > 42.7.

Table 4:

Multivariate Model for Risk of SC-LBP With Peak Task CLI (PCLI) as Continuous Variable

Variable (Overall p Value)^a	Category/Function	n (Cases)	HR	95% CI	p Value
PCLI (p = .02)^a	Spline terms
	Per unit PCLI ≤ 2.6	141 (12)^b	3.14^c	[1.17, 8.38]	.02
	Per unit PCLI >2.6	117 (12)^b	1.03^c	[0.69, 1.55]	.06^d
	HR for PCLI = 2.6		3.14^2.6 = 19.59
	HR for PCLI > 2.6		19.59*1.03^(PCLI–2.6)
Covariates
Age (years) (p = .07)^a	Spline terms
	Per unit age ≤ 42.7	176 (15)	0.92^e	[0.85, 1.00]	.05
	Per unit age > 42.7	82 (9)	1.11^e	[1.01, 1.22]	.03^d
Gender (p < .01)^a	Male	173 (13)	1.0	—	—
	Female	85 (11)	3.35	[1.38, 8.13]	<.01
BMI (p = .14)^a	Continuous
	Per unit score	258 (24)	0.95	[0.87, 1.02]	.17
History of LBP (p < .001)^a	No	193 (10)	1.0	—	—
	Yes	65 (14)	4.84	[2.12, 11.04]	<.001

Note. SC-LBP = seeking care for low-back pain; HR = hazard ratio; CI = confidence interval; BMI = body mass index.

Overall significance associated with including each variable in the model using the likelihood ratio test.

For cases, n is based on exposure at the time a worker became a case. For noncases, n is based on exposure at baseline.

For PCLI, ${\hat{b}}_{1} = 1.1433$ , and ${\hat{b}}_{2} = - 1.1097$ . $HR = e^{b_{1} \cdot CLI}$ and $HR = e^{b_{1} \cdot PCLI + b_{2} (PCLI - 2.6)}$ for score ≤ 2.6 and > 2.6, respectively. HR per unit increase in PCLI is $e^{b_{1}}$ for score ≤ 2.6 and $e^{b_{1} + b_{2}}$ for PCLI > 2.6.

For age, ${\hat{b}}_{1} = - 0.0780$ , and ${\hat{b}}_{2} = 0.1841$ . $HR = e^{b_{1} \cdot age}$ and $HR = e^{b_{1} \cdot age + b_{2} (age - 42.7)}$ for age ≤ 42.7 and > 42.7, respectively. HR per unit increase in age is $e^{b_{1}}$ for age ≤ 42.7 and $e^{b_{1} + b_{2}}$ for age > 42.7.

Table 5:

Multivariate Model for Risk of SC-LBP With Peak Task Composite Lifting Index (PCLI) as Categorical Variable With Cut Point Equal to 25th Percentile of Cases (PCLI ≤ 2.2)

Variable (Overall p Value)^a	Category/Function	n (Cases)	HR	95% CI	p Value
PCLI (p < .01)^a	PCLI ≤ 2.2	112 (6)^b	1.0	—	—
	PCLI > 2.2	146 (18)^b	3.91	[1.46, 10.46]	<.01
Covariates
Age (years) (p = .03)^a	Spline terms
	Per unit age ≤ 42.7	176 (15)	0.92^c	[0.85, 0.99]	.03
	Per unit age > 42.7	82 (9)	1.23^c	[1.04, 1.46]	.02^d
Gender (p < .01)^a	Male	173 (13)	1.0	—	—
	Female	85 (11)	3.27	[1.38, 7.75]	<.01
BMI (p = .26)^a	Continuous
	Per unit score	258 (24)	0.95	[0.88, 1.03]	.25
History of LBP (p < .01)^a	No	193 (10)	1.0	—	—
	Yes	65 (14)	4.61	[2.03, 10.47]	<.001

Note. SC-LBP = seeking care for low-back pain; HR = hazard ratio; CI = confidence interval; BMI = body mass index.

Overall significance associated with including each variable in the model using the likelihood ratio test.

For cases, n is based on exposure at the time a worker became a case. For noncases, n is based on exposure at baseline.

For age, ${\hat{b}}_{1} = - 0.0883$ , and ${\hat{b}}_{2} = 0.2054$ . $HR = e^{b_{1} \cdot age}$ and $HR = e^{b_{1} \cdot age + b_{2} (age - 42.7)}$ for age ≤ 42.7 and > 42.7, respectively. HR per unit increase in age is $e^{b_{1}}$ for age ≤ 42.7 and $e^{b_{1} + b_{2}}$ for age > 42.7.

In Tables 2 through 5, an overall p value is provided for each variable in the multivariate model. This p value represents the significance of each variable in the final multivariate model using the likelihood ratio test. For spline transformed variables, this p value is for the entire range of that variable (i.e., both spline terms).

Table 1 summarizes the results from unadjusted univariate analyses of potential covariates for determining increased risk of SC-LBP. The statistically significant (p ≤ .05) factors were BMI, history of LBP, and swimming. Age and high cholesterol were marginally significant (p ≤ .10). Female gender and exercise had p ≤ .20.

Univariate analyses of biomechanical measures included changes to physical exposure over time (i.e., time-varying covariates). When fitted with linear spline functions, both PLI and PCLI were marginally statistically significant (p = .08 and p = .07, respectively; Table 1). There were no cases for PLI ≤ 1.0 or PCLI ≤ 1.0. Thus, categorical analyses of these variables using the Waters et al. (1994) recommended limits were not possible. As an alternative, we dichotomized PLI and PCLI based on 25th percentile of cases to keep the reference group as close to 1.0 as possible, while providing at least five cases in each reference group. The revised cut points for PLI and PCLI were 1.8 and 2.2, respectively. As a categorical variable (PLI ≤ 1.8 and PLI > 1.8), PLI was not statistically significant (p = .21); however, PCLI as a categorical variable (PCLI ≤ 2.2 and PCLI > 2.2) was statistically significant (p = .03) (Table 1).

Multivariate Results

The covariate PH regression model included: age, gender, BMI, and LBP history. When introduced into the multivariate model of covariates, PLI—treated as a continuous variable using a linear spline function—showed a statistically significant trend for increased risk of SC-LBP (p = 0.03; Table 2). The linear spline function showed that risk (HR) increased with an increase in PLI up to PLI = 2.2 (HR = 4.2 per unit increase, 95% CI = [1.1, 15.5], p = .03). With a further increase in PLI (i.e., PLI > 2.2), there was evidence that risk remained elevated, although it began to decrease from its peak (Table 2). The HR relative to unexposed for PLI > 2.2 is calculated as 23.01*0.89^(PLI–2.2) where 23.01 is the HR at PLI = 2.2 (knot) and 0.89 is the HR per unit increase in PLI above the knot (with respect to a reference point at or above the knot).

As a categorical variable, PLI showed marginal statistical significance for increased risk of SC-LBP for PLI > 1.8 (p = .06, HR = 2.4, 95% CI = [0.90, 6.21]; Table 3).

When the PCLI was introduced into the multivariate model of covariates and treated as a continuous variable using a linear spline function, it showed a statistically significant association with increased risk of SC-LBP (p = .02). The multivariate model showed a sharp increase in risk of SC-LBP up to PCLI = 2.6 (HR = 3.1 per unit increase, 95% CI = [1.2, 8.4]) and then rose at a much slower rate for PLI > 2.6 (Table 4). The HR relative to unexposed for PCLI > 2.6 is calculated as 19.59*1.03^(PCLI–2.6) where 19.59 is the HR at PCLI = 2.6 (knot) and 1.03 is the HR per unit increase in PCLI above the knot (with respect to a reference point at or above the knot). In the multivariate model, PCLI treated as a categorical variable had a statistically significant association with increased risk of SC-LBP (p = .01) with HR of 3.9 for PCLI > 2.2 (Table 5).

Estimated risk of SC-LBP for different levels of physical exposure was calculated as HR = $e^{{\hat{b}}_{1} x + {\hat{b}}_{2} {(x - K)}_{+}}$ . In the multivariate analyses, ${\hat{b}}_{1}$ and ${\hat{b}}_{2}$ were 1.4262 and −1.5411, respectively, for PLI spline terms (note c of Table 2) and 1.1433 and −1.1097, respectively, for the PCLI spline terms (note c of Table 4). These HRs are summarized in Table 6 for different levels of PLI and PCLI. Estimated peak HR for PLI was 23.0 at PLI = 2.2. Similarly, the estimated peak HR for PCLI was 21.9 at PCLI = 6.0 (Table 6). Workers exposed to biomechanical stressors with PLI or PCLI > 1.0 had at least 3 times the risk of SC-LBP as compared with those unexposed (p ≤ .03; Table 6). Those workers exposed to PLI or PCLI > 3.0 had at least 19 times the risk of SC-LBP as compared with unexposed and at least 5 times the risk as those with exposure to PLI or PCLI = 1.0.

Table 6:

Hazard Ratio Estimates for Peak Subtask Lifting Index (PLI) and Peak Task Composite Lifting Index (PCLI) Based on Multivariate Analyses

PLI	Hazard Ratio^a	PCLI	Hazard Ratio^a
0.2	1.3	0.2	1.3
0.4	1.8	0.4	1.6
0.6	2.4	0.6	2.0
0.8	3.1	0.8	2.5
1.0	4.2	1.0	3.1
1.6	9.8	1.5	5.6
2.2^b	23.0	2.0	9.8
2.5	22.3	2.6^b	19.6
3.0	21.0	3.0	19.8
3.5	19.9	3.5	20.1
4.0	18.7	4.0	20.5
4.5	17.7	4.5	20.8
5.0	16.7	5.0	21.2
6.0	14.9	6.0	21.9

Hazard ratio estimates are from results in Tables 2 and 4 and were calculated using the equation $e^{{\hat{b}}_{1} x + {\hat{b}}_{2} {(x - K)}_{+}}$ .

Value represents the knot for the linear spline function.

Comparison of Workers Who Did and Did Not Seek Care for LBP

There were 123 incident cases of LBP (see Part 1). Out of these 123 cases, 24 cases sought care for their LBP. Pain intensity at the time these 24 workers sought care was statistically greater (p = .02) than the peak pain intensity of those who did not seek care (6.1 ± 2.1, range = 3 to 10, for SC-LBP compared with 4.9 ± 1.6, range = 1 to 10, for LBP only).

There was no difference in biomechanical stressors between those who sought care for LBP and those who did not. Mean peak PCLIs for workers who sought care for LBP and those who did not were 3.0 ± 1.5 (range = 1.2 to 8.5) and 2.8 ± 1.4 (0.8 to 7.4), respectively. The difference was not statistically significant (p = .47). Similarly, the mean PLIs for those who sought care for LBP and those who did not were 2.4 ± 1.0 (range = 1.1 to 5.3) and 2.4 ± 1.1 (0.7 to 5.0), respectively.

Discussion

This is likely the first reported prospective cohort study of industrial workers (employed in manufacturing, warehousing, and meat/poultry processing, etc.) in which the association between individually measured biomechanical stressors and risk of SC-LBP was investigated. The results suggest that SC-LBP has a multifactorial etiology among those workers who were LBP free for ≥90 days at baseline. These factors include (a) job physical factors, (b) past history of LBP, and (c) demographic factors (age, gender, and BMI). Regarding job physical factors, both the PLI and PCLI predicted risk of SC-LBP, when treated as both categorical variables and continuous variables. In their continuous forms, PLI and PCLI predicted risk of SC-LBP with maximum HRs of 23.0 and 21.9, respectively (Table 6).

As stated earlier, the cohort used in this study was LBP free for ≥90 days at baseline. Using the same cohort to study LBP and SC-LBP made it possible to compare risk factors for the two outcomes. The results of this study and those from Part 1 suggest that some of the risk factors for SC-LBP are the same as those associated with a new episode of LBP. These include job physical exposure and history of LBP. Our results suggest that although work-related psychosocial factors are associated with a new episode of LBP (see Part 1), they appear to be relatively unimportant in SC-LBP. All three scales of psychosocial factors suggested increased risk for SC-LBP in univariate analyses (i.e., HR > 1.0); however, they did not have sufficient statistical significance to be considered in multivariate models (p = .24 to 0.78; Table 1). Two of these scales, APGAR (Bigos et al., 1991) and Tense-Edge-Nervous (Garg et al., 2014), were statistically associated with increased risk of developing a new case of LBP (see Part 1). Regarding other individual factors, female gender and age were associated with SC-LBP but were not associated with developing a new case of LBP; conversely, housework was associated with increased risk of LBP but not with SC-LBP. Our results appear to be consistent with the findings of Mortimer (2003), who reported that neither psychosocial factors nor lifestyle factors were associated with SC-LBP. It should be noted that this study had relatively few cases of SC-LBP (n = 24 cases) and thus might be underpowered to adequately address psychosocial factors and other individual factors.

Job Physical Factors

We found strong evidence that biomechanical stressors are associated with increased risk of SC-LBP. Peak job demands measured at both the subtask level (PLI) and the task level (PCLI) showed strong evidence of association with incident cases of SC-LBP (p ≤ .03). The PCLI showed a continuous increase in risk with an increase in PCLI. For PLI, the risk increased with an increase in PLI, and then risk began to decrease somewhat with a further increase in PLI, although risk remained elevated as compared with unexposed. This tapering of risk may suggest survival and/or selection bias. The PCLI had a relatively a narrower confidence interval as compared with PLI, and both had about the same p values (p = .02 and .03, respectively). Our data suggest that PCLI might be a better measure of biomechanical stressors in determining risk for SC-LBP.

There are no previously reported studies on the association between RNLE and risk of SC-LBP. Nevertheless, our findings are consistent with past studies that that have shown job physical demands are associated with increased risk of SC-LBP (Alexopoulos et al., 2006; Cote et al., 2005; IJzelenberg & Burdorf, 2005; Ozguler et al., 2000; Vingard et al., 2000).

The total number of studies on SC-LBP is relatively small (n = 23), and most of these studies are population based (Adamson et al., 2011). We could find only five studies that were conducted in occupational settings, and two of those five studies were limited to specific occupational settings: shipyard workers and scaffolders (Alexopoulos et al., 2006; Cote et al., 2005; IJzelenberg & Burdorf, 2004; Molano, Burdorf, & Elders, 2001; Ozguler et al., 2000). In all past studies, either job physical factors were not considered or they were based on job titles or worker self-reports. In this study, biomechanical stressors were carefully measured for each individual worker. It may be because of these individualized measurements of biomechanical stressors, and inclusion of LBP-free workers at baseline, that stronger associations were found between biomechanical stressors and incident cases of SC-LBP than those reported in many prior investigations.

Individual Factors

We found that females were at a higher risk for SC-LBP than males. This finding is consistent with the prior reports of SC-LBP (Cote et al., 2005; IJzelenberg, & Burdorf, 2005; Papageorgiou et al., 1997; Walker et al., 2004) as well as the common perception that women display a greater willingness to seek care for health issues than do men (Kent & Keating, 2005).

In this study, the relationship between age and SC-LBP was U shaped and was marginally statistically significant (p = .07). The risk for SC-LBP decreased up to an age of 43 and then increased. There are few studies on the relationship between age and seeking care, and the results are inconsistent (Admanson et al., 2011; IJzelenberg & Burdorf, 2005; Ozguler et al. 2000).

We found decreasing risk for SC-LBP with an increase in BMI; however, the association was not statistically significant (p ≥ .14). The cohort in this study was exclusively employed in manual materials handling jobs and had a high level of participation in hobbies and physical activities outside of work. It is possible that some of the high BMIs reported in this study may reflect high muscle mass rather than adipose tissue, which may be confounding the association between obesity and SC-LBP. There are few studies on relationship between BMI and SC-LBP, and the results are inconsistent. Whereas Ozguler et al. (2000) reported increased risk with an increase in BMI, Mortimer (2003) found no association between body weight and care-seeking behavior.

We found that history of LBP is associated with increased risk of seeking health care for LBP. History of LBP was defined as (a) LBP resulting in workers’ compensation, (b) lost time, (c) light duty, or (d) a visit to a care provider and/or (e) LBP radiating down the leg(s). Most past studies on SC-LBP were conducted on individuals experiencing either LBP or chronic LBP. Thus, other than intensity of pain, the role of history of LBP in SC-LBP has not been investigated. Prior reports have suggested that whereas leg pain and pain onset at work are associated with increased risk of SC-LBP (Carey et al., 1996), previous episodes of LBP were not statistically associated with SC-LBP (Cote et al., 2005).

We did not find statistically significant evidence that work-related psychosocial factors were associated with increased risk of SC-LBP. Our results are consistent with Cote et al. (2005), who found no relationship between job satisfaction and SC-LBP. Similarly, others have reported that job strain, poor job satisfaction, or psychological distress did not affect care-seeking behavior (Mortimer, 2003; Papageorgiou et al., 1997). Contrary to these reports, IJzelenberg and Burdorf (2005) found that high job strain and low social support at work were associated with health care use.

From the foregoing discussion, it appears that at present, the association between individual factors and SC-LBP is not clear, especially in occupational settings. There are conflicting findings. On the basis of a systematic review of 23 studies, Adamson et al. (2011) concluded that there does not appear to be an association between age, gender, or socioeconomic condition and SC-LBP. The roles of age, gender, BMI, LBP history, and psychosocial factors in SC-LBP require further investigation.

Pain Intensity and Seeking Health Care

In this study, 19.5% (24 out of 123) of workers with LBP sought care for their LBP. Past studies suggest that most individuals do not seek medical care immediately when they experience LBP (Croft, Macfarlane, Papageorgiou, Thomas, & Silman, 1998). Only a small percentage of individuals (5% to 44%) seek care because of a new LBP episode (IJzelenberg, & Burdorf, 2004; Kent & Keating, 2005; Vingard et al., 2000; Walker et al., 2004). Thus, our results appear to be in agreement with the past studies.

In the case of care utilization, it is assumed that the LBP is severe enough to prompt medical consultation (de Vet et al., 2002). Authors of a few studies have reported high pain intensity as one of the factors for SC-LBP (Carey et al., 1996; IJzelenberg & Burdorf, 2004; Mortimer, 2003; Walker et al., 2004). Contrary to these findings, Ozguler et al. (2000) reported that only 22.3% of workers seeking care had pain intensity >3 on a 0-to-7 visual analog scale, and Mortimer (2003) reported that numerous individuals with low pain intensity also sought care. In this study, the mean pain intensity was 6.2/10 for those who sought care for LBP (pain at the time they sought care). Although this intensity was statistically greater pain than among those who did not seek care, the difference was very small (1.2 units) and suggests there was no practical difference in pain between these two groups. Thus, our results appear to be consistent with those reported by Ozguler et al. (2000). The differences in cohorts between past studies and this study (chronic LBP versus LBP free at baseline, and community-based studies versus occupational settings) might explain some of these discrepancies. SC-LBP needs further investigations in occupational settings.

Strengths and Weaknesses of the Study

Two major strengths of this study are that (a) the cohort consisted of those workers who were LBP free for ≥90 days at baseline and (b) biomechanical stressors were measured for individual workers. With a few exceptions, practically all studies on SC-LBP have included only those individuals who were currently experiencing LBP or suffering from chronic LBP and have relied on either job titles or self-reports from workers for physical exposure assessments.

Data collection for this study did not include recording of point or 3-month prevalence of SC-LBP. Thus, we are unable to compare our results with those prior reports whose cohorts included those with LBP but not SC-LBP. Other strengths and limitations regarding general study design are summarized in Part 1.

Conclusion

This study suggests a multifactorial etiology for SC-LBP. Factors include job physical exposure (PLI and PCLI), LBP history, female gender, and age. Our finding that job physical demands play a role in SC-LBP is consistent with the conclusions of past studies. In this study, one in five workers experiencing LBP sought care, and risk of seeking care increased substantially at higher levels of physical exposure. Both the PLI and PCLI showed statistically significant exposure-response relationships for risk of SC-LBP (p ≤ .03) and are useful metrics for estimating exposure to job physical stressors. It is recommended that jobs be designed to keep both PLI and PCLI low to reduce the risk of SC-LBP.

Key Points

Seeking care for low-back pain (LBP) has a multifactor etiology.

Risk factors for seeking care for LBP are likely biomechanical stressors, LBP pain history, age, and female gender.

Peak job demands are associated with increased risk of seeking care for LBP.

Lifting index and composite lifting index are useful metrics for estimating exposure to biomechanical stressors.

Materials handling jobs should be designed to keep lifting index and composite lifting index low to reduce risk of seeking care for LBP.

Footnotes

Acknowledgements

The authors wish to acknowledge the major contributions of the subjects and employers who allowed this team to measure and assess them for several years of their lives. The authors also wish to recognize the contributions of the research team, which follows:

University of Wisconsin-Milwaukee:

James C. Foster, MD, MPH; David L. Drury, MD, MPH; Suzanna Tomich, MS, CPE, OTR; Gail Groth, MS, OTR; Karen Wahlgren, MS, OTR; Melissa Lemke, MS; Jessica Gin, MS; Prithima Reddy Mosaly, PhD; Vivek Kishore, MS; Priyank Gupta, MS; Meenu Sagar, MS; Christopher Hoge, BA; and Bridget Fletcher, BS

University of Utah:

Richard Sesek, PhD, CSP, MPH; Xiaoming Sheng, PhD; Richard Kendall, DO; Eric Wood, MD, MPH; Hannah Edwards, MD, MPH; Jeremy Biggs, MD, MSPH; William Mecham, MS, CPE, CSP; Ulrike Ott, MS; Steven J. Oostema, MS; Riann Robbins, MS; Atim Effiong, MS; and Richard Holubkov, PhD

Texas A&M University:

Gordon Vos, PhD, and John Paul Stevens, PhD

This study was funded, in part, by grants from the National Institute for Occupational Safety and Health (NIOSH/CDC), 1 U 01 OH008083-01, and NIOSH Education and Research Center Training Grant T42/CCT810426-10

Author(s) Note:

The author(s) of this article are U.S. government employees and created the article within the scope of their employment. As a work of the U.S. federal government, the content of the article is in the public domain.

Arun Garg is a professor and the director of the Center for Ergonomics at the University of Wisconsin-Milwaukee. He is a certified professional ergonomist and has developed several job analysis methods: The Revised NIOSH Lifting Equation, 3-D Static Strength Biomechanical Model, The Energy Expenditure Model, The Strain Index, and The Human Strength Prediction Model. His areas of expertise include ergonomics, biomechanics, work physiology, and design of workplace to reduce musculoskeletal injuries and illnesses.

Jay M. Kapellusch is an assistant professor of occupational science and technology at the University of Wisconsin-Milwaukee. His interests and expertise are in studying the effects of job physical exposure on incidence of musculoskeletal injuries, job analysis methods, and job design. He has more than 15 years of research and consulting experience in ergonomics and has analyzed in excess of two thousand jobs in more than 150 companies.

Kurt T. Hegmann is a professor and the director of the University of Utah’s NIOSH-sponsored Rocky Mountain Center for Occupational and Environmen¬tal Health, and holds the Dr. Paul S. Richards Endowed Chair in Occupational Safety & Health. He is active in education, having founded three graduate degree programs, as well as in MSD epidemiological research and clinical care of patients. He chairs the American College of Occupational and Environmental Medicine’s Evidence-Based Guidelines.

J. Steven Moore developed and coauthored the Strain Index and has numerous publications on workplace MSDs. He retired from his positions of professor of occupational health and safety and executive associate dean at the School of Rural Public Health in 2010.

Sruthi Boda is a postdoctoral researcher in the Industrial and Manufacturing Engineering Department at the University of Wisconsin-Milwaukee. She received her PhD in engineering in 2011 from UWM. Her research interests and expertise include studying risk factors for occupational injuries and illnesses, development of safety and ergonomics training for utility companies, and ergonomic workplace design in manufacturing settings.

Parag Bhoyar is a doctoral student in industrial and manufacturing engineering at the University of Wisconsin-Milwaukee. He received his MTech in industrial design from IIT Kanpur.

Matthew S. Thiese is an assistant professor at the University of Utah’s Rocky Mountain Center for Occupational and Environmental Health. He earned a PhD in occupational injury prevention in 2008. He has extensive experience in musculoskeletal epide¬miological research and has included multiple field studies, including three large cohort studies and two large cross-sectional studies.

Andrew Merryweather is an assistant professor in the Department of Mechanical Engineering at the University of Utah. Over the past 10 years he has managed significant research projects investigating musculoskeletal injuries in the workplace, assistive technologies for persons with disabilities, and many other projects involving computer simulation modeling and 3D human movement analysis.

Gwen Deckow-Schaefer is a retired research specialist from the University of Wisconsin-Milwaukee. She received her master’s degree in occupational therapy from UWM in 2002.

Donald Bloswick is a professor of mechanical engineering at the University of Utah. He directs research in ergonomics, occupational biomechanics, and rehabilitation engineering and directs the Ergonomics/Safety Program at the Rocky Mountain Center for Occupational and Environmental Health. He is a professional engineer and certified professional ergonomist and serves as an ergonomics trainer and consultant to industry, OSHA, and the legal community throughout the United States.

Elizabeth J. Malloy is an associate professor of statistics at American University. She received her PhD in statistics and then completed a two-year postdoctoral fellowship in biostatistics at the Harvard School of Public Health. She has expertise in modeling nonlinear exposure-response relationships in occupational and environmental settings and in functional data analysis methods for functional response and functional predictor models.

References

Adamson

Hunt

Nazareth

(2011). The influence of socio-demographic characteristics on consultation for back pain: A review of the literature. Family Practice, 28, 163–171.

Alexopoulos

E. C.

Tanagra

Konstantinou

Burdorf

(2006). Musculoskeletal disorders in shipyard industry: Prevalence, health care use, and absenteeism. BMC Musculoskeletal Disorders, 7, 88. doi:10.1186/1471-2474-7-88

Bigos

S. J.

Battié

M. C.

Spengler

D. M.

Fisher

L. D.

Fordyce

W. E.

Hansson

T. H.

Nachemson

A. L.

Wortley

M. D.

(1991). A prospective study of work perceptions and psychosocial factors affecting the report of back injury. Spine, 16, 1–6.

Bureau of Labor Statistics. (2011). Nonfatal occupational injuries and illnesses requiring days away from work, 2010 (USDL-11-1612). Washington, DC: Author. Retrieved from http://www.bls.gov/iif/oshcdnew.htm

Carey

T. S.

Evans

A. T.

Hadler

N. M.

Lieberman

Kalsbeek

W. D.

Jackman

A. M.

Fryer

J. G.

NcNutt

R. A.

(1996). Acute severe low back pain: A population based study of prevalence and care seeking. Spine, 21, 339–344.

Collett

(2003). Modeling survival data in medical research. London, UK: Chapman & Hall/CRC.

Cote

Baldwin

M. L.

Johnson

W. G.

(2005). Early patterns of care for occupational back pain. Spine, 30, 581–587.

Cox

D. R.

(1972). Regression models and life tables. Journal of the Royal Statistical Society: Series B, 34, 187–220.

Croft

P. R.

Macfarlane

G. J.

Papageorgiou

A. C.

Thomas

Silman

(1998). Outcome of low back pain in general practice: a prospective study. British Medical Journal, 316, 1356–1359.

10.

de Vet

H. C. W.

Heymans

M. W.

Dunn

K. M.

Pope

D. P.

van der Beek

A. J.

Macfarlane

G. J.

Bouter

L. M.

Croft

P. R.

(2002). Episodes of low back pain: A proposal for uniform definitions to be used in research. Spine, 27, 2409–2416.

11.

Freburger

J. K.

Holmes

G. M.

Agans

R. P.

Jackman

A. M.

Darter

J. D.

Wallace

A. S.

Castel

L. D.

Kalsbeek

W. D.

Carey

T. S.

(2009). The rising prevalence of chronic low back pain. Archives of Internal Medicine, 169, 251–257.

12.

Garg

Boda

Hegmann

K. T.

Moore

J. S.

Kapellusch

J. M.

Bhoyr

Thiese

Merryweather

Deckow-Schaefer

Bloswick

Malloy

E. J.

(2014). The NIOSH lifting equation and low-back pain, Part 1: Association with low-back pain in the BackWorks prospective cohort study. Human Factors, 56, 6–28.

13.

Garg

Kapellusch

J. M.

Hegmann

K. T.

Wertsch

J. J.

Merryweather

A. S.

Deckow-Schaefer

Malloy

E. J.

(2012). The Strain Index (SI) and Threshold Limit Value (TLV) for Hand Activity Level (HAL): Risk of carpal tunnel syndrome (CTS) in a prospective cohort. Ergonomics, 55(4), 396–414.

14.

Hurvich

C. M.

Tsai

C-L.

(1989). Regression and time series model selection in small samples. Biometrika, 76, 297–307.

15.

IJzelenberg

Burdorf

(2004). Patterns of care for low back pain in a working population. Spine, 29, 1362–1368.

16.

IJzelenberg

Burdorf

(2005). Risk factors for musculoskeletal symptoms and ensuing healthcare use and sick leave. Spine, 30, 1550–1556.

17.

Kent

P. M.

Keating

J. L.

(2005). The epidemiology of low back pain in primary care. Chiropractic & Osteopathy, l3, l3. Retrieved from http://www.chiromt.com/content/13/1/13

18.

Molano

S. M.

Burdorf

Elders

L. A.

(2001). Factors associated with medical-care seeking due to low-back pain in scaffolders. American Journal of Industrial Medicine, 40, 275–281.

19.

Mortimer

(2003). To seek or not to seek? Care-seeking behaviour among people with low-back pain. Scandinavian Journal of Public Health, 31, 194–203.

20.

Ozguler

Leclerc

Landre

Pietri-Taleb

Niedhammer

(2000). Individual and occupational determinants of low back pain according to various definitions of low back pain. Journal of Epidemiology and Community Health, 54, 215–220.

21.

Papageorgiou

A. C.

Macfarlane

G. J.

Thomas

Croft

P. R.

Jayson

M. I.

Silman

A. J.

(1997). Psychosocial factors in the workplace-do they predict new episodes of low back pain? Evidence from the South Manchester Back Pain Study. Spine, 22, 1137–1142.

22.

Plouvier

Leclerc

Chastang

Bonenfant

Goldberg

(2009). Socioeconomic position and low-back pain-the role of biomechanical strains and psychosocial work factors in the GAZEL cohort. Scandinavian Journal of Work, Environment, and Health, 35, 429–436.

23.

R Development Core Team. (2011). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.R-project.org/

24.

Vingard

Alfredsson

Hagberg

Kilbom

Theorell

Waldenstrom

Hjelm

E. W.

Wiktorin

Hogstedt

, & the MUSIC-Norrtalje Study Group. (2000). To what extent do current and past physical and psychosocial occupational factors explain care-seeking for low back pain in a working population? Results from the Musculoskeletal Intervention Center–Norrtajle Study. Spine, 25, 493–500.

25.

Waddell

Burton

A. K.

(2001). Occupational health guidelines for the management of low back pain at work: Evidence review. Occupational Medicine, 51, 124–135.

26.

Walker

B. F.

Muller

Grant

W. D.

(2004). Low back pain in Australian adults: Health provider utilization and care seeking. Journal of Manipulative & Physiological Therapeutics, 27, 327–335.

27.

Waters

T. R.

Putz-Anderson

Garg

(1994). Applications manual for the revised NIOSH lifting equation (Pub. No. 94-110). Cincinnati, OH: U.S. Department of Health and Human Services, National Institute for Occupational Safety and Health.