UCLA Study

Largest-ever study of back supports shows the devices
may reduce low-back injuries by one-third

Workers who wear back supports can reduce the number of low-back injuries by about one-third, according to findings from the largest-ever study of the increasingly popular –yet unproven—devices.

Researchers form the UCLA School of Public Health studied the workplace injury history of 36,000 workers of The Home Depot, a national home center chain, over a six year period and found that low-back injuries fell by about one-third after the company imposed a consistent policy on back support use.

The study, scheduled to be published in the November edition of the International Journal of Occupational and Environmental Health, is the largest study thus far to examine the effectiveness of back supports. No funding from manufacturers of back supports was used for the research.

"We found compelling evidence that back supports can play an important role in helping to reduce back injuries among workers who do a lot of lifting," said Jess Kraus, an epidemiologist and director of the UCLA-based Southern California Injury Prevention Research Center. "Along with worker training and proper workplace ergonomic design, back supports can be part of an overall back injury prevention program."

More than one million workers suffer back injuries each year, accounting for one out of every five workplace injuries and illnesses. Moreover, back injuries account for one-fourth of all workers compensation claims, costing businesses billions of dollars each year.

Back supports have become standard issue for a wide variety of workers over the past several years, despite there being little scientific inquiry into whether the devices help prevent injuries.

Several smaller studies have found uncertain results about the effectiveness of back supports. The National Institute of Occupational Safety and Health reviewed the scientific findings and issued a report in 1994 that concluded the benefit of the back supports remained unproven and did not recommend that they be used by uninjured workers.

The UCLA researchers examined the effectiveness of back supports by analyzing worker injury data collected by the home improvement retailer Home Depot on 36,000 people who worked at its 77 California stores from the start of 1989 to the end of 1994. The company imposed a consistent back support use policy that was phased in between 1990 and 1992.

Analyzing injury reports and other worker information, UCLA researchers found that Home Depot workers sustained about 31 back injuries per 1 million work hours without the supports, compared to about 20 injuries per 1 million work hours after a consistent back support use policy was imposed.

"I went into the study very skeptical about claims that these back supports could help reduce back injuries," Kraus said. "I suspected we would not find any positive effect, so I was very much surprised by our findings."

The benefits of using the back supports was seen in both men and women workers, in young and older workers, and among workers engaged in both low and high levels of lifting, according to the UCLA researchers.

The biggest benefit was seen among the groups of workers at highest risk of back injury—men who were 25 and younger or over age 55, had worked for the company for one to two years, and had jobs that required the highest intensity of lifting.

Home Depot employees logged more that 100 million work hours during the study period. Other workplace training and safety measures adopted by Home Depot during the study period were taken into account by researchers.

"People need to be careful about generalizing these findings to workers who are not engaged in material handling jobs," Kraus said. "There needs to be further research examining occupations such as construction, agriculture and mining to see if back supports prevent injuries to those workers."

The Southern California Injury Prevention Research Center is one of 10 injury prevention research centers sponsored nationally by the U.S. Centers for Disease Control and Prevention. The center emphasizes study of injuries in high-risk, ethnic/racial minorities and other traditionally underserved populations.

Funding for the back support study was provided by the Southern California Injury Prevention Research Center, the UCLA Center for Occupational and Environmental Health, the California State Department of Industrial Relations and the 3-E Company, a San Diego-based industrial safety consulting firm.

Other authors of the study are Kathryn A. Brown, David L. McArthur, Corinne Peek-Asa, and Lei Zhou, all of UCLA, and Lupe Samaniego and Chris Kraus of the 3-E Co.

Jess F. Kraus, MPH,PHD, Kathryn A. Brown, MPH,
David L. McArthur, PHD,MPH, Corinne Peek-Asa, MPH, PHD,
Lupe Samaniego, MPH, Chris Kraus, Lei Zhou

The objective of this study was to determine the effect of a change in back support use policy on the occurrence of work related low back injuries among a large cohort of employees in the retail trade home improvement industry. Working hours of exposure, back support use, and intensity of materials lifting requirements were collected from 1989 through 1994. Records of injury related claims were reviewed for all documented injuries to the lower back among members of the cohort during the same period. Over 101,000,000 working hours were recorded by nearly 36,000 employees; 2152 employees reported an acute low back injury occurring during working hours as a first report of episode, with medical physician diagnosis and acute/abrupt onset. Incidence density rates were calculated for persons wearing and not wearing the back support. Rate ratios and prevented fractions were evaluated. Before implementation of a company wide back support policy, the employees had a rate of acute low back injuries of 30.6 per million working hours. After implementation, this rate fell to 20.3 per million working hours, a significant reduction of 34.0%. This effect was seen in both genders, in younger workers and in those aged 55+, with low levels of lifting as well as high lifting intensities, and in persons with one to two years of employment with the company. The authors conclude that uniform mandatory implementation of a back support use policy significantly reduces the incidence of acute low back injuries incurred in the workplace. Key words: back injury; back supports; musculoskeletal system; low back; materials handling; epidemiology; effectiveness.

The occupational health literature on back injury (or pain) is extensive (ref 1-5) and provides consistent evidence that back injuries are a serious public health problem because of their high incidence and prevalence, the costs of treatment, the impact on business, and limited success in prevention.
    Compensation claims, which are the best current source of information about the occurrence of low back injuries, show varying rates by industry. Reported high risk occupations include truck drivers (ref 5-11), nurses and hospital workers (ref 12-16), miners (ref 17), farmers (ref 18,19), material handlers (ref 16,20), and laborers (ref 21), among many others (ref 19,22,23), Regardless of occupation, a number of specific work tasks appear to be associated with back injuries, including lifting (ref 8,13,17,20,21,24-26), bending (ref 25,26), pulling (ref 8,21,24), twisting (ref 27-29), carrying (ref 8,21,24), and reaching (ref 8), either alone or in combination with one another.
    There is much variation in the ways in which back injuries have been defined, classified, and described. In some studies (ref 30-32), "low back pain" is the outcome of interest while in others "back sprains and strains," (ref 33), "back injuries" (ref 1) "low back injuries" (ref 21), "back pain" (ref 34), and "low back pain syndrome" (ref 35), are the points of focus. Non-specificity of the injury (or pain) has also hampered diagnostic precision and clinical management (ref 36), which in turn has influenced the degree of compensation for the injury.
    A number of work exposures have been reported to be associated with low back injuries, including heavy manual labor (ref 19,21,25,37-42), exposure to vibrations (ref 5,7,8,11,23,39), exposure to cold (ref 43), and work tasks involving repetitive activities. (ref 44-46), Age (ref 4,10,31,47), gender (ref 10,31,47), race (ref 14), education (ref 14), height and weight (ref 10,31,47), smoking (ref 9-11,28,31,47), alcohol use (ref 44,47), medical history, and psychological stress (ref 4,8,9,10,26,40), have also been reported to be associated with low back injuries.
    A number of preventive measures have been introduced over time to prevent work related back injuries. Worker training, job screening, and ergonomic modification are currently recommended by the U.S. National Institute of Occupational Safety and Health (NIOSH) (ref 48), but objective evidence of their effectiveness alone or in combination has been elusive and subject to many methodologic problems. Back supports (also called back belts, outhouses, corsets, or braces) have become widely used in recent years in occupations involving material handling. Their use followed medical application for treatment of low back pain (ref 49) and use among weightlifters (ref 50).
    The biomechanical basis for assuming that back supports might reduce the occurrence or severity of low back injury is beyond the scope of this paper, but has been extensively reviewed by Perkins and Bloswick (ref 51), Although the theoretical basis for effectiveness of the back support is quite appealing, there is little epidemiologic research to demonstrate a positive effect in reducing the incidence or severity of low back injury in usual working environments. Walsh and Schwartz (ref 52) studied the effects of training and back supports in a random sample of 90 warehouse workers for a Texas grocery company. Thirty workers were assigned to a one-hour training intervention, 30 to a one-hour training intervention plus a back support, and 30 to a group receiving no intervention. All but nine workers were followed for six months. While there were no differences in back injury rates among the groups, there was a significant decrease in severity (as measured by lost days from work) among those wearing the device, but only if they had previously had back injuries. The small sample size, loss to follow-up, and inability to separate the effect of training from back-support use limited the value of the findings of this study.
    Reddell and colleagues (ref 53), reported on 896 American Airlines baggage handlers who were randomly assigned to four study groups: those receiving a weightlifting belt, those receiving a weightlifting belt plus training, those receiving training only, and those left as "controls." After eight months of follow-up, there was no difference in any injury outcome measure among the groups; however, the study had very low compliance and 28% dropouts, leading to possible selection bias in outcome assessment.
    Mitchell and associates (ref 54) retrospectively surveyed 1,316 workers who performed lifting activities at Tinker Air Force Base in Oklahoma. Although some positive effects were identified, the retrospective nature of the survey and the reliance on worker recall for reporting of exposures, belt use, and injury episodes hamper the interpretation of the findings for this study.
    NIOSH, after a review of all published evidence on back supports and low back injury, concluded "that the effectiveness of using back belts to lessen risk of back injury among uninjured workers remains unproven" (ref 48, p.1). The NIOSH report went further by neither recommending the use of back belts to prevent injuries among uninjured workers nor considering them to be personal protective equipment.
    The increasing use of back supports among workers in many industries, the lack of scientific evidence of their effectiveness in real-world work settings, and the theoretical biomechanical and physiologic basis for expecting some protective effect prompted the current study. Its purpose was to determine the effect of a change in back-support-use policy on the occurrence of work-related low back injuries among material handlers in the retail-trade home improvement industry.

The study involved a historical cohort (ref 55) followed longitudinally from January 1, 1989, through December 31, 1994. Exposure risk hours and outcomes (acute low back injury while at work) were measured in the cohort. Back support use was determined by documenting a uniform mandatory belt-support policy introduced during this period.
The cohort consisted of all workers employed by The Home Depot (a national home improvement retail chain) assigned to any retail store in California in the study period. Employees whose work was centered in administrative offices or distribution warehouses were excluded. To be included in the study cohort, workers must have been compensated for at least one day during the study period. There was no other exclusion to the cohort, which numbered almost 36,000 different employees over the six-year period.

Employee working hours were classified according to pre-and post-back-support-use policy period. Hours per person were so classified by age, gender, length of employment (LOE), and lifting intensity determined from job title. This information, plus name, social security number, and store number, was provided on computer tapes by The Home Depot management for each quarter from 1990 through 1994. Since back supports were not worn by any Home Depot employee in California during 1989, all working hours in this year was classified as non-back-support-use. Personal identifiers were used to link individual employee working hours from one quarter to the next and to each of the qualifying variables mentioned above. Once working hours were aggregated across all variable domains, the computer tapes were returned to The Home Depot, for reasons of confidentiality.

Company policy mandating the use of back supports was introduced early in 1990, and implemented at varying times in different retail stores well into 1992. Figure 1 schematically illustrates the nature of back-support-use implementation by store. Prior to the institution of this company policy, back supports had not been supplied by the company or used by any employee. Individual store managers implemented the company policy on back support use at various times depending on availability of the device, shipping times from the vendor, and training needs. Stores in service in early 1990 received back supports in staggered shipments, necessitating delays in the implementation of their use. Stores that opened after the policy was implemented received back supports prior to the opening dates, and all working hours during the store start-up phases and after opening were with back support use. With knowledge of the vendor's shipping date, the store opening (or closing) date, and individual employee working hours by store and date, it was possible to allocate all working hours by back-support-use status as well as individual descriptors such as age and gender. The only exception was for those retail stores open before the company policy on back support use was put into effect. During the transition from non-use to full use in these stores (never more than one month), there was uncertainty with respect to back support use; hence all working hours during that one-month period were excluded from the analyses.

All store employees are required to lift and/or carry materials as part of their routine job tasks. There are, however, considerable variations in the frequencies and weights or the materials to be moved, according to job title. The identical job titles are used in all stores in California. For purposes of these analyses, job titles were categorized into one of three lifting or carrying levels, based on corporate safety management determination, ergonomic analyses undertaken in 1991, and the investigators' empirical observations in 1993 or 1994. The amounts or intensities of exposures and the most common job titles were:

Age and gender were derived from personnel records. Hire date and termination date, if applicable, were used to determine length of employment (LOE) for each employee, which was limited to time spent with The Home Depot. Data such as race, ethnicity, prior medical history, and prior injury history were not recorded in company personnel files.

All computerized injury-claim records with date of occurrence during the study period and for any injury to the musculoskeletal elements of the trunk were provided by Safety Management of The Home Depot. These included claims for injuries to the neck, shoulders, upper back, chest, ribs, thorax, mid-back, lower back, hip, pelvis, abdomen, groin, and buttocks. Each record was reviewed by two investigators to ensure conformity to the inclusion criteria: date of injury occurrence in the study period, a first report of the episode, a medical-physician diagnosis, acute/abrupt onset, and occurrence while at work in a Home Depot retail store.

Injury-reporting forms, which follow the format of the Cal/OSHA Employer's Report of Occupational Illness or Injury, record a number of descriptive variables, including body part affected, type of injury (strain, sprain, pull, tear, etc.), name of worker, social security number, age, gender, LOE, job title, store number, date and time of injury, and days lost from work.

All included cases were double-checked to verify the inclusion criteria and to eliminate any duplicate case records that might have been generated from multiple physician visits for follow-up treatment for the same injury. All injury cases were included for rate derivation, but for a small number, data for some factors were missing. No contact was made with the employee or corporate safety management to secure the missing information because of the historical nature of the study and pledges of confidentiality.

The back supports provided by The Home Depot to their employees were commercially available, made of Lycra material, most with (but some without) straps (figure 2). Back support use or non-use was verified in three ways: 1) implementation dates of company policy, 2) district manager's loss-prevention notations for each store, and 3) an independent store survey. The latter was an unannounced walk-through assessment of all floor personnel (which includes about 95% of all store personnel) undertaken in all 77 California stores form late 1993 into early 1994. The prevalence of back support use was in excess of 98 % at that time; most of the exceptions were observed to be employees advanced in their pregnancies or employees on break, neither of whom would be involved in more than trivial lifting.

The effect of the use of back supports on the incidence of low back injuries was evaluated in several ways. First, incidence density rates (IDRs) per one million working hours were calculated overall and according to working hours for employees using and not using back supports. Second, incidence density rate ratios (IDRRs) and 95% confidence intervals (CIs) were derived using standard methods (ref56). An IDRR greater than 1 (with a 95% CI above 1) was deemed protective; that is the IDR for those not wearing back supports was significantly higher than that for those using back supports. Third, the amount of protection was calculated as a prevented fraction, which is the percentage excess in IDR attributable to not using the back support or the percentage of the low back injury rate that might be expected to be eliminated if those not using back supports had instead worn them.

Direct adjustment methods (ref57) were used to account for one, two, and three factor stratification. This approach was necessary since no known statistical technique evaluates interaction effects with many cells having zero frequency. In our dataset, many cells representing the multi-way crossing of factors of age, gender, lifting intensity, length of employment, back support use, and low back injury were empty.

In 1989 there were 31 Home Depot retail stores in California, with approximately 7,500 employees. By the end of 1994, the number of stores had increased to 77 and the number of employees to over 19,000. During the six year study period, 101,023,308 working hours were recorded by the employee cohort. In the same period, 2,152 employees reported experiencing an acute low back injury, for an overall rate of 21.5 per million (figure 3). Overall, low back injury rates were about two-thirds higher for males than for females, and the rates were highest for those under the age of 25.

Slightly more than 1.1 million working hours and 18 reported low back injuries were recorded in stores during the period of transition from pre-policy to company policy on implementing back support use. These hours and injuries were excluded from all analyses. Table 1 shows that employees not wearing back supports had an acute low back injury incidence density rate (IDR) 1.52 times higher than that of those wearing back supports (30.6 vs. 20.2 per million, respectively). The relative difference in rates between the two groups was highly statistically significant (p<0.0001) and the prevented fraction of the back support use in its reduction in the rate of low back injuries was 34%.
Incidence density rates, rate ratios, and the prevented fractions for back support use by gender and age are summarized in Table 2. The IDR was higher for men compared with women. Within gender, the rates were significantly higher for those not using back supports (1.32 times higher for women and 1.56 times higher for men). The prevented fractions for the use of the back supports were 24.0% for women and 36.2% for men.
The highest rates of low back injury were observed in employees less than 25 or more than 55 years of age who did not use back supports (Table 2). Their injury rates were 2.0 and 2.5 times higher, respectively, than those for employees of the same age using back supports. The prevented fraction for back support use among employees under age 25 was more than 50%, and that for those aged 55 and older, 60%.

Table 1

Numbers of employee working hours, numbers injured, incidence density rates, rate ratios, 95% Confidence Intervals (CIs), and prevented fraction for back-support use, Home Depot Stores, California, 1989-1994

* Excludes hours and injuries occurring during store transition to back support use policy implementation.

Numbers of working hours, numbers injured, incidence density rates, rate ratios, 95% Confidence Intervals (CIs), and prevented fractions for back-support use by gender and age, Home Depot Stores, California, 1989-1994

-	Working Hours (x 1000)		Number Injured		Rate per Million Hours		-	-
-	No Back Support	Back Support	No Back Support	Back Support	No Back Support (a)	Back Support(a)	Rate Ratio and 95% CI	Prevented Fraction % (a-b/a)x100
Gender	-	-	-	-	-	-	-	-
Woman	4,056	28,380	78	414	19.2	14.6	1.32(1.03-1.68)	24.0
Men	8,756	58,698	314	1,346	35.9	22.9	1.56(1.38-1.77)	36.2
Age Group (years)	-	-	-	-	-	-	-	-
<25	4,197	25,594	184	555	43.8	21.7	2.02(1.72-2.40)	50.5
25-34	4,888	33,869	127	683	26.0	20.2	1.29(1.06-1.50)	22.3
35-44	2,121	15,588	43	330	20.3	21.2	0.96(0.69-1.31)	-
45-54	993	7,562	15	125	15.1	16.5	0.91(0.53-1.55)	-
55+	612	4,465	23	67	37.5	15.0	2.50(1.64-4.16)	60

Table 3 Numbers of working hours, numbers injured, incidence density rates, 95% Confidence Intervals (CIs), and prevented fraction for back-support use by lifting intensity and length of employment (LOE), Home Depot Stores, California, 1989-1994

-	Working Hours (x 1000)		Number Injured		Rate per Million Hours		-	-
-	No Back Support	Back Support	No Back Support	Back Support	No Back Support (a)	Back Support(a)	Rate Ratio and 95% CI	Prevented Fraction % (a-b/a)x100
Lifting Intensity	-	-	-	-	-	-	-
Low	2,565	15,874	49	72	19.1	4.5	4.21(2.93-6.06)	76.4
Moderate	2,403	17,576	43	284	17.9	16.2	1.11(0.08-1.53)	-
High	2,845	53,628	300	1,404	38.2	26.2	1.46(1.29-1.65)	31.4
LOE(Years)	-	-	-	-	-	-	-	-
<1	5,390	32,345	146	901	27.1	27.9	0.97(0.82-1.16)	-
1-2	2,914	19,864	177	376	60.7	18.9	3.21(2.68-3.84)	68.9
2-3	1,789	12,951	35	188	19.6	14.5	1.35(0.94-1.93)	-
3-4	1,170	8,675	21	146	17.9	16.8	1.07(0.67-1.68)	-
4+	1,549	13,243	13	149	8.4	11.3	0.73(0.42-1.31)	-

Intensity of lifting frequency and LOE are two interrelated factors potentially associated with low back injury. The highest rates of low back injury observed within sub-groups of the cohort were recorded for those whose job tasks seldom required lifting or related physical activities (Table 3). Those whose jobs had low lifting requirements and who did not use back supports recorded a rate of low back injuries that was 4.2 times higher than that of those within the same job titles but using back supports. Over 61% of all working hours recorded by the cohort were in job titles requiring high-level material lifting activities. Within this subgroup of the employee cohort, the injury rate was 38.2 per million for those not using back supports-1.46 times higher than that of those with the same job titles who used the supports (rate of 26.2 per million). The effectiveness of back supports for those with moderate lifting intensity requirements, while positive, did not reach statistical significance.
Employees with one to two years experience at California Home Depot stores and not using back supports were at highest risk of low back injury (Table 3). Their rate was 3.2 times higher than that of those with the same LOE but using back supports, and the prevented fraction was almost 69%.

Table 4 Incidence density rate ratios (IDRRs), 95% Confidence Intervals (CIs), and prevented fractions for back-support use by gender and age, gender and lifting intensity, and gender and length of employment (LOE), Home Depot Stores, California, 1989-1994

-	IDDR	95% CI	Prevented Fraction%
Gender and Age	-	-	-
Men, <25 years old	2.30	1.90-2.77	56.4
Men, 25-34 years old	1.30	1.05-1.61	23.0
Men, 35-34 years old	0.82	0.57-1.19	-
Men, 45-54 years old	0.90	0.48-1.69	-
Men, 55+ years old	2.41	1.45-4.01	58.6
-	-	-	-
Women, <25 years old	1.29	0.89-1.88	-
Women, 25-34 years old	1.22	0.80-1.84	-
Women, 35-34 years old	1.60	0.86-2.97	-
Women, 45-54 years old	0.94	0.33-2.64	-
Women, 55+ years old	3.16	0.84-11.91	-
-	-	-
Gender and lifting intensity	-	-
Men, low intensity	5.10	3.27-7.96	80.5
Men, moderate intensity	1.63	0.93-2.83	-
Men, high intensity	1.45	1.26-1.65	30.8
-	-	-	-
Women, low intensity	2.37	1.20-4.71	57.9
Women, moderate intensity	0.94	0.64-1.40	-
Women, high intensity	1.57	1.11-2.23	36.6
-	-	-	-
Gender and LOE	-	-	-
Men<1 years	1.01	0.84-1.23	-
Men 1-2 years	3.25	2.65-3.99	69.2
Men 2-3 years	1.58	1.07-2.33	36.7
Men 3-4 years	1.09	0.65-1.81	-
Men 4+ years	0.73	0.40-1.36	-
-	-	-	-
Women <1 years	0.80	0.53-1.20	-
Women 1-2 years	3.05	2.11-4.40	67.2
Women 2-3 years	0.63*	0.23-1.75	-
Women 3-4 years	0.95*	0.34-2.70	-
Women 4+ years	0.88*	0.46-1.69	-

Table 5 Incidence density rate ratios (IDRRs), 95% Confidence Intervals (CIs), and prevented fractions for back-support use by age and lifting intensity, Home Depot Stores, California, 1989-1994

Age Years	Lifting Intensity	IDDR	95% CI	Prevented Fraction %
-	-	-	-	-
<25	Low	11.62	5.44-24.83	91.3
-	Moderate	1.27	0.81-2.01	-
-	High	1.96	1.62-2.37	49.0
-	-	-	-	-
25-34	Low	2.69	1.40-5.16	63.0
-	Moderate	0.92	0.51-1.68	-
-	High	1.29	1.04-1.59	22.2
-	-	-	-	-
35-44	Low	2.95	1.39-6.27	66.3
-	Moderate	0.93*	0.37-2.35	-
-	High	0.80	0.54-1.18	-
-	-	-	-	-
45-54	Low	3.78+	1.11-12.93	73.7
-	Moderate	0.79*	0.18-3.36	-
-	High	0.73*	0.37-1.44	-
-	-	-	-	-
55+	Low	3.67+	0.67-20.05	-
-	Moderate	2.71+	0.55-13.43	-
-	High	2.38	1.42-4.00	57.9

*Rates in non-back-support-use group based on fewer than ten cases.
+Rates in non-back-support-use and non-use groups based on fewer than 10 cases.

Two-way stratifications are presented below as IDRRs with 95% confidence intervals and prevented fractions. The analyses of back-support-use status and factor interactions beyond two-way stratification was limited to men because of very small or zero cell sizes for women.
    Gender and age. A statistically significant prevented fraction for back support use was observed for men less than 35 years old and also for men aged 55 or older (Table 4). While there was a protective effect of back supports overall among women, there does not appear to have been a significant differential protective effect in any one of the age groups. Statistical power was less due to smaller sample sizes among women.
    Gender and lifting intensity or LOE. Men and women (ignoring age) with low or high job material-lifting requirements appeared to benefit from back support use. Men showed more benefit than women at low and moderate lifting intensities, while women at high lifting intensities showed more benefit than men. Men and women with one to two years of employment showed the greatest positive effect from back support use (Table 4). Among women, no other length of employment was associated with a protective effect.
    Age and lifting intensity. Back supports appeared to be protective in preventing low back injury for some age-lifting intensity categories (Table 5). Back supports were protective for those less than 25 years of age with either low or high job lifting requirements. For those aged 35 to 54, a protective effect was seen only for those with low job lifting requirements. In the oldest age group (55 + years) a protective effect was seen for all three lifting-intensity groups, But significance was reached for those in the high-lifting-intensity category only. Small case sample sizes were evident in the low and moderate categories.
    Age and LOE. Twenty-six of 60 cells from the stratification of age group by LOE category had fewer than ten cases in either one or both of the back-support-use groups. Hence, IDRRs could be calculated with reasonable precision for only ten combinations of age and LOE. Employee hours for those less than 25 years of age with one to two or two to three years of employment showed a protective effect from use of back supports (Table 6). Employees aged 25-34, 35-44, or 55 + years and with one to two years' LOEs, benefited from the back support. The prevented fractions, varied by age and LOE group, ranging from zero to a high of almost 79% for employees less than 25 years old with one to two years of work experience and those aged 55 and older also with one to two years of experience.

Table 6 Incidence density rate ratios (IDRRs), 95% confidence intervals (CIs), and prevented fractions for back-support use by age and length of employment (LOE), Home Depot Stores, California, 1989-1994.

Age (Years)	LOE(Years)	IDRR	95% CI	Prevented Fraction %
-	-	-	-	-
<25	<1	1.10	0.84-1.44	9.4
-	1-2	4.75	3.64-6.20	78.9
-	2-3	2.15	1.26-3.66	53.4
-	-	-	-	-
25-34	<1	0.91	0.67-1.23	-
-	1-2	2.62	1.91-3.60	61.9
-	3-4	1.20	0.63-2.28	16.7
-	-	-	-	-
35-44	<1	0.68	0.42-1.10	-
-	1-2	1.27	0.69-2.35	21.6
-	-	-	-	-
55+	<1	2.55	1.28-5.08	60.9
-	1-2	6.11	2.62-14.14	83.6

*Excludes age-LOE categories with fewer than ten cases in either back-support-use or non-use group.

The crude low back injury rates during the pre-policy (non-use) period covered January 1989 well into 1992 in some stores. That is, in 1990,1991, and 1992, some stores implemented the policy while others did not. To assess whether the factors of age, gender, lifting intensity and LOE working hours were differentially distributed over the pre- versus post-policy periods, the crude rates were adjusted for these factors. Because of very small or zero cell numbers in many of the subcategories of age, lifting intensity, and LOE for women, direct adjustment of the crude rates was restricted to those of men. The crude rates among men were 35.9 for non-use and 22.9 for back support use per one million working hours (Table 2). Rates adjusted for age, lifting intensity, and LOE were 35.9 and 23.0 respectively, which are almost identical to the non-adjusted rates.

This study of the effect of the use of back supports to prevent low back injury used the largest group of study participants to appear in the scientific literature thus far, and was exclusively focused on material handlers in retail stores of the home improvement industry. Although a randomized controlled trial would have been the research design of choice, the nature of most businesses prohibits randomized controlled trial would have been the research design of choice, the nature of most businesses prohibits randomized trials. Nonetheless, the historical-cohort approach of this study has merit, in that all instances of employee-reported low back injuries were discovered and all working hours according to back support use were available for analyses. In addition, it was possible to closely approximate the level of lifting exposure to low back stress on exertion while on the job.
    This study showed that a back-support-use policy leading to almost complete back support use reduced back injuries. Younger and older male workers, but not females, showed significantly reduced rates using the back supports. Those with low- and high-lifting-intensity jobs benefited from the back supports regardless of gender or age (when sample size was sufficient). The results of this study do not indicate that the use of back supports was harmful to any of the groups of employees examined.
    The historical nature of the study design precluded changes in injury reporting practices because of knowledge of the ongoing study. However, the six year study period offered opportunities for change in work practices (other than introduction of back supports), which could have reduced the risk of low back injury. The only change in work practices noted by corporate safety management was an increased use of pallets and forklifts or "reach tracks" for stacking and storage of material. The overall impact of this change (which was introduced in 1992-93) on overall level or frequency of lifting is unknown, but it would have affected those with high-job-lifting-intensity activities the most. No other change in work or hiring practices was reported by safety management, and there was no change in personnel recruitment, job assignments, tasks, injury reporting, or recording mechanisms during the six year study period.
    Another advantage of the current design was the availability of compensated working hours by each of the primary factors studied, namely age, gender, lifting intensity, and length of employment. Changes in job title for each employee were traced by social security number, and all hours were aggregated over the six year period by each descriptive subgroup.
    It was not possible in this historical cohort design to observe use or non use of back supports for all employees over the entire six year period in all stores. The prevalence surveys conducted in 1993 and 1994, however, showed a very high rate of compliance with company policy on the use of back supports. The investigators were informed by safety management that prior to the implementation of the store policy on back support use employees had not been provided with nor had they used back supports.
    One outcome that unfortunately was not evaluated was the severity of the low back injuries and its relationship to back support use. Time lost from work was not recorded consistently on the Employer's Report of Occupational Illness or Injury form. This factor should be explored in more detail in concert with the earlier finding regarding differences in injury severity reported by Walsh and Schwartz (ref 52).
    A shortcoming of this study was the inability to simultaneously evaluate back support use, injury status, gender, age, job lifting intensity, and length of employment. These and other factors may be important cofounders or effect modifiers that delimit (or could further expand) the potential effect of back supports observed in this study. However, adjustments for the effects of age, lifting intensity, and LOE yielded almost no difference form the crude rates for males, indicating that these variables did not appreciably affect the difference in crude rates by back support use or non-use. Future studies, preferably with randomized trials stratified on these factors, will be essential to ensure that adequate exposure time for all effect modifiers is available in the study population.
    Back supports appear to offer some beneficial effect: Those using back supports showed a 34% reduction in low back injury rates compared with those who did not. Although this finding reflects a before-and-after-policy comparison, factors associated with work activities, lifting frequency, recruitment practices, injury-reporting processes, and related concerns did not change over the six year period with the one exception of the increased use of forklifts and pallets noted above. In the present study, misclassification may have occurred if an employee was injured while performing a task outside normal job definitions, but in the Home Depot setting, degree of lifting has always been tightly linked to job type.
    Several subgroups of the cohort were at highest risk before the introduction of the company's back support policy, namely, those less than 25 years old or more than 55, men, those with the highest job lifting intensities, and those with one to two years of employment. All of these groups showed a significant beneficial effect from back support use. Even some subgroups with the lowest rates of low back injury benefited from the use of back supports, namely women and those with low job lifting requirements.
    The explanation of the protective effect of the back-support-use policy in the low-lifting-intensity group of employees is not known. It is possible that employees who had histories of back injury may have been selected for or transferred to these job titles, and thus constitute a special class of back support users; however, this is not likely since most transfers to these job titles were due to promotions made without reference to an employee's injury history. As might be expected, those engaged in high-intensity material lifting and carrying work had the highest pre-policy low back injury rates irrespective of gender, and both men and women with these heavy-lifting job requirements showed a protective effect of back support use.
    The protective effect of the back support for men was similar to that overall, that is, those aged less than 35 and aged 55 or more benefited the most from the back support-use policy. A protective effect by age group was not observed for women. The reason for the differential positive effect on low back injury with back support use in some age groups and not others is not understood, but may illustrate an interaction of training, experience, previous injury, work history, height/weight, general fitness, or other related factors. Except for age, LOE, gender, and lifting intensity, we were not able to evaluate these important factors.
    One investigator (ref 58) has suggested that chronic repetitive stress to low back muscles or spinal discs may occur over time, and that with an acute exertion event, the threshold of injury is exceeded. This chronic-onset model might partly explain why those with one to two years of experience (regardless of gender) sustained some of the highest low back injury rates observed in this cohort. The model suggests that the first year on the job is a period of cumulative pre-threshold stress of all ages, genders, and lifting intensity levels. During this initial year, low back musculoskeletal components are weakened; with an acute exertional event, the injury threshold is exceeded and an acute low back strain or sprain results. The beneficial effect of the low back support was most dramatic during years one to two on the job, with prevented fractions of 69% for men and 67% for women. There are no changes in work activities within job titles during the first two years on the job that might explain this observation. For a chronic-injury-onset model to hold merit, all age, gender, and lifting-intensity groups should show elevated low back injury rates with one to two years on the job. This assumes that the effects of prior work history or prior back injury would be nil or similar for all newly recruited employee a very unlikely profile.
    Future research needs to address the effects of back supports among subpopulations of employees to further elucidate which employees benefit most from back support use, and investigate causal models as they relate to back support use and low back injury.

Special thanks to John Howard (Cal/OSHA) and Rob Ward (3-E Company) for their commitment to seeing this work to completion. This research would not have been possible without the complete support and cooperation of Steven Taylor, David Bryant, Edward Edmunds, and Laura Ferguson of The Home Depot. Additional thanks to Keith Rosenblum (formerly of The Home Depot), who helped to get the project started, and to Bonnie Blanker (UCLA) and Eugene Lai (3-E Company) for their work in data processing and library searches. Finally, the authors thank the management and employees of all California Home Depot stores who cooperated so well in the data-collection phases of this research.

AMERICAN INDUSTRIAL SUPPLY, INC.
PH. 865-687-5022 FAX 865-687-4926
Copyright � 2008

Back Support Use	Working Hours*	Number Injured*	Rate per Million Hours	Rate Ratio and 95% CI	Prevented Fraction % (a-b/a)x100
NO	12,812,726	392	30.6 (a)	1.52 (1.36-1.69)	34.0
YES	87,078,438	1,760	20.52 (b)	-	-
Total	99,891,164	2,152	21.5	-	-