Content uploaded by Maura D Iversen
Author content
All content in this area was uploaded by Maura D Iversen
Content may be subject to copyright.
Northeastern University
Bouvé Faculty Publications Bouvé College of Health Sciences
July 31, 1997
A controlled trial of an educational program to
prevent low back injuries
Lawren H. Daltroy
Brigham and Women's Hospital; Harvard Medical School; Harvard School of Public Health
Maura D. Iversen
Brigham and Women's Hospital; Northeastern University
Martin G. Larson
Brigham and Women's Hospital; Harvard School of Public Health; Boston University School of Medicine; The Framingham
Heart Study
Robert Lew
Brigham and Women's Hospital; Harvard Medical School; Boston University School of Medicine
Elizabeth Wright
Brigham and Women's Hospital; Harvard Medical School
See next page for additional authors
This work is available open access, hosted by Northeastern University.
Recommended Citation
Daltroy, Lawren H.; Iversen, Maura D.; Larson, Martin G.; Lew, Robert; Wright, Elizabeth; Ryan, James; Zwerling, Craig; Fossel,
Anne H.; and Liang, Matthew H., "A controlled trial of an educational program to prevent low back injuries" (1997). Bouvé Faculty
Publications. Paper 18. http://hdl.handle.net/2047/d20000944
Special Article
322
July 31, 1997
The New England Journal of Medicine
A CONTROLLED TRIAL OF AN EDUCATIONAL PROGRAM TO PREVENT
LOW BACK INJURIES
L
AWREN
H. D
ALTROY
, D
R
.P.H., M
AURA
D. I
VERSEN
, B.S.P.T., S.D., M
ARTIN
G. L
ARSON
, S.D., R
OBERT
L
EW
, P
H
.D.,
E
LIZABETH
W
RIGHT
, P
H
.D., J
AMES
R
YAN
, M.D., M.P.H., C
RAIG
Z
WERLING
, M.D., P
H
.D., A
NNE
H. F
OSSEL
,
AND
M
ATTHEW
H. L
IANG
, M.D., M.P.H.
A
BSTRACT
Background
Low back injuries are common and
costly, accounting for 15 to 25 percent of injuries
covered by workers’ compensation and 30 to 40 per-
cent of the payments made under that program. The
high costs of injury, the lack of effective treatment,
and the evidence that there are behavioral risk fac-
tors have led to widespread use of employee educa-
tion programs that teach safe lifting and handling.
The effectiveness of those programs, however, has
received little rigorous evaluation.
Methods
We evaluated an educational program
designed to prevent low back injury in a randomized,
controlled trial involving about 4000 postal workers.
The program, similar to that in wide use in so-called
back schools, was taught by experienced physical
therapists. Work units of workers and supervisors
were trained in a two-session back school (three
hours of training), followed by three to four rein-
forcement sessions over the succeeding few years.
Injured subjects (from both the intervention and the
control groups) were randomized a second time to
receive either training or no training after their re-
turn to work.
Results
Physical therapists trained 2534 postal
workers and 134 supervisors. Over 5.5 years of fol-
low-up, 360 workers reported low back injuries, for a
rate of 21.2 injuries per 1000 worker-years of risk.
The median time off from work per injury was 14
days (range, 0 to 1717); the median cost was $204
(range, zero to $190,380). After their return to work,
75 workers were injured again. Our comparison of
the intervention and control groups found that the
education program did not reduce the rate of low
back injury, the median cost per injury, the time off
from work per injury, the rate of related musculoskel-
etal injuries, or the rate of repeated injury after re-
turn to work; only the subjects’ knowledge of safe
behavior was increased by the training.
Conclusions
A large-scale, randomized, controlled
trial of an educational program to prevent work-
associated low back injury found no long-term ben-
efits associated with training. (N Engl J Med 1997;
337:322-8.)
©1997, Massachusetts Medical Society.
From the Robert Breck Brigham Multipurpose Arthritis and Musculo-
skeletal Diseases Center (L.H.D., M.D.I., M.G.L., R.L., E.W., A.H.F.,
M.H.L.) and the Department of Rheumatology–Immunology (L.H.D.,
M.G.L., R.L., M.H.L.), Brigham and Women’s Hospital; the Department
of Medicine, Harvard Medical School (L.H.D., R.L., E.W., M.H.L.); Har-
vard School of Public Health (L.H.D., M.G.L., J.R., M.H.L.); the Depart-
ment of Physical Therapy, Boston Bouvé College of Pharmacy and Health
Sciences, Northeastern University (M.D.I.); the Evans Department of
Medicine, Boston University School of Medicine (M.G.L., R.L.); and the
U.S. Postal Service (J.R., C.Z.) — all in Boston; the Framingham Heart
Study, Framingham, Mass. (M.G.L.); and the University of Iowa Injury
Prevention Research Center, Iowa City (C.Z.). Address reprint requests to
Dr. Daltroy at the RBB Multipurpose Arthritis and Musculoskeletal Dis-
eases Center, Brigham and Women’s Hospital PBB-B2, 75 Francis St., Bos-
ton, MA 02115.
OW back pain affects 70 to 80 percent of
adults at some time.
1
In the United States
and Canada,
2-4
low back injuries constitute
15 to 25 percent of the injuries covered by
workers’ compensation and account for 30 to 40
percent of workers’ compensation payments. Most
compensation claims related to back injury (87 per-
cent) are for strains and sprains, most of which (72
percent) are due to overexertion, as in lifting and
handling.
4
Back schools are educational programs
developed by physical therapists for patients with
back pain.
5-7
Most such programs include informa-
tion on back anatomy and physiology, the mecha-
nisms of pain, pain management, good posture, safe
techniques of lifting and handling, and muscle
strengthening and stretching; the training is given in
small groups.
8
Despite the variable and inconclusive
results of controlled clinical trials of training,
1,8-10
the
high cost of back injuries, the lack of effective treat-
ment, and evidence of the existence of behavioral
risk factors have led to widespread adoption of em-
ployee education programs that teach safe lifting
and handling as a form of primary prevention.
1,11
The first reports of the success of back schools in
reducing rates of industrial low back injury came
from trials using historical controls
12,13
and have not
been supported by more broadly based comparisons
of companies with and without such programs.
14
La-
had et al.
15
identified several randomized, controlled
trials of education and exercise programs in the
L
The New England Journal of Medicine
Downloaded from nejm.org at NORTHEASTERN UNIVERSITY on June 13, 2011. For personal use only. No other uses without permission.
Copyright © 1997 Massachusetts Medical Society. All rights reserved.
CONTROLLED TRIAL OF AN EDUCATIONAL PROGRAM TO PREVENT LOW BACK INJURIES
Volume 337 Number 5
323
workplace.
16-21
Programs including exercise reduced
back pain and the number of sick days, but no pro-
gram reduced injury rates. The generalizability of
these studies was limited by the small samples and a
focus on volunteers or subjects with a history of back
problems. Lahad et al.
15
concluded that although
there is some evidence that exercise prevents low
back pain, the justification for other prevention strat-
egies is insufficient. Several recent studies of back
schools have found improved intermediate outcomes
associated with training, such as increased knowl-
edge, safer behavior, and fewer visits to a doctor, but
no reductions in injury rates or sick leave.
22-25
We developed a back school for the primary pre-
vention of low back injury and evaluated it in a large,
randomized, controlled trial in an industrial setting,
which included all employees of the two facilities in-
volved. The program’s design and its effect on work-
ers’ knowledge and behavior have been described
elsewhere
26
; the effect on rates of low back injury is
described here.
METHODS
Population
The study population consisted of approximately 4000 U.S.
postal workers at two mail-processing facilities. Before the study,
low back injury occurred at a rate of 2.4 percent per year and ac-
counted for 17 percent of workers’ compensation injuries and 35
percent of workers’ compensation payments, figures similar to na-
tional averages. The workers studied included mail handlers (and
some maintenance workers), who do heavy lifting — such as han-
dling 16-to-32-kg (35-to-70-lb) bags — and clerks, who do light
work, such as manual or mechanized mail sorting. In the study
population, 70 percent of low back injuries in the five years before
the study began were related to lifting and handling.
Intervention
The intervention, described in detail elsewhere,
26
included all
elements of typical employee-education programs on low back
safety
5,6,27
but was adapted to a Postal Service setting (Table 1)
and was designed with extra features according to health-educa-
tion principles.
28
Workers and supervisors, in groups of 10 to 12,
were taught principles of back safety, correct lifting and handling,
posture, exercises, and pain management; the instructors were
physical therapists. The therapists examined each work station
with workers and supervisors and suggested physical and proce-
dural modifications, such as the adjustment of shelf heights, the
use of lumbar supports on chairs, the installation of rollers on
chairs, and changes in the pacing of mechanized operations. Work-
ers and line supervisors discussed ways to facilitate implementa-
tion. Protective equipment, such as lifting belts, was not part of
the program. Subjects were trained in work-unit groups in order
to allow workers and supervisors to establish and reinforce norms
of proper lifting and handling and to reduce the likelihood that
the workers given training would interact with and influence
the control group. The therapists provided additional reinforce-
ment training six months after the first sessions and yearly there-
after.
Because the Postal Service considered the program to be an em-
ployee-safety initiative, attendance at training sessions was manda-
tory, on paid company time, for workers and supervisors in the in-
tervention group. Protocols for data collection and safeguards for
the protection of the workers’ privacy and rights were negotiated
with union representatives and approved by our hospital’s Com-
mittee to Protect Human Subjects from Research Risks.
Design
The study was a randomized, controlled trial lasting 5.5 years.
The unit of analysis was the work unit: an administrative grouping
of employees, usually on the same shift, working under a com-
mon general supervisor in a shared work environment. We ar-
ranged 34 work units into 17 pairs, matched on the basis of craft
(clerk or mail handler [the latter category included a small num-
ber of maintenance workers]) and job characteristics (machine or
manual work), and we randomly assigned 1 unit of each pair to
receive training and the other to serve as a control group. All the
workers (including controls) received the standard Postal Service
training in back-injury prevention (a film shown at the orienta-
tion of new employees) as well as periodic safety talks given at
their supervisors’ discretion. Control workers transferring into in-
tervention-group units were also trained, to reduce any diminu-
tion of the intervention’s effect. Data on exposure to risk and in-
jury were not collected for a matched pair of units until training
began in the intervention-group unit.
If a worker had a back injury, we calculated the number of cal-
endar days between the injury and the subject’s return to work.
The individual worker was the unit of analysis for the assessment
of the effect of pre-injury training on time off from work. Deci-
sions about return to work were made independently of the
study. When they returned to work, injured subjects (from both
the intervention group and the control group) were randomly as-
signed to participate or not in the ongoing back-education class-
room sessions and were stratified according to the original status
of their work unit (intervention group or control group). Injured
subjects from control units received training, on their return to
work, in primary-prevention sessions with a neighboring inter-
vention-group unit, but they continued working in their own
original work units. This secondary randomization allowed us
to assess, in a two-by-two factorial design, the differential effect
of classroom training (intervention-group training vs. secondary
training of controls) on time until further injury for individual
workers.
Measured Variables and Data Management
The data collected from computerized Postal Service records
included the subject’s craft category (clerk or mail handler [in-
cluding maintenance workers]), sex, age, duration of employ-
T
ABLE
1.
E
LEMENTS
OF
THE
P
ROGRAM
TO
P
REVENT
B
ACK
I
NJURY
.
Teaching staff
Physical therapists
Classroom methods
Two 90-minute sessions
12 workers and supervisors per class
Lecture
Discussion
Film and slides
Pamphlets
Demonstration and practice
Program content
Safe lifting and handling
Posture while sitting, standing, and lying down
Pain management
Stretching and strengthening exercises
Group discussion of barriers to implementation
On-site work-station ergonomic analysis
Reinforcement
Physical therapists observe work stations, followed
by review with supervisor, workers, or both
Supervisors trained to provide reinforcement to
workers, with in-person or videotape review of
supervisor’s skills annually
Pamphlets, paycheck stuffers (1 or 2 a year)
The New England Journal of Medicine
Downloaded from nejm.org at NORTHEASTERN UNIVERSITY on June 13, 2011. For personal use only. No other uses without permission.
Copyright © 1997 Massachusetts Medical Society. All rights reserved.
324
July 31, 1997
The New England Journal of Medicine
ment, work unit, shift, and hours worked (as estimated quarterly
for each work unit). The occurrence of a back injury, its cause,
and its cost were determined on the basis of workers’ compensa-
tion claims and Postal Service accident reports; either data source
was sufficient to identify a case, but 89 percent of the injuries
were identified from both sources. Diagnoses (acute low back
pain, with or without radiculopathy; chronic low back pain; and
disk or inflammatory disease) were taken from the workers’ com-
pensation claim forms filled out by the workers’ physicians. Data
on individual injured workers were cross-checked by hand.
Statistical Analysis
In the main hypothesis test we used extended log-linear mod-
els
29
to compare rates of low back injury in the intervention and
control groups. Only a worker’s first injury was counted. In sec-
ondary analyses, we examined the intervention’s effect on the
rates of other musculoskeletal injuries (not low back) related to
lifting and handling. Time off per injury and time until further
injury were modeled with life-table analyses.
30
Given the highly
skewed distributions, we compared the costs of injuries with Wil-
coxon rank-sum statistics. Our main analysis was based on the as-
signment of work units to be trained, regardless of the actual per-
centage of workers trained, but we also modeled workers’ actual
exposure to the education program. We regard an intention-to-
treat analysis as the appropriate one for assessing a policy of train-
ing all workers.
We estimated that we would need data on 390 injuries to have
an 80 percent power to detect a 25 percent reduction in injury
rates (alpha
0.05).
31
If there were 5000 workers in the study and
a base injury rate of 17 per 1000 worker-years (1 worker-year con-
sisting of 2000 hours of work), 390 injuries would accrue in
5 years. This sample size and estimated injury rate would also give
the study an 80 percent power to detect a 25 percent reduction
in days off from work per injury and in time until further injury,
as assessed with survival-analysis techniques.
RESULTS
Program Implementation
Over a period of five years (from September 1985
to September 1990), 12 staff physical therapists and
2 senior therapists trained 2534 workers and 134 su-
pervisors in primary prevention. After the end of
training, we continued to track injuries and their
costs for six months. During the study period, there
were 8886 reinforcement contacts (personal, video,
and written) from the physical therapists (3.5 per
worker), plus an undocumented amount of rein-
forcement by line supervisors. Training sessions were
scheduled so as not to interfere with the work units’
productivity; periodic catch-up sessions maintained
each unit’s training level.
In the last three years of the study, we maintained
an average of 71 percent of workers trained (range,
66 to 86 percent) in intervention-group work units
(Fig. 1); transfers gradually contaminated the con-
trol-group units (20 percent of control-group work-
ers had received training by the end of the study).
The average proportion of trained subjects, during
all 5.5 years of the study, was 61 percent in interven-
tion-group units and 8 percent in control-group
units. Worker turnover prevents any simple charac-
terization of the study groups, but a cross-section of
the population in the fourth quarter of fiscal 1990
showed the two study groups to be similar in age,
sex, craft category, and duration of employment (Ta-
ble 2). A survey conducted at the study’s midpoint
26
found significant increases in knowledge of safe be-
havior among workers in the intervention group, as
compared with the control group, but no significant
improvements in actual behavior, as reported by the
subjects, or significant reductions in the proportion
of workers with tired backs.
Rates of Primary Low Back Injury
Among the subjects, 360 had low back injuries in
16,960 worker-years of exposure to risk (21.2 inju-
ries per 1000 worker-years of exposure), a rate sim-
ilar to that seen in national data and historical data
from the site. Most of the injuries (93 percent) were
characterized by acute low back pain (or strain), and
85 percent were due to lifting and handling (Table
3). Of the injuries, 58 percent caused the loss of at
least one workday beyond the day of injury itself.
Primary Prevention of Low Back Injury
As recorded quarterly, each work unit in the study
had a slowly changing mix of trained and untrained
mail handlers and clerks. We estimated the effect of
the intervention on the rates of primary injury with
log-linear models, examining the influence of craft,
season, fiscal year, and proportion of trained workers
in the work unit for each quarter. Only the craft cat-
egory had a statistically significant effect on out-
come; mail handlers had a higher injury rate than
clerks (rate ratio, 1.24; 95 percent confidence inter-
val, 1.01 to 1.53). The variation in outcome accord-
ing to fiscal year was nearly significant (chi-square
8.96, 5 df; P
0.11). Intervention-group units had
Figure 1.
Percentage of Postal Service Workers Trained, Accord-
ing to Study Group.
Data points are for the midpoint of each quarter of the fiscal year.
0
100
90
80
70
60
50
40
30
20
10
1986 1987 1988 1989 1990 1991
Fiscal Year
Workers Trained (%)
Intervention
group
Control group
The New England Journal of Medicine
Downloaded from nejm.org at NORTHEASTERN UNIVERSITY on June 13, 2011. For personal use only. No other uses without permission.
Copyright © 1997 Massachusetts Medical Society. All rights reserved.
CONTROLLED TRIAL OF AN EDUCATIONAL PROGRAM TO PREVENT LOW BACK INJURIES
Volume 337 Number 5
325
a higher rate of injury than control-group units (rate
ratio, 1.11; 95 percent confidence interval, 0.90 to
1.37), but the difference was not significant. Simi-
larly, trained workers had a higher rate of injury than
untrained workers (rate ratio, 1.12; 95 percent con-
fidence interval, 0.49 to 2.55), but the difference
was not significant. The intervention’s effect re-
mained nonsignificant, even when we controlled for
craft category and year. The comparison of trained
and untrained workers within the study groups, ac-
cording to craft category (Table 4), showed that
trained workers had lower rates of injury in three of
the four comparisons, and that untrained workers in
intervention-group units had a higher rate of injury
than untrained workers in control-group units.
There were 177 lifting-and-handling injuries that
caused lost workdays (10.4 such injuries per 1000
worker-years). These are the most consequential of
the injuries that might be influenced by an interven-
tion to change behavior. Mail handlers had higher
rates of this type of injury than did clerks (rate ratio,
1.34; 95 percent confidence interval, 0.99 to 3.61).
In a pattern consistent with historical data, there
were significant seasonal variations in the rate of
these injuries (chi-square
15.02, 2 df; P
0.001);
the rate was almost twice as high in spring and sum-
mer as in fall and winter. Although the rate of major
lifting-and-handling injuries was higher in interven-
tion-group units than in control-group units (rate
ratio, 1.29; 95 percent confidence interval, 0.96 to
1.73), the difference was not significant, even when
we controlled for craft category and season. Trained
and untrained workers had similar rates of this type
of injury (rate ratio, 0.97; 95 percent confidence in-
terval, 0.74 to 1.30).
Other Musculoskeletal Injuries
We used accident-report data to examine the ef-
fect of the program on lifting-and-handling injuries
to other major muscle groups (neck, abdomen, chest,
hip, shoulder, trunk, and thigh) that would presum-
ably be protected, like the back, by changes in be-
havior. Since the intervention focused on injury to
the back, we expected the data on other injuries to
be less susceptible to reporting bias and Hawthorne
effects (responses to administrative concern over
working conditions rather than the intervention it-
self).
There were 359 accident reports of musculoskel-
etal injuries not involving the back that were due to
lifting and handling and related unsafe practices
(21.2 such injuries per 1000 worker-years). The rate
of these injuries was 15.4 per 1000 worker-years for
control-group clerks, 19.8 per 1000 for interven-
tion-group clerks, 28.0 per 1000 for control-group
mail handlers, and 27.2 per 1000 for intervention-
group mail handlers. Mail handlers had a higher rate
of such injuries than clerks (rate ratio, 1.58; 95 per-
cent confidence interval, 1.24 to 2.02), but there
was no significant difference between the interven-
tion group and the control group (rate ratio, 1.15;
95 confidence interval, 0.93 to 1.41), even when we
controlled for craft category.
Time off from Work
There was little difference in the proportions of
total injuries that resulted in lost workdays in the in-
tervention group (61 percent) and the control group
*The workers who were not clerks were mail han-
dlers or maintenance workers.
T
ABLE
2.
C
HARACTERISTICS
OF
I
NTERVENTION
-
G
ROUP
AND
C
ONTROL
-G
ROUP
W
ORK
U
NITS
AS
OF
THE
F
OURTH
Q
UARTER
OF
F
ISCAL
1990.
C
HARACTERISTIC
I
NTERVENTION
C
ONTROL
No. of workers 1703 1894
Mean (
SD) age — yr 43.0
12.0 42.0
12.5
Average (
SD) duration of
employment — yr
6.2
4.0 5.1
3.8
Male sex — % 75 74
Clerks — %* 60 65
*The workers who were not clerks were mail han-
dlers or maintenance workers.
†Numbers of days shown are for quartiles of work-
ers returning after losing work time due to back in-
jury. Data on 10 workers were censored because the
study ended before they returned to work.
T
ABLE
3.
C
HARACTERISTICS
OF
360 W
ORKERS
WITH
L
OW
B
ACK
I
NJURIES
.
C
HARACTERISTIC
V
ALUE
Male sex — % 66
Clerks — %* 58
Age — yr
Median
Range
33
18–73
Duration of employment — mo
Median
Range
33
1–378
History of back injury in past 3 yr — % 6
Acute low back pain, with or without
radiculopathy — %
93
Chronic low back pain — % 3
Disk or inflammatory disease — % 4
Lifting-and-handling–related injury — % 85
Lost
1 workday after injury — % 58
No. of calendar days until return to work
for those losing work time†
25% back at work
50% back at work
75% back at work
100% back at work
8
14
40
1717
The New England Journal of Medicine
Downloaded from nejm.org at NORTHEASTERN UNIVERSITY on June 13, 2011. For personal use only. No other uses without permission.
Copyright © 1997 Massachusetts Medical Society. All rights reserved.
326 July 31, 1997
The New England Journal of Medicine
(56 percent). The only factor significantly associated
with the loss of workdays was a history of a workers’
compensation claim for a non-back injury in the
previous three years (rate ratio, for those with such
a history as compared with those without, 3.4; 95
percent confidence interval, 1.28 to 9.03). For the
210 workers with lost workdays, the median time off
from work was 14 days (range, 1 to 1717). Survival-
analysis (log-rank) models of time elapsed until re-
turn to work showed no significant effect associated
with assignment to an intervention-group unit or
with training before the injury. Only the cause of in-
jury reliably predicted time off from work; the me-
dian number of days off was 13 for lifting-related
injuries and 18 for injuries not related to lifting (chi-
square 3.99, P 0.046).
Cost
The median total cost per back injury, as accrued
through the end of the study, was $204 (range, zero
to $190,380). The median cost was $103 in the
control group (range, zero to $190,380), and $309
in the intervention group (range, zero to $122,145).
Workers with a history (before the study) of low
back injury had higher median total costs than did
workers without such a history ($1,300 vs. $192;
P0.005), higher median medical costs ($150 vs. ze-
ro, P 0.03), and higher median personnel-replace-
ment costs ($965 vs. zero, P 0.004). No other var-
iable, including study group or training status, was
significantly associated with cost.
Training after Injury
We evaluated whether participation in an ongoing
primary-prevention program, after returning to work
following a low back injury, prevented further inju-
ries. There were 183 primary injuries among inter-
vention-group workers; 90 of these injured workers
were assigned to training and 93 to no training on
their return to work. Of the 90 assigned to training,
62 percent actually received training; of the 93 as-
signed to no training, 26 percent in fact received
subsequent training. The 26 percent training rate in
this secondary control group reflects subjects’ par-
ticipation in their units’ ongoing primary-prevention
sessions.
There were also 177 primary injuries among con-
trol-group workers; 84 were assigned to training and
93 to no training on their return to work. Of the 84
assigned to training, 69 percent actually received it;
of the 93 designated controls, 5 percent in fact re-
ceived subsequent training. The median time from
return to work until the start of training, for all in-
jured subjects, was five months, because subjects had
to wait for the scheduled classes on their shifts.
Of all the injured subjects, 75 (21 percent) had at
least one further injury. The median time between
return to work and another injury was 4 three-
month quarters (range, 0 to 21). Time until further
injury was analyzed with Cox regression models,
30
with the quarter as the time unit. The likelihood of
repeated injury was higher if the worker had lost
more workdays because of the initial injury (rate ra-
tio for 15 days vs. 1 to 14 days vs. none, 1.05; 95
percent confidence interval, 1.01 to 1.09); if the
worker had a more serious initial injury, defined as
a disk problem, inflammatory disease, or chronic low
back pain as compared with acute low back pain alone
(rate ratio, 1.62; 95 percent confidence interval,
1.05 to 2.51); or if the worker was a man (rate ratio,
1.66; 95 percent confidence interval, 0.98 to 2.83).
Age, craft category, duration of employment, and a
history of a low back injury before the start of the
study had no effect on the likelihood of repeated in-
jury. When we controlled for the seriousness of the
initial injury, the time off from work resulting from
*The overall crude rate was 21.2 injuries per 1000 worker-years of exposure. The individual rates
shown have been fitted to proportional exposure time. A small number of maintenance workers are
included in the group of mail handlers.
TABLE 4. FITTED RATES OF PRIMARY LOW BACK INJURIES AMONG BOSTON POSTAL
WORKERS, ACCORDING TO CRAFT CATEGORY, WORK-UNIT ASSIGNMENT TO
THE INTERVENTION OR CONTROL GROUP, AND ACTUAL TRAINING STATUS.*
CATEGORY OF
WORKERS MAIL HANDLERS CLERKS
CONTROL
-
GROUP
UNITS
INTERVENTION
-
GROUP
UNITS
COMBINED
UNITS
CONTROL
-
GROUP
UNITS
INTERVENTION
-
GROUP
UNITS
COMBINED
UNITS
rate of injury per 1000 worker-years
Untrained workers 21.6 27.4 23.4 19.4 23.2 20.4
Trained workers 27.2 25.0 25.4 16.4 18.0 17.8
All workers 22.2 25.8 24.2 19.2 20.0 19.5
The New England Journal of Medicine
Downloaded from nejm.org at NORTHEASTERN UNIVERSITY on June 13, 2011. For personal use only. No other uses without permission.
Copyright © 1997 Massachusetts Medical Society. All rights reserved.
CONTROLLED TRIAL OF AN EDUCATIONAL PROGRAM TO PREVENT LOW BACK INJURIES
Volume 337 Number 5 327
the initial injury, and sex, we found that the study-
group assignment, assignment to training or no train-
ing after injury, or whether the subject was actually
trained had no significant effect on the likelihood
of repeated injury. However, the power of our anal-
ysis to detect a treatment effect was reduced to 30
percent because of the low rates of actual training
among workers assigned to training and contamina-
tion of the control group by the inclusion of workers
who actually received training.
DISCUSSION
Employers, eager to reduce the illness, lost work,
and cost associated with low back injuries, have
adopted back-education programs without clear ev-
idence of their effectiveness. The results of our large,
randomized trial indicate that back schools are not
by themselves an effective intervention for the pri-
mary prevention of industrial low back injury. We
found that workers’ being in a unit assigned to train-
ing or actually being trained had no significant effect
on rates of primary low back injury, on time off
from work, on costs associated with injury, or on
time elapsed until a further injury.
The failure to detect any effect of a program may
be due to inadequate sample size, inadequate imple-
mentation of the program, or a fundamental lack of
efficacy of the program. Our analysis had adequate
power (78 percent) to detect meaningful differences
between groups, but contrary to our expectations,
injury rates were higher in the intervention group
than among controls. It is unlikely that our interven-
tion caused injuries; the techniques of lifting and
handling that were recommended, which are biome-
chanically correct in theory and taught in a large
number of back schools, have been shown to be ca-
pable of reducing the number and severity of back
symptoms.
5,7
Our data suggest that the elevated rate of claims
in our study may be due to an increased acceptability
of reporting injuries among the intervention-group
units (Table 4). In three of four comparisons of
trained and untrained workers (in groups defined by
craft category and study-group assignment), trained
workers had lower injury rates than untrained work-
ers, which suggests that training was effective. More-
over, untrained workers in intervention-group units
had higher injury rates than untrained workers in
control-group units, which suggests a reporting bias.
Finally, the number of reports of lifting-and-han-
dling injuries involving muscle groups other than
the back were also higher in the intervention-group
units (data not shown).
The education program successfully imparted
knowledge and skills related to safe lifting and han-
dling, but despite this training and, for most inter-
vention-group workers, its regular reinforcement,
the increased practice of desirable behavior did not
take place. This failure to change behavior reflects
complex factors such as the workers’ level of job sat-
isfaction and negative perceptions of what supervi-
sors and coworkers were doing to improve back safe-
ty.
26
Our program included substantial efforts to
maintain and reinforce the practice of learned be-
havior. If such reinforcement is in fact ineffective, it
represents a key weakness in educational programs
directed at individual workers and small groups at
the work site. Larger economic and social factors,
and management–labor issues, may ultimately deter-
mine the success or failure of such programs.
The failure of our program to reduce the number
of repeated injuries may be due in part to the in-
complete and delayed training of returning workers,
which significantly reduced the power of the analysis
to detect differences between groups. Programs that
focus on educating injured workers before their re-
turn to work
32
and programs emphasizing exercise
for workers after their return
15
may be more effective
than our approach. In this trial, we could not study
the effect of offering training to freshly injured work-
ers to speed their recovery, because workers resisted
participation in such programs while workers’ com-
pensation claims were pending.
We believe that by dealing realistically with the
challenges of training a changing work force in a
large industry, our program has provided a fair test
of the ability of back-education programs to limit in-
juries. Our results are consistent with the findings of
the few other randomized, controlled studies of back
schools, which have generally found no significant
effect of the schools on injury rates, despite some ev-
idence of better pain management and reductions in
the number of sick days.
7
Back-education programs
may still have a place in reducing back injury and
disability, but they appear to be ineffective when un-
dertaken alone.
Supported in part by a grant (AR36308) from the National Institutes of
Health.
We are indebted to the management and employees of the U.S.
Postal Service, the American Postal Workers Union, Boston Metro
Area Local, and Mailhandlers Local 301 for their help and cooper-
ation, and to therapists from the Department of Rehabilitation Serv-
ices, Brigham and Women’s Hospital, for help in training workers.
REFERENCES
1. Deyo RA, Cherkin D, Conrad D, Volinn E. Cost, controversy, crisis:
low back pain and the health of the public. Annu Rev Public Health 1991;
12:141-56.
2. Spengler DM, Bigos SJ, Martin NA, Zeh J, Fisher L, Nachemson A.
Back injuries in industry: a retrospective study. I. Overview and cost anal-
ysis. Spine 1986;11:241-5.
3. Bombardier C, Baldwin J-A, Crull L. The epidemiology of regional
musculoskeletal disorders: Canada. In: Hadler NM, Gillings DB, eds. Ar-
thritis and society: the impact of musculoskeletal diseases. London: But-
terworths, 1985:104-18.
4. Klein BP, Jensen RC, Sanderson LM. Assessment of workers’ compen-
sation claims for back strains/sprains. J Occup Med 1984;26:443-8.
5. Forssell MZ. The back school. Spine 1981;6:104-6.
The New England Journal of Medicine
Downloaded from nejm.org at NORTHEASTERN UNIVERSITY on June 13, 2011. For personal use only. No other uses without permission.
Copyright © 1997 Massachusetts Medical Society. All rights reserved.
328 July 31, 1997
The New England Journal of Medicine
6. Mattmiller AW. The California Back School. Physiotherapy 1980;66:
118-21.
7. Ayoub MA. Control of manual lifting hazards. III. Preemployment
screening. J Occup Med 1982;24:751-61.
8. Keijsers JFEM, Steenbakkers MWHL, Meertens RM, Bouter LM, Kok
G. The efficacy of the back school: a randomized trial. Arthritis Care Res
1990;3:204-9.
9. Schlapbach P. Back school. In: Schlapbach P, Gerber NJ, eds. Physio-
therapy: controlled trials and facts. Vol. 14 of Rheumatology: the interdis-
ciplinary concept. Basel, Switzerland: Karger, 1991:25-33.
10. Snook SH. Approaches to the control of back pain in industry: job de-
sign, job placement, and education/training. Occup Med 1988;3:45-59.
11. Fielding JE, Piserchia PV. Frequency of worksite health promotion ac-
tivities. Am J Public Health 1989;79:16-20.
12. Videman T, Kosunen J, Asp S, Cedercreutz G. Low back pain in the
Saab-Valmet car manufacturing plant in 1976-1979: a survey of factors re-
lated to the reduction of absences. Sosiaalilaalcetieteellinen Aikakauslehti J
1983;20:160-70.
13. Fitzler SL, Berger RA. Chelsea Back Program: one year later. Occup
Health Saf 1983;52:52-4.
14. Snook SH, Campanelli RA, Hart JW. A study of three preventive ap-
proaches to low back injury. J Occup Med 1978;20:478-81.
15. Lahad A, Malter AD, Berg AO, Deyo RA. The effectiveness of four in-
terventions for the prevention of low back pain. JAMA 1994;272:1286-91.
16. Donchin M, Woolf O, Kaplan L, Floman Y. Secondary prevention of
low-back pain: a clinical trial. Spine 1990;15:1317-20.
17. Linton SJ, Bradley LA, Jensen I, Spangfort E, Sundell L. The second-
ary prevention of low back pain: a controlled study with follow-up. Pain
1989;36:197-207.
18. Walsh NE, Schwartz RK. The influence of prophylactic orthoses on
abdominal strength and low back injury in the workplace. Am J Phys Med
Rehabil 1990;69:245-50.
19. McCauley M. The effect of body mechanics instruction on work per-
formance among young workers. Am J Occup Ther 1990;44:402-7.
20. Gundewall B, Lijeqvist M, Hansson T. Primary prevention of back
symptoms and absence from work: a prospective randomized study among
hospital employees. Spine 1993;18:587-94.
21. Kellett K, Kellett DA, Nordholm LA. Effects of an exercise program
on sick leave due to back pain. Phys Ther 1991;71:283-93.
22. Shi L. A cost-benefit analysis of a California county’s back injury pre-
vention program. Public Health Rep 1993;108:204-11.
23. Woodruff SI, Conway TL, Bradway L. The U.S. Navy Healthy Back
Program: effect on back knowledge among recruits. Mil Med 1994;159:
475-84.
24. Feldstein A, Valanis B, Vollmer W, Stevens N, Overton C. The Back
Injury Prevention Project pilot study: assessing the effectiveness of Back At-
tack, an injury prevention program among nurses, aids, and orderlies. J Oc-
cup Med 1993;35:114-20.
25. Weber M, Cedraschi C, Roux E, Kissling RO, Von Kanel S, Dalvit G.
A prospective controlled study of low back school in the general popula-
tion. Br J Rheumatol 1996;35:178-83.
26. Daltroy LH, Iversen MD, Larson MG, et al. Teaching and social sup-
port: effects on knowledge, attitudes, and behaviors to prevent low back
injuries in industry. Health Educ Q 1993;20:43-62.
27. Porter RW, Paris SV. The Back School. Atlanta: Back School, 1984.
28. Green LW, Kreuter MW, Deeds SG, Partridge KB. Health education
planning: a diagnostic approach. Palo Alto, Calif.: Mayfield Publishing,
1980.
29. Laird N, Olivier D. Covariance analysis of censored survival data using
log-linear analysis techniques. J Am Stat Assoc 1981;76:231-40.
30. Cox DR. Regression models and life-tables. J R Stat Soc [B] 1972;34:
187-220.
31. Schoenfeld DA, Richter JR. Nomograms for calculating the number
of patients needed for a clinical trial with survival as an endpoint. Biomet-
rics 1982;38:163-70.
32. Bergquist-Ullman M, Larsson U. Acute low back pain in industry: a
controlled prospective study with special reference to therapy and con-
founding factors. Acta Orthop Scand 1977;170:1-117.
RECEIVE THE JOURNAL’S TABLE OF CONTENTS EACH WEEK BY E-MAIL
To receive the table of contents of the New England Journal of Medicine by e-mail
every Thursday morning, send an e-mail message to:
listserv@massmed.org
Leave the subject line blank, and type the following as the body of your message:
subscribe TOC-L
You can also sign up through our website at: http://www.nejm.org
The New England Journal of Medicine
Downloaded from nejm.org at NORTHEASTERN UNIVERSITY on June 13, 2011. For personal use only. No other uses without permission.
Copyright © 1997 Massachusetts Medical Society. All rights reserved.
