See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/279804362

Exercise-based rehabilitation for injured workers: Programme efficacy and

identificaiton of factors predicting programme completion and outcome

Article  in  International Journal of Rehabilitation Research · March 2000

DOI: 10.1097/00004356-200023010-00002

CITATIONS READS

12 24

1 author:

Dianna Theadora Kenny

The University of Sydney
336 PUBLICATIONS   3,732 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Developmental and educational psychology View project

Workplace injury, prevention, management, and rehabilitation View project

All content following this page was uploaded by Dianna Theadora Kenny on 20 April 2015.

The user has requested enhancement of the downloaded file.

Exercise-based rehabilitation for injured workers: Programme
efficacy and identification of factors predicting programme
completion and outcome
Dianna T. Kenny

International Journal of Rehabilitation Research, 2000, 23, 7-17

A sample of 355 injured workers presenting to a tertiary referral agency for supervised physical activity programmes were assessed for
programme completion and changes in work status at the conclusion of the programme. Seventy-five percent of the sample comprised
long-term (i.e. greater than six months) injured workers, of whom 45% were unemployed at the commencement of the programme.
There was a 15% dropout rate and 49% of completers improved their work status at the end of the programme. A series of stepwise
logistic regression analyses were conducted to identify predictors of dropout and improved work status. Joint pathology was the only
predictor of dropout. Preprogramme work status, referral source, intervertebral pathology, and time between injury and commencement
of the programme predicted post-programme work status. Those workers who were employed, referred by sources other than rehabili-
tation providers, such as doctors, employers or insurers, did not have a diagnosis including intervertebral pathology and who began
their programmes within six months of injury were more likely to improve their work status at the conclusion of the programme. Other
factors induding age, gender, occupation, insurer category, location of injury (i.e. lumbar or other), and neurological signs predicted
neither programme completion nor change in work status.

Keywords: active rehabilitation; exercise; injured workers; programme outcome; prediction

INTRODUCTION treatment; nor was there evidence that discriminated

The burgeoning cost of workers’ compensation
in terms of treatment, compensation for lost wages,
and absenteeism/loss of productivity signals an
urgent need to identify effective rehabilitation
approaches for workers injured in the course of their
duties. A large number of therapeutic interventions
is available for the treatment of common work-
related injuries, particularly those to the lower back,
which are the most common cause of disability for
persons tinder the age of 45 years (Oldridge and
Stoil, 1997). A major challenge for researchers and
practitioners is to identify which treatments are most
effective for specific client groups with specific
types of injuries. In addition to demonstrating the
overall effectiveness of rehabilitation programmes,
studies are needed to identify the situations in which
certain types of rehabilitation interventions are most
effective, in terms of client, injury, and workplace
characteristics. By identifying factors that are
associated with more positive out- conies, it will be
possible to optimize interventions by targeting them
at individuals most likely to show improvements.
In recent years, exercise therapy has become
an increasingly widely used treatment modality,
particularly for chronic low back pain (van Tulder
et al., 1996). However, a review of 16 randomized
controlled trials of physiotherapy exercises for back
pain (Koes e: al., 1991) reported that there were
insufficient data to conclude that exercise therapy
was more effective than other forms of conservative
International Journal of Rehabilitation
Research
amongst exercise types in terms of efficacy. These
conclusions have been reiterated more recently by
Oldriclge and Stoil (1997), who stated that exercise
to increase strength, range of motion, and endurance
are common practice but scientific support for their
use is lacking. In a recent review of the efficacy of
functional restoration progranimes for chronic low
back pain. Teasdell and Harth (1996) concluded that
support for functional restoration was lacking, and
although several studies had reported statistically
significant improvements in return to work rates
of treated participants, the studies were flawed
because of the use of inappropriate comparison
groups, selection bias, and incomplete follow-
up. The question as to whether exercise therapy is
more effective than reference treatments for non-
specific low back pain with regard to pain intensity,
functional status, overall improvement, and return
to work is the subject of a systematic Cochrane
review currently being undertaken by van Tulder
and colleagues (van Tulder et al., 1998).
Despite these inconclusive reviews, physical
activity programmes have been reported to produce
good outcomes for some people undergoing
rehabilitation (Lindstrom et al., 1992; Faas et al.,
1995; Frost et al., 1995; Manniche, 1996; Bendix
et al., 1997). However, reported programmes have
varied along a number of dimensions, including
the type and intensity of activities included in the
programme, the number and frequency of supervised
sessions, and the duration of the programme. There
Exercise-based rehabilitation for injured workers 1
have been few reported studies on the efficacy of
physical activity programmes for injuries other than
back injuries and few that have reported return to
work outcomes for workers with compensable work-
related injuries (Marshall, 1997). One randomised
controlled study (Lindstrom et al., 1992) reported
earlier return to work and lower utilization of
physiotherapy services and sick Leave at two year
follow-up for workers with low back pain who had
undertaken a supervised graded activity progranime
compared to a control group.
A major shortcoming of the few studies that include
return to work as an outcome, report return to work
as a dichotomous variable; that is, the worker was
deemed to have returned to work or not. However,
in the workers’ compensation arena, return to work
can be graded in terms of duties and/or the number
of hours worked. More specific outcome data in
terms of return to work are required in light of
recent findings that work restrictions following an
episode of low back pain were associated with a
reduced probability of returning to pre-injury work
duties, that they did not reduce absenteeism, and
that they did not significantly reduce recurrences
(Symonds et al., 1995, in Burton (1997)]. One study
(Marshall, 1997) that distinguished among levels of
return to work reported that of 43 injured workers
who completed a supervised physical activity
rehabilitation programme for low back injury, 51.2%
were employed in their pre-injury work duties, and
a further 25.6% were performing alternate duties
at 12 month follow-up. The remaining 23.2% were
not working. These results indicate that 37% of the
original group of 58 participants were working full
time in their pre-injuiy work duties. Unfortunately,
the study did not report the return to work
characteristics for the matched controls.
Dropout is a significant problem in rehabilitation.
Dropout rates for supervised physical activity
programmes are considerably lower than those
for unsupervised programmes, but have been
consistently reported to be in the vicinity of 20% to
25% of those who commence (Burke et al., 1994;
Niemmeyer et al., 1994; Petersen, 1995; Bendix
et al., 1997; Marshall, 1997). A major research
task is to predict those people who are unlikely to
complete their rehabilitation programme. This is an
important consideration, given that attrition from
rehabilitation is a predictor of long-term chronicity
and lower return to work (Snook and Webster, 1992).
Related to dropout are the criteria that are used, in
addition to the physical parameters of the injury, to
select participants into supervised physical activity
programmes.
Exercise-based rehabilitation for injured workers 2
This study examined the return to work outcomes
of a structured, 12 session, supervised, individualized
physical activity programme conducted by
qualified exercise therapists under the direction of
physiotherapists, for workers with compensable
work related injuries, who had been referred by
rehabilitation providers, medical practitioners,
and insurers to a private tertiary referral agency
that specialized in active rehabilitation and work
conditioning.
The aims of the study were twofold. The first
aim was to identify the client, injury, work, and/
or rehabilitation factors that reliably predicted
those clients who completed their rehabilitation
programme. The second aim was to identify the
client, injury, work, and/or rehabilitation factors that
reliably predicted those clients who improved their
work status concomitant with their participation in
the rehabilitation programme.
MATERIAL AND METHODS
Participants
The initial sample consisted of 439 consecutively
presenting clients of a tertiary rehabilitation agency
who fulfilled the following criteria: (1) participants
had been injured after January 1, 1989; (2)
participants were less than 60 years of age at the
time of their referral; and (3) participants were not
insured by compulsoiy third party insurers. A total
of 56 clients were excluded under these criteria, 10
because they had been injured before January 1,
1989, 10 because they were 60 years of age or older
and 36 because they were insured by compulsory
third party insurance.
The initial sample was subsequently reduced
to 5 clients by excluding 84 clients who were not
assigned in the standard rehabilitation programme
(see the description below). Of these 84 clients, 46
completed an Exercise Aptitude Assessment (EAA).
six supervised sessions combined with an exercise
therapist report (n = 5), six supervised sessions
combined with a musculoskeletal assessment (n = 5),
a functional capacity evaluation (n = 1), and ‘other’
programmes (n = 27). The 46 EAA clients were
considered by the assessing physiotherapist to be
least likely to respond favourably to the 12-session
programme. The remaining 38 clients comprised
those whose insurers had refused to pay for the full
programme or who had been referred simply for an
evaluation rather than rehabilitation.
Kenny D.T., 2000
Measures
Information was obtained retrospectively by
extracting data from client records on the following
measures; age, gender, occupation, location of
injury, diagnosis of injury, duration between injuiy
and commencement of rehabilitation, presence of
neurological signs, source of referral, type of insurer,
and work status before and after rehabilitation.
Occupation was scored as a categorical variable
with five levels: unemployed, professional, trade,
clerical, and labourer. Location of injury was scored
on three levels: head or upper body, lumbar or lower
body, and multiple sites of injury. Diagnosis of injury
was scored as a categorical variable with seven levels:
intervertebral pathology, soft tissue injury, fracture,
post-surgery, occupational overuse syndrome, joint
pathology, and ‘other’ diagnoses. Duration between
injury and commencement of rehabilitation was
transformed into a categorical variable with three
approximately equal groups by classifying duration
as either under 6 months, between 6 and 12 months,
or more than 12 months. Presence of neurological
signs was scored as a dichotomous yes/no variable,
where neurological signs were defined as specific
patterns of disturbance to sensation, reflexes, and/
or strength. The source of referral was classified as
either rehabilitation provider or ‘other’. The type
of insurer was classified as self-insurer or managed
funds.
Work status was scored on a 12-point ordinal scale.
This scale ranked clients according to whether they
were working or not, whether they were working
full- or part-time, whether they were perfonning full
or selected duties, whether they were employed or
unemployed, whether they were re-training or not,
and whether their status had changed in the past 6
months. For example, clients who were working
full-time with full duties were ranked as a ‘1’,
whereas those who were not work. ing, who were not
participating in re-training, who were unemployed,
and whose work status had not changed for 6
months or more were scored as a ‘12’. An outcome
variable for change in work status over the course of
the programme was calculated by subtracting work
status after the programme from work status before
the programme and then classifying negative values
and zero as ‘worse/no change’ and positive values
as ‘improved’ work status.
A limitation of the outcome variable as it is
described above was that clients who were working
full-time with full duties both before and after
their rehabilitation programme (n 27) would be
scored as ‘worse/no change’. However, given
International Journal of Rehabilitation
Research
that improvement in work status was not possible
for these clients, and that they were referred for
rehabilitation because it was felt that they would
not have been able to maintain their work status
otherwise, it was decided to credit this group of
clients with ‘improved’ work status.
Procedure
The rehabilitation programme was conducted
over 12 training sessions, usually dispersed over
a period of approximately one to three months.
Training sessions were supervised on an individual
basis and conducted in a gym located conveniently
for the client. As part of the programme, most
clients (n = 290; 81.7%) also received a 3-month
gym membership, which they were at liberty to
use outside of the supervised training sessions. The
remaining clients were not given gym membership
because they were judged to be unlikely to use it
independently, and that for finandal reasons it would
be better to pay for the 12 supervised gym sessions
of these clients on a casual basis. Rehabilitation
was tailored to meet the specific requirements of
the client, and included exercises and functional
activities aimed at strengthening and stretching
muscles, improving cardiovascular fitness, and
endurance.
Statistical analyses
Logistic regression analysis with stepwise forward
entry of predictor variables was employed to identify
the client, injury, work and/or rehabilitation factors
that reliably predicted: (1) clients who completed
their rehabilitation programme; (2) clients whose
work status improved over the course of their
programme; and (3) clients who were not working
before their rehabilitation programme and who
returned to work. The same set of predictor variables
was used for each analysis: age, gender, occupation,
pre-programme work status, diagnosis, location of
injury, presence of neurological signs, insurer type,
source of referral, duration between injury and
commencement of rehabilitation, and whether or
not clients received gym membership. To evaluate
the significance of predictor variables with more
than two levels (occupation, diagnosis, and location
of injury), dummy variables were created for entry
into the logistic regression analyses.
Logistic regression analysis is recommended for
statistical analyses predicting dichotomous group
membership where there are a mixture of continuous
and discrete variables (Tabachnik and Fidell,
1996). Logistic regression was more suitable than
discriminant function analysis for the data under
Exercise-based rehabilitation for injured workers 3
investigation, given that in logistic regression the
predictors do not have to be normally distributed,
linearly related, or of equal variance within each
group. All the statistical analyses were performed
using the SPSS statistical package.
RESULTS
Sample characteristics
The demographic, work, and injury details of the
clients in the sample are displayed in Table 1. The
sample consisted of 355 participants, of whom 227
(63.9%) were male and 128 (36.1%) were female.
The mean age of the participants was 37.56 years
(SD = 9.55). There was no statistically significant
differences between the mean age of the males (37.34
years; SD = 9.23) and the mean age of the females
(37.96 years; SD = 10.11). The majority of clients
described their occupation as either clerical (30.7%)
or labouring (31.3%), and most were involved in the
manufacturing (15.5%), trade (13.2%), healthy and
community service (11.3%) or construction (8.5%)
industries.
Injuries were most frequently reported to the
lumbar spine or lower body (62.5%) and were
commonly diagnosed as either soft tissue injuries
(39.4%) or intervertebral pathology (27.6%). Over
sixty percent (61.7%) of the injuries were associated
with the presence of neurological signs. Almost
two-thirds (63.1%) of the clients were referred by
rehabilitation providers, although there was also
a significant number of referrals from doctors
(14.4%). The majority of clients were insured by
managed funds (87.3%) rather than self- insurers
(12.7%). In terms of the duration between injury
and commencement of rehabilitation, less than 6
months had elapsed for 25.1% of clients; between 6
and 12 months had elapsed for 34.6% of the sample,
and more than 12 months (range 1–10 years) had
elapsed for 40.3% of the sample.
Work status
At the commencement of their rehabilitation,
193 (54.4%) of the clients were working in some
capacity. Of these, 35(18.1%) were working full-
time with full duties, 71(36.8%) were working full-
time with selected duties, 7 (3.6%) were working
part- time with full-duties, and 80 (4 1.5%) were
working part-time with selected duties. Ninety-
nine (27.9%) of the clients reported that they
were employed, but were not working at the time.
Sixty-three (15.5%) clients reported that they were
unemployed, of whom 58 had been unemployed for
Exercise-based rehabilitation for injured workers 4
more than 6 months. Only 6 of the clients who were
not working (2 who were employed and 4 who were
unemployed) were involved in re-training.
Programme completion
Out of the 355 clients who began rehabilitation,
53 (14.9%) did not complete their programme.
A logistic regression was performed on whether
clients completed their rehabilitation programme or
not. On step number one of the analysis, diagnosis
of joint pathology entered the regression equation,
and led to a statistically significant improvement
in prediction of completion status compared to the
constant-only model, χ2 (1, n = 355) = 4.42; P<0.05.
On step number two, professional occupation
entered the regression equation, and again led to a
further statistically significant improvement in the
prediction of completion status compared to the
constant-only model, χ2 = 5.71; P < 0.05. No more
predictors were entered into the model. That is, the
remaining factors of age, gender, time between injury
and commencement, neurological signs, location of
injury (lumbar or other), presence of intervertebral
pathology, insurer type, or pre-programme work
status were not significant predictors of programme
completion.
A test of the model with both significant predictors
in the equation against a constant-only model was
statistically reliable, χ2 (2, n = 355) 9.85; p < 0.01,
indicating that the predictors, as a set, reliably
distinguished between those who completed the
rehabilitation programme and those who did not.
The model was limited in its utility as a means
for predicting completion: although 100% of the
completers were identified correctly and overall
prediction was 85.1%, none of the noncompleters
were predicted. In the final model, diagnosis of joint
pathology was significant (ß = 0.96, Wald test =
4.47; P < 0.05) and professional occupation was not
significant (ß = –1.85, Wald test = 3.21; F> 0.05).
This suggests that diagnosis of joint pathology was
the only reliable predictor of programme completion,
with clients who had a diagnosis of joint pathology
being less likely to complete their programme.
Programme outcomes – work status
Outcome information was not available for the 53
clients who did not complete their programmes. At
the completion of their rehabilitation, 206 (68.2%)
of the 302 clients who completed their programmes
were working in some capacity. Of these, 66
(32.0%) were working full-time with full duties,
57 (27.7%) were working full-time with selected
duties, 8 (3.9%) were working part-time with full
Kenny D.T., 2000
 Table 1.Demographic, work, and injury details for the participants (n = 355)

Characteristic Level Frequency Percentage

Gender Male 227 63.9

Female 128 36.1

Occupation Labourer 111 31.3

Clerical 109 30.7
Trade 57 16.1
Unemployed 44 12.4
Professional 34 9.6

Industry Manufacturing 55 15.5

Wholesale and retail trade 47 13.2
Health and community services 40 11.3
Construction 30 8.5
Government admin./Defence 27 7.6
Transport and storage 24 6.8
Property and business 22 6.2
Other 110 31.0

Injury location Lumbar spine or lower body 222 62.5

Head or upper body 81 22.8
Multiple 52 14.6

Diagnosis Soft tissue injury 140 39.4

Intervertebra! pathology 98 27.6
Post-surgery 40 11.3
Joint pathology 27 7.6
Occupational overuse syndrome 23 6S
Fracture 17 4.8
Other 10 2.8

Presence of neurological signs 219 61.7

Referral source Rehabilitation providers 224 63.1

Doctors 51 14.4
Employers 24 6.8
Managed fund organizations 22 6.2
Government businesses 16 4.5
Physiotherapists 15 4.2
Compulsory third party insurer 3 0.8

Insurer type Managed funds 310 87.3

Self-insurer 45 12.7

duties, and 75 (36.4%) were working part-time with
selected duties. Thirty-seven (12.2%) of the clients
were employed, but were not working at the time.
Fifty- nine (19.6%) of the clients were unemployed,
of whom 37 had been unemployed for more than
6 months. Twenty of the clients who were not
working (7 who were employed and 13 who were
unemployed) were involved in re-training.
As indicated by the change in work status
(measured before and after participation in
International Journal of Rehabilitation
Research
the rehabilitation programme on a 12-point
ordinal scale), there appeared to be an overall
improvement for the participants during the course
of the programnie. Of those who completed their
rehabilitation programmes, the work status of 16
(5.3%) deteriorated, there was no change in work
status for 112 (37.1%), and 174 (57.6%) improved
their work status. If the assumption is made that all
of the people who did not complete their programme
deteriorated in terms of their work status, of the 355
Exercise-based rehabilitation for injured workers 5
 Table 2. Logistic regression coefficients, Wald statistics, and significance levels for the prediction of work outcome
Predictor variables
Unemployed status
Diagnosis of intervertebral pathology
Source of referral
Duration of injury - rehabilitation programme
who began rehabilitation programmes, 69 (19.4%)
demonstrated a deterioration in work status, 112
(31.5%) demonstrated no change, and 174 (49%)
demonstrated an improvement.
A logistic regression was performed to predict
those clients whose work status improved during
rehabilitation (n = 174), with forward entry of
predictor variables. On step number one of the
analysis, unemployed status entered the regression
equation, and led to a statistically significant
improvement in prediction of work outcome
compared to the constant-only model, χ2 (1, n = 302)
16.20, P < 0.001. On step number two, diagnosis
of intervertebral pathology entered the regression
equation, and led to a further statistically significant
improvement in the prediction of work outcome
compared to the constant-only model, χ2 = 7.83,
P < 0.01. On step number three, source of referral
entered the equation, and led to a further statistically
significant improvement in the prediction of work
outcome, χ2 = 4.87, P < 0.05. On step number four,
duration between injury and the commencement of
rehabilitation entered the regression equation, and
led to a further statistically significant improvement
in the prediction of work outcome, χ2 = 6.18, P
<0.05. No more predictors were entered into the
model. Once again, age, gender, insurer category,
location of injury (lumbar or other), neurological
signs, or occupation other than ‘unemployed’ were
not significantly associated with improvement in
work status at programme end.
A test of the model with the four significant
predictors included against a constant-only model
was statistically reliable, χ2 (5, n = 302) = 35.07;
P < 0.001, indicating that the predictors, as a set,
reliably distinguished between those who improved
their work status during the programme and those
who did not. Prediction success was reasonable,
with correct identification of 85.6% of those who
improved their work status and 4.0.6% of those
who did not improve their work status. The overall
success rate of prediction was 66.6%.
Table 2 shows the regression coefficients, Wald
statistics, and associated significance levels for each
Exercise-based rehabilitation for injured workers 6
B Wald test P
1.19 8.39 0.004
0.79 7.92 0.005
-0.56 4.62 0.032
0.79 5.62 0.018
of the four predictors that entered the regression
equation. Clients whose work status improved were
more likely to be employed at the commencement
of their programme, not to have intervertebral
pathology, to be referred by someone other than
a rehabilitation provider, and to commence their
rehabilitation programme within 6 months of their
injury.
Programme outcomes – return to work
There was an overall improvement for the
participants when work status was measured as
working or not working. Fifty-two clients returned
to work, which represents more than a third (37.6%)
of the 138 clients not working at the start of their
programme. Only 10 (3.3%) of the clients who were
working at the start of their programme were not
working at its completion. A logistic regression was
performed to predict those clients who returned to
work, with forward entry of predictor variables.
On step number one of the analysis, unemployed
status entered the regression equation, and led to a
statistically significant improvement in prediction of
return to work compared to the constant-only model,
χ2 (1, n = 138) 13.06; P <0.001. On step number
two, age entered the regression equation, and again
led to a further statistically significant improvement
in the prediction of return to work compared to the
constant-only model, χ2 = 6.60; P < 0.05. No more
predictors were entered into the equation.
A test of the model with both significant predictors
in the equation against a constant-only model
was statistically reliable, χ2 (2, n = 138) = 19.66;
P <0.001, indicating that the predictors, as a set,
reliably distinguished between those who completed
the rehabilitation programme and those who did
not. Prediction success was reasonable, with correct
identification of 80.23% of those who re. turned to
work and 46.15% of those who did not return to
work. The overall success rate of prediction was
67.39%.
In the final model, unemployed status (ß = 1.73;
Wald test = 10.50; P <0.01) and age (ß = –0.05;
Wald test = 6.22; P < 0.05) were both significant
Kenny D.T., 2000
predictors. People who returned to work were more
likely to be employed and to be younger.
DISCUSSION
The principal findings of the study were that the
supervised physical activity programme resulted in
substantial improvements in work status for both
employed and unemployed injured workers with
a range of work-related injuries of up to ten years
duration. Although the programme was demanding in
terms of effort and time, dropout before programme
completion was not a major problem for this sample.
Unemployed status, the presence of intervertebral
pathology, referral by a rehabilitation provider (as
opposed to doctors, employers, or insurers), and
longer time between injury and commencement
of the programme predicted poorer work status
at programme conclusion. The presence of joint
pathology was the only predictor of programme
completion. There were no differences on any of the
outcome measures for those workers with low back
injuries compared to workers with other injuries.
Of the 355 clients who began the programme,
15% did not complete the 12 sessions. This figure is
somewhat lower than those previously reported. This
may be due to the screening procedures undertaken
by the agency during the initial physiotherapy
assessment. Those clients who demonstrated what
the physiotherapist considered to be excessive pain
behaviours, somatization, or lack of motivation
or interest in participating in an exercise-based
rehabilitation programme were either excluded or
given an exercise aptitude assessment to determine
their suitability for the programme. These screening
procedures appear to be effective in maintaining
dropout rates at a relatively Low level. It is also
noteworthy that there was no relationship between
dropout and pre-progranime work status: only five
of the 55 clients who were not working at the start
of the programme dropped out.
Using the variable set available, joint pathology
was the only reliable predictor of programme
completion, with those diagnosed with joint
pathology less likely to complete their programmes.
Given that only 7.6% of the sample had a diagnosis
of joint pathology, there was no reliable predictor of
compieters amongst the other 92.4% of the sample
who did not have such a diagnosis.
Excluding the 53 clients who failed to complete
their programme for whom work status information
was not available, 68% of the 206 coinpleters were
working in some capacity at programme end, 32% of
International Journal of Rehabilitation
Research
whom were working full time/normal duties. If the
53 clients who dropped out are included, and if it be
assumed that dropouts did not have improvements
in work status, then the figure drops to 58%.
However, it is conceivable that people dropped out
because they got better and returned to work before
programme completion, and therefore felt that they
no longer required rehabilitation, so the true post-
programme work status lies somewhere between
these two estimates. Either way, there does not
appear to have been a substantial increase in return
to work rates over the baseline of 54.4% of the
sample who were working at the commencement of
the programme.
However, comparison of change in work status
from the beginning to the end of the rehabilitation
programme paints a more optimistic picture. Even
if it be assumed that all dropouts got worse, the
percentages for the total sample are that work status
deteriorated for 19.4%, remained the same for 31.5%,
and improved for 49%. In this regard, the findings of
this study were similar to those of Petersen (1995),
who used a similar retrospective case record study
design to evaluate a work hardening programme for
100 injured workers with low back pain. He reported
a 24% non-completion rate and a 50% return to work
rate. Niemmeyer et al. (1994). in a comprehensive
review of 36 work hardening programmes. reported
return to work rates that ranged from 50% to 88%
with a 24.6% non- completion rate, and return to
pre-injury duties for 48.2% of clients, with a further
30.5% returning to alternative or modified duties at
programme end. Outcome was not related to client
age, gender, location of injury, or physical demand
level of the job. Duration of injury was a significant
predictor of decreased return to work.
Return to work per se may not be an appropriate
way to assess the efficacy of a rehabilitation
programme. Level of return or change in work status
may be a more sensitive measure of programme
efficacy. The development of a work status index
for this study is a methodological improvement on
other studies reporting return to work outcomes,
providing as it does a much more sensitive outcome
measure that incorporates incremental changes
in work status that take account of work status at
programme entry and exit. Further, many of the
studies reporting on return to work outcomes do not
provide pre-programme work status. This makes it
difficult to judge the degree of improvement that
occurs at the end of the programme. Further studies
should incorporate these more sensitive measures of
pre- and post- programme work status.
Exercise-based rehabilitation for injured workers 7
 In this study, improvement in work status was
predicted by employed status at programme
commencement, the absence of intervertebral
pathology, referral by a source other than a
rehabilitation provider, and less than six months
between injury and commencement of the programme.
There is ample evidence in the literature supporting
early intervention and early referral to rehabilitation
(Strautins and Hall, 1989; Gardner, 1991, Lusted,
1993; Lechner, 1994; Ryan et al., 1995; Ehrmann-
Feldman et al., 1996). The concomitant reduction
in time lost from work as a predictor of successful
rehabilitation is also supported by a number of
studies (Snook and Webster, 1992; Robert et al.,
1995). Employed status at the commencement of a
rehabilitation programme has also been associated
with better outcomes for workers with nonspecific
low back pain (Malmivaara et al., 1995; Burton,
1997). In addition, employed status suggests an
explicit focus on functional work goals in the
rehabilitation programme and this factor has been
associated with successful outcome (Carosella et
al., 1994).
The sample’s demographic characteristics indicates
that the sample under study were a challenging
group. For example, more than 6 months had elapsed
between the date of injury and commencement of
the supervised activity programme for 75% of the
sample; in 40% of cases, the elapsed time was greater
than 12 months. Delays of this magnitude have
been strongly associated with poorer rehabilitation
outcomes (Boschen, 1989; Haig et al., 1990; Ryan et
al., 1995). Secondly, at least two-thirds of the group
had experienced some other form of rehabilitation
that had been unsuccessful, prior to referral to the
exercise therapy programme. Experiences of failure,
in either programme efficacy or in achieving return
to work, often result in a multiplicity of secondary
problems such as demoralization, depression,
anger, hostility, and decreased motivation (Kessler
et al., 1987; Price, 1992; Kenny, 1995b). Forty-six
percent of the group were not working at the time
of commencement; 16% had been unemployed for
more than six months. Both of these factors have
been associated with poorer rehabilitation outcomes
(Johnson and Baldwin, 1993; Kenny, 1995a). Given
the nature of the study group, that is, a group of
compensable, long term work injured clients, a
result indicating improved work status in half of the
group is encouraging.
The results of this study and other recent studies
suggest that the common stereotype of more
frequent treatment failures amongst compensable
as compared to non-compensable clients following
Exercise-based rehabilitation for injured workers 8
such programmes are overly pessimistic. Return
to work rates following such interventions are, in
some circumstances, comparable for compensable
and non-compensable groups (Tollison, 1993;
Ambrosius et al., 1995; Rainville et al., 1997; Mayer
et al., 1998). Another study reported differences in
self- reported pain, depression, and disability in
compensation clients, but comparable outcomes
in terms of physical parameters such as flexibility,
strength, and lifting ability at the conclusion of the
rehabilitation programme (Rainville et al., 1997).
In addition to the quality and appropriateness of
the rehabilitation intervention, a number of factors
outside the control of the treating professional
in a tertiary referral agency may influence return
to work. These factors include availability of
work, flexibility of return to work arrangements,
management support of the worker returning to the
workplace, financial arrangements, and status of
the workers’ compensation claim, among others.
The tertiary referral agency refers the ‘work ready’
worker back to the rehabilitation provider whose
task is to assist the worker to return to the workplace.
Successful completion of an individualized,
supervised exercise programme is not a sufficient
condition for restoration to the workplace, and
assessment of programme efficacy in terms of
return to work only may result in an underestimate
of programme efficacy in terms of other outcomes,
such as functional status, utilization of health care
services, and use of medication, among others.
Since unemployed status at programme entry did
not predict programme completion, but predicted
work status at programme end, the factors militating
against return to work outlined above may be
associated with these outcomes.
It is acknowledged that the exploratory,
retrospective nature of the study is a design weakness,
as is the lack of a suitable control group. This makes
interpretation of the results difficult because there is
no way of knowing how many people with similar
characteristics to the study group would have
spontaneously recovered without the intervention
and returned to work. Given the duration of the injury
and the latency between injury and commencement
of the programme, it is not unreasonable to assume
that the period during which spontaneous recovery
was most likely to occur had passed for a significant
proportion of the group prior to commencement of
the programme.
Another limitation of the present study was
the variables coded on the database that could be
included as predictors. Most of the variables were
Kenny D.T., 2000
structural in nature, and did not index the crucial
subjective mediating factors that are known to
influence return to work. Subsequent prospective
randomized controlled trials would need to
incorporate other potential individual predictors
such as presence of behavioural signs (Werneke et
al., 1993), return to work expectations, perceived
disability, pain, and somatic focus (Carosella et al.,
1994; Petersen, 1995; Rainville et al., 1997), anxiety
and fear avoidance (McCracken et al., 1998) and
motivation and compliance (Friedrich et al., 1998);
workplace predictors such as the nature of the work,
employer support for the injured worker, flexibility
of return to work arrangements, and work-related
problems; and rehabilitation factors such as quality
and suitability of the programme, interpersonal fit
between therapist and client, programme length
and duration of sessions. Inclusion of these factors,
together with the change in work status and the
graded return to work outcome variables developed
and trialled in this study, has the potential for further
clarifying both the efficacy of supervised physical
activity programmes for compensable work
injured clients, and for assisting in the selection
of those clients who are most likely to complete
the programme and improve their work status at
programme end.
REFERENCES
Ambrosius, F.M., Kremer, A.M., Herkner, P.B., DeKraker,
M. and Bartz, S. (1995). Outcome comparison of
workers’ compensation and noncompensation low
back pain in a highly structured functional restoration
programme. Journal of Orthopaedic and Sports
Physical Therapy, 21(1), 7–12.
Bendix, A.F., Bendix, T., Lund, C., Kirkbak, S. and
Ostenfeld, S. (1997). Comparison of three intensive
programmes for chronic low back pain patients: a
prospective, randomized, observer-blinded study
with one year follow-up. Scandinavian Journal of
Rehabilitation Medicine, 29, 81–9.
Boschen, K (1989). Early intervention in vocational
rehabilitation. Rehabilitation Counseling Bulletin, 32,
34–45.
Burke, S.A., Harms-Constas, C.K et al. (1994). Return
to work/work retention outcomes of a functional
restoration programme. Spine, 19, 1880–6.
Burton, AK. (1997). Spine Update: Back injury and work
loss – Biomechanical and psychosocial influences.
Spine, 22(21), 2575–80.
Carosella, A.M., Lackner, J.M. and Feuerstein, M.
(1994). Factors associated with early discharge from
a multidisciplinary work rehabilitation programme for
chronic low back pain. Pain, 57(1), 69–76.
International Journal of Rehabilitation
Research
Ehrmann-Feldnian, D., Rossignol, M., Abenhaim, L and
Gobeille, D. (1996). Physician referral to physical
therapy in a cohort of workers compensated for low
back pain. Physical Therapy, 76(2), 150–6.
Faas, A., van Eijk, J.T.M., Chavannes, AW. and Gubbels,
J.W. (1995). A randomized trial of exercise therapy in
patients with acute low back pain. Spine, 20, 941–7.
Friednch, M., Gittler, G., Halberstadt, Y., Cermak, T. and
Heiller, I. (1998). Combined exercise and motivation
programme: Effect on the compliance and level of
disability of patients with chronic low back pain:
A randomized controlled trial. Archives of Physical
Medicine and Rehabilitation, 79(5), 475–87.
Frost, H., Kiaber Moffett, l.A., Moser, .LS. and
Fairbank, J.C.T. (1995). Randomized control trial for
the evaluation of fitness programme for patients with
chronic low back pain. British Medical Journal, 310,
151–4.
Gardner, l.A. (1991). Early referral and other factors
affecting vocational rehabilitation outcome for
the workers’ compensation client. Rehabilitation
Counseling Bulletin, 34(3), 197–209.
Haig, A., Linton, P., McIntosh, L., Moneta, R. and Mead,
M. (1990). Aggressive early medical management by
a specialist in physical medicine and rehabilitation:
Effect on lost time due to injuries in hospital employees.
Journal of Occupational Medicine, 32, 241–4.
Johnson, W.G. and Baldwin, M. (1993). Returns to work
by Ontario workers with permanent partial disabilities.
Ontario: Research and Evaluation Branch, Strategic
Policy and Analysis Division, Ontario Workers’
Compensation Board.
Kenny, D.T. (1995a). Occupational rehabilitation
assessed: The verdict of employers. Paper presented to
the WorkCover Authority of New South Wales.
Kenny, D.T. (1995b). Occupational rehabilitation
assessed: The verdict of injured workers. Paper
presented to the WorkCovcr Authority of New South
Wales.
Kessler, R.C., Turner, J.B. and House, J.S. (1987)
Intervening processes in the relationship between
unemployment and health. Psychological Medicine,
17, 949–61.
Koes, B.W., Bouter, L.M., Beckennan, H., van der
Heijden, GJ.M.G. and Knipschild, P.O. (1991).
Physiotherapy exercises and back pain: A blinded
review. British Medical Journal, 302, 1572–6.
Lechner, D.E. (1994). Work hardening and work
conditioning interventions: Do they affect disability?
Physical Therapy, 74(5), 47 1–93.
Lindstrom, I., Ohiund, C. et al. (1992). The effect of
graded activity on patients with sub-acute low back
pain: A randomized prospective clinical study with an
operant-conditioning behavioural approach. Physical
Therapy, 72, 279–93.
Lusted, M. (1993). Predicting return to work after
Exercise-based rehabilitation for injured workers 9
 rehabilitation for low back injury. Australian Journal
of Physiotherapy, 39(3), 203–10.
Malmivaara, A., Hakkinen, U., Aro, T., Heinridis, M.L,
Koskenniemi, L, Kuosma, E., I.appi, S., Paloheimo, R.,
Servo, C., Vaaranen, V. and Hernberg, & (1995). The
treatment of acute low back pain: bed rest, exercises, or
ordinary activity? New England Journal of Medicine,
332, 35 1–5.
Manniche, C. (1996). Clinical benefit of intensive
dynamic exercises for low back pain. Scandinavian
Journal of Medicine and Science in Sports, 6,82–7.
Marshall, E. (1997). The Effect of a Supervised Physical
Activity Programme on Persons with Work-related
Low Back Injuries. Final Report: Number 12. Sydney:
Author.
Mayer, T., McMahon, Mi., Gatchel, R.J., Sparks, B.,
Wright, A. and Pegues, P. (1998). Socioeconomic
outcomes of combined spine surgery and functional
restoration in workers’ compensation spinal disorders
with matched controls. Spine, 23(5), 598–605.
McCracken, LM., Goctsch, V.L and Semenchuk, E.M.
(1998). Coping with pain produced by physical activity
in persons with chronic low back pain: Immediate
assessment following a specific pain event. Behavioural
Medicine, 24(1), 29-34.
Niemmeyer, L.O., Jacobs, K., Reynolds-Lynch, K.,
Bettencourt, C. and Lang, S. (1994). Work hardening:
past present, and future – the work programs special
interest section national work-hardening outcome
study. American Journal of Occupational Therapy,
48(4), 327–9.
Oldridge, N.B. and Stoll, i.E. (1997). Low back pain
syndrome. In ACSM’s Exercise Management for
Persons with Chronic Diseases and Disabilities (edited
by the American College of Sports Medicine), pp.
155–60. Champaign, IL: Human Kinetics.
Petersen, M. (1995). Nonphysical factors that affect
work hardening success: A retrospective study. Journal
of Orthopaedic and Sports Physical Therapy, 22(6),
238–46.
Price, R.H. (1992). Psychosocial impact of job loss
on individuals and families. Current Directions on
Psychological Science, 1(1), 9–11.
Rainville, J., Sobel, 3., Hartigan, C. and Wright, A.
(1997). The effect of compensation involvement on the
reporting of pain and disability by patients referred for
rehabilitation of chronic low back pain. Spine, 22(17),
2016–24.
Robert, JJ., Blide, R.W. et al. (1995). The effects of a
Exercise-based rehabilitation for injured workers 10
View publication stats
work hardening program on cardiovascular fitness and
muscular strength. Spine, 20(10), 1187–93.
Ryan, W.E., Krishna, M.K. and Swanson, C.E. (1995).
A prospective study evaluating early rehabilitation
in preventing back pain chronicity in mine workers.
Spine, 20(4), 489–91.
Snook, S.H. and Webster, B.S. (1992). The effectiveness
of a standardised treatment protocol in reducing
disability from low back pain. International Scientific
Conference of Work-Related Musculoskeletal
Disorders, Stockholm, Sweden.
Strautins, P. and Hall, W. (1989). Does early referral to
an on-site rehabilitation programme predict an early
return to work? Journal of Occupational Health and
Safety, 5(2), 137–43.
Symonds, T.L, Burton, A.K, Tillotson, KM. and Main,
C.J. (1995). Do attitudes and beliefs influence work
loss due to low back trouble? Occupational Medicine,
46, 25–32.
Tabachnik, B.G. and Fidel, LS. (1996). Using
Multivariate Statistics (Third Edition). New York, NY:
HarperCollins College Publishers.
Teasell, R.W. and Harth, M. (1996). Functional restoration.
Returning patients with chronic low back pain to work:
Revolution or fad? Spine, 22(3), 358–60.
Tollison, C.D. (1993). Compensation status as a predictor
of outcome in nonsurgically treated low back injury.
Southern Medical Journal, 83(11), 1206–9.
van Tulder, M.W., Koes, B.W. and Bouter, LM. (Eds.)
(1996). Low back pain in primary care: Effectiveness
of diagnostic and therapeutic interventions, pp. 1-285.
Amsterdam, The Netherlands: EMGO Institute.
van Tuider, M.W., Malmivaara. A., E.smail, R. and Koes,
B.W. (1998). The effectiveness of exercise therapy
for low back pain. In Back Review Group for Spinal
Module of the Cochrane Database of Systematic
Reviews [database on disk, CDROM and outline;
updated 2 December, 19971 (edited by C. Bombardier,
A. Nachemson, R. Deyo, R. de Bie, L Bouter, P.
Shekelle, G. Waddell, M. Roland and F. Guillemin).
The Cochrane Collaboration; Issue 1. Oxford: update
Software; Updated quarterly.
Werneke, M.W., Harris, D.E. and Lichter, R.L (1993).
Clinical effectiveness of behavioral signs for screening
chronic low-back pain patients in a work-oriented
physical rehabilitation programme. Spine, 18. 2412–
18.
Kenny D.T., 2000