Spine J. 2001 (Mar); 1 (2): 138–147 ~ FULL TEXT
Marjorie L. Baldwin, PhD, Pierre Côté, DC, MSc, John W. Frank, MD, MSc, William G. Johnson, PhD
Department of Economics,
East Carolina University,
A433 Brewster, Greenville, NC 27834, USA
BACKGROUND CONTEXT: Back pain is the single most costly work-related injury. Chiropractors and physicians are the main primary care providers for occupational low back pain (OLBP), but there is no consensus regarding the relative cost-effectiveness of these two modes of care.
PURPOSE: To critically appraise and synthesize recent literature on the cost-effectiveness of medical and chiropractic care for OLBP, and to propose a cost-effectiveness methodology that integrates epidemiologic and economic methods for future studies.
STUDY DESIGN: Literature review. MEDLINE was searched from 1990 through 1999. Nine articles that met the inclusion criteria were reviewed. The methodological quality of the articles was critically appraised independently by two epidemiologists using standardized review criteria. Two health economists reviewed the studies on cost-effectiveness.
RESULTS: The current literature suggests that chiropractors and physicians provide equally effective care for OLBP but that chiropractic patients are more satisfied with their care. Evidence on the relative costs of medical and chiropractic care is conflicting. Several methodological deficiencies limit the validity of the reviewed studies. No studies combine high-quality cost data with adequate sample sizes and controls for confounding factors.
CONCLUSION: Existing studies fail to clarify whether medical or chiropractic care is more cost effective. We suggest that future studies must combine epidemiologic and economic methods to answer the question adequately.
From the FULL TEXT Article:
Back problems are the single most costly work-related
injury. In 1992, back cases represented 24% of US workers’
compensation claims and 31% of costs . Back problems
are also common outside the workplace, and the distinction
between work-related and non–work-related back pain is
often uncertain. Approximately 11% of the adult population
experience a disabling episode of back pain in any 6-month
period, and nearly 80% of adults, including persons outside
the labor force, experience one or more episodes of back
pain during their lifetime [2,3].
Physicians and chiropractors are the primary caregivers
for individuals with back pain [4–7]. Together they provide
80% of the care delivered per episode of pain. Despite the
fact that physicians and chiropractors account for most of
the costs of treating back pain, and treat the condition with
distinctly different modes of care, there is a lack of convincing evidence regarding the relative cost-effectiveness of the
care they provide .
In the economics literature, cost-effectiveness analysis is a
mechanism for quantifying the values of alternative policies
or treatments to determine if the expected benefits outweigh
the expected costs, or to identify the policy or treatment that
yields the greatest benefits for a particular cost. In cost-effectiveness
analysis, expected benefits are expressed in health
units (e.g., lives saved, hospital days saved, days of work absence
saved) in the numerator of the cost-effectiveness ratio,
while expected costs are expressed in dollars in the denominator.
The resulting ratio indicates the benefit of a particular
policy or treatment, in health units, per dollar expenditure.
A limitation found in the existing literature on medical
and chiropractic care is the tendency to focus on only one
dimension of the research question. Epidemiological studies
focus on the effects of care on patients’ disability status,
functional limitations, and symptoms (the numerator of the
cost-effectiveness ratio), while economic studies focus on
costs (the denominator). In either case, the other dimension
of the issue, either health outcomes or costs, is ignored or
This article has two main objectives. First, we critically review
and synthesize the recent literature on the cost-effectiveness
of medical and chiropractic care for occupational low
back pain. Second, we propose an integrated epidemiologic
and economic methodology that will improve the design of
cost-effectiveness studies of occupational low back pain.
Search of epidemiological literature
A MEDLINE search covering the period January 1990
through April 1999 was conducted using Medical Subject
Headings and text words, and the keywords low back pain,
chiropractic medicine, and treatment. Titles and abstracts of
the retrieved articles were screened to determine if a study
met the following criteria for review: it is a randomized trial
or cohort study; it compares the outcomes of patients receiving
medical versus chiropractic care for low back pain;
it includes measures of pain, functional status, or return to
work; and it is written in English.
Seven of the 148 articles retrieved from MEDLINE met
our inclusion criteria. Three randomized trials compare chiropractic
care with physiotherapy, and four cohort studies
compare chiropractic care with medical care. Studies of
physiotherapy are included in medical management, because
physiotherapy typically follows medical referrals but
is not usually provided in concert with chiropractic care.
The randomized trials are appraised using criteria described
by Van Tulder et al. (Appendix Table A1) . The cohort studies
are appraised using criteria developed at the Institute for
Work and Health (IWH; unpublished; Appendix Table A2).
The guidelines necessary to operationalize the IWH criteria are
presented in Appendix Table A3. Two of the authors (PC and
JWF) independently reviewed all papers. Disagreements between
the reviewers were resolved by consensus.
Search of economics literature
The articles reviewed for economic content were retrieved
from a search of the Journal of Economic Literature,
subject heading I1 (Health), from 1990 onward, and the
MEDLINE search described above. A study must satisfy the
following criteria to be included in the review: it is a comparison
of medical and chiropractic care for low back pain,
it includes information on the costs of care, and it satisfies
accepted standards for controlling for confounding variables
in the economics literature.
Five articles satisfy the inclusion criteria and were reviewed
independently by the two economists (MLB and
WGJ). The economic evaluations assess whether a study satisfies
the following criteria for general methodology: it has
adequate sample sizes, it uses appropriate methods to control
for confounding variables, and it controls for differences in
the case mix and severity of injury treated by physicians and
chiropractors. In addition, the following criteria are used to
evaluate the cost data in each study: the costs of health care
are based on payments, rather than provider charges; the cost
data account for patient-borne costs, including co-payments
and deductibles; and the cost comparisons include disability
costs and the costs of injury-related work absences.
Epidemiologic literature: chiropractic versus physiotherapy
The reviews of three randomized trials comparing chiropractic
care and physiotherapy are summarized in Table 1.
Two high-quality studies suggest that the two modes of care
are equally effective in reducing symptoms and improving
function for patients with low back pain. In a Swedish trial,
individuals with acute and chronic back and neck conditions
were randomized to receive chiropractic care or physiotherapy
[9,10]. The choice of treatment modalities, and the frequency
and duration of treatment, were at the discretion of
each clinician. According to Skargren et al. , no differences
in pain intensity, functional status, or self-reported
general health status were reported at the end of treatment,
or at 6 or 12 months after treatment.
In a 2-year follow-up study comparing chiropractic manipulation,
McKenzie exercises, and an educational booklet,
Cherkin et al.  report no differences between comparison
groups with regard to bothersomeness of symptoms,
disability status, days of work-related disability, or recurrence
of pain. Patients receiving chiropractic manipulation
or McKenzie exercises were, however, more satisfied with
their care than those who received the educational booklet.
In contrast, a British trial suggests that chiropractic care
is more effective in treating back pain than hospital-based
outpatient physical therapy . In this study, chiropractic
patients show greater improvement in functional status in
the short term and long term (2 years) than physiotherapy
patients, but the size of the difference is small (7 points on a
100-point Oswestry scale). The results of the trial must be
interpreted cautiously, however, because of possible selection
and expectation biases associated with the accrual and
randomization strategy. As acknowledged by Meade et al.
, chiropractic treatments were more effective for patients
who selected this type of care before the study assigned
them to a chiropractor than for chiropractic patients
who had initially selected a different type of provider.
Epidemiologic literature: chiropractic versus medical care
The reviews of four cohort studies comparing chiropractic
and medical care are presented in Table 2. Two cohort
studies use workers’ compensation claims data to compare
chiropractic and medical care for OLBP [13,14]. The descriptive
results are similar in substance, but the methods of
analysis and conclusions differ.
Studying claims data from injured Australian workers,
Ebrall  compared between-group differences in mean
outcome measures for 998 chiropractic patients and 998
physician patients (the latter drawn as a random sample
from 3712 physician patients). Ebrall implicitly assumes
that the random sample of physician patients has the same
unmeasured characteristics as the chiropractic patients. The
validity of this assumption is not, however, empirically
tested. The study concludes that chiropractic care is associated
with a larger proportion of medical only claimants who
receive no disability benefit payments (a medical only claim
is defined as a claim in which the worker returns to a job in
less than the 5-day waiting period for wage loss benefits);
fewer total compensation days for back pain; and fewer
claimants developing chronic back pain.
Using workers’ compensation claims data from California,
Johnson, et al.  also report a higher proportion of medical
only claims among chiropractic patients. The authors’ approach
differs from Ebrall’s study in that they analyze differences
within claim types, comparing medical only claims
treated by chiropractors with medical only claims treated by
physicians. They then test the hypothesis that chiropractic
care increases the probability of return to work within the 3-
day waiting period by estimating a multivariate regression
model with controls for age, gender, and other potential confounders.
Lacking direct measures of severity, the authors estimate
a variant of the model with controls for unobserved
heterogeneity, that is, controls for systematic differences between
the patients treated by physicians and chiropractors that
are unobservable to the investigator. The authors assume the
heterogeneity controls include a component that represents
confounding differences in the severity of back pain.
The results with no controls for unobserved heterogeneity
show that chiropractic patients have significantly higher
probabilities of return to work within California’s 3-day
waiting period than physician patients. The statistical significance
of the effect disappears, however, in the models that
include controls for heterogeneity. In brief, the conclusions
of the Johnson et al. study agree with Ebrall that chiropractic
patients return to work sooner than physician patients,
but the Johnson et al. study attributes the difference in durations
of work absence to unobserved differences in the characteristics
of chiropractic and physician patients, rather than
to the effects of chiropractic care. When controls for heterogeneity
are included in the models, the authors find no statistically
significant associations between type of care and
1) the probability of return to work within the 3-day waiting
period; or 2) the incidence of permanent impairments; or 3)
the duration of temporary disability benefits.
These two studies suffer from important methodological
problems, and their conclusions must be interpreted accordingly.
Many of the limitations are inherent in the use of
claims data. The main flaws are difficulty in determining if
the subjects are representative of the underlying population,
because the characteristics of the underlying population are
unknown; imprecise definition of back pain; inadequate
measurement of the subjects’ baseline characteristics; questionable
validity and reliability of the data on type of care
received, because services are identified from codes on the
billing data of health-care providers; inadequate controls for
confounders, such as severity of injury, because claims data
do not include clinical measures of severity and often omit
ICD9 codes; and no description of the longer-term followup,
because claims data do not provide information after a
claim is closed (Table 2).
Using different outcomes, two other studies suggest that
there is no difference in the relative effectiveness of chiropractic
and medical care. In a cohort study conducted on
North Carolina residents, Carey et al.  report that the
time to functional recovery is similar for chiropractic and
general-medical-practice patients, but that chiropractic patients
are more satisfied with their care. The samples of patients
were recruited from medical, orthopedic, and chiropractic
offices. One third of the cohort was composed of
patients with back pain with a workers’ compensation claim.
One important problem with the North Carolina study is
that the nature of the underlying population is unknown.
Approximately 65% of the invited primary care physicians,
and 87% of the invited health management organizations
agreed to participate in the study. Patients of the participating
providers were then invited to join the study, but no information
is provided on the patients or providers who refused
to participate. No analysis of selection bias is included
in the study. Carey et al. controlled for several confounding
factors through multivariate survival analyses, but the potential
for residual confounding because of self-selection or
comorbidities remains an issue.
Finally, in a study of members of a health-care network
in California, Hurwitz found  that the functional status
of chiropractic and medical patients is similar 3 months after
their initial visit. As shown in Table 2, however, several
methodological flaws limit the validity of the results. Of
particular concern is the low power of the statistical measures
and the difficulty in identifying the inception period
for this cohort.
Economics literature: health-care costs
This section reviews five studies that compare medical
and chiropractic care for back pain and estimate differences
in the costs of the two modes of care. Two of the studies
also analyze differences in the costs of work disability. The
studies, with critiques summarized in Table 3, include the
Carey et al., Johnson et al., and Ebrall articles described in
the previous section, and additional studies by Nyiendo
(published in three parts) and Stano [13–15,17–19]. A recent
study by Williams et al.  is not included in our review,
although it includes information on both health-care
costs and durations of work loss for patients treated by a variety
of providers. The study is excluded, because the results
focus on the relationships between costs and differences in
durations rather than on comparisons among providers. A
number of earlier studies (see the articles cited by Nyiendo
and the review by Manga) are also excluded from the discussion
because of inadequate attention to methodological
standards widely applied to economic evaluations of healthcare
Carey’s study of North Carolina back patients, described
in the previous section, concludes that outcomes are similar
for patients of physicians and chiropractors, but primary
care physicians are lower-cost providers than either chiropractors
or orthopedic surgeons . The conclusion regarding
costs is uncertain for at least two reasons. First, the
data used to estimate the costs of health care are based on
provider charges rather than payments. Provider charges are
a biased measure of health-care costs, because payments for
health-care services are typically discounted from charges,
and the discounts vary among payers and providers. The Zenith
Project, for example, shows that there are differences
between the discounts applied to chiropractic charges and
the discounts applied to physician charges . The second
problem with the cost comparisons is that the data on
charges are not the charges submitted by the providers participating
in the study, but average charges for services from
a single insurer in North Carolina. Because of the missing
data on health-care payments, and the likely errors in the
data on charges, Carey’s conclusion that primary care physicians
are the low-cost providers is extremely uncertain.
Stano’s article on the costs of chiropractic care is based
on the largest number of cases studied to date . The results
are based on health insurance claims data for 7077 patients
with 9314 episodes of care over a 2-year period. Responsibility
for the outcomes of care is assigned to either
physicians or chiropractors according to which type of provider
delivered the first service in an episode of care. The
cost data are appropriately measured in terms of payments
rather than charges. The payments do not, however, include
co-payments or deductibles borne by patients. The author
uses categorical variables for co-payments and deductibles
to adjust, as much as possible, for the effects of differences
in patient-borne costs on the cost comparisons between chiropractors
and physicians. Comparisons are made within
ICD9-CM codes to control for potential differences in the
average severity of injuries among patients in the respective
provider groups. The health insurance data do not measure
work loss days or indemnity payments for work absences.
The overall result is that average costs of care are lower
for chiropractic patients than for physician patients. Much
of the average cost differential is attributed to differences in
the costs of hospital inpatient care. The importance of inpatient
costs to the final results is addressed by the author,
who indicates that “some (or even all) of the cost differentials
could be explained by the hospitalization variable”
. In other words, the inclusion of a control for surgical
or nonsurgical inpatient care would virtually eliminate statistically
significant differences in the average costs of care.
Stano suggests that he does not include a variable to control
for hospitalization because, in his opinion, the evidence
suggests that nearly all surgery for back pain is inappropriate.
Nonsurgical methods of care are preferred to unnecessary
surgery, regardless of provider type. The findings do
not evaluate the relative cost effectiveness of chiropractors
and physicians in the provision of nonsurgical care for back
pain. That comparison is important, because surgery is not a
treatment provided by chiropractors.
One remarkable feature of the results is that hospital inpatient
costs represent a relatively large portion of total healthcare
payments for chiropractic patients as well as physician
patients, suggesting that assigning responsibility to providers
based on first contact is not equivalent to receiving services
from only one type of provider. Nearly 25% of adjusted total
payments for chiropractic patients with ICD9846 (sprains
and strains of the sacroiliac region), for example, is for hospital
inpatient care. More than 40% of total payments for
chiropractic patients with ICD9724.3 (sciatica, other, and
unspecified disorders of the back) is for inpatient care.
The practice of assigning total costs to the type of provider
first seen by a patient could introduce a systematic
bias in the costs attributed to chiropractors. Consider cases
who first see a chiropractor and eventually undergo back
surgery. The average cost of surgical cases are, as the previous
results suggest, quite high relative to nonsurgical care
whether provided by chiropractors or physicians. The attribution
of some costs of surgery to chiropractors because of
the first provider assumption is not offset by a symmetrical
assignment of high-cost chiropractor cases to physicians
simply because chiropractors do not perform surgery. The
extent to which the bias exists is, of course, an empirical
question that can be answered only by an evaluation of the data.
Economics literature: disability costs
Only a few articles, namely Ebrall , Johnson et al.
, and Nyiendo [17,18], include information on both
health-care costs and work-related absences for a study population.
The results for all three agree that chiropractic patients
are more likely to return to work within 1 week than
physician patients. Nyiendo describes this finding as typical
of nearly all previous studies of employed populations .
Nyiendo compares health-care costs and durations of
work absence for workers’ compensation patients treated by
chiropractors or physicians. Days of work absence are computed
from beginning and ending dates of work loss periods
recorded on workers’ compensation files. One of the most interesting
features of the research is that the author has access
to patients’ medical records. The principal limitation of the
study is the small number of cases available for analysis. The
results are based on 188 cases in the state of Oregon, with
equal numbers of physician and chiropractic patients but no
controls for differences in severity of back pain between the
two groups. The author carefully details the limits imposed
by the data and states that “the findings of this study emphasize
the need for prospective studies of treatment outcome
using similar patients in two or more provider groups” .
The costs of chiropractic care are found to be higher than
the costs of medical care, primarily because the duration of
treatment is longer for chiropractic patients. The author attributes
the cost differentials to differences in case mix and
severity of injury between patients in the two comparison
groups. Specifically, Nyiendo reports that chiropractic patients
are more likely to have chronic or recurrent problems,
more likely to have risk factors that delay recovery, and more
likely to be female than physician patients. The author does
not estimate the extent to which the cost differentials change
when differences in case mix and severity are controlled.
Nyiendo compares work loss time for chiropractic and
physician patients but does not estimate the costs of work absences.
The results do not measure, therefore, the extent to
which differences in the costs of work loss time offset or increase
the cost differentials for health care. Nor, as the author
acknowledges, do the comparisons of work loss time
control for other factors that influence return to work, such
as the economic incentives associated with disability benefit
payments. Given these limitations, the results show that chiropractic
patients are more likely to return to work within 1
week than patients treated by physicians, but the difference
is not statistically significant. One intriguing result is that
work loss time for chiropractic patients with chronic problems
is shorter than work loss time for patients with acute
conditions . The samples identified as “chronic problems”
are not, however, restricted to persons with chronic
pain, because the definition of “chronic” cases includes any
worker who ever filed a previous job-related low back claim.
Ebrall compares average total costs (the sum of health
care and disability costs) for workers’ compensation back
cases treated by physicians and chiropractors . He reports
that average costs per case are lower for chiropractic patients
than for physician patients, but the conclusion is driven by
the relatively large proportion of chiropractic patients who do
not receive disability benefits because they return to work
within the waiting period. Ebrall’s conclusion, that chiropractors
are the lower-cost provider, is questionable, because
his approach implicitly assumes that chiropractic and physician
patients are identical in the characteristics, other than
health care, that influence durations of absence from work. In
other words, the study assumes that the higher proportion of
medical only claims among chiropractic patients results from
more effective care, when it could result from confounding
factors, such as differences in the severity of back cases
treated by physicians and chiropractors.
One important question raised by the Nyiendo and Ebrall
studies is whether chiropractic patients return to work more
rapidly than physician patients, on average, because their
capacity to function in the workplace is sustained by care
that continues after a return to work. The Johnson et al. 
study, discussed next, reports that average durations of
work absence are shorter for chiropractic patients than for
physician patients, but chiropractic patients are more likely
to continue to receive care after they return to work. If continuing
care permits earlier returns to work, the relevant research
question is whether the additional costs of care are
offset by the cost reductions associated with shorter durations
of work absence.
Johnson et al. consider both the costs of health care and
the costs of work loss days associated with back pain. The
study excludes surgical cases, thereby eliminating some of
the costs that were important in the Stano study. The authors
recognize that some back surgeries may be unnecessary, but
they do not have access to the health-care records needed to
determine if unnecessary surgery was performed.
The authors assume that differences in the severity of
back pain are included in the controls for unobserved heterogeneity
in the models . Estimates that include controls
for heterogeneity show no statistically significant difference
between chiropractic and physician patients in the probability
of return to work within California’s 3-day waiting period.
The authors conclude that chiropractors and physician
are close substitutes in the treatment of back pain with neither
provider offering a clear advantage in terms of cost.
The need for a cost-effectiveness methodology integrating epidemiology and economics
The purpose of our critical review is to synthesize current
knowledge on the cost-effectiveness of medical and
chiropractic care for low back pain and to identify conceptual
and methodological deficiencies in the existing studies.
The studies suggest that chiropractic and medical care (including
physiotherapy) are equally effective in promoting
recovery from OLBP but that patients are more satisfied
with chiropractic care. Several methodological deficiencies,
however, limit the validity of the reviewed studies.
The current literature on the effectiveness of medical and
chiropractic care provides conflicting evidence that one
type of care is superior to the other. As suggested by two
high-quality randomized clinical trials conducted from general
population samples, it is likely that both types of care
are equally effective, but the generalizability of the results
to workers’ compensation populations remains unclear
[9,10]. The research consistently shows that chiropractic patients
have shorter durations of work absence, on average,
than physician patients [13,14,17]. There is no convincing
evidence, however, that identifies whether the difference in
durations is related to differences in the types of health care
received or to intrinsic differences in the types of patients
treated by chiropractors and physicians.
The methodological quality of the literature reporting on
costs does not allow us to reach valid conclusions about the relative
costs of medical and chiropractic care for OLBP. As yet,
there are no studies that combine high-quality cost data with
adequate sample sizes and controls for confounding factors.
Our critical appraisal has identified several sources of
bias that jeopardize the internal and external validity of the
observational studies. No study to date has used rigorous
methodology to adequately study the cost-effectiveness of
medical and chiropractic care. The next logical step in attempting
to answer the cost-effectiveness question, therefore,
is to study health outcomes and costs in a workers’
compensation population, with a study that combines epidemiologic
and economic methodologies. While this appears
simple in theory, it implies that two distinct research paradigms
be merged and adapted to address a specific research
question. Specifically, it entails the recruitment of a large
sample of injured workers that will be followed for at least 1
year and for whom valid and reliable health outcomes information
can be combined with accurate and complete cost
data obtained from insurance companies and employers.
We propose that a cohort study is the design of choice to
investigate the relative cost-effectiveness of medical and chiropractic
care for work-related back pain. Although a randomized
controlled trial is more appropriate to study the efficacy
of treatments, a cohort study is best suited to investigate the
effectiveness of already well-accepted and commonly used
methods of care provided to injured workers. Further, there
are two important drawbacks to implementing a randomized
trial in this context. First, the ability to randomize injured
workers to treatment groups is limited by state workers’ compensation
laws. In the US, the choice of provider for the treatment
of an occupational injury may reside with the worker,
the employer, or both, depending on the state. Second, it has
been demonstrated that randomizing patients with low back
pain to types of care that are contrary to their preferences may
result in significant selection and expectation biases .
Therefore, using an inception cohort design with extensive
measurement and adjustment for confounding variables may
offer the best strategy to study cost-effectiveness.
Defining an inception cohort to study occupational low
back pain is challenging. The difficulty lies in ensuring that
all subjects are at the same stage of their condition, in this
case, early in their current episode. Therefore, defining the
appropriate “time zero” is crucial. Given the natural history
of low back pain, it is likely that a high proportion of injured
workers have chronic or recurrent back pain and that they
may be at different stages in the evolution of their condition
when claims are filed. Because this may result in selection
bias if subjects with chronic or recurrent back pain systematically
select physicians or chiropractors, it is essential to
ascertain any previous history of back problems and to control
for this variable in multivariate regression analyses.
The selection of valid and meaningful outcomes measures
is key to providing useful answers when studying low back
pain. Deyo et al.  propose that standardized outcomes
measures be used in this field of research. The proposed outcomes,
which become the numerator of the cost-effectiveness
ratio, include symptoms, functional status, general well-being,
work disability, and satisfaction with care. The authors
emphasize that the instruments selected to measure health
outcomes must meet fundamental methodological criteria.
First, an instrument should discriminate among individuals
with different levels of severity of low back pain. Second, an
instrument should be reliable and yield small variations between
replicate measures. Third, an instrument should provide
reproducible results if no real changes occur between
two or more data collections. Fourth, when improvement or
deterioration of the condition occurs, an instrument should be
sufficiently responsive to detect small, but possibly clinically
important, changes. Finally, an instrument must measure
what it is supposed to measure, that is, it must be valid.
The cost measures, which become the denominator of
the cost-effectiveness ratio, should be accurate and complete
measures of the total costs of an episode of workrelated
back pain from the viewpoint of a particular payer.
From an employer’s viewpoint, for example, the costs of
work-related injuries include the costs of health care and the
costs of indemnity payments to injured workers, taking account
of different insurance coverages and different mechanisms
for experience rating insurance premiums. From a
worker’s perspective, the costs of an episode of workrelated
back pain include net wage losses during injuryrelated
work absences and losses of income that result from
lower productivity after a return to work. The costs of health
care are not relevant in a cost-effectiveness study from the
worker’s viewpoint, because the costs are fully covered by
workers’ compensation insurance.
Regarding the analysis, the key issue in an observational
study of different treatment modalities is whether the statistical
adjustments for confounding variables adequately control
for differences in subjects’ characteristics that affect the
costs and outcomes measures. To implement the statistical
controls, one must measure in all subjects, before treatment,
the various factors that are known to influence the prognosis
of OLBP. The measures are then included as covariates in
multivariate regression models to adjust the observed differences
in treatment effects, if any, for differences in the characteristics
of the self-selected treatment groups that may affect
outcomes. Chiropractors, for example, do not prescribe
strong analgesics, so it might be expected that severity of
back pain and attitudes toward pain will differ systematically
between self-selected physician and chiropractor patients.
Some of these pain variables are known to influence
the outcomes of care, hence unadjusted comparisons of outcomes
between the two treatment groups may be inaccurate,
because the outcomes are confounded by the differences in
pain-related aspects of OLBP.
The data must be analyzed with multivariate models,
making the self-selected treatment groups as comparable as
possible to mimic a randomized controlled trial that assesses
treatment effects in an unbiased fashion. It can still
be argued that some unknown confounder was unmeasured
and, therefore, uncontrolled in the study design, but the
credibility of this argument is greatly weakened when observed
differences in outcomes fail to disappear as additional
confounders are added to the multivariate model.
Once the differences in costs and outcomes of medical
and chiropractic care have been accurately measured, with
controls for confounding variables, the cost-effectiveness
ratios can be computed. The cost effectiveness ratios compare
the difference in health outcomes between physician
and chiropractic patients in the numerator, with the difference
in costs in the denominator. A large cost-effectiveness
ratio indicates that medical care is relatively more costeffective
than chiropractic care, and a small cost-effectiveness
ratio indicates the opposite conclusion.
In sum, we have identified the following minimal criteria
for a cohort study of the cost-effectiveness of medical and
chiropractic care for OLBP: the sample is identified immediately
after onset of back pain; the study obtains data on
prior history of back pain; standardized outcomes measures
are collected; the total costs of an episode of back pain are
measured accurately; costs are evaluated from the viewpoint
of a pre-identified payer; and multivariate models are
used to control for differences in patients’ characteristics.
There is clearly a need for research on the different treatments
for occupational low back pain that considers both
health care and indemnity costs and controls for differences in
severity of back pain at onset. It is virtually impossible, however,
to conduct a truly experimental, randomized controlled
trial of “largely chiropractic” versus “largely physician” care
of back patients without either causing some unwanted effects
by changing patients’ care preferences, or being forced
to study only the minority of OLBP patients who apparently
have no preference regarding type of provider. The great importance
of the OLBP patient’s faith in his or her chosen caregiver,
and the inextricable “placebo” effects of each form of
care, make it difficult to conduct rigorous yet interpretable
randomized controlled trials in this field, where no subject
can be truly blinded to the type of care he or she receives. The
provisions of state workers’ compensation laws also assign
the right of choice of a medical provider to either workers or
employers, and randomized assignments of patients would violate
these rights. Thus, the most practical and generalizable
evaluation design available is a prospective cohort study.
The failure of existing studies to clarify, definitively,
whether medical or chiropractic care is more cost-effective
for occupational low back pain arises, in part, because of
methodological challenges. These challenges can be addressed
only by combining the evaluation approaches of epidemiology
(for health outcomes) and economics (for cost
outcomes), and by capturing the best available measures of
severity of back pain at onset, to the degree possible.
National Safety Council.
Itasca (IL): National Safety Council, 1993.
Cassidy JD, Carroll LJ, Côté P.
The Saskatchewan health and back pain survey: the prevalence of low back pain
and related disability in Saskatchewan adults.
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.
Côté P, Cassidy JD, Carroll L.
The treatment of spinal pain: Who seeks care? Who goes where?
Institute for Work and Health Working Paper 117,
Toronto, Ontario, Canada 2000.
Deyo RA, Tsui-Wu YJ.
Descriptive epidemiology of low back pain and its related medical care in the United States.
Hurwitz E, Morgenstern A.
The effects of co-morbidity and other factors on medical versus chiropractic care for low back pain.
Shekelle P, Markovich PM, Louie R.
An epidemiologic study of episodes of back pain care.
Van Tulder MW, Assendelft WJJ, Koes BW, Bouter LM.
Method guidelines for systematic reviews in the Cochrane collaboration back review group for spinal disorders.
Skargren EI, Oberg BE, Carlsson PG, Gade M.
Cost and effectiveness analysis of chiropractic and physiotherapy treatment for low back and neck pain:
Skargren EI, Carlsson PG, Oberg BE.
One-year follow-up comparison of the cost and effectiveness of chiropractic and physiotherapy as
primary management for back pain: Subgroup analysis, recurrence,and additional health care utilization.
Cherkin DC, Deyo RA, Battle M, Street J, Barlow W.
A Comparison of Physical Therapy, Chiropractic Manipulation, and Provision of an Educational Booklet
for the Treatment of Patients with Low Back Pain
New England Journal of Medicine 1998 (Oct 8); 339 (15): 1021-1029
Meade TW, Dyer S, Browne W, et al.
Low Back Pain of Mechanical Origin: Randomised Comparison of Chiropractic and Hospital Outpatient Treatment
British Medical Journal 1990 (Jun 2); 300 (6737): 1431–1437
Mechanical Low-Back Pain: A Comparison of Medical and Chiropractic Management
Within the Victorian WorkCare Scheme
Chiropractic Journal of Australia 1992 (Jun); 22 (2): 47–53
Johnson WG, Baldwin ML, Butler RJ.
The costs and outcomes of chiropractic and physician care for back pain.
Journal of Risk and Insurance 1999;66:185–206.
Carey TS, Garrett J, Jackman A, et al.
The Outcomes and Costs of Care for Acute Low Back Pain Among Patients Seen by Primary Care
Practitioners, Chiropractors, and Orthopedic Surgeons
New England J Medicine 1995 (Oct 5); 333 (14): 913–917
The relative impact of chiropractic vs. medical management of low back pain
on health status in a multispecialty group practice.
J Manipulative Physiol Ther 1994;17:74–82.
Disabling Low Back Oregon Workers' Compensation Claims Part II: Time Loss
Journal of Manipulative and Physiological Therapeutics 1991;14:231–239
Disabling Low Back Oregon Workers' Compensation Claims. Part III:
Diagnostic and Treatment Procedures and Associated Costs
Journal of Manipulative and Physiological Therapeutics 1991 (Jun); 14 (5): 287-297
The Economic role of chiropractic: Further analysis of relative insurance costs for low back care.
Journal of the Neuromusculoskeletal System 1995;3:139–144.
Williams DA, Feuerstein M, Durbin D, Pezzullo J.
Health care and indemnity costs across the natural history of disability in occupational low back pain.
Manga P, Angus D, Papadopoulos C, Swan W.
The Effectiveness and Cost-Effectiveness of Chiropractic Management of Low-Back Pain
Ottawa: Kenilworth Publishing; 1993.
Johnson WG, Baldwin ML, Burton Jr, JF.
Why is the treatment of work-related injuries so costly? New evidence from California.
Cherkin DC, Deyo RA, Street JH, Barlow W.
Predicting poor outcomes for back pain seen in primary care using patients’ own criteria.
Deyo RA, Battie M, Beurskens AJ, Bombardier C, Croft P, Koes B, Waddell G.
Outcome measures for low back pain research: A proposal for standardized use.
Return to the COST-EFFECTIVENESS Section