THE ASSOCIATION BETWEEN USE OF

CHIROPRACTIC CARE AND COSTS OF CARE AMONG

OLDER MEDICARE PATIENTS WITH CHRONIC LOW
BACK PAIN AND MULTIPLE COMORBIDITIES
William B Weeks, MD, PhD, MBA, a b Brent Leininger, DC, c James M Whedon, DC, MS, d e
Jon D Lurie, MD, MS, f Tor D Tosteson, ScD, g Rand Swenson, DC, MD, PhD, a
Alistair J OMalley, PhD, h and Christine M Goertz, PhD, DC i

ABSTRACT

Objective: The purpose of this study was to determine whether use of chiropractic manipulative treatment (CMT)
was associated with lower healthcare costs among multiply-comorbid Medicare beneficiaries with an episode of
chronic low back pain (cLBP).
Methods: We conducted an observational, retrospective study of 2006 to 2012 Medicare fee-for-service
reimbursements for 72 326 multiply-comorbid patients aged 66 and older with cLBP episodes and 1 of 4 treatment
exposures: chiropractic manipulative treatment (CMT) alone, CMT followed or preceded by conventional medical
care, or conventional medical care alone. We used propensity score weighting to address selection bias.
Results: After propensity score weighting, total and per-episode day Part A, Part B, and Part D Medicare
reimbursements during the cLBP treatment episode were lowest for patients who used CMT alone; these patients had
higher rates of healthcare use for low back pain but lower rates of back surgery in the year following the treatment
episode. Expenditures were greatest for patients receiving medical care alone; order was irrelevant when both CMT
and medical treatment were provided. Patients who used only CMT had the lowest annual growth rates in almost all
Medicare expenditure categories. While patients who used only CMT had the lowest Part A and Part B expenditures
per episode day, we found no indication of lower psychiatric or pain medication expenditures associated with CMT.
Conclusions: This study found that older multiply-comorbid patients who used only CMT during their cLBP
episodes had lower overall costs of care, shorter episodes, and lower cost of care per episode day than patients in the
other treatment groups. Further, costs of care for the episode and per episode day were lower for patients who used a
combination of CMT and conventional medical care than for patients who did not use any CMT. These findings
support initial CMT use in the treatment of, and possibly broader chiropractic management of, older multiply-
comorbid cLBP patients. (J Manipulative Physiol Ther 2016;xx:1-13)
Key Indexing Terms: Manipulation; Chiropractic; Medicare; Retrospective Studies; Propensity Score

a g
Professor, The Geisel School of Medicine at Dartmouth, The Professor, The Geisel School of Medicine at Dartmouth,
Dartmouth Institute for Health Policy and Clinical Practice, Department of Biomedical Data Science, Lebanon, NH.
h
Lebanon, NH. Professor, The Geisel School of Medicine at Dartmouth,
b
Director, Health Services and Clinical Research, Palmer College Department of Biomedical Data Science, The Dartmouth Institute
of Chiropractic, Palmer Center for Chiropractic Research, Davenport, IA. for Health Policy and Clinical Practice, Lebanon, NH.
c i
Assistant Professor, Integrative Health and Wellbeing Research Vice Chancellor, Palmer College of Chiropractic, Palmer
Program, Center for Spirituality and Healing, University of Center for Chiropractic Research, Davenport, IA.
Minnesota, Minneapolis, MN. Submit requests for reprints to: William B Weeks, MD, PhD,
d
Instructor, The Geisel School of Medicine at Dartmouth, The MBA, TDI, 35 Centerra Parkway, Lebanon, NH, 03766.
Dartmouth Institute for Health Policy and Clinical Practice, (e-mail: [email protected]).
Lebanon, NH. Paper submitted September 17, 2015; in revised form
e
Director, Health Services Research, Southern California September 24, 2015; accepted October 13, 2015.
University of Health Sciences, Whittier, CA. 0161-4754
f
Associate Professor, The Geisel School of Medicine at Copyright 2016 by National University of Health Sciences.
Dartmouth, The Dartmouth Institute for Health Policy and Clinical All rights reserved.
Practice, Lebanon, NH. http://dx.doi.org/10.1016/j.jmpt.2016.01.006
2 Weeks et al Journal of Manipulative and Physiological Therapeutics
Medicare Costs of Care Month 2016

ontrolling the growth of healthcare costs continues to recommended for back pain in older adults by a variety of

C be a critical health policy issue. An aging population

and advances in the ability to extend life and manage
chronic disease have conspired to produce approximately
advisory bodies. 31,32 While CMT has been shown to result in
slightly better pain and function outcomes compared to other
active treatments for chronic LBP, a number of researchers have
75 million people in the US who have multiple concurrent questioned the clinical importance of these findings. From a
chronic conditions 1,2: 62% of Americans over age 65 have healthcare system or societal perspective, small differences in
multiple chronic conditions, 3 and 23% of Medicare benefi- clinical outcomes may be important if associated with minimal
ciaries have 5 or more chronic conditions. 4 The care of additional costs, or cost savings. Studies examining differences
individuals with chronic conditions is estimated to account for in healthcare expenditures between CMT users and non-users
78% of US healthcare spending, and Medicare beneficiaries show that CMT users are younger, wealthier, and healthier than
with more than 1 chronic condition account for 95% while non-users. 3335 After propensity score matching to adjust for
those with more than 5 chronic conditions account for 66% such differences, 1 study of the Medical Expenditure Panel
of all Medicare spending. 3 The likelihood that patients will Survey found that chiropractic care was associated with a lower
use expensive health care resources such as hospital care use of medical resources, overall. 36 However, to date,
increases substantially when comorbidities are present, 5,6 propensity score methods have not been applied to claims
and resource consumption increases dramatically if patients data for the purposes of evaluating costs of care for
are also depressed. 7 The Strategic Framework on Multiple multiply-comorbid patients seeking CMT for treatment of LBP.
Chronic Conditions has called for development of new models Therefore, to explore whether older Medicare fee-for-
of care for multiply comorbid Medicare beneficiaries. 8 service beneficiaries with an episode of LBP and multiple
In addition, it has become increasingly evident that comorbidities who obtained CMT during their episode had
chronic pain is associated with high rates of diagnosable lower costs than those who did not, we used Medicare files
psychopathology, 9 and that unrecognized and untreated and a propensity score weighting methodology to adjust for
psychopathology can interfere with rehabilitation. 10 Because confounders and create equivalent groups for comparison. 3740
anxiety can decrease pain thresholds and tolerance, 11 Further, we sought to determine whether, for particular
emotional distress can magnify medical symptoms, 12 and diagnostic mixes within these treatment groups, reduced
depression can worsen chronic pain treatment outcomes, 13 expenditures on psychiatric care or pain medications might be
psychiatric comorbidities may be implicated in perpetuating associated with CMT.
pain-related dysfunction. 14
Among older US adults, back pain is common and
associated with co-morbidities and self-reported difficulty
with most functional tasks. 15 Medicare data from the 1990s METHODS
indicated that low back pain (LBP) diagnoses and related Overview
expenditures increased disproportionately, 16 the use of lumbar We conducted an observational, retrospective study that
and facet injections for LBP increased dramatically, 17 and there used Medicare fee-for-service files from 2006 to 2012 to
was intensive use of pharmaceutical agents among LBP identify older (aged 66 to 99) patients who had a discrete
patients. 18 The escalating prevalence of LBP among Medicare episode of cLBP (lasting at least 90 days 41 and preceded and
beneficiaries, the increasing costs of its treatment, and the high followed by 180 days in which no LBP diagnosis was
use and costs of pharmaceuticals suggest a critical need to recorded) and during which each patient also had an
identify appropriate, cost-effective, and conservative treatments additional musculoskeletal disorder diagnosis and a mental
for older patients with LBP. Further, the tenacity and cost of health disorder diagnosis (details of inclusion and exclusion
chronic pain disorders, and the very high rates of comorbid criteria are provided in the online Appendix). Although we
depression, 19,20 undiagnosed mood disorders (reaching levels had access to 7 years of data, because of a 1-year look-back
of 75% among those with cLBP), 21,22 and anxiety and 1-year look forward period, our analyses included
disorders 19,20,23 with chronic pain disorders suggest that episodes of cLBP that occurred during 5 years, between 2007
exploring ways to disrupt the vicious cycle of pain, stress, and 2011. If an individual patient had multiple cLBP episodes
and emotional dysfunction are warranted. Since simultaneous during our study period, we included only the first.
pharmacological treatment of pain symptoms and major Informed by Liliedahl 42 and Weigel, 43 we identified
depression has led to improved function and quality of life, 24 patients who used 1 of 4 treatments during their episode:
and longitudinal analyses have shown that changes in pain and 1) only CMT; 2) conventional medical care followed by CMT;
depression symptoms influence 1 another, 25,26 it makes sense 3) CMT followed by conventional medical care; and 4) only
that concurrent treatment of both conditions is recommended. 10 conventional medical care.
Most LBP in older adults can be managed non-surgically, 16 For these 4 treatment groups, we examined all un-weighted
and randomized controlled clinical trials have demonstrated that and propensity-score weighted inflation- and price-adjusted
chiropractic manipulative treatment (CMT) is an effective, Medicare Part A, B, and D reimbursements during the episode
conservative treatment option for LBP 2730 that has been (Medicare Part A covers hospital, skilled nursing facility,
Journal of Manipulative and Physiological Therapeutics Weeks et al 3
Volume xx, Number Medicare Costs of Care

Table 1. Demographics and Un-weighted Spending Results That Show the Impact of Serial Application of Exclusion and Inclusion Criteria for Chronic Back Pain Episodes in Older
home health and hospice care expenditures; Medicare Part B

On pain
Total meds
covers doctors services and other outpatient expenditures; and

583
320
273

4574 699
347
Medicare Part D covers prescription medications). We paid

Part D

4104
3235
2614

2956
particular attention to Part B reimbursements for CMT care,
psychiatric care, physical therapy care, and spinal manipula-

On Psychiatric
Mean Medicare Expenditures During the Episode ($)
tion therapy (SMT) provided by doctors of osteopathy (DOs)
as well as to Part D reimbursements for pain medications.

43.12
16.00
12.02

47.54
15.07
Because the length of an average episode of care varied

Total On CMT Care

according to group, we also calculated each of these costs per
episode day for each group.

53.80
79.64
82.26

66.28
98.43
We also examined 2 longer-term patient outcomes:
additional healthcare claims for LBP and rate of spine
surgery within 1 year of the end of the episode. Also, from

Part B

3667
3967
2553

6162 4318
3080
annual cohorts of patients who met our inclusion criteria,
who had episodes of cLBP that lasted 2 years or less, and

5333
4245
2711

3321
Total A, Part
whose episodes began in 2007 to 2010, we calculated annual

B, and D A
compound rates of growth for price and inflation-adjusted
Medicare expenditures. Finally, for particular diagnostic

14 553
9980
15 469

17 779
11 690
groupings, we specifically considered expenditures for CMT
visits and psychiatric visits that were reimbursed through

in an Episode
Medicare Part B as well as expenditures for pain medications

Enrolled Mean Days

reimbursed through Medicare Part D (for this analysis, we
limited our analysis to patients with cLBP whose additional

350
415
409

444
373
diagnostic combinations (ie, depression plus osteoarthritis)
represented at least 2% of the entire treatment group).
White Used a DC Part D

49.8
55.7
48.3

56.3
50.7
Statistics (%)
Age Male Race for CMT

Because we anticipated that the different groups studied

22.6
14.6
21.1

17.6
26.9
(%)

would have different patient characteristics that might

confound results, 36,44,45 to adjust for those confounders
86.0
88.6
88.6

88.1
and create equivalent groups for comparison, 3739 we used 85.9
(%) (%)

multinomial stepwise logistic regression 46 to calculate each

33.4
21.0
38.0

18.6
30.3

subjects propensity to be in a particular treatment group

(with criteria to enter the stepwise regression set at P b .05
Exclusion Mean

75.4
76.5
75.4

76.4
75.5

and criteria to exit set at P N .10). 40 The variables used for

(y)

CMT, chiropractic manipulative treatment; DC, doctor of chiropractic.

propensity score calculation are described in online

% After

Applied

Appendix 1. To avoid complications due to time-varying

Criteria

8.4 19.1
327 914 14.5 32.8

7.2
Medicare Part A and Part B Enrollees in 2007 to 2012

predictors, the propensity score calculation was performed

After application of exclusion criteria 1 000 095 44.1 100

using predictors measured at baseline.

3.2
% of

We then used all cases in linear or logistic regressions to

2 267 643 100
All

compare means or proportions, respectively, of outcome

Those with a mental health diagnosis 190 566

72 326

variables associated with being in 1 treatment group after

inversely weighting each case by the patients propensity to
N

be in that treatment group. In essence, this method

relatively over-weighted patients who were least likely to
All episodes of chronic back pain

be in their respective treatment group given their

musculoskeletal condition

demographics, rendering each individual treatment group

Those with an additional

cohort more like the overall distribution of patients and thus

more similar, demographically. 40 For the analysis that
Those with both

explored whether particular diagnostic combinations might

have different treatment patterns or expenditures, we used
the same methods to calculate a different propensity score
that did not incorporate the categories of comorbidities that
defined our study population.
4 Weeks et al Journal of Manipulative and Physiological Therapeutics
Medicare Costs of Care Month 2016

Table 2. Mean Values and Standard Errors of Key Demographics, Measures of Poverty and Income, the Per-capita Supply of Doctors of
Chiropractic, and Measures of Illness Burden Before and After Propensity Score Weighting for the 4 Treatment Groups
Key Demographics
Less than Enrolled in
Mean White high school Medicare
N age (yrs) Male (%) race (%) education*(%) Part D
Un-weighted Only used CMT 3909 73.7 (0.12) 24.6 (0.69) 96.1 (0.31) 5.61 (0.10) 51.6 (0.80)
Used conventional medical care first, then CMT 3563 74.3 (0.13) 23.4 (0.71) 93.8 (0.40) 5.99 (0.10) 51.4 (0.84)
Used CMT first, then conventional medical care 5235 74.6 (0.11) 24.4 (0.59) 94.0 (0.33) 5.99 (0.09) 50.0 (0.69)
Did not use CMT 59 619 76.9 (0.03) 17.4 (0.16) 86.7 (0.14) 7.08 (0.03) 57.4 (0.20)
Propensity score Only used CMT 3909 76.4 (0.06) 18.2 (0.14) 88.0 (0.12) 6.91 (0.05) 57.8 (0.18)
weighted Used conventional medical care first, then CMT 3563 76.1 (0.06) 18.7 (0.15) 88.4 (0.12) 6.88 (0.05) 56.5 (0.19)
Used CMT first, then conventional medical care 5235 76.4 (0.06) 18.3 (0.14) 87.3 (0.12) 6.92 (0.05) 55.6 (0.19)
Did not use CMT 59 619 76.4 (0.06) 18.5 (0.14) 88.1 (0.12) 6.87 (0.05) 55.9 (0.19)

CMT, chiropractic manipulations therapy; DC, doctor of chiropractic; HRR, hospital referral region.

Prior to performing the study, we correctly anticipated Comparison of Demographics of Our Study Cohorts
that the smallest group of patients in our study would be Most patients that we studied did not use chiropractic
those who were enrolled in Medicare Part D with pain services (Table 2, top). As anticipated, before propensity
medication costs. We assumed that the medication score weighting, patients who used any chiropractic care
categories we studied would account for 25%, or about were younger, more likely to be white, more educated, less
$500, of all pharmaceutical costs, with a standard deviation likely to have indicators of poverty, had less illness burden,
of $200. We determined that, with a sample size of 17 000 had generated fewer Medicare expenditures in the year
patients, we would be able to discern a $25 difference in preceding their cLBP episode, and lived in regions with a
Medicare Part D costs across the groups at the 0.001 level higher per-capita supply of DCs than those who did not.
(two-sided) with a power of 0.99. Our pharmaceutical cost Somewhat surprisingly, in this multiply-comorbid cohort,
estimates were accurate and our sample size for this analysis we found that males were more likely to use chiropractic
was the 56.3% of the 72 326 patients who were enrolled in care and that patients who used CMT lived in lower income
Medicare Part D, or 40 720 patients, ZIP Codes. A higher proportion of patients who did not use
CMT were enrolled in Medicare Part D. After propensity
score weighting, differences in these patient characteristics
Ethics Considerations
were mitigated across treatment groups (Table 2, bottom).
The Dartmouth College Committee for the Protection of
Patients using only CMT had the shortest back pain
Human Subjects approved the study (CPHS #24 094).
episodes and patients who obtained CMT followed by
medical care had the longest back pain episodes (Table 3,
top). Before propensity score weighting, patients who used
RESULTS only CMT had the lowest Medicare expenditures for every
Comparison of Patients in Our Study to the Medicare Fee-for-Service category of expenditures examined with the exception of
Population With Chronic Low Back Pain expenditures for CMT. Application of inverse propensity
Table 1 shows the impact of application of our exclusion score weighting resulted in similar, albeit less dramatic,
and inclusion criteria on demographics, patient characteristics, differences between treatment groups. Nonetheless, even
and un-weighted spending for cLBP episodes in older after propensity score weighting, patients who obtained
Medicare beneficiaries. The multiply-comorbid population only CMT had Medicare Part A expenditures that were
that we studied was generally older, less likely to be male, less about 80% lower, and Part B and D expenditures that were
likely to use CMT, more likely to be enrolled in Medicare Part about 50% lower, than those of patients who did not use any
D, endured longer cLBP episodes, and experienced higher CMT during their cLBP episode. When comparing patients
Medicare costs in all expenditure categories (except for CMT who used only CMT to those who used no CMT, higher
expenditures) than the overall Medicare fee-for-service expenditures for CMT were offset by lower expenditures
beneficiary population that had a cLBP episode. Our study for psychiatric, physical therapist (PT), or DO services and
population represented only 7.2% of the broader Medicare with substantially lower pharmaceutical (and particularly
population that had cLBP episodes and to which the same pain medication) costs. While back surgery rates were
criteria for participation in the study were applied. Most of the substantially lower among patients who used only CMT as
differences between our study cohort and the broader compared to the other groups, the proportion of patients
Medicare population appeared to be attributable to the who had additional claims for LBP after their episode ended
presence of a mental health diagnosis. were substantially higher for this treatment group.
Journal of Manipulative and Physiological Therapeutics Weeks et al 5
Volume xx, Number Medicare Costs of Care

Table 2. Mean values and standard errors of key demographics, measures of poverty and income, the per-capita supply of doctors of
chiropractic, and measures of illness burden before and after propensity score weighting for the 4 treatment groups
Measures of Poverty and
Income Indicators of Illness Burden
The per-10000
Part D poverty Medicaid Charlson Iezzoni Disability was original reason Total Medicare spending in capita supply of
flag (%) enrolled (%) score score for Medicare eligibility (%) year prior to episode ($) DCs in the HRR
11.8 (0.52) 8.4 (0.44) 1.30 (0.03) 0.90 (0.02) 7.8 (0.43) 4718 (168) 2.68 (0.016)
13.2 (0.57) 10.2 (0.51) 1.51 (0.03) 1.03 (0.02) 9.7 (0.50) 5420 (188) 2.49 (0.016)
14.1 (0.48) 10.6 (0.43) 1.35 (0.03) 0.95 (0.02) 8.6 (0.39) 4425 (140) 2.58 (0.013)
29.9 (0.19) 24.0 (0.17) 2.13 (0.01) 1.38 (0.01) 12.9 (0.14) 6586 (51) 2.24 (0.004)
27.0 (0.16) 21.5 (0.15) 1.93 (0.02) 1.30 (0.01) 12.6 (0.12) 6277 (91) 2.28 (0.007)
25.5 (0.16) 20.5 (0.15) 1.95 (0.02) 1.30 (0.01) 12.3 (0.12) 6089 (91) 2.29 (0.007)
25.7 (0.16) 20.7 (0.15) 1.99 (0.02) 1.29 (0.01) 11.6 (0.12) 6326 (92) 2.31 (0.007)
26.9 (0.17) 21.4 (0.15) 1.99 (0.02) 1.31 (0.01) 12.1 (0.12) 6280 (92) 2.30 (0.007)

When examining treatment groups that used both CMT well as price- and inflation-adjusted Medicare Part A, B,
and conventional medical care, patients who used CMT first and D reimbursements during a cLBP episode that lasted
had higher Part A, Part B, and Part D expenditures but less than 2 years and began in the year indicated, across the
lower rates of back surgery and lower rates of use of 4 treatment groups. Over time, episode lengths trended
additional care for back pain in the year following their upward for all treatment groups while measures of
cLBP episodes than patients who use conventional medical reimbursement fluctuated year to year. When examining
care first. total episode costs, patients who used only CMT had lower
When examining expenditures per episode day (Table 3, compound annual growth rates than any other treatment
bottom), after propensity score weighting, Part A expenditures group for all variables except overall Part B spending and
were similar for patients who used both CMT and spending on CMT. Part A, B, and D expenditure growth
conventional medical care, regardless of the order in which was highest for patients who used CMT first, and then used
that care was obtained. However, Part B and D expenditures conventional medical care, but episode lengths increased
were higher for patients who used conventional medical care most for this group as well. When examining reimburse-
first. Daily expenditures for CMT were higher in the groups ments per episode day, the same patterns generally held.
that actually used chiropractic; however, daily expenditures on Over the 4 annual cohort years, spine surgery rates
pharmaceuticals were lower for patients who used only CMT remained stable, but the proportion of patients who
and for those who use CMT first. Despite lower pharmaceu- obtained additional healthcare for back pain decreased for
tical costs, there was no decrease in daily expenditures for all groups (data not shown).
psychiatric treatment amongst patients using CMT.

Exploration of Psychiatric or Pain Medication Cost-Offsets

Examination of Care Provided by DCs, Psychiatrists, PTs, and DOs Across The distribution of patients across diagnostic group
Treatment Groups combinations differed according to treatment group (Table 6).
For each treatment group, Table 4 shows the mean Regardless of their psychiatric diagnoses, patients with
number of visits, and their distribution, to a DC for CMT, a osteoarthritis were somewhat more likely to be in the treatment
psychiatrist, a PT, or a DO for spinal manipulation during a group that did not use CMT while patients with other back and
cLBP episode. Interestingly, the plurality of patients who neck pain or with some other NMS condition were more likely
used any CMT (including 57% of those who used only to be in the treatment group that used only CMT. Episode
CMT) had 13 or more visits during their episode. The vast lengths were invariably shorter for patients in the treatment
majority of patients that we examined had no psychiatric, group that only used CMT; with 1 exception (patients with
PT, or DO visits, and very few patients obtained SMT anxiety and another NMS condition), patients who obtained
provided by DOs. While there was little overlap of CMT first and then conventional medical care had the longest
providers among patients using only CMT, there was episodes. Overall Part A and Part B expenditures were
substantial overlap of providers among patients using both invariably lowest for patients who used only CMT. With 1
medical care and CMT. exception (depression and osteoarthritis) Part A expenditures
were highest for patients who did not obtain any CMT.
Overall part B expenditures varied considerably by diagnos-
Compound Annual Growth Rate of Medicare Expenditures tic mix but, for all treatment groups that did not only use
Table 5 shows the annually calculated and inverse CMT, they were considerably higher when patients had both
propensity score weighted average length of an episode as depression and anxiety.
6 Weeks et al Journal of Manipulative and Physiological Therapeutics
Medicare Costs of Care Month 2016

Table 3. Unweighted and Propensity Score Weighted Spending and Outcomes Results (With Standard Errors in Parentheses) for the
Entire Episode of Care (Top) and per Episode Day (Bottom) for Older Medicare Part A and Part B Enrollees in 20072012 With
Chronic Low Back Pain Episodes Multiple Comorbidities Across 4 Treatment Groups
Medicare Expenditures During
the Episode ($)

Mean Days in Total Part A

N an Episode and Part B Part A
Expenditures Un-weighted Only used CMT 3909 298 (4.9) 3010 (144) 1168 (94)
during the Used conventional medical care first, then CMT 3563 367 (5.6) 8277 (251) 4070 (185)
entire episode Used CMT first, then conventional medical care 5235 481 (5.3) 8993 (216) 4843 (166)
Did not use CMT 59 619 455 (1.5) 11 231 (71) 6730 (58)
Propensity Only used CMT 3909 287 (2.1) 3581 (66) 1429 (45)
score weighted Used conventional medical care first, then CMT 3563 369 (2.4) 8721 (105) 4351 (80)
Used CMT first, then conventional medical care 5235 486 (2.8) 10 271 (113) 5567 (90)
Did not use CMT 59 619 454 (2.7) 11 039 (117) 6593 (98)
Expenditures Un-weighted Only used CMT 10.10 (0.50) 3.92 (0.34)
per episode Used conventional medical care first, then CMT 22.57 (0.70) 11.10 (0.53)
day Used CMT first, then conventional medical care 18.69 (0.46) 10.06 (0.37)
Did not use CMT 24.69 (0.16) 14.80 (0.14)
Propensity Only used CMT 12.50 (0.22) 4.99 (0.16)
score weighted Used conventional medical care first, then CMT 23.64 (0.28) 11.79 (0.22)
Used CMT first, then conventional medical care 21.13 (0.23) 11.45 (0.19)
Did not use CMT 24.32 (0.25) 14.52 (0.22)

CMT, chiropractic manipulation therapy.

a
Part D expenditures are calculated from only those patients who were enrolled in Medicare Part D.

CMT expenditures were highest for patients who only who had a discrete episode of cLBP. After propensity score
obtained CMT, and they were similar in groups who weighting that addressed differences in demographics across
obtained both CMT and conventional medical care during the treatment groups (including the finding that CMT patients
their cLBP episode. Psychiatric visit expenditures varied had lower illness burdens), we found that patients who used
considerably, but with 1 exception (depression and only CMT during their cLBP episodes had lower overall costs
osteoarthritis) were highest when patients had both of care, shorter episodes, and lower cost of care per episode day
depression and anxiety. There was no indication of reduced than patients in the other treatment groups. Further, costs of care
psychiatric expenditures associated with CMT. However, for the episode and per episode day were lower for patients who
there was evidence of lower overall and pain medication used a combination of CMT and conventional medical care than
expenditures associated with CMT: with 1 exception for patients who did not use any CMT, although most cost
(anxiety and osteoarthritis) patients who used CMT only differences were due to differences in inpatient care cost;
had the lowest overall and pain medication expenditures. Medicare part B and D expenditures for patients who used any
When examining expenditures per episode day, with 1 conventional medical care for their cLBP were similar, as were
exception (anxiety and osteoarthritis), patients who used compound annual expenditure growth rates. While costs of
only CMT had the lowest Part A and Part B expenditures care, and annual growth of healthcare costs, were generally
per day (Table 7). Among patients who used any CMT, lower for patients who used only CMT, that advantage might be
with 1 exception (anxiety plus depression and osteoarthri- offset somewhat by higher rates of later treatment for chronic
tis), mean daily expenditures of CMT were highest for LBP within a year of the episodes completion in this group.
patients who used only CMT, next highest for those who A high proportion of CMT users had 13 or more chiropractic
used conventional medical services followed by CMT, and visits. Others have found substantial variation in the number and
lowest for those who used CMT first and then conventional duration of episodes of chiropractic care and the number of visits
medical services. We found no indication that CMT was associated with those episode. 43,47 However, the potential
associated with lower psychiatric, overall medication, or overall (and particularly pain medication) cost savings that we
pain medication expenditures per episode day. found when only DCs provided back pain treatment for patients
with cLBP might warrant further exploration of a new role for
DCs in managing such multiply-comorbid patients.. 48,49
In contrast to other studies results, 42,43 we did not find that
DISCUSSION order of treatment was associated with large differences in
We examined 4 clinical treatment patterns for older, treatment costs of care when both CMT and conventional
Medicare fee-for-service enrolled, multiply-comorbid patients medical care were used during an episode of care. This might be
Journal of Manipulative and Physiological Therapeutics Weeks et al 7
Volume xx, Number Medicare Costs of Care

Table 3. Unweighted and Propensity Score Weighted Spending and Outcomes Results (with Standard Errors in Parentheses) for the
Entire Episode of Care (top) and per Episode Day (bottom) for Older Medicare Part A and Part B Enrollees in 20072012 with Chronic
Low Back Pain Episodes Multiple Comorbidities Across 4 Treatment Groups
Medicare Expenditures During
the Episode ($) 1 Year Post-Episode Events
Part B Part D a Future claims
Back for back pain
Total On CMT On Psychiatric care On PT care For DO SMT Total On pain meds surgery (%) treatment (%)
1842 (50) 478.51 (4.01) 19.82 (2.70) 6.82 (1.24) 0.04 (0.07) 1905 (164) 300.47 (37.59) 0.10 (0.051) 31.8 (0.75)
4207 (80) 329.93 (3.49) 32.71 (3.64) 71.47 (4.19) 2.73 (0.61) 3054 (209) 497.95 (48.79) 1.21 (0.183) 31.3 (0.78)
4150 (65) 332.17 (2.89) 35.29 (3.12) 41.95 (2.65) 1.52 (0.38) 3802 (200) 532.66 (43.28) 0.86 (0.128) 20.0 (0.55)
4501 (20) 0.00 (0.00) 51.32 (1.11) 34.08 (0.71) 1.49 (0.11) 4882 (64) 747.54 (14.44) 0.54 (0.030) 12.6 (0.14)
2152 (25) 522.24 (3.97) 31.85 (1.67) 6.62 (0.57) 0.03 (0.02) 2381 (75) 355.25 (15.74) 0.04 (0.008) 31.9 (0.17)
4370 (37) 323.68 (3.18) 37.52 (1.85) 72.22 (1.91) 2.67 (0.23) 3751 (95) 537.19 (19.58) 0.81 (0.034) 30.3 (0.17)
4704 (38) 342.72 (3.25) 44.28 (1.99) 43.14 (1.47) 1.58 (0.18) 4387 (106) 576.08 (21.01) 0.65 (0.030) 17.0 (0.14)
4446 (37) 0.00 (0.00) 49.06 (2.09) 35.88 (1.34) 1.61 (0.18) 4744 (112) 734.27 (24.09) 0.62 (0.029) 13.1 (0.13)
6.18 (0.15) 1.61 (0.01) 0.07 (0.009) .02287 (.00486) .00014 (.00029) 6.30 (0.37) 0.99 (0.10)
11.47 (0.20) 0.90 (0.01) 0.09 (0.010) .19480 (.01338) .00745 (.00198) 8.50 (0.40) 1.39 (0.11)
8.63 (0.12) 0.69 (0.01) 0.07 (0.006) .08718 (.00645) .00317 (.00093) 7.98 (0.29) 1.12 (0.07)
9.90 (0.04) 0.00 (0.00) 0.11 (0.002) .07492 (.00182) .00328 (.00029) 10.67 (0.10) 1.63 (0.02)
7.51 (0.07) 1.82 (0.01) 0.11 (0.005) .02311 (.00210) .00010 (.00010) 8.57 (0.18) 1.28 (0.04)
11.84 (0.08) 0.88 (0.01) 0.10 (0.004) .19570 (.00547) .00722 (.00080) 10.15 (0.17) 1.45 (0.04)
9.68 (0.06) 0.71 (0.01) 0.09 (0.003) .08874 (.00318) .00325 (.00046) 9.12 (0.14) 1.20 (0.03)
9.79 (0.07) 0.00 (0.00) 0.11 (0.004) .07905 (.00311) .00355 (.00050) 10.39 (0.16) 1.61 (0.04)

attributable to the fact that we examined a multiply-comorbid fee-for-service beneficiaries with a cLBP episode who had an
cohort of patients. We did find modestly higher daily Part B and additional NMS diagnosis and anxiety, depression, or both.
Part D costs for patients who used conventional medical care However, we found no reductions in psychiatric expenditures
before they used CMT, but this cost advantage was offset by associated with CMT. While we found reductions in overall
longer episode lengths when CMT was obtained first. and pain medication expenditures associated with CMT at the
We sought evidence that CMT might reduce expenditures episode level, these disappeared when examining those
for psychiatric care or pain medications among older Medicare expenditures on a per-episode-day basis. We did find evidence

Table 4. Mean Number of Visits (and Standard Errors), and Visit Distribution for Care Provided by Doctors of Chiropractic and
Psychiatrists (Top) and Physical Therapists and Doctors of Osteopathy (Bottom) During an Episode for the 4 Treatment Groups
Numbers of Visits per Episode
% of Patients in the Cohort with % of Patients in the Cohort with
the Following Number of Visits the Following Number of Visits
Mean Mean
(SE) 0 1-2 3-5 6-9 10-12 13 + (SE) 0 1-2 3-5 6-9 10-12 13 +
To a DC for CMT To a psychiatrist
Only used CMT 20.76 (0.34) 0.0 6.8 12.3 13.8 9.4 57.6 0.43 (0.08) 92.6 3.6 1.9 0.9 0.4 0.6
Used conventional medical 13.75 (0.32) 0.0 21.5 18.6 16.1 8.2 35.7 0.58 (0.08) 90.7 4.5 2.2 1.0 0.6 1.0
care first, then CMT
Used CMT first, then conventional 14.23 (0.29) 0.0 20.0 20.9 17.0 8.6 33.5 0.63 (0.05) 90.0 4.6 2.2 1.6 0.5 1.2
medical care
Did not use CMT 0.00 100 0.0 0.0 0.0 0.0 0.0 0.93 (0.02) 85.7 6.4 3.3 1.9 0.8 1.9

To a PT To a DO for SMT
Only used CMT 0.26 (0.04) 97.4 0.5 0.5 0.5 0.2 0.8 0.00 (0.002) 99.9 0.1 0.0 0.0 0.0 0.0
Used conventional medical 2.56 (0.17) 80.7 3.1 3.6 3.9 2.0 6.7 0.09 (0.015) 97.6 1.4 0.5 0.3 0.0 0.2
care first, then CMT
Used CMT first, then conventional 1.55 (0.09) 86.3 2.3 2.9 3.1 1.8 3.6 0.05 (0.010) 98.7 0.8 0.3 0.1 0.0 0.1
medical care
Did not use CMT 1.23 (0.02) 89.0 2.1 2.1 2.1 1.3 3.3 0.05 (0.003) 98.8 0.8 0.2 0.1 0.0 0.1
CMT, chiropractic manipulation therapy; DC, doctors of chiropractic; DO, doctor of osteopathy; PT, physical therapist; SMT, spinal manipulation therapy.
 8
Medicare Costs of Care
Weeks et al
Table 5. Compound Growth Rate Analysis.
Year in Which Episode Started Compound Growth Rate Compound Growth Rate
(20072010) Based on (20072010) Based on
2007 2008 2009 2010 Total Episode Costs Episode Costs per Day
Episode days Only used CMT 236 (2.4) 241 (2.7) 242 (2.8) 244 (2.7) 1.2%
Used conventional medical care first, then CMT 279 (2.9) 291 (3.0) 284 (3.2) 293 (3.1) 1.7%
Used CMT first, then conventional medical care 313 (3.3) 319 (3.4) 321 (3.6) 350 (3.4) 3.8%
Did not use CMT 290 (3.2) 304 (3.4) 322 (3.4) 321 (3.2) 3.4%
Part A spending ($) Only used CMT 1127 (77) 1179 (88) 1088 (89) 1065 (104) 1.9% 3.0%
Used conventional medical care first, then CMT 3471 (147) 3613 (159) 3977 (184) 4311 (218) 7.5% 5.8%
Used CMT first, then conventional medical care 4183 (178) 3907 (178) 4171 (196) 5513 (249) 9.6% 5.6%
Did not use CMT 5737 (207) 5771 (220) 5730 (220) 5669 (246) 0.4% 3.7%
Part B spending ($) Only used CMT 1739 (34) 1932 (42) 1756 (43) 1928 (56) 3.5% 2.4%
Used conventional medical care first, then CMT 3724 (54) 3670 (60) 3966 (70) 3946 (84) 1.9% 0.3%
Used CMT first, then conventional medical care 3345 (56) 3594 (64) 3415 (67) 4476 (91) 10.2% 6.2%
Did not use CMT 3542 (57) 3700 (66) 3802 (68) 3818 (82) 2.5% 0.9%
Spending on CMT ($) Only used CMT 477.08 (6.27) 463.86 (6.95) 483.31 (7.16) 564.73 (8.78) 5.8% 4.6%
Used conventional medical care first, then CMT 313.50 (5.51) 315.35 (5.90) 315.55 (6.27) 287.81 (6.54) 2.8% 4.4%
Used CMT first, then conventional medical care 306.97 (6.00) 315.79 (6.36) 286.54 (6.20) 280.95 (6.52) 2.9% 6.5%
Did not use CMT 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) 0.0% 0.0%
Spending on psychiatric care ($) Only used CMT 18.28 (1.60) 16.59 (1.88) 13.20 (1.66) 13.04 (1.76) 10.7% 11.6%
Used conventional medical care first, then CMT 23.37 (1.96) 20.20 (2.14) 25.02 (2.48) 17.44 (2.12) 9.3% 10.8%
Used CMT first, then conventional medical care 28.60 (2.39) 21.62 (2.38) 33.75 (2.99) 32.23 (2.91) 4.1% 0.3%
Did not use CMT 32.99 (2.55) 44.93 (3.50) 46.83 (3.37) 44.39 (3.33) 10.4% 6.7%
Part D a spending ($) Only used CMT 2279 (92) 2305 (111) 1938 (140) 2271 (122) 0.1% 1.2%
Used conventional medical care first, then CMT 2620 (111) 2675 (128) 2734 (183) 2889 (131) 3.3% 1.6%

Journal of Manipulative and Physiological Therapeutics

Used CMT first, then conventional medical care 2651 (127) 2821 (145) 4833 (279) 3099 (159) 5.3% 1.5%
Did not use CMT 3298 (140) 3119 (158) 3328 (211) 3459 (166) 1.6% 1.8%
Pain medication spending ($) Only used CMT 491.44 (30.34) 251.22 (22.32) 312.42 (30.15) 256.83 (24.76) 19.5% 20.3%
Used conventional medical care first, then CMT 402.09 (30.70) 529.86 (34.81) 581.79 (45.27) 351.16 (27.43) 4.4% 6.0%
Used CMT first, then conventional medical care 534.96 (40.27) 329.90 (30.29) 549.99 (50.33) 357.97 (32.46) 12.5% 15.7%
Did not use CMT 616.37 (42.94) 534.94 (39.87) 549.43 (46.00) 499.42 (38.00) 6.8% 9.9%
CMT, chiropractic manipulation therapy.
Data are for Patients who had A Chronic Low Back Pain Episode that Lasted Less than 730 Days and Whose Episode Started in A Particular Year Between 2007 and 2010. Compound Annual Growth Rates in
Expenditures during the Episode and Expenditures per Day are Shown.
a
Part D expenditures are calculated from only those patients who were enrolled in Medicare Part D

Month 2016
 Volume xx, Number
Journal of Manipulative and Physiological Therapeutics
Table 6. Numbers of Patients, Proportions of Patients of Each Diagnostic Mix Within A Treatment Group, Episode Lengths, and Medicare Expenditures During the Chronic Low Back
Pain Episode for the Different Medicare Spending Categories
Medicare Expenditures ($)
Part B Part D a
% of the Episode
Treatment Length For Psychiatric For Pain
N Group in Days Part A Total For CMT Care Total Medications
Anxiety Osteoarthritis Only used CMT 172 4.4% 260 (7.4) 1417 (150.7) 1868 (78.8) 502 (12.5) 8.09 (2.11) 2357 (195.9) 392 (44.9)
Used medical services first, then CMT 212 6.0% 372 (8.0) 3713 (219.5) 4107 (105.2) 323 (9.0) 7.65 (1.85) 1804 (147.0) 220 (28.9)
Used CMT first, then medical care 303 5.8% 531 (9.7) 4465 (245.0) 4304 (109.6) 350 (9.5) 16.44 (2.76) 2874 (206.9) 244 (33.9)
Did not use CMT 6313 10.6% 458 (7.1) 5117 (206.2) 3743 (80.4) 0 (0.0) 23.11 (2.57) 3868 (195.6) 413 (35.9)
Other neck and back pain Only used CMT 413 10.6% 300 (8.1) 1670 (245.5) 1870 (96.8) 450 (15.9) 21.78 (4.60) 1748 (177.4) 117 (23.8)
Used medical services first, then CMT 306 8.6% 322 (8.9) 3292 (366.9) 3464 (140.3) 301 (13.8) 11.78 (3.60) 2198 (205.4) 251 (35.8)
Used CMT first, then medical care 476 9.1% 457 (10.6) 3391 (374.6) 3378 (139.4) 320 (14.3) 8.22 (3.03) 3099 (222.6) 226 (31.1)
Did not use CMT 2789 4.7% 403 (12.7) 4295 (535.8) 3536 (181.2) 0 (0.0) 16.98 (5.53) 3044 (311.0) 265 (47.3)
Any other NMS Only used CMT 508 13.0% 267 (6.0) 812 (98.5) 2059 (73.9) 483 (10.4) 7.43 (1.60) 2289 (159.1) 125 (19.9)
Used medical services first, then CMT 378 10.6% 347 (7.7) 4048 (248.9) 3972 (116.2) 333 (9.7) 12.05 (2.30) 2915 (203.4) 215 (29.6)
Used CMT first, then medical care 544 10.4% 424 (8.6) 3387 (229.6) 3551 (110.8) 295 (9.2) 6.87 (1.75) 2747 (206.2) 325 (38.0)
Did not use CMT 4802 8.1% 436 (9.4) 4621 (290.2) 3801 (124.0) 0 (0.0) 14.13 (2.72) 3714 (255.3) 439 (47.1)
Depression Osteoarthritis Only used CMT 285 7.3% 306 (18.2) 1708 (134.9) 2576 (72.0) 563 (10.2) 106.02 (6.76) 3063 (258.3) 491 (46.5)
Used medical services first, then CMT 290 8.1% 378 (20.1) 4362 (226.6) 4566 (100.7) 339 (8.3) 60.16 (5.35) 4878 (378.5) 814 (69.5)
Used CMT first, then medical care 458 8.7% 506 (18.5) 6965 (265.4) 4486 (92.5) 345 (7.7) 42.01 (4.15) 4953 (358.6) 731 (61.9)
Did not use CMT 11 613 19.5% 450 (3.5) 6452 (187.4) 4222 (65.9) 0 (0.0) 49.24 (3.29) 5183 (267.3) 787 (46.8)
Other neck and back pain Only used CMT 621 15.9% 301 (11.8) 1386 (135.5) 1819 (55.1) 463 (10.8) 14.53 (2.09) 3270 (269.6) 444 (49.1)
Used medical services first, then CMT 473 13.3% 351 (14.5) 3929 (236.8) 3500 (79.4) 318 (9.3) 16.42 (2.31) 3681 (286.5) 464 (50.2)
Used CMT first, then medical care 635 12.1% 400 (13.4) 5149 (293.8) 3647 (87.8) 288 (9.6) 28.99 (3.32) 4853 (411.7) 531 (67.3)
Did not use CMT 4480 7.5% 393 (5.0) 6303 (386.4) 4075 (110.4) 0 (0.0) 36.61 (4.44) 4104 (443.1) 611 (84.4)
Any other NMS Only used CMT 821 21.0% 254 (4.4) 798 (76.7) 1870 (55.2) 529 (11.5) 20.34 (2.64) 2031 (177.0) 261 (39.5)
Used medical services first, then CMT 519 14.6% 338 (6.1) 3232 (187.3) 3793 (95.3) 339 (11.2) 15.48 (2.80) 3423 (239.1) 560 (60.3)
Used CMT first, then medical care 839 16.0% 467 (6.9) 4991 (221.6) 4642 (100.4) 340 (10.7) 50.83 (4.83) 4492 (280.6) 604 (64.1)
Did not use CMT 8108 13.6% 418 (6.8) 5957 (252.3) 4110 (98.4) 0 (0.0) 37.15 (4.30) 4647 (308.8) 814 (80.5)
Anxiety and Osteoarthritis Only used CMT 98 2.5% 287 (9.4) 1121 (166.6) 2386 (124.6) 563 (16.3) 44.81 (7.72) 3997 (358.7) 613 (85.3)
depression Used medical services first, then CMT 105 2.9% 463 (12.0) 8105 (450.5) 5888 (196.9) 230 (10.5) 80.32 (10.40) 9603 (513.7) 1010 (101.2)
Used CMT first, then medical care 199 3.8% 599 (11.8) 8618 (400.3) 7132 (186.7) 398 (11.9) 136.56 (11.68) 4758 (324.1) 395 (56.7)
Did not use CMT 5507 9.2% 530 (8.2) 9054 (304.5) 5288 (119.3) 0 (0.0) 105.98 (7.64) 6163 (300.3) 1033 (74.7)
Other neck and back pain Only used CMT 182 4.7% 313 (12.3) 1418 (227.5) 2003 (169.3) 459 (26.6) 26.48 (12.08) 1753 (394.4) 287 (75.4)
Used medical services first, then CMT 175 4.9% 384 (12.9) 3515 (339.0) 5138 (256.6) 399 (23.5) 136.04 (25.92) 2783 (525.6) 576 (112.9)
Used CMT first, then medical care 268 5.1% 514 (14.8) 6306 (447.6) 4878 (246.5) 381 (22.6) 88.55 (20.62) 5239 (608.1) 833 (114.5)
Did not use CMT 1924 3.2% 471 (17.6) 7846 (622.1) 4755 (303.3) 0 (0.0) 83.89 (25.00) 5963 (939.2) 1012 (182.8)
Any other NMS Only used CMT 821 21.0% 323 (9.6) 1707 (210.9) 2340 (116.4) 607 (16.5) 46.08 (7.51) 1817 (217.5) 266 (47.6)

Medicare Costs of Care

Used medical services first, then CMT 519 14.6% 441 (10.6) 5206 (347.0) 4998 (160.2) 344 (11.7) 102.69 (10.56) 4849 (366.3) 812 (85.8)
Used CMT first, then medical care 839 16.0% 550 (11.3) 7867 (406.9) 5431 (159.3) 280 (10.0) 78.76 (8.82) 4555 (339.9) 866 (84.7)
Did not use CMT 8108 13.6% 481 (11.0) 8004 (429.2) 5075 (161.0) 0 (0.0) 79.47 (9.27) 4782 (355.4) 841 (85.2)

Weeks et al
CMT, chiropractic manipulation therapy; NMS, neuro-musculoskeletal condition.
a
Part D Expenditures are calculated from only those patients who were enrolled in Medicare Part D.

9
 10
Medicare Costs of Care
Weeks et al
Table 7. Expenditures per Day for the Different Medicare Spending Categories
Medicare Expenditures per Episode Day ($)
Part B Part D a
Part A Total For CMT For Psychiatric Care Total For Pain Medications
Anxiety Osteoarthritis Only used CMT 5.46 (0.57) 7.19 (0.26) 1.93 (0.05) 0.031 (0.010) 8.80 (0.55) 1.46 (0.14)
Used medical services first, then CMT 9.97 (0.58) 11.03 (0.24) 0.87 (0.02) 0.021 (0.006) 5.32 (0.33) 0.65 (0.07)
Used CMT first, then medical care 8.40 (0.45) 8.10 (0.18) 0.66 (0.02) 0.031 (0.006) 6.03 (0.33) 0.51 (0.06)
Did not use CMT 11.18 (0.44) 8.18 (0.15) 0.00 (0.00) 0.051 (0.007) 8.46 (0.32) 0.90 (0.06)
Other neck and back pain Only used CMT 5.56 (0.79) 6.22 (0.23) 1.50 (0.04) 0.073 (0.012) 5.98 (0.49) 0.40 (0.09)
Used medical services first, then CMT 10.22 (1.10) 10.75 (0.31) 0.94 (0.04) 0.037 (0.009) 7.44 (0.56) 0.85 (0.14)
Used CMT first, then medical care 7.43 (0.79) 7.40 (0.22) 0.70 (0.03) 0.018 (0.005) 6.79 (0.39) 0.50 (0.08)
Did not use CMT 10.66 (1.28) 8.78 (0.32) 0.00 (0.00) 0.042 (0.011) 7.28 (0.60) 0.63 (0.13)
Any other NMS Only used CMT 3.04 (0.37) 7.70 (0.21) 1.81 (0.04) 0.028 (0.006) 8.49 (0.48) 0.46 (0.07)
Used medical services first, then CMT 11.66 (0.72) 11.44 (0.26) 0.96 (0.03) 0.035 (0.007) 8.77 (0.50) 0.65 (0.08)
Used CMT first, then medical care 7.98 (0.54) 8.37 (0.20) 0.70 (0.02) 0.016 (0.004) 5.83 (0.36) 0.69 (0.07)
Did not use CMT 10.60 (0.67) 8.72 (0.22) 0.00 (0.00) 0.032 (0.006) 8.25 (0.46) 0.97 (0.09)
Depression Osteoarthritis Only used CMT 5.38 (0.44) 8.12 (0.19) 1.77 (0.03) 0.334 (0.018) 10.86 (0.52) 1.74 (0.13)
Used medical services first, then CMT 11.48 (0.62) 12.01 (0.22) 0.89 (0.02) 0.158 (0.012) 13.28 (0.58) 2.22 (0.15)
Used CMT first, then medical care 13.93 (0.55) 8.97 (0.15) 0.69 (0.01) 0.084 (0.007) 10.23 (0.42) 1.51 (0.10)
Did not use CMT 14.33 (0.43) 9.38 (0.12) 0.00 (0.00) 0.109 (0.006) 11.46 (0.34) 1.74 (0.08)
Other neck and back pain Only used CMT 4.76 (0.49) 6.25 (0.18) 1.59 (0.04) 0.050 (0.009) 9.67 (0.54) 1.31 (0.12)
Used medical services first, then CMT 11.15 (0.71) 9.93 (0.22) 0.90 (0.03) 0.047 (0.008) 11.00 (0.58) 1.39 (0.13)
Used CMT first, then medical care 12.96 (0.78) 9.18 (0.21) 0.73 (0.02) 0.073 (0.011) 11.74 (0.68) 1.28 (0.14)
Did not use CMT 16.06 (1.04) 10.38 (0.27) 0.00 (0.00) 0.093 (0.015) 10.68 (0.79) 1.59 (0.19)
Any other NMS Only used CMT 3.14 (0.31) 7.37 (0.18) 2.08 (0.04) 0.080 (0.009) 8.40 (0.45) 1.08 (0.12)
Used medical services first, then CMT 9.57 (0.57) 11.23 (0.24) 1.00 (0.03) 0.046 (0.007) 10.37 (0.44) 1.70 (0.14)
Used CMT first, then medical care 10.69 (0.49) 9.94 (0.18) 0.73 (0.02) 0.109 (0.008) 8.98 (0.34) 1.21 (0.10)
Did not use CMT 14.24 (0.62) 9.83 (0.20) 0.00 (0.00) 0.089 (0.008) 11.01 (0.44) 1.93 (0.14)
Anxiety and depression Osteoarthritis Only used CMT 3.91 (0.60) 8.31 (0.36) 1.96 (0.05) 0.156 (0.024) 11.48 (0.67) 1.76 (0.19)

Journal of Manipulative and Physiological Therapeutics

Used medical services first, then CMT 17.50 (1.01) 12.72 (0.35) 0.50 (0.02) 0.173 (0.020) 15.56 (0.54) 1.64 (0.13)
Used CMT first, then medical care 14.38 (0.69) 11.90 (0.26) 0.66 (0.02) 0.228 (0.017) 9.75 (0.43) 0.81 (0.09)
Did not use CMT 17.08 (0.60) 9.98 (0.19) 0.00 (0.00) 0.200 (0.013) 11.50 (0.36) 1.93 (0.11)
Other neck and back pain Only used CMT 4.54 (0.77) 6.40 (0.38) 1.47 (0.06) 0.085 (0.027) 6.11 (0.85) 1.00 (0.18)
Used medical services first, then CMT 9.16 (0.93) 13.38 (0.47) 1.04 (0.05) 0.354 (0.047) 9.94 (1.16) 2.06 (0.28)
Used CMT first, then medical care 12.26 (0.92) 9.48 (0.34) 0.74 (0.03) 0.172 (0.028) 10.98 (0.78) 1.75 (0.17)
Did not use CMT 16.66 (1.39) 10.10 (0.46) 0.00 (0.00) 0.178 (0.037) 11.80 (1.14) 2.00 (0.25)
Any other NMS Only used CMT 27.47 (4.69) 7.25 (0.31) 1.88 (0.05) 0.143 (0.024) 6.26 (0.54) 0.92 (0.14)
Used medical services first, then CMT 18.17 (2.48) 11.34 (0.32) 0.78 (0.03) 0.233 (0.024) 10.93 (0.59) 1.83 (0.17)
Used CMT first, then medical care 10.64 (1.46) 9.88 (0.25) 0.51 (0.02) 0.143 (0.016) 8.93 (0.48) 1.70 (0.14)
Did not use CMT 19.61 (2.72) 10.55 (0.29) 0.00 (0.00) 0.165 (0.020) 10.29 (0.55) 1.81 (0.16)
CMT, chiropractic manipulation therapy; NMS, neuro-musculoskeletal condition.

Month 2016
a
Part D expenditures are calculated from only those patients who were enrolled in Medicare Part D.
Journal of Manipulative and Physiological Therapeutics Weeks et al 11
Volume xx, Number Medicare Costs of Care

that patients with osteoarthritis were less likely, and those with expenditures suggests that cost savings might be found in
other NMS diagnoses were more likely, to use CMT; however, other areas. Future work should examine broader and younger
this may reflect that doctors of chiropractic and doctors of populations, where such cost savings might be more readily
medicine have different coding practices. found. Finally, future studies should attempt to examine
patient centered health outcomes so that cost-effectiveness
analyses could be conducted.
Clinical and Policy Application
Our findings suggest that, from a Medicare cost
standpoint, CMT may be a cost-efficient first line treatment CONCLUSIONS
choice for older, multiply-comorbid patients with cLBP. If We found that older multiply-comorbid patients who used
policymakers encouraged DCs to have a greater role in only CMT during their cLBP episodes had lower overall costs
initially managing such patients, patients may have of care, shorter episodes, and lower cost of care per episode
episodes of care that were shorter and less costly (both day than patients in the other treatment groups. Further, costs
overall and per episode day), and they might have lower of care for the episode and per episode day were lower for
pharmaceutical expenditures for pain medications. Further, patients who used a combination of CMT and conventional
should such management require the addition of conven- medical care than for patients who did not use any CMT.
tional medical care after an initial course of CMT, These findings support initial CMT use in the treatment of,
policymakers might expect that overall costs might be and possibly broader chiropractic management of, older
similar to those for episodes wherein CMT was added after multiply-comorbid cLBP patients.
conventional medical care.

FUNDING SOURCES AND POTENTIAL CONFLICTS

Limitations and Future Studies OF INTEREST
Our study has several limitations. First, findings from the
multiply-comorbid group that we examined may not be This research was funded by the National Institutes of
generalizable either to the larger Medicare fee-for-service Health: National Center for Complementary and Integra-
population or to the US population. Second, we were tive Health (previously the National Center for Comple-
constrained by the use of large Medicare datasets. While mentary and Alternative Medicine) (R21AT008287 and
these datasets generated relatively large numbers of patients F32AT007507). No conflicts of interest were reported for
in the 4 treatment groups and reflect actual care utilization this study.
patterns, we were unable to determine whether care provided
was justified or resulted in better health outcomes, as
determined by patients. Third, when patients choose a
CONTRIBUTORSHIP INFORMATION
particular treatment, there exists the potential for selection Concept development (provided idea for the research):
bias due to unmeasured confounders. While we attempted to W.B.W., T.D.T., J.M.W., J.D.L., R.S., C.M.G.
address selection bias through inverse propensity score Design (planned the methods to generate the results):
weighting, ours is not a randomized controlled trial, and so W.B.W., T.D.T., J.M.W., B.L., J.D.L., R.S., C.M.G.,
there is no guarantee that the distributions of any unmeasured A.J.O.
risk factors do not vary between the weighted groups. Supervision (provided oversight, responsible for orga-
Therefore, all findings are referred to as associations and do nization and implementation, writing of the manuscript):
not necessarily imply causality. Finally, we analyzed discrete, W.B.W.
defined episodes of cLBP; analyses that use other definitions Data collection/processing (responsible for experiments,
of cLBP may generate different results. patient management, organization, or reporting data):
Studies such as ours provide initial evidence that CMT use W.B.W.
is associated with lower expenditures among Medicare Analysis/interpretation (responsible for statistical analy-
beneficiaries with cLBP and multiple co-morbidities. While sis, evaluation, and presentation of the results): W.B.W.,
the study design limits our ability to make strong conclusions, T.D.T., J.M.W., B.L., J.D.L., R.S., C.M.G., A.J.O.
future exploration of causation through randomized con- Literature search (performed the literature search): W.B.W.,
trolled trials is warranted; such studies might be a reasonable T.D.T., J.M.W., B.L., J.D.L., R.S., C.M.G., A.J.O.
next step in determining the most effective and efficient Writing (responsible for writing a substantive part of the
treatment for this multiply-comorbid and costly group of manuscript): W.B.W.
patients. Also, while we did not find reductions in Critical review (revised manuscript for intellectual
expenditures for psychiatric or pain medication associated content, this does not relate to spelling and grammar
with CMT in this population, that patients who used only checking): W.B.W., T.D.T., B.L., J.D.L., R.S., C.M.G.,
CMT had lower overall and per day Part A and Part B A.J.O.
12 Weeks et al Journal of Manipulative and Physiological Therapeutics
Medicare Costs of Care Month 2016

10. Gatchel RJ. Psychological disorders and chronic pain: cause

Practical Applications and effect relationships. In: Gatchel RJ, Turk DC, editors.
Psychological approaches to pain management: a practioner's
Among older, multiply-comorbid Medicare handbook. New York: Guilford Publications; 1996. p. 33-54.
beneficiaries with a chronic low back pain 11. Cornwal A, Doncleri DC. The effect of experimental induced
episode, chiropractic manipulative treatment anxiety on the experience of pressure pain. Pain 1988;35.
was associated with lower overall episode 12. Katon W. The impact of major depression on chronic medical
costs and lower episode costs per day. illness. Gen Hosp Psychiatry 1996;18:215-9.
13. Burns J, Johnson B, Mahoney N, Devine J, Pawl R. Cognitive
Most multiply-comorbid chronic low back and physical capacity process variables predict long-term
patients who used any chiropractic manipu- outcome after treatment of chronic pain. J Clin Consult
lative treatment had at least 6 chiropractic Psychiatry 1998;66:434-9.
visits; most of those who exclusively used 14. Holzberg AD, Robinson ME, Geissner ME, Gremillion HA.
The effects of depression and chronic pain on psychosocial
chiropractic manipulative treatment had more
and physical functioning. Clin J Pain 1996;12:118-25.
than 12 visits. 15. Weiner DK, Haggerly CL, Kritchevsky SB, et al. How does low
Use of chiropractic manipulative treatment was back pain impact physical function in independent, well-
associated with lower total Part A and Part D functioning older adults? Evidence from the Health ABC Cohort
Medicare cost growth for multiply-comorbid and implications for the future. Pain Med 2003;4:311-20.
16. Weiner DK, Kim YS, Bonino P, Wang T. Low back pain in
patients with chronic low back pain episodes
older adults: are we utilizing healthcare resources wisely? Pain
over the time period examined. Med 2006;7:143-50.
While we found overall Medicare cost-savings 17. Friedly J, Chan L, Deyo R. Increases in lumbosacral injections in
associated with use of chiropractic care, we the Medicare population: 19942001. Spine 2007;32:1754-60.
found no evidence of lower psychiatric or pain 18. Solomon DH, Avorn J, Wang PS, et al. Prescription opiod use
among older adults with arthritis or low back pain. Arthritis
medication expenditures associated with chiro-
Rheum 2006;55:35-41.
practic manipulative treatment within diagnostic 19. Gore M, Sadosky A, Stacey B, Tai K, Leslie D. The burden of
subgroups. chronic low back pain: clinical comorbidities, treatment
patterns, and health care costs in usual care settings. Spine
2012;37:E668-77.
20. Gerrits MM, Vogelzangs N, van Oppen P, van Marwijk HW,
van der Horst H, Penninx BW. Impact of pain on the course of
depressive and anxiety disorders. Pain 2012;153:429-36.
REFERENCES 21. Salazar A, Duenas M, Mico JA, et al. Undiagnosed mood
disorders and sleep disturbances in primary care patients with
1. Vogeli C, Shields AE, Lee TA, et al. Multiple chronic chronic musculoskeletal pain. Pain Med 2013;14:1416-25.
conditions: prevalence, health consequences, and implications 22. Olaya-Contreras P, Styf J. Biopsychosocial function analyses
for quality, care management, and costs. J Gen Intern Med changes the assessment of the ability to work in patients on
2007;22:391-5. long-term sick-leave due to chronic musculoskeletal pain: the
2. Parekh AK, Barton MB. The challenge of multiple comor- role of undiagnosed mental health comorbidity. Scand J
bidity for the US health care system. JAMA 2010;303:1303-4. Public Health 2013;41:247-55.
3. Partnership for Solutions Chronic Conditions: Making the Case 23. Ritzwoller DP, Crounse L, Shetterly S, Rublee D. The association
for Ongoing Care. Princeton: Robert Wood Johnson Foundation; of comorbidities, utilization and costs from patients identified with
2002. low back pain. BMC Musculoskelet Disord 2006;7:72.
4. Richardson W, Nerwick D, Bisgard J, et al. Crossing the 24. Wygant EG. Relief of depression in patients with chronic
Quality Chasm: A New Health System for the 21st Century. diseases. Curr Ther Res Clin Exp 1966;8:363-72.
Washington, DC: National Academies Press; 2001. 25. Kroenke K, Wu J, Bair MJ, Krebs EE, Damush TM, Tu W.
5. Braunstein JB, Anderson GF, Gerstenblith G, et al. Noncar- Reciprocal relationship between pain and depression: a 12-month
diac comorbidity increases preventable hospitalizations and longitudinal analysis in primary care. J Pain 2011;12:964-73.
mortality among Medicare beneficiaries with chronic heart 26. Myhr A, Augestad LB. Chronic pain patientseffects on
failure. J Am Coll Cardiol 2003;42:1226-33. mental health and pain after a 57-week multidisciplinary
6. Niefeld M, Braunstein JB, Wu AW, Saudek CD, Weller W, rehabilitation program. Pain Manag Nurs 2013;14:74-84.
Anderson GF. Preventable hospitalization among elderly 27. Rubinstein SM, van Middelkoop M, Assendelft WJ, de Boer
Medicare beneficiaries with type 2 diabetes. Diabetes Care MR, van Tulder MW. Spinal manipulative therapy for chronic
2003;26:1344-9. low-back pain. Cochrane Database Syst Rev 2011;
7. Himelhoch S, Weller W, Wu AW, Anderson GF, Cooper L. 2CD008112.
Chronic medical illness, depression, and use of acute medical 28. Furlan AD, Yazdi F, Tsertsvadze A, et al. Complementary and
services among Medicare beneficiaries. Med Care 2004;42: alternative therapies for back pain II. Evid Rep Technol
512-21. Assess (Full Rep) 2010:1-764.
8. Parekh AK, Kronick R, Tavenner M. Optimizing health for 29. Walker BF, French SD, Grant W. A Cochrane review of
persons with multiple chronic conditions. JAMA 2014;312: combined chiropractic interventions for low-back pain. Spine
1199-200. 2011;36:230-42.
9. Dersh J, Polatin PB, Gatchel RJ. Chronic pain and 30. Walker BF, French SD, Grant W, Green S. Combined
psychopathology: research findings and theoretical consider- chiropractic interventions for low-back pain. Cochrane
ations. Psychosom Med 2002;64:773-86. Database Syst Rev 2011;14CD005427.
Journal of Manipulative and Physiological Therapeutics Weeks et al 13
Volume xx, Number Medicare Costs of Care

31. Chou R, Qaseem A, Snow V, et al. Diagnosis and treatment of of Chiropractic Users and Nonusers in Older, Multiply
low back pain: a joint clinical practice guideline from the Comorbid Medicare Patients With Chronic Low Back Pain.
American College of Physicians and the American Pain J Manipulative Physiol Ther 2015;38:620-8.
Society. Ann Intern Med 2007;147:478-91. 41. Rozenberg S. Chronic low back pain: definition and
32. American Geriatrics Society. The management of chronic treatment. Rev Prat 2008;58:265-72.
pain in older persons: AGS Panel on Chronic Pain in Older 42. Liliedahl RL, Finch MD, Axene DV, Goertz CM. Cost of care
Persons. J Am Geriatr Soc 1998;46:635-51. for common back pain conditions initiated with chiropractic
33. Ndetan HT, Bae S, Evans MW, Rupert RL, Singh KP. doctor vs medical doctor/doctor of osteopathy as first
Characterization of health status and modifiable risk behavior physician: experience of one Tennessee-based general health
among United States adults using chiropractic care as insurer. J Manipulative Physiol Ther 2010;33:640-3.
compared with general medical care. J Manipulative Physiol 43. Weigel PAM, Hockenberry JM, Bentler SE, Kaskie B,
Ther 2009;32:414-22. Wolinsky FD. Chiropractic episodes and the co-occurrence
34. Stano M, Smith M. Chiropractic and medical costs of low of chiropractic and health services use among older Medicare
back care. Med Care 1996;34:191-204. beneficiaries. J Manipulative Physiol Ther 2012;35:168-75.
35. Haas M, Sharma R, Stano M. Cost-effectiveness of medical 44. Weeks WB, Whedon JM, Toler A, Goertz CM. Medicare's
and chiropractic care for acute and chronic low back pain. J Demonstration of Expanded Coverage for Chiropractic Ser-
Manipulative Physiol Ther 2005;28:555-63. vices: limitations of the Demonstration and an alternative direct
36. Martin BI, Gerkovich MM, Deyo RA, et al. The association of cost estimate. J Manipulative Physiol Ther 2013;36:468-81.
Complementary and Alternative Medicine use and health care 45. Coulter ID, Hurwitz EL, Adams AH, Genovese BJ, Hays R,
expenditures for back and neck problems. Med Care 2012;50: Shekelle PG. Patients using chiropractors in North America:
1029-36. who are they, and why are they in chiropractic care? Spine
37. Patorno E, Glynn RJ, Hernandez-Diaz S, Liu J, Schneeweiss (Phila Pa 1976) 2002;27:291-6 [discussion 78].
S. Studies with many covariates and few outcomes: selecting 46. Rassen JA, Glynn RJ, Brookhart MA, Schneeweiss S.
covariates and implementing propensity-score-based con- Covariate selection in high-dimensional propensity score
founding adjustments. Epidemiology 2014;25:268-78. analyses of treatment effects in small samples. Am J
38. Rassen JA, Shelat AA, Franklin JM, Glynn RJ, Solomon DH, Epidemiol 2011;173:1404-13.
Schneeweiss S. Matching by propensity score in cohort 47. Whedon J, Song Y, Davis MA. Trends in the use and cost of
studies with three treatment groups. Epidemiology 2013;24: chiropractic spinal manipulation under Medicare Part B. Spine
401-9. J 2013;13:1449-54.
39. Suh HS, Hay JW, Johnson KA, Doctor JN. Comparative 48. Davis MA, Whedon JM, Weeks WB. Complementary and
effectiveness of statin plus fibrate combination therapy and alternative medicine practitioners and Accountable Care
statin monotherapy in patients with type 2 diabetes: use of Organizations: the train is leaving the station. J Altern
propensity-score and instrumental variable methods to adjust Complement Med 2011;17:669-74.
for treatment-selection bias. Pharmacoepidemiol Drug Saf 49. Paskowski I, Schneider M, Stevans J, Ventura JM, Justice BD.
2012;21:470-84. A hospital-based standardized spine care pathway: report of a
40. Weeks WB, Tosteson TD, Whedon JM, et al. Comparing multidisciplinary, evidence-based process. J Manipulative
Propensity Score Methods for Creating Comparable Cohorts Physiol Ther 2011;34:98-106.
13.e1 Weeks et al Journal of Manipulative and Physiological Therapeutics
Medicare Costs of Care Month 2016

APPENDIX 1. DETAILS OF THE METHODS USED To evaluate patients with multiple comorbidities, we
Patient Inclusion Criteria limited the study population to patients who had an
ICD-9-CM code indicating at least 1 of the following
We studied fee-for-service Medicare beneficiaries using additional musculoskeletal diagnoses during the episode or
100% Medicare Part A and Part B files and 40% Medicare Part in the 12 months preceding episode onset: other back and
D files for 20062012. The population was derived from neck pain (307.89, 722.3, 722.32, 722.8, 722.83, 846 series,
Medicare beneficiaries with at least 1 evaluation and 847.2, 847.3, 847.9, 996.41), non-back rheumatism
management visit in a calendar year with the following annual (729.0), arthritis and other arthropathies (716.19, 716.68,
exclusions: a) any Medicare Advantage; b) less than full Part 716.95) osteoarthritis (715.00, 715.09, 715.18, 715.21,
A and B enrollment for the entire period (or from the month 715.22, 715.28, 715.98), and diffuse diseases of the
turning 65 to month of death); c) residence outside the 50 connective tissue (728.89, 728.9, 729.1, 729.3, 729.4).
United States or Washington, DC; d) residence in an We further limited the study population to patients who had
unidentifiable county; and f) ageb66 (because of a required an ICD-9-CM code indicating at least 1 of the following
1-year look back period for exclusion criteria detailed below) psychological comorbidities during the episode or in the 12
or N 99. months preceding episode onset: depression (296.3, 300.4,
Following Cherkins methodology, 50 we eliminated 309.0, 309.2, 309.8, 311), anxiety (293.84, 300.00, 300.09)
patients with ICD-9-CM codes that are unlikely to be or insomnia (327.01, 327.02).
responsive to non-surgical treatment: neoplasms (CPT-9-CM Patients who used chiropractic manipulative therapy
codes 140239.9), intra-spinal abscess (324.1), osteomyelitis (CMT) were identified as having as at least 1 visit to a
(730730.99), vertebral fractures with spinal cord injury doctor of chiropractic (identified as code 35 in the
(806.0-806.9), open vertebral fractures without spinal cord provider specialty field) and generating at least 1 CPT
injury (805.1, 805.3, 805.5, 805.7, 805.9), vertebral disloca- based charge for CMT (code 98940 (1 or 2 regions),
tions (839839.59), chordotomy (03.2-03.29), and cervical 98941(3 to 4 regions), or 98942 (5 regions)) during the
and dorsal fusions (81.01-81.03). Because Doctors of episode of cLBP. We excluded costs of treatments that
Chiropractic (DCs) often treat patients for vertebral disloca- were expanded for the purposes of conducting The
tions (ICD-9-CM diagnostic codes 839839.59) we did not Demonstration of Expanded Coverage of Chiropractic
exclude patients who had these codes if the code was Services under Medicare. 44
recorded during a chiropractic visit. Informed by Liliedahl 42 and Weigel, 43 we identified 4
Guided by Martins analysis of the Medical Expenditures treatment groups: 1) Those in which only CMT was used;
Panel Survey (MEPS) prevalence of diagnostic categories of 2) Those in which the initial treatment for the episode
back problems, 51 we defined low back pain (LBP) patients as consisted of non-chiropractic medical care, but where CMT
those who had a primary ICD-9-CM diagnosis code in the was also obtained during the episode; 3) Those in which the
724 series: 724.2 (lumbago), 724.3 (sciatica), 724.4 (thoracic initial treatment for the episode consisted of CMT, but where
or lumbosacral neuritis or radiculitis, unspecified), 724.5 non-chiropractic medical care was also obtained during the
(backache, unspecified), 724.8 (other symptoms referable to episode; and 4) Those in which no CMT was obtained during
back), and 724.9 (other unspecified back disorders). But the episode.
because DCs frequently use additional ICD-9-CM codes, we
also included patients with a primary ICD-9-CM diagnosis
code in the following series if such diagnoses were made Measures
during a chiropractic visit: 722 (intervertebral disc disorders), Our primary outcome measures were inflation- and
739 (non-allopathic subluxation), and 839 (other multiple and price-adjusted Medicare Part A, Part B, and Part D costs
ill-defined dislocations). for a cLBP episode of care for each treatment group.
Chronic low back pain (cLBP) lasts at least 3 months. 41 Because the length of an average episode of care varied
We therefore defined a cLBP episode as beginning with the according to group, we also calculated each of these costs
recording of a low back pain diagnosis at least twice over at per episode day.
least a 90-day period following a period of 6 months during We were interested in whether the multiply-comorbid
which no such diagnosis was recorded; the episode was patients who obtained CMT might use psychiatric care,
defined to end on the day that the last such diagnosis was other care for back pain, or pain medications differently
followed by at least 180 days without any such diagnosis. than those who did not, with the potential for seeing an
Although we had access to 7 years of data, because of the offset effect. Therefore, for each episode, we calculated
required 1-year look-back and 1-year look-forward periods, expenditures for CMT visits (described above) and
our analyses included cLBP episodes that occurred during psychiatric visits (defined as any visit wherein the provider
5 years, between 20072011. If an individual patient had has an AMA specialty code of ADP, CHP, PFP, P, or PYA).
multiple episodes during our study period, we counted only Further, anticipating that the 2 other types of providers that
the first 1. Medicare reimburses for spinal manipulative therapy
Journal of Manipulative and Physiological Therapeutics Weeks et al 13.e2
Volume xx, Number Medicare Costs of Care

(SMT) doctors of osteopathy (DOs) and physical groups using the following variables (all of which were
therapists (PTs) might provide care to any of the treatment included in the model):
groups, we used CPT codes (for DOs, CPT codes 98925
98928 and for PTs, CPT code 97140) with a concurrently a. From Medicare files:
documented LBP diagnosis to identify, calculate the i. Socio-demographic variables (patient sex, patient race,
reimbursements for, and calculate the number of visits per whether the patient was concurrently enrolled in
episode for such care. Medicaid, whether the patient was originally enrolled in
We also examined reimbursements for pain medications Medicare by virtue of disability or age, and whether the
(including opiate analgesics, non-steroidal anti-inflammatory beneficiary has a low-income subsidy for Medicare Part
drugs, Cox-2 inhibitors, steroid injections, and other D (a measure of patient poverty)); and
ii. Diagnostic codes which we used to identify patients who had
medications commonly used to treat recalcitrant low back
the aforementioned categorically defined comorbidities that
pain (including gabapentin, other anti-seizure medications,
defined our study population and from which we calculated
oral steroids, and non-benzodiazepine muscle relaxants (as
modified Deyo-Charlson 52 scores and Iezzoni 53 scores.
benzodiazepines are not reimbursed by the Medicare Part D
program) that were identified by the relevant Part D drug b. From ZIP Code-linked datasets:
reimbursement codes). i. An estimate of patient income (ZIP Code specific mean
In addition, we examined 2 secondary measures that annual household income in 2010) as well as the
occurred within a year of the end of the episode. First, proportion of the population that lived under the federal
we considered future healthcare use, as measured by poverty level;
ii. The per-capita supply of DCs in the Dartmouth Atlas
claims, for LBP after a 180-day period wherein no low
Defined Hospital Referral Region. 54 We included this
back pain diagnoses were recorded, but within a year of
in the propensity score because the supply of DCs has
the end of the cLBP episode. Second, we examined
been associated with CMT utilization in the Medicare
admission for back surgery (defined as admission for
population; 55,56 therefore, the supply of DCs represents
DRG 028, 029, or 030 before 2008; admission for MS
a confounder that should be included in propensity score
DRG 491, 473, or 460 since) within 1 year of the end of
generation as it might be associated with both treatment
the cLBP episode.
exposure and overall CMT utilization.
iii. The regional carrier that was used by Medicare to
process CMT claims in the region where the patient
Analyses lived as a variable in generating propensity scores.
Because we anticipated that we would find that the
different treatment groups would have different distribu- We then used all cases in linear or logistic regressions to
tions of patient characteristics, 36,44,45 we used the following compare means or proportions, respectively, of outcome
propensity score methodology to adjust for confounders variables before and after inversely weighting each case by the
and create equivalent groups for comparison. 3739 First, we propensity to be in the treatment group to which they
used multinomial stepwise logistic regression 46 (with self-assigned. In essence, this method relatively over-weighted
criteria to enter the stepwise regression set at pb0.05 and results for patients who were least likely to be in that treatment
criteria to exit set at p N 0.10) to calculate the estimated group given their demographics, rendering the cohorts more
probability of each patient to be in each of the 4 treatment similar on observed variables, demographically.

FURTHER READING 53. Iezzoni L. Risk Adjustment for Measuring Healthcare

Outcomes. 4th ed. Chicago, IL: Health Administration Press;
2012.
50. Cherkin DC, Deyo RA, Volinn E, Loeser JD. Use of the 54. The Dartmouth Atlas Project. Accessed October 1, 2012, at ,
International Classificatio of Diseases (ICD-9-CM) to identify http://www.dartmouthatlas.org2012.
hospitalization for mechanical low back problems in admin- 55. Whedon JM, Song Y. Geographic variations in availability
istrative databases. Spine 1992;17:817-25. and use of chiropractic under Medicare. J Manipulative
51. Martin BI, Deyo RA, Mirza SK, et al. Expenditures and health status Physiol Ther 2012;35:101-9.
among adults with back and neck problems. JAMA 2008;299:656-64. 56. Whedon JM, Song Y, Davis MA, Lurie JD. Use of
52. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical chiropractic spinal manipulation in older adults is
comorbidity index for use with ICD-9-CM administrative strongly correlated with supply. Spine (Phila Pa 1976) 2012;
databases. J Clin Epidemiol 1992;45:613-9. 37:1771-7