referral source and outcomes of physical therapy care in patients with low back pain _ journal of...

8/13/2019 Referral Source and Outcomes of Physical Therapy Care in Patients With Low Back Pain _ Journal of Orthopaedic &

1/11

journal of orthopaedic&sports physical therapy | volume 42 | number 8 | august 2012 | 705

Low back pain (LBP) is

among the most prevalentand costly musculoskeletal

disorders in the UnitedStates.1,27,28 There is considerable

variation in how LBP is managedwithin the current healthcare

environment.36,43 Primary care physi-

cians (PCPs) are the most commonly uti-

lized provider for individuals with LBP,34

and 2 of the most common providers towhich PCPs refer patients with LBP are

orthopaedic surgeons and physical thera-

pists.9,34,36Patients with spine dysfunction

comprise a substantial portion of outpa-

tient physical therapy caseloads.11Over-

all, however, PCPs manage the majority

of LBP cases without referral to other

healthcare providers.9Primary care prac-

tice guidelines for LBPabsent red flags,

such as systemic or radicular signsrec-

ommend an initial trial of self-care and

appropriate pharmaceutical agents be-fore a referral is made for nonpharmaco-

logical management.5

STUDY DESIGN:Retrospective longitudinal

cohort.

OBJECTIVES:To describe the clinical charac-

teristics of patients with low back pain according

to physician referral source, and to identify as-

sociations between referral source and discharge

functional status, as well as number of physical

therapy visits.

BACKGROUND:Little is known about as-

sociations between physician referral source and

outcomes of physical therapy care for patients with

low back pain. Exploring these associations can

contribute to better understanding of physicianphysical therapist relationships and may lead to

improved referral patterns.

METHODS:Data from a proprietary clinical

database were examined retrospectively. Physician

referral source was classified as primary care,

specialist, or occupational medicine. Outcomes

were overall health status at discharge and number

of physical therapy visits. Descriptive statistics

and bivariate associations between referral source

and each outcome were assessed by calculating

differences and 95% confidence intervals (CIs) in

means and proportions. To account for potential

confounding, multilevel linear regression was

used to adjust for baseline clinical covariates,effects related to clustering of patients treated by

individual clinicians, and clinicians working within

individual clinics.

RESULTS:Bivariate and multilevel analyses

revealed significant associations between referral

source and discharge overall health status, as well

as number of visits. After multilevel adjustment for

covariate and clustering effects, primary care and

occupational medicine referrals were associated,

on average, with point increases of 1.6 (95% CI: 0.7,

2.6) and 4.8 (95% CI: 2.7, 6.9) in discharge overall

health status scores, respectively, compared to

specialist referral. Similarly, primary care and

occupational medicine referrals were associated,on average, with 0.44 (95% CI: 0.27, 0.61) and

0.83 (95% CI: 0.44, 1.22) fewer visits, respectively,

compared to specialist referral.

CONCLUSION:After accounting for clinical

covariates and clustering, patients with low back

pain who were referred by occupational medicine

and primary care physicians tended to have better

functional outcomes and required fewer physical

therapy visits per episode of care.

LEVEL OF EVIDENCE:Prognosis, level 2c.

J Orthop Sports Phys Ther 2012;42(8):705-715,

Epub 8 March 2012. doi:10.2519/jospt.2012.3957

KEY WORDS:lumbar spine, physician referral,practice-based evidence

1Associate Professor, Department of Physical Therapy Education, College of Health Professions, SUNY Upstate Medical University, Syracuse, NY.2Assistant Professor, Department

of Physical Therapy Education, College of Health Professions, SUNY Upstate Medical University, Syracuse, NY. 3Supervisor of Physical Therapy, St Camillus Health and

Rehabilitation Center, Syracuse, NY. 4(Deceased) Director of Consulting and Research, Focus On Therapeutic Outcomes, Inc, Knoxville, TN. This research was supported by a

grant from the Section on Health Policy & Administration of the American Physical Therapy Association. This study was determined to be exempt from review by the Institutional

Review Board for the Protection of Human Subjects of SUNY Upstate Medical University, Syracuse, NY. Drs Brooks, VanBeveren, and Dolphin affirm that they have no financial

affiliation, including research funding, or involvement with any commercial organization that has a direct financial interest in any matter included in this manuscript. Dr Hart was

an employee of, and investor in, Focus On Therapeutic Outcomes, Inc (FOTO), the database management company that managed the data analyzed in the manuscript. Analyses

of data like the analyses presented in this article were part of Dr Harts daily w ork activities. Addresscorrespondence to Dr Gary Brooks, SUNY Upstate Medical University,

250 East Adams Street, Room 2231, Silverman Hall, Syracuse, NY 13210. E-mail: [email protected] 2012 Journal of Orthopaedic & Sports Physical Therapy

GARY BROOKS, PT, DrPH, CCS1 MICHELLE DOLPHIN, PT, DPT, OCS2

PATRICK VANBEVEREN, PT, DPT, OCS, GCS3 DENNIS L. HART, PT, PhD4

Referral Source and Outcomesof Physical Therapy Care in Patients

With Low Back Pain

[RESEARCHREPORT]

Copyright2012JournalofOrthopaedic&S

po

rtsPhysicalTherapy.Allrightsreserved.
mailto:[email protected]:[email protected]:[email protected]


2/11

706 | august 2012 | volume 42 | number 8 | journal of orthopaedic &sports physical therapy

Little is known about the patterns and

sources of referral to physical therapy for

LBP in the United States. Freburger and

colleagues11 found that physical therapy

referral for patients with spine disor-ders was associated with factors such as

medical diagnosis and physician deci-

sion to order more diagnostic tests or to

consult another physician, in addition to

other patient-level variables. An inves-

tigation of individuals receiving work-

ers compensation found that referral

to physical therapy was associated with

greater baseline disability, as measured

by the Roland-Morris questionnaire and

the Short-Form 12-Item Health Survey.6

For patients with chronic LBP, physicaltherapy referral was positively associated

with specialist physician providers and

workers compensation insurance, and

negatively associated with having no in-

surance.10With regard to referral source,

referrals from PCPs and orthopaedists

were associated with better discharge

functional status outcomes in patients

with lumbar or cervical involvement.33

Another investigation noted that referral

source was not associated with functional

outcome; however, there was a trend to-ward an association between specialist

physician and lower number of visits.3

Patients referred from specialist phy-

sicians may have clinical characteristics

different from those of patients referred

by PCPs; for example, they may be more

likely to have chronic LBP or more com-

plicated clinical presentations, or to

have failed to respond to treatment by

a PCP. By the time a patient with LBP

is referred to a physical therapist, she/

he might have been seen by a PCP andperhaps additional specialist physicians,

and might also have undergone diagnos-

tic procedures, such as imaging studies.

During this process, patients may experi-

ence delays in receiving physical therapy

care, thus increasing their symptom du-

ration at initiation of physical therapy

treatment and influencing outcomes of

care.7,33

Identifying associations between

physician referral source and physical

therapy outcomes can inform policy and

practice regarding physician-physical

therapist referral trends, and may influ-

ence communication between the physi-

cal therapist and referring physicians.These associations can provide insights

into the nature of the physicianphysical

therapist relationship and suggest further

exploration of the optimal referral path-

way among PCPs, specialist physicians,

and physical therapists for patients with

LBP. The purpose of this investigation

was to describe the characteristics of

patients with LBP according to physi-

cian referral source, including specialist,

primary care, and occupational medicine

physicians. An additional purpose was toidentify associations between referral

source and functional status at discharge

from physical therapy, as well as utili-

zation of physical therapy, as indicated

by number of physical therapy visits. In

identifying associations, we sought to ad-

just for potential confounding by patient

characteristics such as age, sex, symptom

duration, medication use, payer source,

and baseline functional status, and also

to adjust for clustering of patients within

clinicians and clinics.

METHODS

Data Source

Data were extracted from the

Focus On Therapeutic Outcomes,

Inc (FOTO, Knoxville, TN) clini-

cal database. The FOTO database has

been used for outcomes research in

a number of clinical populations, in-

cluding patients with musculoskeletal

impairments,

7,22

spine pathology,

24,25

knee pathology,23 and upper extremity

impairments. 19,26 In addition, research-

ers have used the FOTO database to

examine clinical expertise in physical

therapy,31 orthopaedic clinical special-

ization in physical therapy,18and physi-

cal therapy clinic performance.33 The

FOTO database has also been useful in

comparative-effectiveness research7and

pay-for-performance methodology.17We

acquired a data set from FOTO that in-

cluded all records in the database of pa-

tients who were treated for nonspecific

LBP syndromes between the years 2003

and 2005. From this initial data set, we

derived an analysis data set, as describedsubsequently.

Subjects

Selection of observations to be used in the

study analyses was based on nonmissing

values for several inclusion and exclusion

criteria. FIGURE 1 illustrates the selection

of observations from the initial, full data

set to the final sample used in the analy-

sis data set. Observations were included

in analyses if they were classified as hav-

ing orthopaedic and lumbar impairment,treated by a physical therapist or physi-

cal therapist assistant, and referred by a

physician. Observations were excluded

from analyses if they had surgery for the

current condition or had missing values

for both dependent variables. To main-

tain Health Insurance Portability and

Accountability Act compliance, observa-

tions of those aged 90 years or older were

also excluded.

VariablesThe primary independent variable, re-

ferral source, was physician status, cat-

egorized as primary care, specialist, or

occupational medicine. Physicians listed

as internal medicine, obstetrician/gy-

necologist, pediatrician, or family prac-

tice were coded as PCPs. Orthopaedic

surgeons, neurologists, neurosurgeons,

rheumatologists, physiatrists, and plastic

surgeons were coded as specialist physi-

cians. Occupational medicine physician

status was retained and coded as a thirdreferral-source category.

Discharge overall health status (OHS)

and number of visits were the 2 depen-

dent variables. The OHS is a patient self-

report measure of physical function that

adapts to different patient conditions, the

development of which has been described

in detail elsewhere.15-17,20Briefly, the OHS

was developed from health-related qual-

ity-of-life measures in the widely used

Medical Outcomes Study 36-Item Short-

[RESEARCHREPORT]


po



3/11

journal of orthopaedic &sports physical therapy | volume 42 | number 8 | august 2012 | 707

Form Health Survey and Short-Form 12-

Item Health Survey.39,42These measures

establish 8 constructs within the OHS,

including general health, physical func-

tioning, emotional and physical roles,bodily pain, mental health, vitality, and

social functioning. Three items assessing

upper extremity function not measured

by the Medical Outcomes Study 36-Item

Short-Form Health Survey or Short-

Form 12-Item Health Survey were added

to the physical functioning component of

the instrument.15,16

The OHS is scored according to pub-

lished algorithms,40,41resulting in a total,

summary score ranging from 0 to 100,

with higher values representing higher-level functioning. Internal consistency

reliability coefficients (Cronbach alpha)

for OHS constructs represented by 2 or

more questions within the instrument

varied between 0.57 and 0.89, with the

highest coefficient for physical function

in a sample of adults in industrial reha-

bilitation. 15 Test-retest reliability (intra-

class correlation coefficient [ICC2,1

]) for

the OHS was 0.90 in a sample of adults

in outpatient musculoskeletal rehabili-

tation, and 0.92 for outpatients withchronic symptoms.16The instrument has

also demonstrated strong sensitivity to

change, with a standardized response

mean of 0.87 and effect size of 0.83, in

a sample of patients with lumbar spine

dysfunction.31

Number of visits was tallied and re-

corded for each episode of care. To elimi-

nate implausible values for analyses

involving number of visits, observations

were also excluded if the number of visits

exceeded the duration of care, as mea-sured in days.

Patient-level variables recorded on

initial visit included age, sex, baseline

OHS score, whether the patient was tak-

ing prescription medications, exercise

status, and symptom duration. Symptom

duration was based on the number of days

between the onset of LBP symptoms and

initial physical therapy examination. This

item was coded such that the patient was

considered acute if duration of symptoms

was 21 days or fewer, subacute if between

22 and 90 days, and chronic if greater

than 90 days. This methodology, which

classifies patients into 3 clinically usefulcategories, has been utilized in previous

studies that employed the FOTO data-

base to investigate outcomes in patients

with LBP.7,17,33,44Exercise status on initial

visit was classified into 1 of 3 categories:

at least 20 minutes 3 or more times per

week, 1 to 2 times per week, or seldom

or never. Payer source was classified in 8

categories, representing fee-for-service,

health maintenance organization or pre-

ferred provider organization, litigation,

Medicaid, Medicare, self-pay, workerscompensation, and other. Also included

in the analyses was duration of the epi-

sode of care, measured in calendar days

between the date of initial evaluation and

discharge.

Data Analysis

To assess potential bias that might have

been introduced during the data-selec-

tion process, variables in the analysis

data set were compared to those of ob-

servations excluded from the analyses

(FIGURE 1). These comparisons used inde-

pendent-samples ttests to calculate mean

differences and 95% confidence intervals(CIs) for numeric variables, and for cat-

egorical variables used contingency-table

analyses to calculate risk differences and

95% CIs. Results of these analyses are

displayed in TABLE 1.

Prior to the study, physician referral

source was classified as 2 levels, primary

care and specialist, with occupational

medicine classified within the specialist

category. During preliminary analyses,

however, it became apparent that ad-

ditional stratification of referral sourcewould better fit the data. Accordingly, we

elected to classify referrals from occupa-

tional medicine physicians as a third cat-

egory. Using 1-way analyses of variance

and contingency-table analyses to deter-

mine differences and 95% CIs in means

and proportions, respectively, we com-

pared baseline characteristics, as well as

number of visits and discharge OHS, ac-

cording to referral source.

Associations between referral source

66 524 observations (initial dataset)

64 814 observations were classified as

having orthopaedic and lumbar

impairment.

47 126 observations were treated by a

physical therapist or physicaltherapist assistant.

10 017 observations were referred by aphysician and had nonmissing values

for referral source.

7971 observations (analysis dataset)

2046 observations excluded:

Had surgery for the primary

condition, n = 1505 Had missing values for discharge

OHS and number of visits, n = 592

Age 90 or older, n = 22

FIGURE 1. Data selection procedure. Abbreviation: OHS, overall health status.


po



4/11


[RESEARCHREPORT]

and each outcome were assessed for

confounding through a 2-step process

using multilevel modeling. Multilevel

modeling is appropriate for hierarchi-

cally structured data, in which individu-

al observations are nested within larger

units. In the current investigation, pa-

tients were nested within treating cli-

nicians, who were nested within clinics

where they were employed. Factors at

each level may influence outcomes and

should be controlled in analyses of hi-

erarchical data.32In step 1 of the analy-

sis, separate linear regression models

for each of the 2 outcomes were fitted

for individual-level variables. Variables

entered at this stage included referral

source (specialist physician was the ref-

erence value), baseline OHS score, age,

sex, symptom duration (acute status was

the reference value), taking medications

on admission, exercise status (no exercise

was the reference value), payer source

(fee-for-service was the reference value),

and duration of care. For step 2, random-

intercept models were fitted for each out-

come by entering the unique identifiers

for clinician and clinics, respectively, as

random-effects variables. These models

tested whether associations found in

individual-level analyses persisted after

adjustment for higher-level effects of

clinician and clinic. All variables in the

TABLE 1Nonmissing Values of Patient Characteristics in Observations Excluded

From, and Observations Included in, the Analysis Dataset

Abbreviations: CI, confidence interval; HMO, health maintenance organization; OHS, overall health status; PPO, preferred provider organization.

*Values are mean (95% CI) differences for numeric variables, or differences in proportions (95% CI) for categorical variables.Values are meanSD.Values are n (%).

n Value n Value Difference*

Age, y 58541 49.116.2 7970 48.416.5 0.7 (0.3, 1.1)

Baseline OHS score 58553 48.011.6 7971 48.911.0 0.9 (1.2, 0.7)

Discharge OHS score 23234 60.616.6 5163 64.117.2 3.5 (4.1, 3.1)

Number of visits, n 32724 8.46.4 7943 7.45.0 1.0 (0.8, 1.1)

Duration of care, d 31987 31.825.9 7690 29.322.9 2.5 (1.9, 3.1)

Male 58519 23140 (39.5) 7969 3158 (39.6) 0.1 (1.2, 1.1)

Symptom duration 58515 7970

Acute 12508 (21.4) 2182 (27.4) 6.0 (7.0, 5.0)

Subacute 14241 (24.3) 2107 (26.4) 2.1 (3.1, 1.1)

Chronic 31766 (54.3) 3681 (46.2) 8.1 (7.0, 9.3)Exercise 34589 7762

3 or more times per wk 11706 (33.8) 2634 (33.9) 0.1 (1.3, 1.1)

1 to 2 times per wk 8624 (24.9) 2027 (26.1) 1.2 (2.3, 0.1)

Seldom or never 14259 (41.2) 3101 (39.9) 1.3 (0.1, 2.5)

Took prescription medications on admission 34622 24111 (69.6) 7783 5353 (68.8) 0.8 (0.3, 2.0)

Referral source 8359 7971

Primary care 3737 (44.7) 4515 (56.6) 11.9 (13.5, 10.4)

Specialist 3436 (41.1) 2392 (30.0) 11.1 (9.6, 12.6)

Occupational medicine 498 (6.0) 1064 (13.4) 7.4 (8.3, 6.5)

Nonphysician 688 (8.2) 0 (0.0) 8.2 (7.6, 8.8)

Payer source 34661 7952

Fee-for-service 2071 (6.0) 579 (7.3) 1.3 (1.9, 0.7)Litigation 220 (0.6) 52 (0.6) 0.0 (0.1, 0.2)

Medicaid 1867 (5.4) 297 (3.7) 1.7 (1.2, 2.1)

Medicare 7629 (19.4) 1289 (16.2) 3.2 (2.3, 4.1)

Self-pay 279 (0.8) 81 (1.0) 0.2 (0.5, 0.0)

HMO or PPO 17422 (50.3) 3765 (47.4) 2.9 (1.7, 4.1)

Workers compensation 4882 (14.1) 1589 (20.0) 5.9 (6.9, 4.9)

Other 1191 (3.4) 300 (3.7) 0.3 (0.8, 0.1)

Excluded Observations (n = 58 553) Included Observations (n = 7971)


po



5/11


individual-level models were retained in

the multilevel models. All analyses were

performed with SAS Version 9.1 (SAS

Institute Inc, Cary, NC), using PROC

MIXED for multilevel analysis with= .05.

RESULTS

The acquired FOTO data set in-

cluded 66 524 observations, of

which 7971 met inclusion criteria for

analysis. TABLE 1 displays the descriptive

variables in the excluded observations

and those included for analyses. Two vari-

ables for which there was no significant

difference between observations includedin and excluded from the analyses were

sex and taking prescription medications

on admission. For all other variables, the

95% CIs indicated that there were sig-

nificant differences in the characteristics

for those included in and excluded from

analyses. Due to the large sample size,

some of these differences were small and

of no practical importance. Potentially

important differences were seen in dis-

charge OHS (included observations had

higher mean scores), symptom duration(included observations were more likely

to be acute and less likely to be chronic),

referral source (referrals from included

observations were more likely to be from

PCPs or occupational medicine physi-

cians and less likely to be from specialist

physicians), and workers compensation

payer source (included observations were

more likely to have workers compensa-

tion insurance).

Baseline and discharge OHS scores,

according to referral source, are illus-trated in FIGURE 2. Results of bivariate

analyses are presented in TABLES 2and 3.

The 95% CIs for the differences in means

(TABLE 2) and in proportions (TABLE 3) of

variables indicated that there were sig-

nificant differences according to referral

source in all variables. Of the 2 primary

outcomes, mean discharge OHS scores

were lowest for referrals from specialist

physicians and highest for referrals from

occupational medicine physicians. Mean

number of visits was lowest in referralsfrom occupational medicine physicians

and highest in referrals from specialist

physicians.

TABLE 4displays linear and multilevel

analyses for discharge OHS score. After

adjustment for individual-level covari-

ates and for clustering of patients within

clinician and clinic, referral source con-

tinued to be associated with discharge

OHS score. Compared to referral from

specialist physicians (reference category),

referral from PCPs was associated, onaverage, with nearly a 1.7-point increase

(95% CI: 0.73, 2.6) in discharge OHS,

and referral from occupational medi-

cine was associated, on average, with a

4.8-point increase (95% CI: 2.7, 6.9) in

discharge OHS. Other variables were as-

sociated with discharge OHS score after

multilevel analysis. Higher admission

OHS scores, exercise 3 or more days per

week, and payment by a health mainte-

nance organization or preferred provider

organization were all associated withhigher discharge OHS scores. Older age,

longer duration of care, chronic and sub-

acute symptom duration, and Medicaid

payment were all associated with lower

discharge OHS scores.

TABLE 5presents linear and multilevel

analyses for number of visits. After ad-

justment for individual-level covariates

and for clustering of patients within cli-

nician and clinic, referral source was also

associated with number of visits. Com-

pared to referral from specialist physi-

cians, referrals from primary care and

occupational medicine physicians were

associated, on average, with 0.44 (95%

CI: 0.27, 0.61) and 0.83 (95% CI: 0.44,1.2) fewer visits, respectively. Other as-

sociations with a lower number of visits

included higher admission OHS scores

and Medicaid payment. Older age, lon-

ger duration of care, use of prescription

medication on intake, and workers com-

pensation payer source were all associ-

ated with a higher number of visits.

DISCUSSION

The current findings indicatethat physician referral status is as-

sociated with functional status on

discharge and with number of visits.

Compared to referrals from specialist

physicians, referrals from both primary

care and occupational medicine physi-

cians were associated with higher dis-

charge OHS scores, indicating better

function, and fewer visits. Associations

seen in bivariate analyses persisted after

adjustment for important individual-,

therapist-, and clinic-level variables, andwere stronger for occupational medicine

physicians than for PCPs relative to spe-

cialist physicians.

Our results may be compared with

previous investigations that included

referral source in analyses of physical

therapy outcomes using FOTO data.

Deutscher and associates7found that re-

ferral from general practitioners was as-

sociated with higher discharge functional

status relative to other physicians in Is-

raeli patients with lumbar spine impair-ment. Our results are consistent with this

finding, despite the fact that Deutscher et

al7 used a different referral-source clas-

sification. Our findings are also in agree-

ment with Resnik and colleagues,33who

also found that referral from PCPs was

associated with better discharge function

(higher OHS scores). In contrast to our

study, Resnik and colleagues33also found

that referral from orthopaedists was as-

sociated with higher discharge OHS

40

Baseline OHS

Primary care

Discharge OHS

45

50

55

60

65

70

75

Occ med

Specialist

FIGURE 2. Unadjusted baseline and discharge OHS

scores according to referral source. Abbreviations:

Occ med, occupational medicine; OHS, overall health

status.


po



6/11


[RESEARCHREPORT]

scores, and that referral from occupation-

al medicine physicians was not associated

with discharge OHS scores. Regarding

number of visits, only Resnik and associ-

ates33modeled this variable, and they did

not find an association between referral

source and number of visits per treat-

ment episode. These previous studies

included referral source as explanatory

variables; however, the meaning of re-

lationships between referral source and

outcomes was not discussed. Variations

between the current findings and previ-

ous studies should be explored in future

research.

TABLE 2

Baseline Characteristics and Outcomes and Differences

by Referral Source: Numeric Variables

Abbreviations: Occ Med, occupational medicine; OHS, overall health status.

*Values are n, meanSD.Values are mean (95% confidence interval).

Primary Care

(n = 4515) Specialist (n = 2392) Occ Med (n = 1064)

Primary Care,

Specialist Primary Care, Occ Med Specialist, Occ Med

Age, y 4514, 48.816.7 2392, 51.716.4 1064, 39.411.4 2.9 (3.8, 1.9) 9.4 (8.1, 10.7) 12.3 (10.9, 13.7)

Baseline OHS score 4515, 49.711.2 2392, 48.611.0 1064, 46.59.7 1.1 (0.5, 1.8) 3.2 (2.3, 4.1) 2.1 (1.1, 3.0)

Discharge OHS score 2865, 64.816.5 1549, 60.616.5 749, 68.919.6 4.2 (2.9, 5.4) 4.1, (5.8, 2.5) 8.3 (10.1, 6.6)

Number of visits 4365, 7.14.6 2310, 8.45.4 974, 6.64.3 1.3 (1.6, 1.0) 0.5 (0.1, 0.9) 1.8 (1.4, 2.2)

Duration of care, d 4374, 29.422.7 2324, 33.223.5 992, 19.919.7 3.8 (5.1, 2.4) 9.5 (7.7, 11.4) 13.3 (11.3, 15.3)

Differences Among Referral SourcesTotal Nonmissing*

TABLE 3

Baseline Characteristics and Differences

by Referral Source: Categorical Variables

Abbreviations: HMO, health maintenance organization; Occ Med, occupational medicine; PPO, preferred provider organization.

*Values are n (%).Values are percentages (95% confidence interval).

Differences Among Referral SourcesTotal Nonmissing*

Primary Care

(n = 4515) Specialist (n = 2392) Occ Med (n = 1064)

Primary Care,

Specialist

Primary Care,

Occ Med Specialist, Occ Med

Sex (female) 2888 (64.0) 1482 (62.0) 441 (41.5) 2.0 (0.4, 4.4) 22.5 (19.3, 25.8) 20.5 (17.0, 24.1)

Symptom duration

Acute 1109 (24.6) 300 (12.5) 773 (72.7) 12.1 (10.2, 13.9) 48.1 (51.0, 45.1) 60.1 (63.1, 57.1)

Subacute 1267 (28.1) 638 (26.7) 202 (19.0) 1.4 (0.8, 3.6) 9.1 (6.4, 11.8) 7.7 (4.7, 10.6)

Chronic 2138 (47.4) 1454 (60.8) 89 (8.4) 13.4 (15.9, 11.0) 39.0 (36.8, 41.2) 52.4 (49.9, 55.0)

Exercise

3 or more times per wk 1529 (34.6) 827 (35.5) 278 (27.3) 0.9 (3.3, 1.5) 7.4 (4.3, 10.4) 8.2 (4.9, 11.6)

1 to 2 times per wk 1173 (26.6) 572 (24.6) 282 (27.7) 2.0 (0.2, 4.2) 1.1 (4.1, 1.9) 3.1 (6.4, 0.1)

Seldom or never 1712 (38.8) 930 (39.9) 459 (45.0) 1.1 (3.6, 1.3) 6.3 (9.6, 2.9) 5.1 (8.8, 1.5)

Took prescription medications

on admission

2879 (65.1) 1589 (68.0) 885 (86.7) 2.9 (5.2, 0.5) 21.6 (24.1, 19.1) 18.7 (21.5, 15.9)

Payer source

Fee-for-service 398 (8.8) 180 (7.6) 1 (0.1) 1.2 (0.1, 2.6) 8.7 (7.9, 9.6) 7.5 (6.4, 8.5)

Litigation 43 (1.0) 8 (0.3) 1 (0.1) 0.6 (0.3, 1.0) 0.9 (0.5, 1.2) 0.2 (0.1, 0.5)

Medicaid 236 (5.2) 58 (2.4) 3 (0.3) 2.8 (1.9, 3.7) 4.9 (4.2, 5.7) 2.1 (1.5, 2.9)

Medicare 764 (16.9) 524 (22.0) 1 (0.1) 5.1 (7.1, 3.1) 16.8 (15.7, 18.0) 21.9 (20.2, 23.6)

Self-pay 52 (1.1) 29 (1.2) 0 (0.0) 0.1 (0.6, 0.5) 1.2 (0.8, 1.5) 1.2 (0.8, 1.7)

HMO or PPO 2523 (56.0) 1221 (51.3) 21 (2.0) 4.7 (2.2, 7.1) 54.0 (52.3, 55.7) 49.3 (47.2, 51.5)

Workers compensation 301 (6.7) 252 (10.6) 1036 (97.5) 3.9 (5.3, 2.5) 90.8 (92.0, 89.6) 86.9 (88.4, 85.3)

Other 192 (4.3) 108 (4.5) 0 (0.0) 0.2 (1.3, 0.7) 4.3 (3.7, 4.8) 4.5 (3.7, 5.4)


po



7/11

journal of orthopaedic &sports physical therapy | volume 42 | number 8 | august 2012 | 711

The parameter estimates for discharge

OHS and number of visits associated

with referral source represent differences

in outcomes that may appear small whenviewed through the lens of clinical treat-

ment effects for individual patients; how-

ever, from an epidemiologic perspective,

these differences may reflect important

benefits in function and resource utiliza-

tion across a population. Furthermore,

this investigation seeks to inform about

the potential impact of an important

element of the physical therapy referral

process rather than treatment effective-

ness. From this health services utilization

perspective, the magnitude of the asso-

ciations disclosed by our findings may be

important, as little is currently known

about the impact of referral source onphysical therapy outcomes. Future stud-

ies may build on these findings, further

illuminating the referral pathway and

factoring in current trends such as direct

access to physical therapy.

In the current study, differences in

the distribution of symptom duration

among the 3 physician referral sources

are evident. Most referrals from occu-

pational medicine physicians (72.7%)

were in the acute phase of recovery and

thus more likely to make gains during

treatment. Most referrals from specialist

physicians (60.8%), however, were in the

chronic phase of recovery and would notbe expected to achieve similar functional

gains.4,12,33Referrals from PCPs occupied

a middle ground in terms of the distribu-

tion of symptom duration as well as func-

tional improvement.

Symptom duration on admission is an

important predictor of functional status

outcomes in LBP.7,12,33Delays in physical

therapy referral may shift acute episodes

of LBP to subacute or chronic episodes.

Early referral to physical therapy, that is,

TABLE 4 Individual-Level and Multilevel Models for Discharge OHS*

Abbreviations: HMO, health maintenance organization; OHS, overall health status; PPO, preferred provider organization.

*The individual-level model included the variables referral source, baseline OHS, age, sex, symptom duration, taking medications on admission, exercise

status, payer source, and duration of care. The multilevel model also included clinician and clinic effects as random effect variables.Reference category is specialist physician.Reference category is acute duration of symptoms.Reference category is no exercise.Reference category is fee-for-service payer source.

Variable Estimate SE PValue Estimate SE PValue

Intercept 42.3800 1.8114


8/11


[RESEARCHREPORT]

when patients are in the acute phase of

symptom duration, has been associated

with improved outcomes in individuals

with LBP

2,8,13,29,30

and associated withdecreased recurrence of LBP episodes.21

Symptom duration was also related

to referral source in the current study,

which suggests that the timing of physi-

cal therapy referral may vary according

to referral source. Optimal referral path-

ways for patients with LBP have not been

identified. PCPs manage the majority of

their patients with LBP without refer-

ral,9but they may also refer patients to

other specialist physicians, rather than

to physical therapists.35,36Whether a pa-

tient is referred to physical therapy for

LBP may depend on factors such as sex,

age, and patient expectations for physi-cal therapy services.11 Future investiga-

tions should identify patients who may

benefit from early physical therapy refer-

ral from a PCP, and those who should see

a specialist physician prior to physical

therapy referral. Distinguishing patients

who may directly access physical therapy

from those who need prior physician con-

sultation is also a topic that should be ex-

plored in future research.

The current investigation is informa-

tive regarding the relationship between

payer source and physician referral

source. Over 97% of occupational medi-

cine referrals were paid through work-ers compensation (TABLE 3). In multilevel

analyses, occupational medicine physi-

cian status was associated with fewer

visits, yet workers compensation payer

source was associated with more visits

(TABLE 5). This apparently paradoxical

result was explored by examining the

number of visits within the workers com-

pensation stratum. This analysis revealed

that referrals by occupational medicine

physicians were treated, on average, in

TABLE 5 Individual-Level and Multilevel Models for Number of Visits*

Abbreviations: HMO, health maintenance organization; OHS, overall health status; PPO, preferred provider organization.

*The individual-level model included the variables referral source, baseline OHS, age, sex, symptom duration, taking medications on admission, exercise

status, payer source, and duration of care. The multilevel model also included clinician and clinic effects as random effect variables.Reference category is specialist physician.Reference category is acute duration of symptoms.Reference category is no exercise.Reference category is fee-for-service payer source.

Variable Estimate SE PValue Estimate SE PValue

Intercept 3.3323 0.3228


9/11


6.5 visits, whereas referrals by PCPs and

specialists were treated, on average, in

8.5 and 10.6 visits, respectively (analysis

available on request). Thus, occupation-

al medicine physicians, referring mostlythrough the workers compensation sys-

tem, utilized fewer physical therapy visits

when compared to PCPs and specialists.

These findings suggest that occupational

medicine physicians may function under

different reimbursement circumstances

from those of primary care or specialist

physicians, and that they may also have

different clinical obligations, such as

to minimize lost work time for injured

workers. In contrast, reimbursement for

primary care and specialist physicianservices was more broadly distributed

among the various payer sources, though

the majority of the reimbursements were

through health maintenance organiza-

tions or preferred provider organizations.

The extent to which bias might have

been introduced during the selection of

observations included in the analyses

warrants discussion. Because of the large

sample size, some differences in baseline

characteristics or outcomes between in-

cluded and excluded were statisticallysignificant but small and clinically unim-

portant. For other characteristics, differ-

ences in included and excluded data sets

may be meaningful. Among the 2 out-

come variables, the observations includ-

ed in the analyses had mean OHS scores

that were, on average, 3.5 points higher

than those of observations excluded from

the analyses. Because no minimal clini-

cally important improvement has been

established for the OHS, we are unable to

use that as a benchmark to assess the im-portance of this difference. Furthermore,

because minimal clinically important

improvement for a given outcome may

be sensitive to baseline value, as well as

comorbidity status and demographic at-

tributes,38it is less likely to be informa-

tive as an index of clinical importance in

a broad population. In addition, included

observations had, on average, 1 fewer visit

than excluded observations. It is possible

that included observations had fewer vis-

its because they had shorter episodes of

care, or were less likely to have chronic

symptom duration. Differences in the

distribution of referral source and other

predictor variables, such as symptom du-ration and workers compensation payer

source, were also seen in the included and

excluded data sets. The degree to which

these factors may be interrelated as po-

tential confounders may be explored in

future research.

Another potential source of selection

bias resulted from including only those

observations with complete data at both

admission and discharge. For example, it

is possible that clinicians or clinics chose

better-performing patients to completethe discharge assessment, resulting in

selective inclusion of higher performers

in analyses of OHS outcome. In our ex-

perience with the FOTO administration,

however, it is more likely that noncomple-

tion was a random rather than a system-

atic occurrence, which would neutralize

any bias that might have been introduced

by exclusion of noncompleters. In addi-

tion, selection bias that might have been

introduced by characteristics of clinics

or clinicians would have been mitigatedby multilevel adjustment for clustering

within clinicians and clinics in the analy-

sis. The observed differences in distribu-

tion of predictor and outcome variables

in the included and excluded data sets

might also have affected external valid-

ity, limiting generalizability of the current

findings to, for example, those patients

who completed a course of care for LBP.

The explanatory power of the analyses

is indicated by the coefficient of determi-

nation (R2

) of the individual-level modelsfitted for each outcome. For number of

visits, R2 was 0.523, and for discharge

OHS scores, R2was 0.262, meaning that

each model explained 52.3% and 26.2%

of the variability in the respective out-

comes. The strongest predictor of the

number of visits was duration of the epi-

sode of care, based on the change in R2

(decreased to 0.073) when the variable

for duration of care was removed from

the model. Thus, referral source, though

it remained significant in the individual-

level multivariate analysis for number

of visits, was a relatively weak predictor.

As a predictor, referral source was more

robust in the model for discharge OHSscore, as the value of R2 changed rela-

tively little when other variables were re-

moved from the model (analysis available

upon request). We reported only the R2

values for the individual-level models be-

cause there is no comparable and readily

interpretable analog to R2 in multilevel

analyses.

Limitations

This study has several limitations. The

current findings reveal associations be-tween referral source and important out-

comes of physical therapy care but do not

fully explain the reasons for these asso-

ciations, nor do they establish cause and

effect. Other factors not measured in the

current study have been shown to contrib-

ute to variation in outcomefor example,

patient expectation of improvement, and

clinic and therapist characteristics.2,7,37

Data regarding the length of time from

physician visit to initiation of physical

therapy were not available, nor were weable to determine whether a patient saw

more than 1 physician (eg, a PCP, then a

specialist) prior to physical therapy refer-

ral. Knowing this would have permitted a

better description of the physical therapy

referral. Number of visits has been used

as an indicator of resource utilization by

previous investigators,2,17,33but number of

visits per episode of care is a crude mea-

sure of physical therapy resources and

does not capture other aspects of patient/

client management, such as administra-tive time and expense or the length and

cost of interventions or examination pro-

cedures. Also, patients who underwent

surgery for the current episode of care

were excluded from our analyses; there-

fore, our findings cannot be generalized

to patients who had surgery immediately

prior to their current episode of care. We

also did not know if observations rep-

resented patients who had undergone

prior surgery, or if the current episode


po



10/11


11/11

journalof orthopaedic& sports physical therapy | volume 42 | number 8 | august 2012 | 715

MORE INFORMATIONWWW.JOSPT.ORG@

19. Hart DL, Tepper S, Lieberman D. Changes

in health status for persons with wrist or

hand impairments receiving occupational

therapy or physical therapy.Am J Occup Ther.

2001;55:68-74.

20. Hart DL, Wright BD. Development of an index of

physical functional health status in rehabilita-

tion.Arch Phys Med Rehabil. 2002;83:655-665.

21. Hides JA, Jull GA, Richardson CA. Long-term

effects of specific stabilizing exercises for first-

episode low back pain. Spine (Phila Pa 1976).

2001;26:E243-248.

22. Jette AM, Delitto A. Physical therapy treatment

choices for musculoskeletal impairments. Phys

Ther. 1997;77:145-154.

23. Jette DU, Jette AM. Physical therapy and health

outcomes in patients with knee impairments.

Phys Ther. 1996;76:1178-1187.

24. Jette DU, Jette AM. Physical therapy and health

outcomes in patients with spinal impairments.

Phys Ther. 1996;76:930-941; discussion 942-945.25. Jewell DV, Riddle DL. Interventions that increase

or decrease the likelihood of a meaningful

improvement in physical health in patients with

sciatica. Phys Ther. 2005;85:1139-1150.

26. Jewell DV, Riddle DL, Thacker LR. Interventions

associated with an increased or decreased likeli-

hood of pain reduction and improved function in

patients with adhesive capsulitis: a retrospec-

tive cohort study. Phys Ther. 2009;89:419-429.

http://dx.doi.org/10.2522/ptj.20080250

27. Luo X, Pietrobon R, Sun SX, Liu GG, Hey L.

Estimates and patterns of direct health care

expenditures among individuals with back pain

in the United States. Spine (Phila Pa 1976).

2004;29:79-86.http://dx.doi.org/10.1097/01.BRS.0000105527.13866.0F

28. Martin BI, Deyo RA, Mirza SK, et al. Expenditures

and health status among adults with back and

neck problems. JAMA. 2008;299:656-664.

http://dx.doi.org/10.1001/jama.299.6.656

29. Nordeman L, Nilsson B, Moller M, Gunnarsson

R. Early access to physical therapy treatment for

subacute low back pain in primary health care: a

prospective randomized clinical trial. Clin J Pain.

2006;22:505-511. http://dx.doi.org/10.1097/01.

ajp.0000210696.46250.0d

30. Pinnington MA, Miller J, Stanley I. An evalua-

tion of prompt access to physiotherapy in the

management of low back pain in primary care.

Fam Pract. 2004;21:372-380. http://dx.doi.

org/10.1093/fampra/cmh406

31. Resnik L, Hart DL. Using clinical outcomes to

identify expert physical therapists. Phys Ther.

2003;83:990-1002.

32. Resnik L, Liu D, Hart DL, Mor V. Benchmarking

physical therapy clinic performance: statistical

methods to enhance internal validity when using

observational data. Phys Ther. 2008;88:1078-

1087. http://dx.doi.org/10.2522/ptj.20070327

33. Resnik L, Liu D, Mor V, Hart DL. Predictors ofphysical therapy clinic performance in the treat-

ment of patients with low back pain syndromes.

Phys Ther. 2008;88:989-1004.http://dx.doi.

org/10.2522/ptj.20070110

34. Ritzwoller DP, Crounse L, Shetterly S, Rublee D.

The association of comorbidities, utilization and

costs for patients identified with low back pain.

BMC Musculoskelet Disord. 2006;7:72.http://

dx.doi.org/10.1186/1471-2474-7-72

35. Robert G, Stevens A. Should general practitio-

ners refer patients directly to physical thera-

pists? Br J Gen Pract. 1997;47:314-318.

36. Starfield B, Forrest CB, Nutting PA, von Schrader

S. Variability in physician referral decisions. J Am

Board Fam Pract. 2002;15:473-480.37. Swinkels IC, Hart DL, Deutscher D, et al.

Comparing patient characteristics and treat-

ment processes in patients receiving physical

therapy in the United States, Israel and the

Netherlands: cross sectional analyses of

data from three clinical databases. BMC

Health Serv Res. 2008;8:163. http://dx.doi.

org/10.1186/1472-6963-8-163

38. Wang YC, Hart DL, Stratford PW, Mioduski JE.

Baseline dependency of minimal clinically

important improvement. Phys Ther. 2011;91:675-

688. http://dx.doi.org/10.2522/ptj.20100229

39. Ware J, Jr., Kosinski M, Keller SD. A 12-Item

Short-Form Health Survey: construction of scales

and preliminary tests of reliability and validity.

Med Care. 1996;34:220-233.

40. Ware JE, Kosinski M, Keller SD. SF-12: How to

Score the SF-12 Physical and Mental Health

Summary Scales. 2nd ed. Boston, MA: The

Health Institute, New England Medical Center;

1995.

41. Ware JE, Kosinski M, Keller SD. SF-36 Physical

and Mental Health Summary Scales: A Users

Manual. 5th ed. Boston, MA: The Health Institute,

New England Medical Center; 1994.

42. Ware JE, Jr., Sherbourne CD. The MOS 36-item

short-form health survey (SF-36). I. Concep-

tual framework and item selection. Med Care.

1992;30:473-483.

43. Weiner DK, Kim YS, Bonino P, Wang T.

Low back pain in older adults: are we

utilizing healthcare resources wisely?

Pain Med. 2006;7:143-150. http://dx.doi.

org/10.1111/j.1526-4637.2006.00112.x

44. Werneke MW, Hart DL, Cutrone G, et al. Associa-

tion between directional preference and central-

ization in patients with low back pain. J Orthop

Sports Phys Ther. 2011;41:22-31. http://dx.doi.

org/10.2519/jospt.2011.3415

BROWSECollections of Articles onJOSPTs Website

The Journals website (www.jospt.org) sorts published articles into morethan 50 distinct clinical collections, which can be used as convenient entrypoints to clinical content by region of the body, sport, and other categoriessuch as differential diagnosis and exercise or muscle physiology. In eachcollection, articles are cited in reverse chronological order, with the mostrecent first.

In addition, JOSPToffers easy online access to special issues and features,including a series on clinical practice guidelinesthat are linked to theInternational Classification of Functioning, Disability and Health. Pleasesee Special Issues & Featuresin the right-hand column of the Journal

websites home page.


po

http://www.jospt.org/http://dx.doi.org/10.2522/ptj.20080250http://dx.doi.org/10.1097/01.BRS.0000105527.13866.0Fhttp://dx.doi.org/10.1097/01.BRS.0000105527.13866.0Fhttp://dx.doi.org/10.1001/jama.299.6.656http://dx.doi.org/10.1097/01.ajp.0000210696.46250.0dhttp://dx.doi.org/10.1097/01.ajp.0000210696.46250.0dhttp://dx.doi.org/10.1093/fampra/cmh406http://dx.doi.org/10.1093/fampra/cmh406http://dx.doi.org/10.2522/ptj.20070327http://dx.doi.org/10.2522/ptj.20070110http://dx.doi.org/10.2522/ptj.20070110http://dx.doi.org/10.1186/1471-2474-7-72http://dx.doi.org/10.1186/1471-2474-7-72http://dx.doi.org/10.1186/1472-6963-8-163http://dx.doi.org/10.1186/1472-6963-8-163http://dx.doi.org/10.2522/ptj.20100229http://dx.doi.org/10.1111/j.1526-4637.2006.00112.xhttp://dx.doi.org/10.1111/j.1526-4637.2006.00112.xhttp://dx.doi.org/10.2519/jospt.2011.3415http://dx.doi.org/10.2519/jospt.2011.3415http://dx.doi.org/10.2519/jospt.2011.3415http://dx.doi.org/10.2519/jospt.2011.3415http://dx.doi.org/10.1111/j.1526-4637.2006.00112.xhttp://dx.doi.org/10.1111/j.1526-4637.2006.00112.xhttp://dx.doi.org/10.2522/ptj.20100229http://dx.doi.org/10.1186/1472-6963-8-163http://dx.doi.org/10.1186/1472-6963-8-163http://dx.doi.org/10.1186/1471-2474-7-72http://dx.doi.org/10.1186/1471-2474-7-72http://dx.doi.org/10.2522/ptj.20070110http://dx.doi.org/10.2522/ptj.20070110http://dx.doi.org/10.2522/ptj.20070327http://dx.doi.org/10.1093/fampra/cmh406http://dx.doi.org/10.1093/fampra/cmh406http://dx.doi.org/10.1097/01.ajp.0000210696.46250.0dhttp://dx.doi.org/10.1097/01.ajp.0000210696.46250.0dhttp://dx.doi.org/10.1001/jama.299.6.656http://dx.doi.org/10.1097/01.BRS.0000105527.13866.0Fhttp://dx.doi.org/10.1097/01.BRS.0000105527.13866.0Fhttp://dx.doi.org/10.2522/ptj.20080250http://www.jospt.org/

referral source and outcomes of physical therapy care in patients with low back pain _ journal of...

Documents