Stagnant Physical Therapy Referral Rates Alongside Rising Opioid Prescription Rates in Patients With Low Back Pain in the United States 1997–2010

Abstract

Study Design.

A cross-sectional observational study utilizing the National Ambulatory and National Hospital Ambulatory Medical Care Surveys between 1997 and 2010.

Objective.

The aim of this study was to characterize national physical therapy (PT) referral trends during primary care provider (PCP) visits in the United States.

Summary of Background Data.

Despite guidelines recommending PT for the initial management of low back pain (LBP), national PT referral rates remain low.

Methods.

Race, ethnicity, age, payer type, and PT referral rates were collected for patients aged 16 to 90 years who were visiting their PCPs. Associations among demographic variables and PT referral were determined using logistic regression.

Results.

Between 1997 and 2010, we estimated 170 million visits for LBP leading to 17.1 million PT referrals. Average proportion of PCP visits associated with PT referrals remained stable at about 10.1% [odds ratio (OR) 1.00, 95% confidence interval (95% CI) 0.96–1.04)], despite our prior finding of increasing number of visits associated with opioid prescriptions in the same timeframe.

Lower PT referral rates were observed among visits by patients who were insured by Medicaid (OR 0.48, 95% CI 0.33–0.69) and Medicare (OR 0.50, 95% CI 0.35–0.72). Furthermore, visits not associated with PT referrals were more likely to be associated with opioid prescriptions (OR 1.69, 95% CI 1.22–2.35).

Conclusion.

Although therapies delivered by PTs are promoted as a first-line treatment for LBP, PT referral rates remain low. There also exist disparately lower referral rates in populations with more restrictive health plans and simultaneous opioid prescription. Our findings provide a broad overview to PT prescription trend and isolate concerning associations requiring further explorations.

Low back pain (LBP) has a lifetime incidence of 11% to 80% in Americans.^1,2 It accounts for one of the leading reasons for physician visits, health care utilization,^3,4 and disability reported globally.⁵ Concerningly, the health care burden of LBP is likely to rise, as studies show that the overall prevalence of LBP has been increasing in the last decades.^3,6

Multiple national and international guidelines have recommended education, exercise therapy, spinal manipulation, acupuncture, massage, and psychological therapies as initial management strategies for LBP, especially given their benign safety profiles.⁷ Physical therapists (PTs) are among those able to prescribe the recommended tailored exercises, manual therapy (massage, mobilization, manipulation) as well as individualized education, though individual PTs’ treatment choices can vary from evidence-based practice recommendations.⁸ Evidence suggests that earlier referral to PT can lead to sooner return to work,⁹ and PT referral in general might decrease utilization of prescriptions, magnetic resonance imaging (MRI), and injections resulting in lower overall medical costs in the treatment of LBP.^10–13 As such, multiple national guidelines have recommended forms of PT for the management of LBP.^14,15

Despite such evidence in support of PT, national referral rates are estimated to be only 7% to 20%.^10,11 Recently, Mafi et al.¹⁶ found that although the number of visits for LBP is increasing, the overall national PT referral trends from 1999 to 2010 have remained stable. This is in sharp contrast to increasing opioid prescription in the same population,^16,17 despite the lack of evidence and guidelines supporting the use of opioids for the treatment of chronic LBP. However, the influence of demographic and socioeconomic variables on PT referral patterns remains unexplored. Here, we used the same national surveys to further characterize patterns of PT referrals during clinic visits with primary care providers (PCPs).

MATERIALS AND METHODS

Data Source

Our analysis utilized data from the national surveys National Hospital and Ambulatory Medical Center Survey (NHAMCS) and National Ambulatory Medical Center Survey (NAMCS) administered by the National Center for Health Statistics (NCHS) as well as sample survey data obtained from Emergency Departments (EDs). These two surveys are designed to collect nationally representative information about ambulatory medical care services in the United States. Data are obtained for a sample of visits to nonfederal-employed, office-based physicians (NAMCS), and outpatient departments (NHAMCS). Specially trained interviewers from the Center for Disease Control and Prevention (CDC) visit the physicians and conduct training before survey administration. Each physician is randomly assigned to a 1-week reporting period, during which a random sample of visits are recorded by the physician or office staff. Data obtained include the patients’ symptoms, physicians’ diagnoses, and medications ordered or provided; demographics; and ambulatory services provided, including diagnostic procedures, patient management, and planned future treatment.

For this study, we analyzed survey data from clinics collected in NHAMCS and NAMCS between the years 1997 and 2010. Only visits for patients between ages 16 and 90 years were considered. Complete data were considered because missing data fields were rare.

Back Pain Visits

For outpatient clinic surveys, ICD-9 codes indicating a diagnosis of back pain were utilized to examine the incidence of back pain related diagnoses as we previously detailed:¹⁷ 353.1, lumbosacral plexus lesions; 722.1, displacement of thoracic or lumbar intervertebral disk without myelopathy; 722.52, degeneration of lumbar or lumbosacral intervertebral disk; 722.83, postlaminectomy syndrome, lumbar region; 724.02, spinal stenosis, lumbar region, without neurogenic claudication; 724.2, lumbago; 724.4, thoracic or lumbosacral neuritis or radiculitis, unspecified; 724.5, backache, unspecified; 738.4, acquired spondylolisthesis; 739.3, nonallopathic lesions, lumbar region; 756.1, congenital anomalies of spine; 756.12, spondylolisthesis; 839.2, closed dislocation thoracic and lumbar vertebra; 846, sprains and strains of sacroiliac region; 847.2, sprain of lumbar; and 847.9, and sprain of unspecified site of back.

Demographic Variables

Patient race and ethnicity were summarized using the provided variables from the databases. Age was broken down into four groups: 16 to 29, 30 to 44, 45 to 59, and 60 to 90 years. This grouping was chosen to create comparably sized groups to maximize descriptive power. Gender was obtained directly from the databases. We created a three-level race/ethnicity variable of white, black, and patients of any race who identified as Hispanic.

Primary Care Physician

This study focused on visits seen by primary care physicians (PCPs). In both NHAMCS and NAMCS, the “PRIMCARE” variable was employed to analyze the subset of ambulatory visits attributable to PCPs.

Insurance Payer Source

The primary insurance payer source is reported in both NHAMCS and NAMCS. There is a slight coding change in the year 2005 that may impact the subpopulation with two or more insurances. A binary variable is created to indicate Medicaid or self-pay as the primary insurance source, and another variable created to indicate Medicare as the primary insurance source. Medicare is the federal United States health insurance for patients 65 years of age or older or younger patients with certain disabilities and end-stage renal disease. Medicaid is a joint federal and state program that helps with medical costs for some people with limited income and resources.

Opioid Prescription

As previously described in more detail,¹⁷ similar procedures were adopted for all ambulatory settings to code for the prescription of opioid medications. As previously explained, for 2006 to 2010, the prescription category variables were used, and values based on Multum’s therapeutic classification “CNS; analgesics; narcotic” and “CNS; analgesics; narcotic analgesic combinations” were employed to indicate the prescription of opioid medications. For 1997 to 2005, National Drug Code Directory, 1995 edition, drug classification was employed, with the value “analgesics-narcotic” indicating prescription of opioid medications. Manual inspection identified occasional discrepancies in coding, which were reconciled before analyses.

Physical Therapy

Physical therapy referral was available and coded as “physioth” between 1997 and 2004 and as “PT” between 2005 and 2010. The surveys did not state specific definitions for physical therapy.

Statistical Modeling

NAMCS and NHAMCS nonrandomly sampled clusters of practices, clinics, and hospitals. Hence, frequency of visits associated with LBP, physical therapy referral, demographic variables, and payer source were analyzed using weighted survey frequency estimation procedures in SAS version 9.2 (SAS Institute Inc., Cary, NC) released by the National Center for Health Statistics. To ensure that no additional bias was added during the analysis of the subpopulations of interest, domain analysis was employed to compute statistics for subpopulations of interest by using the entire sample to estimate the variance of domain estimates. Initially, a univariate analysis was used to estimate the odds ratio (OR) of each covariate—year (as compared with 1997), female gender (as compared with male gender), black race (as compared with white race), Hispanic ethnicity (as compared with non-Hispanic), Medicare (as compared with private pay), and Medicaid (as compared with private pay). Subsequently, a multivariable analysis using weighted logistic regression was used to estimate the adjusted ORs while controlling for other significant covariates (P < 0.05)—year, gender, race, ethnicity, payer status, opioids. The outcome was the probability of a given visit leading to a physical therapy referral.

RESULTS

Between 1997 and 2010, we estimated 170 million visits for LBP leading to 17.1 million PT referrals (Table 1). The average age of patients seen at visits for LBP with PT referral was 50.4 years; 84.5 million (49.3%) were female; 19.8 million were White, 12.4 million were Black, and 2.3 million were Hispanic. About 22.3 million patients had Medicare, and 2.0 million had Medicaid as the primary insurance (Table 2).

TABLE 1.

Trend of PT Referral From PCP Visits for LBP

	# LBP Visits	# PT Referral	Percentage of # LBP
Total	169,666,459	17,147,184	10.1
1997	6,933,323	961,963	13.9
1998	8,589,861	1,120,063	13.0
1999	8,005,355	1,001,426	12.5
2000	8,526,785	890,674	10.4
2001	10,732,439	516,051	4.8
2002	10,527,373	545,027	5.2
2003	13,052,197	1,277,625	9.8
2004	13,304,436	2,015,460	15.1
2005	13,886,533	1,862,488	13.4
2006	13,529,802	1,057,260	7.8
2007	13,853,822	1,272,089	9.2
2008	17,808,220	1,363,498	7.7
2009	16,515,770	1,808,478	11.0
2010	14,400,543	1,455,082	10.1

LBP indicates low back pain; PT, physical therapist.

TABLE 2.

Patient Characteristics of PT Referral From PCP Visits for LBP

	# LBP Visits	# PT Referral	Percentage of # PT
16–29	16,301,616	2,021,779	11.8
30–44	46,829,374	5,653,156	33.0
45–59	60,748,191	6,239,246	36.4
60–90	45,787,278	3,233,003	18.9
Female sex	96,063,528	8,450,777	49.3
Black race	496,053,166	6,405,611	37.4
Hispanic ethnicity	447,611,104	4,984,976	29.1
Medicare insurance	30,563,538	1,828,563	10.7
Medicaid insurance	29,903,167	1,958,836	11.4

LBP indicates low back pain; PT, physical therapist.

Average number of PCP visits associated with PT referrals had remained stable over this time period at about 10.1% (OR 1.00, 95% CI 0.96–1.04; Figure 1; Table 1), despite our prior finding of increasing number of visits associated with opioid prescriptions in this same population.¹⁷ Lower PT referral rates were observed among visits by patients who were insured by Medicaid (OR 0.48, 95% CI 0.33–0.69) and Medicare (OR 0.50, 95% CI 0.35–0.72) (Table 3). Furthermore, visits not associated with PT referrals were more likely to be associated with opioid prescriptions (OR 1.69, 95% CI 1.22–2.35).

Figure 1.

Temporal trends in physical therapy and opioid prescriptions associated with visits for low back pain.

TABLE 3.

Predictors of Physical Therapy Referral

Covariate	Adjusted Odds Ratio
Year (as compared with 1997)	1.00 (0.96–1.04)
Female gender	0.75 (0.57–1.00)
Black race (as compared with white race)	0.91 (0.56–1.47)
Hispanic ethnicity	1.18 (0.78–1.80)
Medicare	0.47 (0.32–0.68)
Medicaid	0.54 (0.37–0.79)
Opioid prescription	0.57 (0.41–0.80)

Odds ratios reported are associated with physical therapy referral (95% confidence interval).

DISCUSSION

The PT referral rates observed during PCP visits between 1997 and 2010 have remained stable. We noted disparities in PT referral rates during PCP visits by Medicaid and Medicare-insured patients and in patients who received simultaneous opioid prescription. Although this population-based study lacks the individual patient granularity needed to differentiate between new versus repeat visits and referrals along with individual patient differences that may explain for the trends observed, we believe that our results prompt further studies to characterize these socioeconomic disparities in PT referral.

The ideal physical therapy referral rate has yet to be defined. The 10.1% noted in our study is in the middle of the range previously described.^10,11 Fritz et al.¹⁰ noted that in a national database of employer-sponsored health plans, about 7% of patients with LBP were referred for physical therapy. However, studies of workers compensated for back injuries, who have been previously documented to be most likely referred to physical therapy, reveal referral rates as high as 18% to 27%.^6,11

One proposed explanation for the stable low rates of PT referral would be increased visits to PCP for less severe back pain. We do not agree this is the case, as in both Mafi’s analysis and prior work by our group on this same population, there has been an increasing number of visits associated with opioid prescriptions.^16,17 This observation of increased opioid rates suggest that additional visits include people with nontrivial pain and counters the argument that the stable low rates of PT referral would be increased visits to PCP for less severe back pain.

Here, we also found that patients who are not referred to PT are more likely to receive opioid prescriptions. Although our study was not equipped to answer why this association exists, this finding needs to be further validated and explored. One can conjecture that cost can be a driving issue. One systematic review suggests that 17% of all direct medical costs for lower back pain was spent on physical therapy.¹⁸ In comparison, the cost of opioid prescriptions can be as low as $22 to $109 in terms of Medicare reimbursement per prescription after visit for each person treated with back pain.¹⁹ However, although opioid prescription may be cheaper in the shorter run, the direct cost of opioids abuse has been estimated to be 8.6 billion dollars.²⁰ Indeed, as aberrant medication-taking behaviors occur in up to 24% of patients receiving opioids for chronic back pain,²¹ the rising rates of opioid prescriptions likely are contributing to the rising opioid analgesic abuse and addiction.²² With more evidence that PT is an effective method of treating back pain,^10–13 there needs to be a push for more insurance coverage of PT referrals to avoid overutilization of cheaper but less effective, and potentially harmful, forms of treatments.

The decreased access based on payer status we noted has been previously described. In other fields of medicine, payer status has been associated with disparate access to laparoscopic surgeries,^23,24 cardiac outcomes,^25,26 and even mortality.²⁷ Fewer studies have looked specifically at how payer status impact management of LBP. Freburger et al.⁶ noted that the rate of referral to physical therapy was 11.6% for Medicaid paid PCP visits as compared with 17.7% for visits paid by private insurance, Medicare, worker’s compensation, or self-pay. Previously, Fritz et al.¹⁰ have shown through analysis of an employer-sponsored health plan that patients insured through a Preferred Provider Organization (PPO) plan were more likely to receive early treatment from a PT (53.4%) as compared with those insured through a Health Maintenance Organization (HMO) plan (44.7%). The reason for these disparities is unclear. Perhaps, PCPs, who often serve as gatekeepers of care, feel obligated to be more cost-effective as historically, PCPs working for HMOs have lower mean total estimated outpatient charges than orthopedic surgeons and chiropractors when caring for patients with LBP.²⁸ These disconcerting economic disparities need to be quantified, elucidated, and addressed.

Our study is limited by the lack patient granularity, which limits our ability to differentiate new versus repeat visits for LBP and PT referrals. Although the visits analyzed were all for a primary diagnosis of LBP, we cannot be sure that patients were referred to physical therapy for other pathologies. More granular data will be needed to further characterize the concerning associations with regard to payer status.

Overall, our findings provide a broad overview of the national trend of PT referral for the management of LBP and isolate concerning associations with regard to payer status and association with opioid prescription. Our study is limited by the lack of individual patient and diagnosis granularity in the dataset, which limits our ability to differentiate new versus repeat referrals and to distinguish among a wide range of spine disorders contributing to LBP. Further research is needed to determine why these disparities in PT referral exist. The continued low rates of physical therapy referrals for patients with LBP suggest that initiatives are needed to educate both providers and patients about the utility of physical therapy.

CONCLUSION

Although therapies delivered by PTs are promoted as a first-line treatment for LBP, PT referral rates continue to be steadily low. There also exist disparately lower referral rates in populations with more restrictive health plans and simultaneous opioid prescription. Our findings provide a broad overview to PT prescription trend and isolate concerning associations requiring further explorations.

Key Points.

• In spite of national guidelines promoting physical therapy as first-line treatment for low back pain, there exist steadily low rates of physical therapy referrals in the United States for patients.

• Populations with more restrictive health plans have disparately lower rates of PT referral.

• Initiatives are needed to educate both providers and patients about the utility of physical therapy.