Access to the PPM Journal and newsletters is FREE for clinicians.
13 Articles in Volume 11, Issue #3
Advances in Cranial Electrotherapy Stimulation
Chronic Migraine: An Interactive Case History, Part 3
Cost-effectiveness Of Treatments for Low Back Pain
Electrical Me
Lessons From The Father of Electromedicine — Dr. Luigi Galvani
Medications for Chronic Pain—Nonopioid Analgesics
Pulsed Radio Frequency Energy As an Effective Pain Treatment
The Role of Body Posture In Musculoskeletal Pain Syndromes
The Role of Body Posture In Musculoskeletal Pain Syndromes
Therapeutic Laser for the Treatment of Chronic Low Back Pain
Tolerance to Opioids
Understanding Electromagnetic Treatments
Update: Clinical Challenges in the Diagnosis And Management of Fibromyalgia

Cost-effectiveness Of Treatments for Low Back Pain

Study examines the duration and cost of second-line therapies in the treatment of patients with idiopathic and herniated disc-related low back pain.
Editor's Note: It’s not often that we practitioners get to review a large database of clinical outcomes from one of the Nation’s premier health plans. Although not the usual multi-center, hard data study we are used to, the study presented here points out that an electric current facilitates resolution of low back pain. This should not come as a surprise to anyone who only has a cursory knowledge of electric current’s history of healing painful medical problems. This unusual and well done study has a simple and practical message: Electric currents can be a cost-effective addition to any treatment regimen to get enhanced results.

Low back pain (LBP) is a common, costly, and recurrent problem in the United States. There are three main classes of LBP: spinal stenosis, disc displacement or herniation, and idiopathic (non-specific), with the last class comprising 85% of the cases.1 The lifetime risk of LBP is estimated at 85% in the United States.2 A recent Medical Expenditure Panel Survey (MEPS) of adults showed that more than a quarter reported LBP within the previous 3 months.3 Recurrences are common: The one-year recurrence rates have been estimated at between 20% and 44%. The emerging picture is that of a chronic condition with acute exacerbations.1,2

In most patients, pain resolves within a few weeks or months regardless of the etiology or treatment, but in about 8% of cases, pain lasts at least 3 months. About 5% of patients have unremitting pain at 22 months.4

The burden to the healthcare system is considerable. LBP is the second-leading symptomatic cause for physician visits, the third-most-common reason for surgical procedures, and the fifth-most-common reason for hospitalizations.2 In 2005, age- and gender-adjusted expenditures among those with LBP was $2580 (73%) higher than among people without LBP.3

Many treatments are available for LBP, including physical and chiropractic therapy, several classes of pharmaceuticals, electrotherapy, surgery, and others. Healthcare spending on imaging and treatments increased by 65% between 1997 and 2005, but notably there was no concomitant increase in patients’ health or functional status.3

We performed a cost-effectiveness analysis of treatments for episodes of idiopathic LBP and LBP caused by disc herniation using managed-care claims data. The therapies included those commonly used with patients whose pain is refractory to rest and over-the-counter (OTC) drug therapy: physical therapy, occupational therapy, chiropractic care, transcutaneous electrical nerve stimulation (TENS), and opiates. We defined “cost” from the payer’s perspective (eg, reimbursements) and effectiveness as the length of the LBP episode. The costs and episode lengths associated with these three treatments were compared separately in the two diagnostic indications.

How the Study Was Performed

The perspective of this cost-effectiveness analysis (CEA) was the payer. We used 2005 managed-care claims data from a commercial vendor allied with a major healthcare system that encompasses health plans across the United States. Inpatient, outpatient, physician, and pharmacy claims were included. Following a clean period, members with the diagnoses of LBP in the following diagnostic groups were included: displaced or herniated disc (inter-vertebral disc disorders, disc degeneration, and sciatica) and idiopathic/non-specific LBP (backache, lumbago, non-allopathic lesions, strains, and sprains). Episodes were assigned to idiopathic LBP if they contained appropriate International Classification of Disease, 9th Revision (ICD-9) codes and contained no ICD-9 codes for herniated disc. Members with concurrent diagnoses of cancer, trauma, intravenous (IV) drug use, or neurological impairment were excluded.

Claims-based episode definitions are not consistent in the literature, so we constructed our own definition using previous work as a starting point.5-10 We adjusted our definition to eliminate any correlation between the number of episodes experienced by a member and the length of episodes. An episode of LBP was defined as the time from the date of the first claim with a relevant primary or secondary diagnosis to the date of the last claim with either: (1) a relevant diagnosis if the claim was for inpatient, outpatient, or physician services, or (2) a relevant National Drug Code (NDC) if the claim was for pharmaceuticals, in which case days’ supply was added to the service date to obtain the end-of-episode date. A clean period had to follow the end-of-episode date; the length of the clean period depended on the type of claim (see Table 1). Episodes with a single claim were considered to have an episode length of 0 days. Episodes with length of 0 days were distributed evenly throughout the year and were not an artifact of the use of a single year of data. (We dealt with this problem in the statistical analysis below.)

Effectiveness was defined as the length in days of the episode. (We assumed that the pain was essentially resolved when the member did not seek further treatment.) Because claims from a single calendar year were included, only episodes with an adequate clean period (indicating that the episode had ended) were included. To eliminate bias in the estimate of the length of episodes due to this requirement, we adjusted for the start date of the episode (relative to January 1, 2005) in the analysis.

For the therapy analysis, therapies during the episodes were identified by Current Procedural Terminology (CPT) codes and assigned to the following treatment groups (episodes could include treatments in more than one group): (1) Outpatient (OUT); (2) TENS/electro-acupuncture (TENS); (3) PT/Osteopathic/chiropractic (PT); (4) Opioids (OPI); and (5) NSAIDs [non-steroidal anti-inflammatory drugs] (NS). Because NSAID claims were so rare, we concluded that most NSAIDs for LBP were non-prescription uses and likely to be common to most episodes of LBP. Episodes with only treatments 1 and/or 5 were eliminated from the analysis; those with 1 and/or 5 combined with 2 and/or 3 and/or 4 were included and analyzed according to the combination of treatments 2, 3, and 4.

The claims data set included allowed charges but not reimbursed amounts (which are payer costs). Therefore, costs per episode were estimated for hospitalizations and physician services with any of the previously listed primary or secondary diagnoses from the Medicare 2007 national average diagnosis-related group (DRG) and CPT reimbursements (based on relative value units [RVUs] or relative weights [RWs] and the 2007 unit payments). Outpatient (OP) hospital services in these commercial claims data contained no CPT codes, so it was not possible to assign an Ambulatory Patient Classification (APC)–based reimbursement to these services. Instead, we applied the Medicare 2007 outpatient cost-to-charge ratio (CCR) to the allowed charges on each OP visit. Drug costs were estimated from the pharmacy claims by first estimating the regression of allowed charges on paid charges in a small sample (n=400) of drugs used to treat LBP in which both allowed and paid charges were provided, and then using the conversion equation to estimate paid amounts from allowed charges in the full data set.

The cost-effectiveness of each treatment group was estimated in a total of 4 linear regression equations. For each diagnosis group, two equations were fit: a Tobit analysis for natural log (LN; effectiveness) (because of the high percentage of episodes with length 0: 14% among herniated disc episodes and 24% among idiopathic episodes) and a multiple regression analysis for LN (cost). Each equation started with the following regressors: episode start date (0=01JAN05), polynomials of start date, age, gender, whether the member had back surgery during the year, Charlson’s comorbidity score, treatment class(es), all 2-factor treatment interaction terms, and all 2-factor treatment by start date interaction terms. The co-variance estimates were adjusted for correlated episodes (ie, those occurring in the same member). In a standard step-down procedure, terms were eliminated in descending order of P value until the significance of each remaining was no larger than .05. The fit of the analyses were evaluated using link tests for the Tobit models and visualization of the residuals in the case of the linear regression models.

The mean cost and effectiveness of TENS, PT, and OPI in each diagnosis class were estimated at the mean covariate vector of the demographics of members with that diagnosis. The episode start date was set to 01JAN05, and the treatments were set to TENS, PT, or OPI so that estimates could be obtained for single-treatment episodes. The unconditional estimates of the mean episode lengths were obtained for the Tobit analyses. (The use of this estimate means that we assumed that the episode lengths of 0 did not represent true single-day episodes but were censored in the claims data. For example, members with shorter-than-average episodes may have found it unnecessary to return to their physician for further treatment.)

The confidence intervals for each treatment’s fitted mean were calculated, along with the differences between each set of treatments and their confidence intervals. There were a total of 6 comparisons: for herniated disc LBP: TENS versus PT, TENS versus OPI, and PT versus OPI; for idiopathic LBP: the same sets. If the difference in episode length was within 7 days, treatments were considered equivalent. If the difference in episode cost was within $100, treatments were considered equivalent. If a treatment was both significantly less expensive and produced significantly shorter episodes than another, the superior treatment was considered to dominate the inferior one. Similarly, if one treatment produced equal cost but shorter episodes, or lower cost but equal episode length, it was considered to dominate the comparison treatment. In comparisons in which neither treatment dominated the other—that is, in which one was both more expensive and more effective (shorter length) than the other—a net monetary benefit (NMB) analysis was conducted. The NMB estimates the savings achieved compared with the amount that the payer or member would have been willing to pay for the increase in effectiveness.11

Table 1. End-of-Episode Criteria

Results of the Study

Predicted episode lengths and payer costs by diagnosis and treatment group are shown in Table 2. Pair-wise comparisons of the treatments’ costs and effects are shown in Table 3.

There were 67,305 episodes of herniated disc LBP in 60,027 patients, approximately 60% of whom were between 35 and 55 years old and 50% of whom were male. Fewer than 3% had surgery during the year. A total of 751 episodes included TENS; 55,482 included PT; 18,947 included OPI.

As shown in Tables 2 and 3, TENS produced significantly shorter episodes (a reduction of 75 days compared with PT and 111 days compared with OPI). TENS was tied for least expensive with PT (a difference of less than $10 per episode) and was significantly less expensive than OPI by almost $300. PT episodes were significantly shorter than OPI by 37 days and also were significantly less expensive by $307.

Table 2. Predicted Treatment Costs and Effects

Table 3. Comparisons of Treatment Costs and Effects

There were 212,917 episodes of idiopathic LBP in 182,340 patients, slightly more than half of whom were aged between 35 and 55 years and 47% of whom were male. A total of 3,022 episodes included TENS; 175,831 included PT; 48,261 included OPI.

As shown in Tables 2 and 3, TENS produced significantly shorter episodes (a reduction of 63 days compared with PT and 48 days compared with OPI). OPI episodes were the least expensive by $207 compared with PT and $215 compared with TENS. TENS was tied in expense with PT (a difference of less than $11 per episode). OPI episodes were significantly shorter than PT by 15 days and (as stated) were also significantly less expensive, the opposite of what was found with herniated discs.

Within the equivalence limits stated in the methods section (7 days and $100), TENS dominated PT in both diagnoses. It dominated OPI in herniated disc diagnoses. PT dominated OPI in herniated disc diagnoses, but the opposite was true in idiopathic diagnoses.

The only comparison that did not have a dominant treatment was idiopathic LBP treated with TENS or OPI. The NMB of TENS versus OPI was estimated in the predicted models and in the sample of idiopathic patients by estimating their costs and effects as if they had been treated with TENS and again as if they had been treated with OPI. The results are shown in Table 4 and Figure 1 using willingness-to-pay (WTP) values ranging from $5 per day to $50 per day. The gray rows show the WTP values at which the mean estimate of the NMB is $0.00 and at which the lower 95% confidence bound on the NMB is $0.00. These are the estimates of the break-even WTP amounts.

The break-even WTP analysis using the models shows that even if payers’ or patients’ willingness to pay for reducing an episode of LBP by 48 days is less than $5 per day, TENS is still a good value compared with OPI. The analysis using the sample shows that the willingness to pay can be as high as $22.26 per day and still have 95% confidence that TENS provides good value compared with OPI (ie, the lower NMB confidence bound is $0 at this WTP). If willingness to pay is $25 per day, the greater effectiveness of TENS is valued at $1000 (model estimate) or $500 (sample estimate) per episode by comparison with actual costs.


Low back pain is now generally regarded as an episodic disease, neither acute nor chronic.7,8,12 The nature of LBP implies that a retrospective analysis of treatment effectiveness should show correlations with the duration of these episodes. Diagnoses, treatment types, patient demographics, and comorbidities were included in the analysis of episode cost and duration.

A total of 280,320 episodes were included in this study. From a patient perspective, 82% had 1 episode, 15% had 2 episodes, 3% had 3 episodes, and fewer than 0.5% had 4 to 6 episodes. This episode distribution is similar to that found by Rizwoller et al7 for a 16,567 episode sample. Their episode frequency was, however, skewed toward more numerous episodes, with 67% having 1 episode, 17% having 2 episodes, and 11% having 3 to 5 episodes. This difference is likely due to the analysis of 1 year of claims data in this report and 5 years of claims data in the case of Rizwoller et al.7 Idiopathic low back pain is estimated to account for about 85% of all cases.1 In this data set, it accounted for 74% of the episodes.

Table 4. Net Monetary Benefit of TENS versus Opioids in Patients with Idiopathic Low Back Pain

Cost of Therapy Relative to Diagnostic Category

The present study defined 2 categories of low back pain: (1) herniated disc and (2) idiopathic. These categories were chosen to be unambiguous in terms of radiographic definition, as this dictates the direction of therapy.1

Also defined were 3 categories of treatments: physical therapy (combining PT, OT, and chiropractic), opiates, and TENS.

Although we originally considered including them, the categories of outpatient visits and NSAIDs were ubiquitous and virtually non-existent, respectively, and so provided little information about treatment effectiveness. Current payer policies encourage patients to obtain NSAIDs OTC without reimbursement; thus, the true use and cost of this treatment category was likely underrepresented in the claims. We believe that the treatments compared in this study are probably used in conjunction with NSAIDs. For these reasons, the current study was limited to the conservative, second-line therapies of TENS, physical therapy, and opiates. The costs and effectiveness of these conservative therapies were analyzed.

Whether the etiology of the low back pain was idiopathic or due to herniated disc, TENS was most effective. Opiates demonstrated the least effectiveness. Opiate use was correlated with increasing the number of LBP episodes up to 42%.7 In this study, the use of opiates was associated with the longest episode durations for herniated disc and second-to-longest episode durations for idiopathic pain (physical therapy had the longest).

On the cost side, opioids were associated with a higher cost for herniated disc episodes than TENS and physical therapy, whereas for idiopathic pain, they were associated with a lower cost. TENS and physical therapy were of equal cost in both diagnoses.

Figure 1. Net Monetary Benefit of TENS Versus Opioids in Patients With Idiopathic Low Back Pain

Although this study attempted to take advantage of good analytic techniques to tease apart the complexities of the actual care that was delivered, there were limitations. Foremost was the inability to directly adjust for LBP severity and duration, which may have been a problem if it was correlated with treatment selection. We believe this potential problem was minimized for two reasons. First, we limited the study to patients who had treatments that were considered second-line therapies. This eliminated all of the episodes for which rest, NSAIDs, and the like were adequate. Second, we made a modest adjustment for disease severity by grouping different etiologies of low back pain and including the Charlson’s comorbidity index along with patient age and gender.

Another potential limitation was the inability of the models to tease apart the effects of the individual therapies. In most episodes, a combination of therapies was used. With the Tobit (length) and regression (cost) models, we attempted to understand the contribution of individual therapies to an episode’s length and cost. The result was that the predicted episode lengths of single therapies were about twice as long as the observed episode lengths, which used multiple therapies. However, the reasonable fit of the models and the correlation of episode length with its start date during 2005 (which was caused by the use of a single year’s data, and for which the model was adjusted) suggest that this prediction is real.

Overall, it appears that TENS has the potential to substantially shorten episode length and, in many cases, decrease the cost as well. It was the least frequently used of the 3 therapies and therefore represents a relatively unexploited opportunity, especially in patients currently treated with PT or opiates. Even in the case where TENS was more expensive than a competitor, its increased effectiveness was available at low incremental cost.

Last updated on: November 30, 2011
close X