The epidemiology and medical management of low back pain during ambulatory medical care visits in the United States

The concept of the National Ambulatory Medical Care Survey (NAMCS) to collect data on medical care provided in physician offices in the United States was developed over 30 years ago [13]. Detailed documentation of the NAMCS instrument, methodology, and data files that served as the basis for this study is available elsewhere [14, 15]. Patient visits were selected using a multistage probability sample design. The first stage included primary sampling units (PSUs) which consisted of counties, groups of counties, county equivalents (e.g. parishes), towns, townships, minor civil divisions, or metropolitan statistical areas (MSAs). These PSUs comprised a probability subsample of those used in the 1985–1994 National Health Interview Surveys [16]. The latter, which covered all 50 states and the District of Columbia, were stratified by demographic and socioeconomic variables and then selected with probability proportional to their size. Stratification was done within four geographic regions by MSA and non-MSA status.

The second stage consisted of a probability sample of practicing physicians selected from the master files of the American Medical Association (AMA) and American Osteopathic Association (AOA). Within each PSU, all eligible physicians were further stratified by specialty. The third stage involved selection of patient visits within the practices of participating physicians. Initially, physicians were randomly assigned to one of the 52 weeks within a calendar year. Then, a systematic random sample of patient visits was selected for each physician during the assigned week. The sampling rate of patient visits varied from a 20% sample for very large practices to 100% for very small practices as determined by a presurvey interview [17]. In this manner, data from about 30 patient visits were recorded by each physician during the assigned week.

The sampling frame for NAMCS included physicians who met the criteria of being: (1) office-based; (2) principally engaged in patient care activities; (3) nonfederally employed; and (4) not in the specialties of anesthesiology, pathology, or radiology. During 2003 and 2004, a total of 6000 physicians were initially screened. Of these, 2032 (34%) did not meet the four inclusion criteria, most commonly because the physician was retired, deceased, or employed in teaching, research, or administration. Of the remaining 3968 eligible physicians, 2779 (70%) participated in NAMCS. However, among these "participating" physicians, 544 (20%) saw no patients during their assigned reporting period because of vacations, illness, or other reasons for being temporarily not in practice. The NAMCS provides data on 25,288 patient visits to 1114 physician offices during the 2003 calendar year and 25,286 patient visits to 1121 physician offices during the 2004 calendar year.

The basic sampling unit for the NAMCS is the physician-patient encounter or "patient visit." The following types of contacts were excluded: telephone calls, visits outside the physician's office (e.g., house calls), visits made in hospital settings (unless the physician had a private office in a hospital), visits made in institutional settings that had primary responsibility for the patient's care (e.g., nursing homes), and visits to the physician's office for administrative purposes only (e.g., to leave a specimen, pay a bill, or pick up insurance forms). Each patient visit was assigned a weight based on four factors: (1) probability of being selected by the three-stage sampling design; (2) adjustment for nonresponse; (3) adjustment for physician specialty group; and (4) weight smoothing to minimize the impact of a few physician outliers whose final visit weights were large relative to those for the remaining physicians. Thus, by applying these weights to each of the 50,574 patient visits in the 2003 and 2004 NAMCS data files, an estimated 1.8 billion physician office visits in the United States were available to derive unbiased national estimates of ambulatory medical care services and to further characterize such services.

Data for the NAMCS were collected by the physician with assistance from office staff when possible. Patient record forms were used to collect the data for each selected visit. The NAMCS field staff performed completeness checks of the patient record forms prior to submission for central data processing. Detailed editing instructions were provided to reclassify or recode ambiguous or inconsistent data entries. Quality control measures, which were used to verify the accuracy of computer data entry, indicated that the mean keying error rate was 0.1% for nonmedical items and that discrepancy rates ranged from 0.0% to 1.1% for required medical items.

Item nonresponse rates were 5% or less for most variables. Major exceptions (nonresponse rate) included: ethnicity (20%), race (18%), tobacco use (30%), and time spent with physician (16%). Missing data for birth year (4%), sex (4%), race (18%), ethnicity (20%), and time spent with physician (16%) were imputed by assigning the value from a randomly selected patient record form representing another patient with similar known characteristics. Such imputations were performed according to physician specialty, geographic region (state was used instead of geographic region to impute ethnicity), and primary diagnosis codes.

This study focused on patient visits for LBP. These were initially identified using the "reason for visit" (RFV) item of the NAMCS patient record form. Specifically, patient visits were included only if back symptoms (RFV classification code number, 1905) or low back symptoms (RFV classification code number, 1910) were reported as one of the three most important reasons for the visit in the patient's own words. Subsequently, any patient visits attributed to a lump, mass, or tumor of the back or low back were excluded. Exploratory analyses of the data stratified according to RFV classification code numbers (1905 vs. 1910) and importance of symptoms (primary reason for visit vs. secondary or tertiary reason for visit) revealed few substantive differences between groups. Consequently, to maximize statistical power, all epidemiological analyses combined RFV classification code numbers 1905 and 1910 to represent prevalent cases of LBP (i.e., back symptoms, other than those attributed to a lump, mass, or tumor, were the primary, secondary, or tertiary reason for the patient visit). However, only those patient visits in which LBP was the primary reason for seeking medical care ("primary LBP patient visits") were included in the medical management analyses to minimize potential confounding by other secondary or tertiary reasons for the patient visit. Patient visits attributed to neck symptoms (RFV classification code number, 1900) exclusive of LBP were not included in the study. National population estimates derived from the NAMCS may be unreliable if they are based on fewer than 30 unweighted patient visits or if the relative standard error (standard error divided by the national population estimate) is greater than 0.30 [14, 15].

Patient sociodemographic characteristics included age, sex, race, ethnicity, geographic region, and MSA status of residence. Patient visit context characteristics included episode of care, chronicity of symptoms, and whether the visit was related to an injury, poisoning, or adverse effect (IPA). Physician provider characteristics included primary care physician status, type of physician provider (Doctor of Medicine or Doctor of Osteopathy), physician specialty, and whether multiple physicians shared responsibility for medical care of the patient. The elements of LBP medical management included any diagnostic tests, patient counseling, drugs, physiotherapy, or surgical procedures that were ordered, scheduled, or performed during the patient visit. Drugs were broadly defined as any medications or injections, including immunizations, allergy shots, anesthetics, or dietary supplements, that were ordered, supplied, administered, or continued during the visit, regardless of prescription or over-the-counter status. Up to eight drugs may have been listed on the NAMCS patient record form during a patient visit. For each drug listed, up to three therapeutic class codes were assigned based on the standard classifications used in the National Drug Code (NDC) Directory [18]. The drugs portion of this study focused exclusively on three common drugs for relief of pain (NDC code, 1700): (1) non-narcotic analgesics (NDC code, 1722); (2) narcotic analgesics (NDC code, 1721); and (3) nonsteroidal anti-inflammatory drugs (NSAIDs) (NDC code, 1727), including antiarthritics (NDC code, 1724). To maintain consistency with the NDC codes used in the NAMCS patient record form, the term "narcotic analgesic" will be used herein rather than "opioid." Physiotherapy consisted of treatments using heat, light, sound, physical pressure, or movement, including manipulative therapy.

To study the epidemiology of LBP, national population estimates of patient visits were derived and stratified according to patient sociodemographic, patient visit context, and physician provider characteristics. Multiple logistic regression was used to compute adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for factors associated with LBP patient visits compared with patient visits for all other reasons. Similar analyses were repeated according to chronicity of LBP: (1) initial visits for acute LBP and (2) follow-up visits for chronic LBP. To study the LBP medical management, national population estimates of the use of diagnostic tests, patient counseling, drugs, physiotherapy, and surgical procedures were derived for patient visits in which LBP was the primary reason for seeking medical care. Simple logistic regression was initially performed to compute crude ORs and 95% CIs for the elements of chronic LBP medical management compared with acute LBP medical management. Multiple logistic regression was subsequently used to compute adjusted ORs and 95% CIs for the most commonly used elements of LBP medical management. All hypotheses were tested at the .05 level of statistical significance.

The electronic files containing the 2003 and 2004 NAMCS data were acquired from the National Center for Health Statistics. The files were merged and analyzed using SPSS Version 14.0 for Windows (SPSS Inc., Chicago, IL). Because the multistage probability design of the NAMCS includes clustering, stratification, and the assignment of unequal probabilities of selection to sample units, all analyses were performed with the SPSS complex samples module to accurately compute estimates of population parameters and their standard errors [19]. A check of these statistical procedures, which involved the entire 2003 and 2004 NAMCS databases, verified that the computed marginal totals for national population estimates were identical to those published by the National Center for Health Statistics [14, 15].

There were an estimated 31 million (3%) patient visits annually attributed to LBP in the United States. For the 2003–2004 period, 1539 patient record forms representing 61.7 million (SE, 4.0 million) LBP patient visits and 1042 patient record forms representing 42.4 million (SE, 3.1 million) primary LBP patient visits were included in the analyses reported herein (Figure 1). The physician specialties most commonly seen during LBP patient visits were family (general) medicine, 25.2 million (SE, 2.3 million); internal medicine, 14.4 million (SE, 3.0 million); and orthopedics, 5.6 million (SE, 0.9 million). In orthopedics, one of every 16 patient visits involved LBP (one of every 11 working-age patients seen for reasons other than preventive or surgery-related care). In family (general) medicine, one of every 17 patient visits involved LBP (one of every 13 working-age patients seen for reasons other than preventive or surgery-related care). Patient visits for LBP were attributed to the following factors: (1) back pain, ache, soreness, or discomfort, 94%; (2) other back symptoms, including cramps, contractures, spasms, limitation of movement or stiffness, or weakness, 2%; and (3) other unspecified back symptoms, 4%. Prior to combining LBP patient visits, they were originally designated as RFV classification code number 1905 in 39.0 million (SE, 2.7 million) patient visits and as RFV classification code number 1910 in 22.8 million (SE, 2.0 million) patient visits. A total of 42.4 million (SE, 3.1 million) patient visits (69%) involved LBP as the primary reason for seeking medical care.

There were few differences in patient visit characteristics according to the importance of LBP (primary reason for seeking medical care vs. secondary or tertiary reasons for seeking medical care) or RFV classification code number (1905 vs. 1910) (Table 1). Neurologists were less likely than other physician specialties to provide care for primary LBP (P = .05). Hispanics were more likely than non-Hispanics (P = .02) and osteopathic physicians were more likely than allopathic physicians (P = .04) to involve RFV classification code number 1910 rather than 1905.

Patient visits for LBP predominantly involved persons aged 25 to 64 years (69%) and females (59%) (Table 2). A majority of patient visits were for follow-up care. Patient visits were about evenly divided between acute and routine chronic LBP, and about one-third of visits were attributed to an IPA. Among the estimated 8.4 million patient visits in which an IPA was specified, 7.4 million (88%) involved injuries (primarily overuse syndromes, motor vehicle accidents, or falls) and 0.9 million (11%) involved adverse effects of surgical procedures, drugs, or environmental agents. Thus, IPAs served as a useful surrogate for injuries. Only slightly more than half (55%) of the patient visits were provided by primary care physicians.

When compared with the 1.8 billion (SE, 7.6 million) patient visits for reasons other than back symptoms, LBP visits were associated with several factors (Table 2). Low back pain was less likely the reason for a patient visit in all younger and older age categories compared with the referent category (25–44 years) (P < .001). There was a geographic variation in patient visits attributed to LBP (P = .001), with fewer visits in the Midwest and South than in the Northeast. Patient visits for LBP were more likely to reflect chronicity of symptoms, either routine ongoing problems or flare-ups, than were patient visits for other reasons (P < .001). However, injuries, as reflected by the surrogate IPA item, were important predictors of LBP patient visits (P < .001). The type of physician provider (P < .001) and physician specialty (P < .001) were associated with LBP patient visits, with osteopathic physicians, family (general) medicine physicians, internal medicine physicians, neurologists, and orthopedic surgeons being more likely to provide medical care during such visits. Similar findings were observed when the analyses involved only primary LBP patient visits rather than all LBP patient visits.

A total of 15.8 million (SE, 1.0 million) LBP patient visits were initial visits for an acute problem (less than three months since onset) and 27.4 million (SE, 2.8 million) LBP patient visits were follow-up visits for chronic LBP (Table 3). Age <25 years (P < .001), injury (P < .001), and being seen by a primary care physician (P = .01) were inversely associated with LBP chronicity, whereas being seen by an osteopathic physician (P < .001) and shared physician care (P = .003) were directly associated with LBP chronicity. Again, similar findings were observed when the analyses involved only primary LBP patient visits rather than all LBP patient visits.

Overall, a majority of primary LBP patient visits included diagnostic tests (89%), patient counseling (53%), and orders for drugs (72%) (Figures 2, 3, 4, respectively). Compared with patient visits in which LBP was absent, primary LBP patient visits were more likely to involve patient counseling (OR, 1.54; 95% CI, 1.17–2.04) and physiotherapy (OR, 7.89; 95% CI, 6.01–10.35); however, they were less likely to involve surgical procedures (OR, 0.56; 95% CI, 0.38–0.83). There were no significant differences in the frequency of diagnostic tests performed or drugs ordered. The most common elements of LBP medical management included radiographs (13%), exercise counseling (20%), NSAIDs (34%), narcotic analgesics (25%), and physiotherapy (20%). There were 14.2 million (SE, 1.2 million) and 10.5 million (SE, 1.1 million) orders, respectively, for NSAIDs and narcotic analgesics during these primary LBP patient visits (Table 4). Except for the variables "exercise counseling," "any patient counseling," "narcotic analgesics," "NSAIDs," and "any drug ordered," the elements of primary LBP medical management were not reported in sufficiently high numbers of patient visits to provide statistically valid and reliable comparisons of initial visits for acute LBP and follow-up visits for chronic LBP. Drugs generally (P = .002), and NSAIDs specifically (P < .001), were ordered less often during follow-up visits for chronic LBP than during initial visits for acute LBP. Similarly, diagnostic tests were generally performed less often during follow-up visits for chronic LBP than during initial visits for acute LBP (P = .003). However, surgical procedures were ordered more often during follow-up visits for chronic LBP than during initial visits for acute LBP (P < .001).

Several other factors emerged in association with the common elements of primary LBP medical management after controlling for potential confounders (Table 5). Patient counseling was less often provided for non-Whites (P = .04) and in geographic regions outside the Northeast (P = .01), although it was provided more often when injuries were reported (P < .001). Specifically, with regard to exercise counseling, there remained a geographic variation (P = .003) and propensity for use following injury (P < .001). However, exercise counseling was less often provided in non-MSAs (P = .04) and by various physician specialties (P < .001), including family (general) medicine physicians and internal medicine physicians.

There was an association between age and drugs ordered (P = .01) during primary LBP patient visits (Table 5). Drugs were ordered less often in the young (<25 years) than in the referent age category (25–44 years). Drugs were ordered more often among non-Whites (P = .01) and by primary care physicians (P = .01), particularly family (general) medicine physicians. Similarly, primary care physicians were more likely to order NSAIDs (P = .03), although osteopathic physicians were less likely to order NSAIDs than allopathic physicians (P = .004). There was an association between age and narcotic analgesic use (P = .05), with less use in older patients (>65 years) compared with the referent age category (25–44 years). Narcotic analgesics were prescribed less often in Hispanics (P = .01) and more often in non-MSAs (P = .03).