The directors of 40 Primary Care Centers of the Spanish National Health Service were sent a letter inviting them to join the study. Those Centers were those in which there was at least one physician having expressed or stated an interest at any time on any kind of research on back pain. The Centers were located in 10 of the 17 Spanish administrative regions and they covered a total population of 693,206 subjects, out of a total Spanish population of 40,499, 792 inhabitants . The Directors of the Centers were asked to forward the invitation to the 361 physicians working in those practices, and 94 physicians (26.0%) accepted to participate in the study, although in the end only 75 (20.8%) actually recruited patients.
Primary care physicians were asked to recruit all subjects who visited them because of low back pain, with or without leg pain, 18 years of age or older. Exclusion criteria were functional illiteracy (inability to answer the questionnaires used to assess pain and functional disability), illnesses affecting the central nervous system, and "red flags" for surgical referral (saddle anesthesia, recent onset of bladder dysfunction or anal sphincter impairment, major or progressive motor weakness, sensory level, or widespread neurological signs), or for possible specific spinal pathology (oncologic disease during the previous 5 years, constitutional symptoms-unexplained weight loss, fever, chills-, recent urinary tract infection, history of intravenous drug use, or immunocompromised host). The study protocol was approved by the institutional review boards of the participating centers and all patients gave written informed consent for use during a 6-month period of his/her data regarding severity of symptoms, health care, and sick leave for low back pain.
This being an observational study, there was no intervention per se. Physicians were given no instructions, but encouraged to treat their patients as they usually do, and no attempt was made to homogenize their practice. They were told that the objective of the study was to contribute to the knowledge of low back pain. Management of non-specific low back pain in the primary care setting of the Spanish National Health Service is based on clinical history, physical examination, recommendation of diagnostic tests, medical counseling, drug treatment, physical therapy, rehabilitation, or referral to specialists. All visits of patients and all the diagnostic procedures, treatments, and referrals that were prescribed by the primary care physicians were registered in structured questionnaires.
The clinical condition of each patient was assessed at the primary care center by his/her general practitioner on the day of inclusion in the study (baseline assessment, day 1) and 15 and 60 days later. Besides these three obligatory follow-up assessments, all visits requested by the patient during the study period because of low back pain were considered additional voluntary visits and data on patients' clinical evolution and on physicians' management of the patients were recorded.
At baseline, the following variables were recorded: age (date of birth); sex; educational level (classified in 5 categories, from lower to higher level); duration of pain (number of days elapsed since the current episode of pain appeared, when the patient was recruited), chronicity of pain when entering the study (classified as acute, subacute or chronic, and defining subacute as a duration of pain between 14 and 90 days) [23, 24]; previous episodes of low back pain (zero, one or two, more than two); current use of drug treatments (yes/no), previous spinal surgery in the dorsolumbar region (number of procedures); living alone, living with someone or institutionalized; usual job; and the number of plain radiographs, CT scans, MRI studies, neurophysiologic or other diagnostic tests, treatments, and referrals to specialists or visits to the emergency room because of low back pain during the 3 months prior to entering the study.
At each assessment, including the first one, the following variables were recorded: current pregnancy (yes/no); work status (potentially active-self-employed or employed by others- or not potentially active-retired, unemployed, student, housewife, other-), sick leave (no, yes-number of days-, not applicable), reason for sick leave (low back pain, other, not applicable – e.g. unemployed population-), severity of low back and leg pain (assessed by means of independent visual analogue scales (VAS)  in which a higher score means a more severe pain, on a 0–10 range), degree of functional disability (assessed by the previously validated Spanish version of the Roland Morris questionnaire  in which a higher score reflects a higher level of disability, on a 0–24 range), and quality of life level (measured by the EuroQol questionnaire , in which a higher score reflects a better quality of life on a 0 to 1 range), number of physician visits by the patient throughout the period (to primary care physicians, public specialists, private specialists and alternative medicine), diagnostic tests (X-rays, CT scans, MRI studies, electromyography, blood tests, other) and drug treatment which were prescribed by the primary care physician, and referrals to rehabilitation, physical therapy, orthopedic surgery, neurosurgery or other services. At each assessment, a straight leg raising test (SLR) was performed at the discretion of the primary care physician (and classified as <30°, between 30° and 60°, and >60°). Results of the SLR test were based on the range of motion triggering referred pain to the leg.
"Key descriptors" of the essential aspects of clinical practice in low back pain were defined, and this information was collected in each of the assessments. Performing the SLR test at any moment between the first and third assessments (considering the SLR and the crossed SLR equally ) was considered as descriptive of the physical examination. The ordering of imaging and neurophysiologic tests at any moment during the study was considered as descriptive of the request for diagnostic tests. Imaging tests were classified as X-rays, or CT scan/MRI, since in some regions the prescription of one or the other depends solely on its availability. Additionally, since in some regions the primary care physician can only prescribe the CT scan or MRI via the orthopedic surgeon or the neurosurgeon, referrals to those specialists with such intention were also considered as prescriptions for a CT scan or MRI. Since a recent history of diagnostic testing can influence its prescription, the realization of these tests was taken into account from the time of recruitment to three months prior.
The prescription of medication in the first visit (NSAIDs, muscle relaxants, analgesics, and "other") or referral (to physical therapy, rehabilitation, orthopedic surgery, neurosurgery, or "other") at any moment during follow-up were considered descriptive of treatment. Physician's counsel was not included as descriptive of treatment because it was understood to be inherent to all medical practice, and because no validated system to classify its content was found ("active management", "postural hygiene", etc.).
The granting of sick leave by the general practitioner at any moment between the first and third assessments was considered descriptive of sick leave.
Data recording and transference to the database were as follows: questionnaires on pain, disability and quality of life were self-administered and completed by the patient on his/her own in the absence of the treating physician, other health care staff, family members, or accompanying persons. Completed self-reported instruments were then given to the auxiliary staff of the primary care center who stapled scales and questionnaires to the patient's data collection form. At the end of the working day, ratings of the scales were transcribed to the data collection form by the treating physician. At each Primary Care Center, concordance of scores in the data collection form and ratings of self-administered instruments were verified by the study coordinator in the Center. Structured data collection forms with information on patients' clinical management were then sent along with patients' self-reported instruments to the central coordination office, where data were entered in a database by two administrative assistants who double-checked data entry. They also double checked that scores on pain, disability and quality of life scales coincided with actual patients' ratings of the VAS, Roland-Morris and EuroQol questionnaires. All throughout the process, patients' answers on the self-administered questionnaires were considered the gold standard, but no inconsistencies were actually found.
The analysis was done by a team of biostatisticians who had no connection with the primary care physicians involved in this study. Frequencies were calculated for categorical variables. For continuous ones, means and standard deviations were calculated for those with a normal distribution, and median and interquartile ranges for the rest. Logistic regression models were developed to explore the influence of different factors on the key descriptors of clinical management of non-specific low back pain. In each model the descriptor was used as a dependent variable. Colinearity of the maximal models was evaluated with the criteria proposed by Belsley . A backward strategy was used, using the value p < 0.05 to eliminate variables from the model.
The following variables were introduced in all maximal models: sex, previous number of episodes of low back pain when entering the study (recoded as "less than three" and "three or more"), intensity of lumbar and leg pain at the first assessment, degree of disability at the first control, and chronicity of pain when entering the study (acute, subacute or chronic) [23, 25]. The variable SLR was also included in all models, categorized as not done, negative (≥60°) or positive (<60°), except in the model that referred to doing the test. Only workers were included in the model on sick leave. Since in the Spanish Social Security System self-employed persons are economically penalized while on sick leave, as opposed to those employed by third parties, workers were classified as "self employed" or "employed by another". In all other models patients in different work situations were included, and the variables "work status" (recoded as "potentially active" and "other") and "sick leave" ("yes" or "no") were included.