Study Population
Study participants were recruited from the outpatient clinics of the Spine Center of New England Baptist Hospital. Consecutive patients age 18 and older with a diagnosis of acute low back pain were evaluated for participation. The majority of patients with LBP presenting to the Spine Center are direct patient self-referrals, although 20-25% of patients are referred by primary care physicians. Inclusion criteria required symptoms of new low back pain ≤ 3 months duration preceded by a substantial pain-free period lasting at least one month; literacy in English or Spanish; daily access to the Internet; basic computer literacy; and a valid email address accessible regularly for a period of 6 weeks. LBP was defined as either lumbar spinal pain or sacral spinal pain, as per recent consensus guidelines[10]. Exclusion criteria included radiating lower extremity pain experienced below the level of the knee; a predominant component of lower extremity pain greater than low back pain; motor weakness, asymmetric reflexes, or decreased sensation of the lower extremities associated with the current episode of LBP; symptoms of neurogenic intermittent claudication; clinical suspicion for 'red flag' conditions including infection, fracture, or neoplasm; recent low back trauma; pregnancy; severe active medical comorbidities or psychiatric illness. These exclusion criteria were chosen primarily to distinguish nonspecific back pain from two other broad categories of back pain: back pain potentially associated with radiculopathy or spinal stenosis, or back pain potentially associated with another specific spinal cause [11]; in addition, pregnant individuals were excluded due to the fact that LBP during pregnancy is likely to be affected by a variety of pregnancy-related factors and is therefore not nonspecific. Severe active medical comorbidities or psychiatric illness were criteria for exclusion because we expected these to make participation with data collection infeasible.
The Institutional Review Board of New England Baptist Hospital approved the conduct of this study and all study materials. Written informed consent was obtained for all individuals prior to study participation.
Baseline Assessment- Demographics, Comorbidities, and Back Pain History
After informed consent was obtained, the examining physician used paper-based methods to record information on participant age, gender, race/ethnicity, medical and psychiatric comorbidity, employment status, and workers' compensation status. Race was categorized as 'Asian', 'Black', 'Hispanic', 'Native American or Alaskan Native', 'Pacific Islander', 'White', and 'Other'. Comorbidities were measured using the Self-Administered Comorbidity Questionnaire (SACQ). The SACQ is widely used in orthopedic research, and has previously demonstrated reliability and validity[12]. Employment status was categorized as part-time employment, full-time employment, student, retired, disabled, and unemployed.
Participants provided information on duration of current back pain symptoms and prior history of LBP episodes, lumbar disk herniation or sciatica, at the baseline evaluation. Current back pain intensity was measured by the numerical pain rating scale (NPRS) for back pain[13–15]. The NPRS is a 11-point scale in which patients rate their pain ranging from 0 (no pain) to 10 (worst imaginable pain). The NPRS is a valid and commonly used measure of back pain intensity that is simple for patients to comprehend and complete[13–15]. Back related disability was measured using the Oswestry Disability Index (ODI). The ODI is a condition-specific measure of disability which is used extensively in studies of low back pain, and has demonstrated validity and reliability in this context[16].
Internet-Based Longitudinal Assessments- Flare Status and Back Pain Characteristics
Participants were sent a secure email link to an online survey tool at an email address they provided, on the day of the initial clinic evaluation. Survey Gizmo online survey software (Widgix, LLC, Boulder, CO) was used for all data collection after the in-clinic baseline assessment. Survey Gizmo is a web-based service that follows the Privacy Rule and Security Rule provisions of HIPAA (Health Insurance Portability and Accountability Act of 1996), and self-certifies adherence to HIPAA. HIPAA is a US federal law that establishes standards for the privacy and security of health information. Further information can be found at: http://www.surveygizmo.com/survey-blog/online-survey-hipaa-safe-harbor-certification. Participants were instructed to complete an initial online questionnaire as soon as possible after their initial clinic evaluation. Further online questionnaires were scheduled at both 3-day intervals and 7-day intervals after the date of completion of the first online questionnaire. The rationale for using 3-day intervals was to obtain more frequent measurements of back pain-related factors, consistent with our primary goal of characterizing short-term pain fluctuations. The 7-day interval was also included due to the fact that follow-up appointments are scheduled at one-week time intervals in standard clinical practice. The 3-day- and 7-day- questionnaires contained the same measures, with the exception of the ODI, which was included only at the 7-day intervals in order to minimize patient burden (see Additional File 1). Because we were aware of no prior information on the Internet-based serial assessment of individuals with acute LBP, we did not know whether it would be feasible to require individuals specific time windows within which to complete questionnaires. Participants were sent questionnaires on their scheduled date, with reminders after 1 day, and after 2 days. Responses were accepted at any time until the due date for completion of the next scheduled questionnaire.
Each online questionnaire assessed the presence or absence of a current pain flare. A flare of pain was defined as 'a period of increased pain lasting at least 2 HOURS , when your pain intensity is distinctly worse than it has been recently'. This definition was adapted for acute LBP from the definition by Von Korff[1]. If reporting a current flare, participants described the duration of the current flare. Current back pain intensity was rated using the NPRS for back pain, and back-related disability was rated using the ODI. The Fear-Avoidance Beliefs Questionnaire[17] was used to assess the strength of participants' fear of back injury and relation to physical activity and work. The FABQ physical activity and FABQ work subscales have been validated for use in back pain patients [17]. Due to the length of the ODI and concerns regarding cumulative participant burden over the repeated assessments in the study, the ODI was administered only at weekly follow-up assessments. The FABQ was completed only at the first online questionnaire. The Internet-based data collection questionnaire items are provided in Additional File 1.
Standard Medical Treatment of Acute LBP
All participants received standard medical care for nonspecific acute LBP as per practice guidelines[11]. Treatment included patient education, encouragement to avoid bed rest and normalize daily activities, use of oral analgesic medications when needed, and non-pharmacologic treatments as indicated[11].
Statistical Analysis
Study participants who did not complete more than the first Internet questionnaire were considered 'non-compliant'. Non-compliant and compliant subjects were compared on demographic, medical/psychiatric, and back pain characteristics using the Student's t-test for quantitative variables and chi-square tests for categorical variables. The cohort of compliant subjects was characterized descriptively using means and standard deviations (SDs) for continuous variables, and frequencies and proportions for categorical variables. Due to the fact that few participants were from racial minority groups, race/ethnicity was categorized as white vs. non-white for all analyses.
Characterization of pain flares during the acute LBP episode was done by determining the prevalence of individuals who reported at least one flare, the number of discrete 'flares' reported by study participants, and the prevalence of flares of different durations. Next, we examined longitudinal changes in pain intensity as a function of time as days since seeking care, using mixed-effects linear regression models. Because longitudinal change in pain intensity during acute LBP has not been well-studied, and because existing knowledge suggests that much improvement occurs within the first 1-2 weeks[18, 19], we first examined longitudinal trends in pain intensity using a linear spline model, including the predictor variables of time since seeking care (in days) and 'knots' (points at which spline slopes are permitted to change) at weekly intervals for six weeks. We then compared and contrasted the linear spline model with simple linear and quadratic models for the predictor variable of time since seeking care. We used the Akaike Information Criteria (AIC) as an indicator of model fit, and chose the most parsimonious model that demonstrated good fit. We then applied this model, while including the time-varying covariate of patient self-report of a current pain flare, to examine the longitudinal association of flare status with pain intensity. We computed ß coefficients and standard errors (SEs). Although change in back pain intensity was the focus of this pilot study, we also characterized longitudinal change in ODI scores using the same analytic methods described above.
Longitudinal analyses were followed by complementary multivariate regression analysis to model associations between flare frequency and the outcome of disability, while adjusting for other possible predictors of disability outcomes. Disability was measured as the final ODI score at study completion. Because some participants missed questionnaires towards the end of the study, the last value for ODI score was carried forward, provided that it was within 1 week of study completion. Flare frequency was defined as the number of flare periods reported by each participant, divided by the total number of questionnaires completed by the participant, expressed as a percentage. We first examined the bivariate association between flare frequency and the outcome of ODI using linear regression. Next, we used separate bivariate linear regression models to examine relationships between potential adjustment factors and the outcome of ODI score. Potential adjustment factors included participant age, gender, medical/psychiatric comorbidity, baseline pain intensity, baseline ODI score, duration of symptoms, work status, FABQ physical activity, and FABQ work. Recognizing our limited sample size, adjustment factors to be included in the final multivariate model were chosen by including only those predictor variables with the strongest bivariate associations with the outcome of ODI score. We included in the final multivariate model the predictor variable of flare frequency, and those adjustment factors that demonstrated p-values < 0.20 in bivariate analyses, up to a maximum of four variables (as limited by our sample size). Adequacy of the fitted model was assessed by scatter plots of studentized residuals. We also conducted a sensitivity analysis using the same analytic approach with an alternate definition of the flare outcome as the total number of reported flares during the six-week follow-up period. Last, analogous linear regression analyses were performed using the same candidate predictor variables, but the outcome of pain intensity at six weeks.