Study design
From April to October 2001, we conducted telephone interviews with 249 patients who were currently suffering from non-specific low back pain that had persisted at least three months. The patients were members of a non-profit managed health care system (Group Health Cooperative in the Puget Sound region of Washington State) and a large multi-specialty group practice (Harvard Vanguard Medical Associates, Boston, Massachusetts). Our goal was to interview 150 patients from Group Health and 100 patients from Harvard Vanguard who were otherwise healthy and spoke English, with 50% of the interviews from adults 65 years or older. The Institutional Review Boards of Group Health Cooperative, Seattle WA and Harvard Pilgrim Health Care, Boston, MA approved the study.
Sample
Using automated visit data, we identified and mailed letters to 787 patients with working phone numbers who visited a Group Health (n = 422) or a Harvard Vanguard (n = 365) primary care provider and had a diagnosis consistent with non-specific low back pain between 12 and 52 weeks previously. Because we were planning to use the results of the survey to help us refine recruitment strategies for two pilot randomized trials, our exclusion criteria for this study were those we planned to use in the subsequent trials. We therefore used the automated visit data to exclude the following individuals from our sample:
• those whose back pain may have been due to a specific disease or condition (i.e., sciatica, herniated disc, spondylolisthesis, spine fracture, vertebral fracture, cancer);
• those who had other pain conditions that could complicate the interpretation of trial results (rheumatoid arthritis, ankylosing spondylitis, fibromyalgia);
• those who were inappropriate candidates for one or more of the clinical trial treatments (aneurism, coagulation disorders, osteoporosis) or who were unlikely to be able to give informed consent or to participate in the baseline and follow-up assessments of the trial (Alzheimers disease, dementia, major psychoses, blindness, deafness).
Before administering the in-person survey, we asked eight screening questions and then excluded individuals who did not have back pain at the time of the interview or who had not had back pain for at least 12 weeks, who had previously had low back surgery, who reported having fractured a vertebrae, who were pregnant, who were involved in back-pain related litigation, who had serious health problems, or who could not speak English.
We tried to contact all mailees, but could not reach 28 (7%) persons from Group Health and 73 (20%) from Harvard Vanguard despite at least seven phone calls. Among the 394 Group Health and 292 Harvard Vanguard patients who were contacted, 57 from Group Health and 81 from Harvard Vanguard refused the interview and we could not determine their eligibility status, 195 from GHC and 104 from Harvard Vanguard were ineligible upon screening, and 142 from Group Health and 107 from Harvard Vanguard were eligible and interviewed. Thus, we were able to screen 70% of all mailees for eligibility (80% from Group Health and 58% from Harvard Vanguard). All screened and eligible mailees were interviewed.
In both areas, individuals aged 65 and older were more likely to refuse screening (in Seattle: 19% of 210 older adults vs. 8% of 212 younger; in Boston: 31% of 184 older adults vs. 13% of 181 younger) and those under 65 were less likely to be contacted by phone (in Seattle: 12% younger vs. 1% older could not be contacted; in Boston: 31% younger vs. 9% older could not be contacted).
Most patients were ineligible because they were not experiencing back pain at the time of the interview (n = 82 from Group Health and n = 68 from Harvard Vanguard) or their pain had not persisted for three months (n = 39 from Group Health and n = 11 from Harvard Vanguard).
Survey questionnaire
We conducted a phone interview that lasted an average of 17.7 minutes (SD = 6.1; range = 8 to 50 minutes). It included questions about demographic characteristics (age, race/ethnicity, education); back pain characteristics (years since first episode of back pain lasting more than two weeks, number of days of pain in the last six months, bothersomeness of pain on a 0 to 10 scale, expectations of pain level one year from the time of interview, and use of medications in the past week); self-reported knowledge (measured on a five point scale) of five CAM treatments or self-care methods (acupuncture, chiropractic, massage, meditation, t'ai chi); previous use of these therapies for any reason and for back pain specifically (and helpfulness of the therapy for back pain relief); perceived harm from previous use of these therapies; expectations of helpfulness of these therapies for current back pain; willingness to try these therapies if offered by the health plan for no additional cost and for a $10 per visit co-pay; willingness to participate in two hypothetical clinical trials, one evaluating acupuncture, chiropractic, and massage and another involving massage, meditation, and t'ai chi. (Respondents were told that the control group in both trials would receive a book about self-management of back pain.) Finally, respondents were asked about which treatment they most preferred among those offered in each trial. Gender and geographic location were obtained for respondents from the enrollment files of each healthplan. General definitions of each therapy were provided only for respondents who informed the interviewers they did not know what a particular therapy was. Acupuncture was described as a system of healing that involved inserting hair thin needles into acupuncture points just beneath the skin or using other methods, such as heat, to stimulate these points, whereas chiropractic was defined as a system of therapy that uses manipulation to adjusts the spine and other body parts to "promote normal nerve functions". Massage was described as the systematic rubbing and manipulation of muscle and other tissues to relieve bodily infirmities, while meditation was defined as a "self-directed practice for relaxing the body and calming the mind". Finally, tai chi was described as a Chinese martial art that uses slow and smooth body movements and is often practiced for its purported health benefits. The survey was pre-tested on a convenience sample of 15 people (both older and younger) in Seattle and 5 people in Boston.
Statistical analyses
We analyzed the data using the SAS statistical software version 6.12 (SAS Institute, Cary, NC). Descriptive data were characterized using percentages or medians. For each of the five CAM therapies, we performed separate exploratory logistic regressions to identify specific characteristics associated with 1) high degree of knowledge (4 or 5 on a 5-point scale) (five separate models), 2) prior use (for any reason and for back pain) (10 separate models), 3) high expectations of helpfulness for current back pain (7 to 10 on a 0 to 10 scale) (five separate models), 4) greatest likelihood of trying therapy for no additional cost (all five therapies) and for a $10 per visit co-pay (acupuncture, chiropractic, massage only) (eight separate models), and 6) greatest likelihood of participating in each of the two hypothetical clinical trials (two separate models). Thus, a total of 30 separate logistic models were created.
For each of the 30 dependent variables, we identified potential predictor variables in advance and evaluated them in preliminary models. In Table 1, the potential predictor variables evaluated in the preliminary models for each of the 28 therapy-specific dependent variables are indicated by an X. In addition, we modeled the likelihood of being "definitely willing" to participate in a hypothetical clinical trial of acupuncture, chiropractic, and massage and of being "definitely willing" to participate in a hypothetical clinical trial of massage, meditation, and t'ai chi. In both models, we evaluated the following 22 variables as potential predictor variables of being "definitely willing" to participate in the hypothetical clinical trial: demographic characteristics (age, gender, race, education, geographic location), prior use of each of the therapies included in the trial (i.e., acupuncture, chiropractic, and massage for one trial and massage, meditation, and t'ai chi for the other trial) for any reason (and for back pain), knowledge of these three included therapies, prior perceived harm from these three included therapies, years since first back pain, symptom bothersomeness, high expectations of each included CAM therapy for current back pain, number of days of back pain in last six months, and medication usage in the week prior to the interview.
Initially, we evaluated potential predictor variables in preliminary models containing five or fewer independent variables. Any independent variable associated with the dependent variable at a p value of 0.15 or less in a preliminary model was a candidate for the appropriate final model. We used a backwards elimination procedure to evaluate candidate predictor variables and to determine the final models [8]. All variables with a p value of 0.01 or less were retained in the final model. Odds ratios (OR) are presented along with 95% confidence intervals (95% CI). Table 4 presents the odds ratios that describe the significant associations (p
Answers to all your Biology Questions
