Randomized trial of a self-management program for primary care patients with acute low back pain: Short-term effects




We evaluated the effect of a self-management program (SMP) on primary care patients with acute low back pain (ALBP) from low income, inner city neighborhood health centers and an emergency department of a public teaching hospital.


We randomized 211 primary care patients who visited a physician for ALBP (<90 days duration) to usual care or an SMP. The SMP consisted of 3 group sessions and telephone followup that focused on understanding back pain, increasing physical activity, and dealing with fears and frustrations.


Of the eligible patients, 52% expressed interest in participation and 39% of all eligible patients were randomized into the study. Among patients in the treatment group, 28.3% attended at least 1 group class, 62.3% received the intervention by mail, telephone, and audiotapes, and 9.4% received no intervention. Interviewers, blinded to the treatment given, collected data at baseline and at 4 months following randomization. Compared with the control group, the intervention group reported significantly better emotional functioning (P < 0.01), increased self efficacy to manage ALBP (P = 0.03), and less fear of movement (P = 0.05) after 4 months.


This SMP produced short-term improvements in emotional functioning and self efficacy to manage symptoms among patients with ALBP living in the inner city. However, methods of program delivery other than group classes are needed to reach a greater portion of the inner city patients.


Low back pain is highly prevalent in the United States: 80–90% of adults experience low back pain during their lifetimes and 50% report symptoms during any given year (1–3). It is the leading cause of visits to orthopedic surgeons, neurosurgeons, occupational medicine physicians, and osteopathic physicians, and is second only to upper respiratory infections as the reason for visits to primary care physicians (PCP) (2). It has been estimated that the direct medical costs of care for low back pain exceeded $24 billion in 1994; if one considers indirect costs (e.g., disability, lost productivity), this estimate may reach between $50 and $100 billion annually (3–4).

Acute low back pain (ALBP), defined as symptoms lasting less than 3 months, accounts for more than 90% of all back pain and nearly half of the costs (1). Current treatment recommendations include patient education, nonsteroidal anti-inflammatory drugs, a gradual return to normal activities (rather than prolonged bed rest), and low-stress aerobic exercise (1). Because ALBP tends to be recurrent, with unpredictable exacerbations and remissions of symptoms, implementation of these recommendations may be facilitated by programs that enhance self efficacy. Self efficacy is the perception of ability to produce a desired outcome (5) and is a strong predictor of health behavior (6–8), including self management (9). Increasing self efficacy requires not only defining specific desired behaviors, but more importantly, enhancing patient's motivation and confidence to implement strategies required to manage their ALBP, thereby improving health-related behaviors and health status (10–12). Emphasizing self efficacy is central to the Arthritis Self-Management Program (ASMP) (13), a program that has consistently demonstrated effectiveness in improving health outcomes and reducing health care utilization among patients with various rheumatic and chronic conditions (13–15), including low back pain (16).

The ASMP has not been evaluated among socioeconomically vulnerable patients with ALBP. This is unfortunate, given that these individuals shoulder a disproportionate burden of disability and morbidity from musculoskeletal conditions and comorbid medical conditions (17–19). Indeed, in a recent assessment of health concerns and needs among randomly selected congregants of 5 urban African American churches, back pain was reported among 41% as a pressing health issue (20). Moreover, data from the National Health Interview Survey indicated that in comparison to whites, African American men and women in the United States reported losing more workdays for greater lengths of time because of low back pain (21).

Despite the prevalence of back pain, there is a dearth of research on strategies to reduce back pain disability in this population. Studies involving other health conditions such as arthritis, asthma, and weight loss have targeted minorities (22–24) and inner city patients (25). These studies incorporated church-based interventions and involved family members. The support system of the church and family may reduce some of the methodologic barriers of conducting interventions among the economically vulnerable. These barriers may include patient relocation without further contact, group class attendance (26), and attrition due to health problems or caregiver responsibilities (27).

A recent qualitative analysis of influences on diabetes self management among southern African American women suggested that church-based and family-centered interventions are the most appropriate to promote disease self management (28). Thus, based on the ASMP, we developed a self-management program for inner city patients with ALBP. Initially, we designed the intervention as a church-based program for primary care patients and their families. After piloting the intervention, patients indicated that they preferred to meet at the local neighborhood primary care health centers. We then evaluated this program among primary care patients using a randomized, controlled trial. We present the short-term (4-month) effects of our program. Our primary outcomes were functional status and patient satisfaction. Secondary outcomes were self efficacy, self-management practices, social support, and fear of movement.


Study site and participants.

Prior to implementing the study, we received approval to conduct this study by the local institutional review board. We recruited from our university-affiliated neighborhood health centers and emergency departments that serve inner city patients during June 1998 through March 2000. Inclusion criteria included the following: 1) age ≥18 years; 2) primary diagnosis reflecting back pain; 3) ALBP (i.e., patient or physician reported current episode <3 months duration and not due to severe trauma); 4) receiving primary care in our clinical venues; 5) deemed eligible for study by their PCP; and 6) access to a working telephone. We excluded patients who met any of the following criteria: 1) prior surgery for back pain; 2) receiving disability insurance payments or in the process of applying for back pain disability; 3) residing in an institution; 4) being incompetent for interview; 5) severely impaired in vision, hearing, or speech; 6) unable to understand and speak English; 7) being pregnant; or 8) judged by their PCP to have a terminal illness (life expectancy <1 year) or severe comorbid condition limiting their functional ability.

Recruitment and baseline data collection.

We sought to identify and enroll patients as quickly as possible after their visit for ALBP. Thus, we used the computerized Regenstrief Medical Record System (RMRS) (29) to identify potentially eligible patients. All of our clinical venues use microcomputer workstations linked to the RMRS to enter all orders (29). When physicians entered the reason for visit in the computer as “low back pain” about a potentially eligible patient, they immediately received a computerized alert describing the study and eligibility criteria, indicating the patient was potentially eligible, asking physicians to confirm eligibility, and requesting their permission to recruit the patient. An e-mail message was automatically sent to our project coordinator about patients for whom physicians confirmed potential eligibility and provided permission to recruit. The coordinator performed chart audits to search for exclusion criteria and then initiated the recruitment protocol. First, she mailed a recruitment letter to patients describing the study (including their physician's recommendation for participation). Next, she called the patient within a few days to confirm eligibility and determine the patient's interest in participating. For patients agreeing to participate, an interviewer, blinded to study hypotheses, obtained informed consent and conducted baseline interviews at either the primary care center, the patient's home, or by telephone. We attempted to schedule all baseline interviews within 7 days of confirming eligibility. Patients received $10 for completing the interview.

Study groups.

Immediately after completing the baseline interviews, we randomized patients to the control or intervention group. Patients in the control group received usual care. Those randomized to the intervention group received usual care and were invited to participate in the Acute Low Back Pain Self-Management Program (ALBPSMP). The program, which was based on the ASMP (13), a chronic back pain program (30), and social-cognitive theory (5), focused on increasing self efficacy and social support to self manage low back pain. Its final structure was based upon a series of focus groups intended to maximize the relevance and feasibility of the ALBPSMP. The program included 3 in-person classes, class handouts, classes on audiotape and a cassette player, physician letters of support, and telephone followup (Figure 1).

Figure 1.

Description of randomization arms. ALBP = acute low back pain; PT = physical therapy; OT = occupational therapy.

In-person classes.

Participants attended 3 classes (once per week) in community rooms of neighborhood health centers. Using standardized, written protocols, these classes had several goals. We provided information about ALBP. We described its intermittent nature to both enhance patients' optimism for a rapid recovery and provide realistic expectations regarding the likelihood of recurrence. Because patients often expected physicians to order diagnostic tests, we reviewed common tests, including the rationale for ordering (or not ordering) them. Patients were encouraged to adhere to their physicians' treatment recommendations, as well as to reduce their body weight when appropriate. Session leaders promoted behaviors supported by current evidence-based treatment recommendations, particularly minimizing bed rest and returning to normal activities as soon as possible.

To facilitate behavioral changes, class leaders presented materials on goal setting, problem solving, and social support. Participants set goals to modify specific behaviors. Leaders facilitated group discussions to identify both potential barriers to these goals and strategies to resolve identified barriers. Social support for their goals was promoted in 2 ways. First, patients were encouraged to bring their spouse or significant other to all sessions of the program. Second, following each session, we developed reports describing progress made by patients towards their goals.

To promote perceived self efficacy, participants received individualized feedback about their progress from their group leaders and other participants when possible. In addition, patients were encouraged to ask questions of their provider and to use positive talk and relaxation, 2 strategies that have been successful in ASMP to manage negative affect (12, 13, 15). Patients recorded their behaviors and progress on a standardized form. We gave one copy of the report to the patient, and kept one in the research office.

Class handouts.

Written education materials showed recommended exercises, including walking, and proper body mechanics.

Classes on audiotape and a cassette player.

When patients missed a class, we provided them with an audiotape of the class, a hand-held cassette player, and copies of the written handouts distributed at the missed class (26). A member of our staff called patients to confirm receipt of materials, verify that the patient reviewed the materials, go through the materials, and answer questions.

Physician letters of support.

With the physicians' permission, we mailed letters with the scanned signature of the primary care provider within 2 days of each session. These letters, tailored to the content of each session, encouraged patients' further participation in the program.

Telephone followup.

To reinforce the class sessions, our research staff made telephone calls to participants at 4, 6, and 8 weeks to discuss ascertainment of goals, assist with problem solving, and set new goals. Thereafter, the staff made telephone calls once a month to continue reinforcing the class sessions and sustain behavioral change.

Followup data collection.

We collected postintervention assessments 4 months from baseline. This period permitted patients to participate in the intervention classes, receive the followup calls, and time to attempt to implement the new skills promoted during the classes. Patients were mailed letters to remind them of their 4-month assessment. Interviewers blinded to group assignment collected these data through either in-person or telephone interviews. The identical measures assessed in baseline were obtained.

Primary outcome measures.

Functional status was assessed with 2 complementary measures. The Arthritis Impact Measurement Scales (AIMS2) is a validated, arthritis-specific measure of functional status (31–33) that can be reliably reduced to a 5-component model. Three of these components were included as our primary outcomes: physical function, affect, and symptoms. Higher scores indicate worse functioning. The Roland Disability Questionnaire is a 23-item back pain-specific measure of functional status (34, 35) that has demonstrated reliability, validity, and sensitivity to change (35). For each question, patients indicate that the statement either “describes me” or “does not describe me.” For example, “I sleep less well because of my back problem.” We used total score, with higher score indicating poorer functioning.

Patient satisfaction was measured with a validated scale developed specifically for patients with low back pain (36). For each of 9 statements, patients rated their satisfaction on a 3-point scale (1 = dissatisfaction, 2 = uncertain, 3 = satisfaction). Thus, overall scores ranged from 9 (least satisfied) to 27 (most satisfied).

Secondary outcome measures.

Self efficacy was assessed with the 6-item version of the Arthritis Self-Efficacy Scale to manage symptoms of low back pain (37). For each item, patients reported their degree of certainty to manage symptoms at the present time on a scale ranging from 1 (very uncertain) to 10 (very certain).

To estimate self-management practices, we asked patients to estimate the number of minutes they spent performing exercise (none, <30 minutes/week, 30–60 minutes/week, 1–3 hours/week, >3 hours/week), cognitive symptom management (never, almost never, sometimes, fairly often, always), and stress management (open ended questions) (38).

Social support was assessed using the social interaction component of the AIMS2 (31–33). The final component of the AIMS2, role functioning, was also assessed as a secondary outcome.

Fear of movement/activity was measured with a 10-item scale that focused on interpretation of pain, fear of activity/exercise, and safety/injury that was adapted for this study (e.g., “I'm afraid that I might injure myself if I exercise”) (39). Response items included a 4-point scale ranging from strongly agree to strongly disagree.

Patient characteristics.

Demographic characteristics were obtained during baseline interviews. Income was assessed with a single item used previously in this patient population. It asked whether patients felt that their household income is “comfortable,” “just able” or “unable” to make ends meet (40). In terms of baseline clinical data, we used RMRS data to calculate Ambulatory Care Group (ACG) (41), a rigorous measure of outpatient comorbidity that has been validated in a number of settings, including inner city Medicaid populations (42). We used this measure to test for between-group differences in comorbidity.

Dose effect of intervention.

As a secondary analysis, we evaluated the dose effect of the intervention. Based upon the premise that staff/participant interaction was the strongest dose of the intervention, we created a scoring system for exposure to the intervention class dose. We scored exposure for those randomized to the intervention in the following manner: attended class = 5; received audiotapes of classes, the class materials, and a telephone call from staff = 4; received the class materials, and a telephone call from staff = 3; receiving audiotapes of classes, but no staff phone call = 2; receiving class materials in mail only = 1; and received no intervention = 0. Similarly, we created a scoring system based on the number of telephone calls the staff made to participants after program delivery.

Statistical analysis.

For all analyses, we performed intention to treat analyses. For the comparisons of baseline characteristics between intervention and usual care, between initially willing and unwilling to participate, and between dropout and completers at 4 months, we conducted t-tests to compare normally distributed continuous variables, and Wilcoxon's rank sum tests for ordinal variables and continuous variables failing to meet the normality assumptions of a t-test. For dichotomous variables, we conducted chi-square tests of independence.

For the analyses of each outcome, we performed an analysis of covariance, with the 4-month outcome being the response variable, and with intervention status, baseline outcome, age, race, sex, body mass index, years with low back pain, severity of back pain in the 4 weeks previous to the baseline interview, and months since onset of last episode of back pain as predictors. This approach is equivalent to modeling the change scores adjusted for baseline.

To control the overall Type I error for multiple hypothesis testing for the primary outcomes of interest (physical function, affect, symptoms from the AIMS2, Roland ALBP specific functional status, and patient satisfaction), we employed Hochberg's sharper Bonferroni procedure (43). In Hochberg's procedure, the m P values of interest (m = 5 in this case) are ranked from smallest (i = 1) to largest (i = 5). Tests were conducted in order from largest P value to smallest, and as soon as 1 test was significant, testing ceases and the remaining tests were declared significant. The ithP value is compared with the nominal level of significance of α/(m-i+1), where α is the family-wise error rate desired. For our sample, we wished to control for a family-wise error rate of α = 0.05, so the largest P value is compared with α/1 = 0.0500, second largest with α/2 = 0.0250, third largest with α/3 = 0.0167, forth largest with α/4 = 0.0125, and smallest with α/5 = 0.0100.

We evaluated the dose-effect of the intervention among those in the intervention group. We conducted analyses similar to the intervention group analysis, except that intervention group status was replaced by intervention class score and telephone call score in the analysis of covariance models.


Of 540 patients who initially fulfilled eligibility criteria, 281 (52%) were willing to participate. There was no significant difference between those initially willing and not willing to participate with respect to gender (68% versus 66%; P = 0.604), but those willing to participate were significantly younger on average (mean age 43 years versus 45 years; P < 0.0001) and more likely to be African American (61% versus 48%; P = 0.003). After excluding 47 patients who could not be reached after multiple attempts and 23 who became ineligible before baseline (i.e., no telephone access), 211 patients were randomized to the intervention (n = 105) or control (n = 106) group. Baseline characteristics are presented in Table 1. There were no significant differences between groups at baseline. In addition, we tested group differences on comorbidity using ACG categories and found no statistically significant differences.

Table 1. Participant characteristics*
 ALBPSMP (n = 105)Usual care (n = 106)P
  • *

    ALBPSMP = acute low back pain self-management program.

Age (range), years45.4 (19–77)45.5 (18–82)0.96
Female, %71.775.20.56
Education, %  0.97
 0–8 years7.77.7 
 9–11 years31.728.9 
 12 years35.636.5 
 >12 years25.026.9 
Married, %31.431.41.00
Living alone, %22.614.30.12
African American, %55.763.80.23
Completed high school, %60.663.50.67
Work for pay, %52.452.50.99
≤ Just making ends meet, %82.478.00.44

Intervention attendance.

Of 106 patients randomized to the intervention program, 30 (28.3%) attended at least 1 class, 66 (62.3%) received the intervention by mail/telephone/audiotape, and only 10 (9.4%) received no intervention. Of those who received the intervention by alternative mode of delivery (i.e., mail or audiotape), at least half received the information in the mail only, without a followup telephone call. The staff attempted to call these participants to confirm receipt, review the materials, answer questions, and assist with goal setting and problem solving.

Followup data.

Of the 211 patients randomized, 163 (77%; 76 intervention, 87 control) completed 4-month interviews (2 persons died, 6 dropped out of the study, 7 could not be contacted, 8 declined due to health problems or other reason, 11 moved with no forwarding address, and 15 had disconnected telephones). Compared with those who completed the study, dropouts were significantly more likely to be in the intervention group, younger, and were seen in the emergency department for the qualifying visit (Table 2). Patients who did not complete the study reported significantly worse physical function as measured by the Roland disability measure and lower self efficacy to manage symptoms than those who remained. There were no differences in any of our secondary measures. In addition, there were no significant interactions between treatment status and dropout.

Table 2. Baseline comparisons between participants who completed and participants who did not complete the followup assessment*
 Completed (n = 163)Dropped out (n = 48)P
  • *

    A higher score on Arthritis Impact Measurement Scales 2 and Roland low back pain-specific functional status indicates poorer functioning.

  • P ≤ 0.05.

  • P ≤ 0.01.

  • §

    P ≤ 0.001.

Randomly assigned to intervention, %46.662.50.05
Age, mean, years47410.01
Enrolled at emergency department, %38.766.70.00§
Roland disability, range 0–2414.316.90.02
Self efficacy, range 1–

Primary outcomes.

Table 3 displays the baseline and 4-month means of the primary outcomes by treatment group: functional status (physical functioning, affect, back pain symptoms), low back pain-specific functioning (Roland disability), and patient satisfaction with care for low back pain. The estimated treatment effects and 95% confidence interval (95% CI) are also shown. Using Hochberg's procedure, only the smallest P value (for affect) is declared to be significant. After 4 months, the intervention group significantly improved their affect (i.e., emotional functioning) compared with the control group (see Table 3). There were no significant changes in the other primary outcomes, although we observed a trend (P = 0.06) toward better low back pain-specific functioning in the intervention group.

Table 3. Baseline and 4-month means on primary outcomes: functional status and patient satisfaction*
 BaselineFour-monthAverage treatment effect
Treatment (n = 76)Control (n = 87)Treatment (n = 76)Control (n = 87)Effect95% CIP
  • *

    Data presented as mean (SD) unless otherwise noted. A higher score on functional status and low back pain (LBP)-specific functional status indicates poorer functioning. A higher score on patient satisfaction indicates higher satisfaction. The treatment effect is estimated from a general linear model adjusting for the baseline outcome and covariates. A negative value indicates that the acute low back pain group had a lower adjusted mean compared with the usual care group. 95% CI = 95% confidence interval.

  • P ≤ 0.0001.

Functional status, range 0–10       
 Symptoms6.2 (2.2)6.0 (2.2)4.7 (2.8)4.9 (2.6)−0.42(−1.21, 0.37)0.29
 Physical function2.5 (1.7)2.6 (1.6)2.0 (1.5)2.0 (1.5)0.06(−0.36, 0.47)0.79
 Affect4.4 (2.0)4.3 (1.8)3.6 (1.9)4.1 (1.9)−0.67(−1.12, −0.22)0.00
LBP specific       
 Functional status, range 0–2414.9 (6.7)13.9 (6.9)11.7 (7.2)12.5 (7.7)−1.80(−3.71, 0.11)0.06
 Patient satisfaction, range 10–3024.4 (5.1)24.4 (5.6)25.5 (5.3)25.6 (5.4)−0.06(−1.58, 1.45)0.94

Secondary outcomes.

Table 4 displays the baseline and 4-month means of the secondary outcomes by treatment group, along with the estimated treatment effect and 95% CI. After 4 months, the intervention group significantly increased their self efficacy to manage their back pain symptoms compared with the control group (P = 0.03). Furthermore, the intervention group increased time spent on average in self-management activities, including minutes spent in stretching (62%), minutes spent in strengthening (104%), and aerobic exercise (59%); however none of these increases approached statistical significance relative to the control group. Finally, the intervention group did significantly decrease their fears and avoidance of activities compared with the control group (P = 0.05).

Table 4. Baseline and 4-month means on secondary outcomes: role functioning and social support, self efficacy, and self management*
 BaselineFour-monthAverage treatment effect
Treatment (n = 76)Control (n = 87)Treatment (n = 76)Control (n = 87)Effect95% CIP
  • *

    Data presented as mean (SD) unless otherwise noted. A higher functional status score indicates poorer functioning. A higher score on self efficacy and fear/avoidance of activities scales indicates greater self efficacy and greater fear/avoidance of activities. Stretching, strengthening, aerobic exercise, and total physical activity time scores are listed as minutes spent per week and are self reported. Positive talk and relaxation score are frequency scores on a 0–5 scale. The treatment effect is estimated from a general linear model adjusting for the baseline outcome and covariates. A negative value indicates that the acute low back pain group had a lower adjusted mean compared with the usual care group. 95% CI = 95% confidence interval.

  • P ≤ 0.05.

Functional status, range 0–10       
 Role function3.2 (2.5)3.2 (2.6)2.2 (2.2)2.4 (2.4)−0.17(−1.12, 0.77)0.71
 Social support3.5 (1.9)3.8 (2.0)3.8 (2.0)3.6 (2.2)0.37(−0.20, 0.94)0.20
Self efficacy, range 1–106.1 (2.0)6.4 (2.2)7.3 (2.0)6.7 (2.2)0.69(0.06, 1.33)0.03
Self management, minutes spent per week       
 Stretching32.4 (52.5)41.3 (58.8)50.7 (56.8)42.4 (55.9)8.01(−9.45, 25.48)0.37
 Strengthening18.4 (42.9)31.3 (56.4)36.5 (53.1)27.5 (50.5)12.15(−3.61, 27.91)0.13
 Aerobic exercise62.8 (74.6)92.3 (88.9)99.9 (99.0)97.0 (108.8)14.91(−19.46, 49.27)0.39
 Total physical activity time116.6 (131.6)161.7 (172.7)186.9 (168.7)165.6 (176.3)39.50(−15.10, 94.09)0.15
Self management, range 0–5       
 Positive talk2.1 (1.9)2.4 (1.7)2.2 (1.9)2.5 (1.8)−0.12(−0.70, 0.46)0.67
 Relaxation1.1 (1.5)1.1 (1.5)1.7 (1.8)1.4 (1.7)0.26(−0.24, 0.76)0.31
Fear/avoidance, range 10–4024.6 (6.9)23.3 (6.6)22.4 (6.7)23.3 (6.1)−1.85(−3.67, −0.04)0.05

Dose effect of intervention.

For those completing the 4-month followup, the intervention dose scores ranged from 0 to 15 with the highest score reflecting attendance at all 3 classes. Approximately 8% received no dose; 24% received a score of 6, which could equal 1 class and 1 call; and 22% received a score 15, which could equal attendance at all 3 classes. The telephone call dose score ranged from 0 to 4 with approximately 17% receiving 1 call and 20% receiving 3 or 4 calls.

Among the primary outcomes, we found a significant dose-response effect for the intervention class score but not for the telephone call score for the Roland score. In this model, for each unit increase in intervention class score, ALBP-specific functional status decreased (improved) by 0.4 units (P = 0.01). Among the secondary outcomes, we found a significant effect for both intervention class and telephone call score on the stretching outcome. In this model, a unit increase in class score led to an increase in minutes spent stretching per week of 3.5 minutes (P = 0.03), and a unit increase in telephone call score led to a decrease in minutes spent stretching per week of 13.1 minutes (P = 0.05). We also found a significant effect for the class but not call score for fear and avoidance of activities, with a unit increase in class score leading to a 0.3 unit decrease in fear and avoidance of activities (P = 0.05).


This randomized trial evaluated the short-term (i.e., 4-month) effects of a self-management program designed for low income, inner city primary care patients with ALBP. The short-term outcomes indicated that the program was effective among these socioeconomically disadvantaged primary care patients; a group that shoulders a disproportionate burden of low back pain (17–21). Specifically, compared with patients receiving usual care, those in the intervention group reported significantly better emotional functioning (i.e., anxiety, depressive symptoms), back pain symptoms, and low back pain functioning. Moreover, we found significant increases in patients' confidence (i.e., self efficacy) to self manage their back pain symptoms and decreases in fear of movement/activity.

Our short-term results are similar to another evaluation of a back pain self-management program that was implemented among a sample of well-educated, employed, and predominantly white members of a health maintenance organization in the Pacific Northwest (16). Their self-management intervention group showed short-term (6 months) improvement in low back pain-specific functioning using the exact instrument (Roland disability) from our study. Again, as our study showed, the between-group difference was not statistically significant because the usual care group also improved (16). It is important to note that this study reported statistically significant improvements in low back pain-specific functioning at 6 months among self- management group members compared with the usual care group, and these differences were marginally significant at 12 months. Thus, low back pain improvements may result only after the patient has practiced self management over a longer period. Another study that offered structured exercise classes along with brief education also reported significant improvements in low back pain-specific functioning using the same instrument, Roland disability, at 6 and 12 months (44).

Our self-management program not only targeted increased physical functioning, but also improved emotional functioning by alleviating anxiety and depressive symptoms. Specifically, patients in the intervention group significantly improved their emotional functioning compared with the usual care group. Thus, exposure to a program that provides information on the basic understanding of the back pain (i.e., sources of back pain, red flag symptoms) and its natural history, and addresses the negative emotions associated with pain appears to effectively reduce patient emotional symptoms that may further exacerbate low back pain. In contrast, participants in the Von Korff et al study (16) had significantly reduced specific fears pertaining to low back pain, but did not significantly improve emotional functioning. The difference in our observations may be a result of different instruments (disease-specific versus generic) used to assess emotional functioning.

In addition to functional status, intervention patients increased self-management processes, i.e., time spent in strengthening activities and their self efficacy to manage future back pain episodes. Thus, it appears that the acute low back pain self-management format was sufficient for implementing the self-management process among inner city patients. However, the penetration of our program was limited. Nevertheless, initially 52% of eligible patients indicated a willingness to participate and less than 10% of those randomized to the treatment group received no form of the intervention. That is, the remaining 90% of the treatment group received a dose of the intervention.

In the dose-effect analyses, we found that higher levels of staff/participant interaction through classes resulted in more desirable outcomes for the functional status (i.e., low back pain specific function) items and fear and avoidance of activity items; however, staff/participant interaction through telephone calls was associated with less desirable outcomes for 1 self management outcome (i.e., stretching time per week). It is possible that this negative association may be a result of patients' health status. That is, those who were home long enough to complete telephone calls may have been in poorer health and less likely to have commitments outside of the home and less likely to attend any classes.

Our promising short-term results are limited by several factors. First, one-third of the self-management intervention group attended the group meetings. More than half of the intervention group received the intervention by an alternative mode (i.e., audiotapes, mail materials, telephone calls). For those who received followup calls, we verified that the participant received the materials and listened to the tapes. Interventions delivered by mail or telephone have shown to be efficacious (45, 46). Second, the data is self reported. However, all measures were validated instruments. Finally, those who dropped out of the study were significantly more likely to be younger and in the intervention group. This is not surprising given that the intervention group required active participation and placed more demands on the participants' time. Younger participants may have other time commitments, such as employment or family responsibilities. Furthermore, these patients were more likely to have visited the emergency department as their place for enrollment, reported worse physical functioning and specific low back pain functioning at baseline, and reported less self efficacy to manage back pain symptoms than those who remained in the study. These data suggest that patients presenting with severe back pain in the emergency department have less confidence to manage their symptoms and are less likely to remain in a self-management program. However, it is worth noting that we found no evidence that subjects first presenting in the emergency department who remained in the program responded differently to the intervention than those not first presenting in the emergency department.

Despite these limitations, the results of this study are encouraging and suggest that the self management model is robust in establishing patient self-management, particularly among urban and economically disadvantaged populations. Screening patients within primary care and obtaining physician consent using our computerized medical record system operated fairly smoothly in this study. With 50% of eligible patients interested in the study, further qualitative studies are needed to develop a more attractive program to this population. Given that one-third of the treatment group attended group classes, alternative methods of program delivery are needed to penetrate a greater proportion of the population. In addition, methods to track inner city patient relocation are necessary because patients frequently relocate without any forwarding information. Obtaining contact information from only 1 family member or friend of the patient is often insufficient. Perhaps patients can be rewarded or earn incentives for checking in with staff monthly.

We plan to analyze long-term outcomes of our acute low back pain self-management program to determine if self- management practices produce further changes in functional status and were cost effective.