The internet-based arthritis self-management program: A one-year randomized trial for patients with arthritis or fibromyalgia


  • Kate R. Lorig,

    1. Stanford University School of Medicine, Stanford, California
    Search for more papers by this author
    • Drs. Lorig and Laurent receive royalties from Da Capo Press for The Arthritis Helpbook. The program software used in this study is owned by Stanford University.

  • Philip L. Ritter,

    Corresponding author
    1. Stanford University School of Medicine, Stanford, California
    • 1000 Welch Road, Suite 204, Palo Alto, CA 94306
    Search for more papers by this author
  • Diana D. Laurent,

    1. Stanford University School of Medicine, Stanford, California
    Search for more papers by this author
    • Drs. Lorig and Laurent receive royalties from Da Capo Press for The Arthritis Helpbook. The program software used in this study is owned by Stanford University.

  • Kathryn Plant

    1. Stanford University School of Medicine, Stanford, California
    Search for more papers by this author

  • identifier: NCT00398255.



To determine the efficacy of an Internet-based Arthritis Self-Management Program (ASMP) as a resource for arthritis patients unable or unwilling to attend small-group ASMPs, which have proven effective in changing health-related behaviors and improving health status measures.


Randomized intervention participants were compared with usual care controls at 6 months and 1 year using repeated-measures analyses of variance. Patients with rheumatoid arthritis, osteoarthritis, or fibromyalgia and Internet and e-mail access (n = 855) were randomized to an intervention (n = 433) or usual care control (n = 422) group. Measures included 6 health status variables (pain, fatigue, activity limitation, health distress, disability, and self-reported global health), 4 health behaviors (aerobic exercise, stretching and strengthening exercise, practice of stress management, and communication with physicians), 5 utilization variables (physician visits, emergency room visits, chiropractic visits, physical therapist visits, and nights in hospital), and self-efficacy.


At 1 year, the intervention group significantly improved in 4 of 6 health status measures and self-efficacy. No significant differences in health behaviors or health care utilization were found.


The Internet-based ASMP proved effective in improving health status measures at 1 year and is a viable alternative to the small-group ASMP.


Arthritis is the most common cause of disability, a major reason for outpatient visits, and one of the most prevalent chronic conditions (1–3). For older people, arthritis is a common comorbidity that often complicates other chronic conditions such as heart disease and diabetes.

For the past 25 years, evidence has been presented for the effectiveness of self-management in the treatment of arthritis-related pain and disability (4–9). The American College of Rheumatology has called for self-management education in its standard of care for osteoarthritis (OA) (10). In addition, as part of its 2010 Goals for the Nation, the Centers for Disease Control and Prevention has called for an increase in the percentage of people participating in arthritis self-management education (11, 12).

Our study seeks to reach a group for whom small group programs are not available or, where available, not acceptable. Seventy percent of Americans use the Internet (13), and as of 2003, ∼93 million Americans (80% of adult Internet users) had searched for health information online (14). Internet-based education provides consistent program delivery, because both the content and process are programmed.

This study reports the results of a 1-year randomized trial of the Internet-based Arthritis Self-Management Program (ASMP). This program, offered via a secure Internet site, was built to replicate the original small-group ASMP as closely as possible. The aim of the study was to examine the 6-month and 1-year outcomes (health status, health behavior, self-efficacy, and health care utilization).

For the past 28 years we have developed, evaluated, and refined low-cost, community-based self-management programs in English and Spanish for people with arthritis. The format of these programs are similar in that they are taught in 2-hour sessions over several weeks in community sites such as churches, senior centers, and community centers. All programs are facilitated by 2 peer leaders who receive training and use a detailed protocol. These programs have demonstrated significant improvements in health status as well as health-related behaviors, and have reduced health care utilization (4–9). These programs have reached a relatively large number of patients, and have been replicated by others in studies with similar results (7, 15–18). The workshops are based on self-efficacy theory and use the processes of skills mastery (action planning and feedback), modeling as supplied by the facilitators, reinterpretation of symptoms, and group persuasion (19).



The Internet ASMP consists of password-protected, interactive, Web-based instruction (The Learning Center); Web-based bulletin board discussion (The Discussion Center); tools that the participants can use individually, such as exercise logs, medication diaries, and tailored exercise programs (My Tools); and the Arthritis Helpbook (20), which contains all of the program content. In addition, it contains discussions of the major types of arthritis and medications, and has drawings of suggested exercises. The book is not a textbook but rather a reference book to which the participants are referred at various times during the program.

The program is focused on reduction of pain and improvement of function. The Learning Center content includes design of individualized exercise programs; use of cognitive symptom management such as relaxation, visualization, distraction, and self-talk; methods for managing negative emotions such as anger, fear, and depression; an overview of medications; aspects of physician–patient communication; healthy eating; fatigue management; action planning; feedback; and methods for solving arthritis-related problems (Table 1). The workshop is taught in an interactive manner designed to enhance self-efficacy (21). In previous studies of the small-group ASMP, improvements in self-efficacy have been found to be associated with improved health status (21, 22).

Table 1. Weekly content of the program
TopicsWeek 1Week 2Week 3Week 4Week 5Week 6
Self-management principlesX    X
Goal setting/action plansXXXXXX
Pain managementX XXXX
Relaxation/cognitive pain managementX XXX 
Problem-solving steps X    
Fitness/exercise XX   
Feedback/problem-solving  X   
Difficult emotions  X   
Healthy eating   X  
Osteoporosis   X  
Fatigue and energy conservation   X  
Medication    X 
Depression    X 
Working with your health care professional    X 
Evaluating treatment plans     X
Sleep     X

A pair of peer moderators leads each workshop. Moderators are trained online by first participating in a workshop and then coleading the workshop with a previously-trained moderator. Moderators assist participants with the program by reminding them to log on, modeling action planning and problem solving, offering encouragement, and posting to the bulletin boards. They also monitor the daily posts of all participants and report inappropriate posts to the investigators. Thus, the moderators act both to facilitate the program and to provide safety. Unlike the small-group program, moderators do not deliver content, as this is programmed.

For each of 6 weeks, participants (approximately 25 per workshop) were asked to log on at least 3 times for a total of 1–2 hours and to participate in the weekly activities. These included reading the week's content on Web pages in The Learning Center (new content is posted each week), posting an action plan on the bulletin boards (The Discussion Center), and participating in self-tests. In addition, any problem a participant wishes to discuss can be posted on the bulletin board (The Discussion Center) and responded to by other group members and the moderators. Intervention subjects were also mailed the Arthritis Helpbook (20). The Internet program closely mirrors the original small-group program, except that it does not require real-time attendance and uses e-mail reminders to encourage nonparticipants to participate. Unlike the small-group program, appropriate tailored exercises are suggested via automated algorithms to each participant based on their answers to an online questionnaire designed to assess problems with each major joint function (23).

Subjects were continually recruited over 18 months in 2004 and 2005 by links to the study Web site placed on established Web sites, online newsletters, and discussion groups. Calendar announcements and articles in newspapers directed subjects to the study Web site. After reaching the study site, potential participants read about the study and left their e-mail address. Approximately 3 weeks before each workshop, the next 70–100 potential subjects on the waiting list were sent e-mails with a link to the study's informed consent and questionnaire. They first completed the online informed consent and then progressed to the questionnaire. The study and consent procedures were approved by the university institutional review board for human subjects. Approximately 50% of those who left their name on the waiting list completed the informed consent and baseline questionnaire.

Following completion of the online questionnaire, participants were randomized to either the intervention group or to a control group. Control subjects continued with their usual care with no intervention, and were sent a $10 certificate for each questionnaire completed. Treatment subjects received their usual care plus the intervention. Data were collected online at baseline, 6 months, and 12 months. Missing data were collected by phone.


Subjects met all of the following criteria: ≥18 years of age; a diagnosis of OA, rheumatoid arthritis (RA), or fibromyalgia; could have other chronic conditions but could not have been in active treatment for cancer for 1 year; had not participated in the small-group ASMP or the Chronic Disease Self-Management Program; had access to a computer with Internet and e-mail capabilities; agreed to 1–2 hours per week of log-on time spread over at least 3 sessions/week for 6 weeks; and were able to complete the online questionnaire.

Outcome measures.

There were 6 health-related quality of life measures. Visual numeric scales were used to measure pain and fatigue. They were developed at Stanford and found to correlate highly with more common visual analog scales and to be easier to use (24). The Health Distress Scale was adapted from the Medical Outcomes Study (25). It focuses on the distress specifically associated with health problems. It had a Cronbach's alpha of 0.87 when measured on 1,130 subjects with various chronic diseases, and a test–retest reliability of 0.87 when readministered to 51 subjects. The Self-Rated Global Health Scale comes from the National Health Survey and has been found to be predictive of future health status (26). The Activities Limitation Scale measures the impact of disease on role activities such as recreation and chores (27). Its internal consistency reliability (Cronbach's alpha) was 0.91. An 8-item version of the Health Assessment Questionnaire measures disability and is based on the measure used in the National Health Survey (23, 26). It had an internal consistency reliability of 0.85. Detailed information about each of the instruments used in this study, including descriptions of their psychometric properties, may be found online (28).

Four health-related behaviors were measured: stretching and strengthening exercise, aerobic exercise, use of cognitive symptom management techniques, and use of techniques to improve communication with health care providers. These instruments were developed and validated by the Stanford Patient Education Research Center for use in previous studies (27, 28).

Five utilization measures were recorded: self-reported outpatient visits to physicians, emergency room visits, nights in the hospital, chiropractic visits, and physical therapy visits. In a previous study, we found correlations between chart audit data and both self-report of outpatient visits (r = 0.70) and days in the hospital (r = 0.83). Discrepancies were consistent over time and across treatment groups (29).

We also measured a perceived Arthritis Self-Efficacy Scale, which reflects participants' confidence to manage their arthritis. The short (8-item) scale was built on an earlier model of the scale (21) and had a Cronbach's alpha of 0.94.

Additional variables.

Demographic variables included age, sex, years of education, marital status, and ethnicity. We also documented type of rheumatic disease, and enumerated the total number of times each treatment participant logged into the class Web site and how many weeks they logged on at least once.

Statistical analyses.

Based on prior studies of the small-group ASMP, it was hypothesized that program participants, in comparison with those randomized to the usual treatment control group, would have better outcomes at 6 months for health indicators, health behaviors, and self-efficacy and would have reductions in health care utilization. It was further hypothesized that these differences in outcomes would be maintained at 1 year after entry into the program. We used t-tests to compare baseline demographic and outcome variables for the intervention participants with those for the usual care controls. Baseline outcome, disease, and demographic measures for 6-month and 1-year noncompleters (those who failed to complete questionnaires) were then compared with the baseline measures of those who completed questionnaires, again using t-tests.

Repeated-measures analyses of variance were used with the 3 time points to determine whether there was a time-randomization interaction. Because there was a slight difference in percentage of married participants at baseline for the intervention versus control groups, marital status was included as a covariate. For all outcome variables with significant time-randomization interactions, post hoc analyses were conducted. Analyses of covariance (ANCOVAs) were used to determine results at 6 months and 12 months.

The data set was then segmented based on type of rheumatic condition. Analyses were rerun separately for those who had RA, for those who had OA but not RA, and finally for all those with fibromyalgia but not RA or OA. All analyses were also run substituting last-reported value for any missing cases (intent-to-treat [ITT] analyses).

Finally, we examined how many of the 6 health-related quality of life outcomes improved by a ≥0.30 effect size (defined as the change score divided by the pooled SD of the baseline score) for each individual. We computed a sum of the number of ≥0.3 improvements (0 to 6), and noted whether each participant had ≥3 improvements of ≥0.3. Treatment participants were compared with the control group participants using t-tests for the sum of improvements and chi-square tests for those who had ≥3 versus <3 effect-size improvements of 0.3.

All data analyses were done using SAS software, version 9.1 (SAS Institute, Cary, NC).



We received completed online informed consent and baseline questionnaires from 855 subjects who were immediately randomized, with 422 in the usual care control group and 433 participating in the intervention. Of these, 331 (78%) usual care and 310 (72%) intervention subjects completed 6-month questionnaires. Similar numbers (344 [82%] usual care group and 307 [71%] intervention group) completed 1-year questionnaires. Of the 651 participants who completed the 1-year questionnaire, 81 (12.4%) had skipped the 6-month questionnaire. Of all the participants, 570 (67%) completed all 3 questionnaires (see Figure 1 for disposition of study subjects).

Figure 1.

Study participants.

Participants could report having more than 1 type of arthritis. Of the participants, 546 (63.9%) reported having OA, 238 (27.8%) reported having RA, 441 (51.6%) reported having fibromyalgia, and 115 (13.5%) reported having other arthritic conditions (Table 2). We requested a confirmation of diagnosis from the participants' physicians, of whom 589 (68.0%) replied to the request and confirmed the patients' self-report of diagnosis in all but 6 cases. These 6 subjects were dropped from analyses.

Table 2. Baseline means*
VariableUsual care (n = 425)Online intervention (n = 441)P (difference)
  • *

    Values are the mean ± SD unless otherwise indicated. The range and direction are given with each variable group, where applicable. An upward arrow indicates that a higher value is desirable, and a downward arrow that a lower value is desirable. RA = rheumatoid arthritis; OA = osteoarthritis; FSM = fibromyalgia.

  • From t-tests comparing the usual care and intervention groups.

  • Participants could report more than 1 disease.

Demographic variables   
 Age, years (range 22–89)52.5 ± 12.252.2 ± 10.90.389
 Female, %90.589.80.776
 Years of education (range 8–23)15.7 ± 3.1115.6 ± 3.090.236
 Married, %
 Non-Hispanic white, %93.790.90.134
 Health-related Web site visits last 6 months2.85 ± 11.682.87 ± 11.20.816
 Diseases, %   
Health indicators ↓   
 Health distress (0–5)2.41 ± 1.192.45 ± 1.170.633
 Self-reported global health (0–5)3.23 ± 0.9093.19 ± 0.9760.477
 Disability (0–3)0.582 ± 0.4560.552 ± 0.4020.312
 Activity limitation (0–4)2.18 ± 1.052.20 ± 1.030.770
 Fatigue (0–10)6.51 ± 2.286.46 ± 2.290.736
 Pain (0–10)6.41 ± 2.236.53 ± 2.230.391
Health behaviors ↑   
 Aerobic exercise (minutes/week)90.4 ± 101100.0 ± 1040.169
 Stretching and strength exercise (minutes/week)42.2 ± 50.747.0 ± 55.70.183
 Communication with physician (0–5)3.11 ± 1.123.13 ± 1.180.836
 Practice stress management (times/week)2.52 ± 4.672.48 ± 4.660.891
Self-efficacy (1–10) ↑4.96 ± 1.965.04 ± 2.120.530
Health care utilization (past 6 months)   
 Physician visits6.00 ± 7.385.60 ± 5.630.373
 Emergency visits0.289 ± 0.9080.238 ± .6880.339
 Days in the hospital0.665 ± 3.380.578 ± 4.970.818
 Chiropractic visits1.95 ± 6.720.80 ± 3.950.040
 Physical therapist visits3.23 ± 8.332.77 ± 8.450.425

Sixty-one percent of participants reported at least 1 nonrheumatic condition. The most common additional diagnosis was hypertension (12.3%). There were no significant differences between treatment and control participants in incidences of comorbid conditions.


Table 2 compares the mean values at baseline for usual care controls and intervention participants. The only significant differences among the outcome measures was for chiropractic visits, with usual care control participants averaging ∼1 more visit in the past 6 months (P = 0.040). There were no significant differences between the groups in type of rheumatic condition or demographic variables, although there was a trend toward a significant difference in the proportion of married participants, with the usual care group being slightly more likely to be married (71.1% versus 65.5%; P = 0.081).


At baseline, those who failed to complete their 6-month questionnaire had significantly higher levels of health distress, activity limitation, and fatigue, and had greater numbers of emergency room visits but fewer hospitalizations (all P < 0.05). They were also younger (49.7 years versus 53.9 years; P < 0.001) and more likely to be married than those who completed their questionnaires (69% versus 63%; P = 0.021).

Those who failed to complete their 12-month questionnaire had significantly higher levels of health distress and activity limitation at baseline and, in addition, did less well on the communication with physician scale (all P < 0.05). They were also younger (mean age 49.2 years versus 54.1 years; P < 0.001), and had fewer years of education (15.4 years versus 15.9 years; P = 0.025).

When the 6-month treatment noncompleter group (n = 123) was compared with the 6-month control noncompleter group (n = 91), the only significant difference in baseline values of outcome variables was hospitalizations (0.11 visits versus 0.03 visits; P = 0.04). Education was also significantly different, with the treatment noncompleter group averaging 15.9 years and the control group 15.0 years (P = 0.03).

When the 1-year treatment noncompleter group (n = 126) was compared with the control noncompleter group (n = 78), the only significant difference at baseline was mean chiropractic visits in the last 6 months (1.9 visits versus 1.2 visits; P = 0.02). There were no significant differences in the demographic variables.

Participation and Web usage.

Of those randomized to the intervention group, 24 never participated (11 dropped out before being assigned a class and 13 never logged in to their assigned class). The remaining 409 participants logged in a mean ± SD of 31.6 ± 24.5 times over 6 weeks (range 1–220). Each workshop of 25 participants generated between 400 and 600 posts to The Discussion Center. Participants reported that they found The Learning Center and The Discussion Center very helpful.

At each time point (baseline, 6 months, and 12 months) there were no significant differences between the intervention group and the usual care control group participants in the frequency of visits to non-ASMP health-related Web sites. The averages for both groups of participants were between 3 and 4 times per month.


Using the ITT approach, 4 of the 6 health status variables had significant time-randomization interactions (health distress, activity limitation, self-reported global health, and pain) (Table 3). In order to correct for multiple outcomes, we applied the Bonferroni correction and used a criteria of 0.008 rather than 0.05 for statistical significance, because we were testing 6 health outcome variables simultaneously. The same 4 outcomes remain statistically significant. All 6 changes were in the predicted direction (with the intervention group improving more than the usual care control group at 6 months and 1 year when compared with baseline). The time-randomization interaction for self-efficacy was also significant. There were no significant time-randomization interactions for health behaviors or health care utilization.

Table 3. Six-month and 1-year outcomes within randomized program (usual care or intervention)*
VariableTime-random interactionUsual care mean ± SDOnline intervention mean ± SD
P (ITT)P (nonmissing)Baseline (n = 375)6 months (n = 331)1 year (n = 344)Baseline (n = 347)6 months (n = 310)1 year (n = 307)
  • *

    The range and direction are given with each variable, where applicable. An upward arrow indicates that a higher value is desirable, and a downward arrow that a lower value is desirable. Probabilities are from repeated-measures tests of time-randomization interaction (Wilks' λ). ITT = intent-to-treat.

  • Baseline means are for all with any followup (either 6 months or 1 year).

Health indicators ↓        
 Health distress (0–5)< 0.001< 0.0012.37 ± 1.192.34 ± 1.162.25 ± 1.192.41 ± 1.202.03 ± 1.182.00 ± 1.18
 Activity limitation (0–4)< 0.001< 0.0013.22 ± 0.9033.42 ± 1.853.29 ± 0.8853.17 ± 0.9733.10 ± 0.9713.09 ± 0.962
 Self-reported global health (0–5)0.0040.0130.569 ± 0.4460.598 ± 0.4830.573 ± 0.4570.547 ± 0.4010.515 ± 0.4560.514 ± 0.445
 Disability (0–3)0.1100.2852.16 ± 1.052.19 ± 1.072.11 ± 1.042.17 ± 1.031.97 ± 1.321.90 ± 1.15
 Fatigue (0–10)0.0800.0386.47 ± 2.306.38 ± 2.196.33 ± 2.206.47 ± 2.366.07 ± 2.515.95 ± 2.63
 Pain (0–10)< 0.0010.0096.37 ± 2.226.34 ± 2.316.10 ± 2.356.53 ± 2.275.86 ± 2.445.77 ± 2.53
Health behaviors ↑        
 Aerobic exercise (minutes/week)0.4870.31988.5 ± 10090.5 ± 93.991.0 ± 92.899.1 ± 104103.3 ± 102.7111.5 ± 109
 Stretching and strength exercise (minutes/week)0.8110.63640.4 ± 49.147.0 ± 55.945.2 ± 55.546.3 ± 55.654.7 ± 56.350.9 ± 55.0
 Communication with physician0.2130.1633.14 ± 1.133.28 ± 1.073.28 ± 1.093.17 ± 1.203.46 ± 1.093.42 ± 1.19
 Practice stress management (times/week)0.4690.2192.48 ± 4.423.11 ± 6.132.77 ± 5.022.37 ± 4.662.93 ± 4.833.19 ± 6.85
Self-efficacy (1–10)0.0180.0034.96 ± 1.985.14 ± 2.055.34 ± 2.065.08 ± 2.135.69 ± 2.075.89 ± 2.09
Health care utilization (past 6 months)        
 Physician visits0.3140.2045.67 ± 6.585.76 ± 6.105.22 ± 5.845.69 ± 5.724.90 ± 6.615.25 ± 5.86
 Emergency visits0.5710.2580.181 ± 0.6610.188 ± 0.5250.243 ± 5.150.175 ± 0.5320.166 ± 0.4590.182 ± 0.529
 Days in the hospital0.9660.6280.667 ± 3.330.504 ± 1.910.632 ± 2.150.634 ± 5.520.471 ± 2.490.609 ± 2.36
 Chiropractic visits0.9490.4051.94 ± 6.411.74 ± 5.551.87 ± 6.741.48 ± 5.601.32 ± 4.411.32 ± 5.52
 Physical therapist visits0.4240.6403.22 ± 8.492.84 ± 7.192.33 ± 7.162.95 ± 8.622.69 ± 7.722.51 ± 7.08

Table 3 also shows the P values when only cases with nonmissing values were used; after correcting for multiple comparisons, 3 of 6 health indicators had statistically significant interactions, as did self-efficacy. Again, none of the health behaviors or health utilization outcome measures had statistically significant time-randomization interactions.

Post hoc analyses for outcome variables with significant time-randomization interaction were conducted using ANCOVAs (again, ITT). At 6 months, all 4 outcome variables had statistically significant differences between intervention and usual care control participants, with the intervention group improving more. These differences were maintained at 1 year compared with baseline and remained statistically significant (using ANCOVAs). Looking at the means at each time point (Table 3), one can see that for the intervention group, there were improvements in health indicators between baseline and 6 months and then either a leveling off or a slight continued improvement from 6 months to 1 year. Control group participants tended to remain the same or to slightly worsen in the first 6 months and then improve slightly over the next 6 months. The differences between intervention and control groups that developed in the first 6 months remained at 1 year.

Changes by effect sizes of ≥0.3.

At 12 months, 44% of the treatment participants had improvements of ≥0.30 effect size for ≥3 of the 6 health indicators, compared with 30% of the usual care control group (P < 0.001, calculated by chi-square test). The mean number of improvements at 1 year of ≥0.30 effect sizes among the 6 health indicators was 2.4 for treatment participants and 1.8 for usual care control participants (P < 0.001).

Changes by diagnosis.

Table 4 reports the baseline to 1-year changes by diagnosis (RA, OA, and fibromyalgia). For this analysis, all people with RA were coded as having RA. All people with only OA or with OA and fibromyalgia were coded as OA, and people whose only diagnosis was fibromyalgia were coded as fibromyalgia. The difference in mean 12-month change scores between the treatment and control groups were generally lower for the fibromyalgia participants than for the RA and OA participants, and thus not an artifact of reduced power due to fewer subjects. The differences in change scores were similar for those with OA versus those with RA but no OA. For example, for health distress, there was a difference of 0.414 between intervention and control participants for those with OA but not RA, and a difference of 0.408 for those with RA (intervention participants showing greater improvement). For those with only fibromyalgia, the difference was only 0.035.

Table 4. One-year changes for treatment and control subjects by disease: health behaviors, health status, and health care utilization*
Treatment (n = 72)Control (n = 72)PTreatment (n = 134)Control (n = 158)PTreatment (n = 40)Control (n = 46)P
  • *

    Values are the mean ± SD unless otherwise indicated. The range and direction are given with each variable, where applicable. An upward arrow indicates that a higher value is desirable, and a downward arrow that a lower value is desirable. Probabilities are from repeated-measures tests of a time-randomization interaction (Wilks' λ), using intent-to-treat. RA = rheumatoid arthritis; OA = osteoarthritis; FM = fibromyalgia.

Health indicators ↓         
 Health distress (0–5)−0.385 ± 1.040.069 ± 0.9730.066−0.455 ± 1.03−0.047 ± 0.908< 0.001−0.388 ± 1.260−0.353 ± 1.080.977
 Self-reported global health (0–5)−0.085 ± 0.8060.250 ± 0.7270.005−0.030 ± 0.7200.095 ± 0.6160.016−0.125 ± 0.822−0.174 ± 0.7390.271
 Disability (0–3)−0.003 ± 0.4060.024 ± 0.3320.850−0.060 ± 0.3290.014 ± 0.2170.0150.013 ± 0.3610.035 ± 0.3590.814
 Activity limitation (0–4)−0.366 ± 1.020.132 ± 0.7340.003−0.258 ± 1.030.002 ± 0.7300.002−0.444 ± 0.924−0.233 ± 0.8430.308
 Fatigue (0–10)−0.361 ± 2.000.056 ± 1.940.925−0.403 ± 2.240.038 ± 1.890.017−0.825 ± 2.37−0.804 ± 1.570.487
 Pain (0–10)−0.514 ± 2.79−0.069 ± 1.690.040−0.806 ± 2.44−0.158 ± 2.180.005−0.875 ± 2.48−0.478 ± 1.870.087
Health behaviors ↑         
 Aerobic exercise (minutes/week)−10.1 ± 98.0−1.18 ± 98.80.96322.6 ± 100.60.316 ± 100.30.2609.75 ± 10627.7 ± 68.220.215
 Stretching and strength exercise (minutes/week)3.99 ± 54.7−1.67 ± 54.20.8776.29 ± 55.88.26 ± 58.00.999−4.25 ± 57.09.26 ± 50.50.205
 Communication with physician (0–5)0.403 ± 1.150.130 ± 0.9520.1250.201 ± 0.9920.040 ± 0.8720.294−0.258 ± 1.160.152 ± 0.9340.187
 Practice stress management (times/week)0.884 ± 3.340.449 ± 3.190.2791.070 ± 8.310.467 ± 5.860.728−0.025 ± 6.59−0826 ± 4.170.234
Self-efficacy (1–10) ↑0.783 ± 1.320.242 ± 1.590.2820.801 ± 2.170.259 ± 1.790.0210.643 ± 2.100.579 ± 1.560.707
Health care utilization (past 6 months)         
 Physician visits−0.875 ± 4.99−0.903 ± 5.920.065−0.776 ± 5.63−0.102 ± 4.310.470−0.125 ± 4.33−0.217 ± 4.560.903
 Emergency visits−0.083 ± 0.5990.014 ± 0.5930.065−0.037 ± 0.5120.102 ± 1.490.4910.0 ± 0.506−0.130 ± 1.470.864
 Days in the hospital0.611 ± 3.27−0.333 ± 3.520.200−0.142 ± 2.81−0.191 ± 2.820.748−2.15 ± 15.8−1.09 ± 7.370.284
 Chiropractic visits−0.169 ± 2.50−0.214 ± 0.8150.0870.641 ± 6.82−0.252 ± 4.880.590−0.425 ± 3.69−1.15 ± 8.760.203
 Physical therapist visits−0.268 ± 7.12−0.714 ± 10.40.302−0.411 ± 11.0−0.510 ± 9.340.302−3.20 ± 12.9−0.717 ± 7.790.310

The significance levels shown in Table 4 are from 3 time-point repeated measures using ITT methodology. When the analyses within diseases were rerun using only cases with nonmissing data, the results were very similar. The same analyses were run using participants with only fibromyalgia, those with only RA, and those with only OA. The results (not shown) again indicated little effect of the intervention on fibromyalgia patients.


Our study had several possible limitations. Participants could not be blinded to the intervention, and thus we cannot rule out the possibility of an attention effect. However, it is unlikely that 6 weeks of attention would have sustained effects for 1 year.

By looking at 6 health indicators and 4 health behaviors, as well as self-efficacy and 5 health care utilization variables, we increased the risk of Type II error. Thus, it is important to consider the overall consistency of the results of the intervention. The differences in change scores all favored the treatment group (except for range of motion exercise, in which the mean for the control group was 1 minute/week more than for the intervention group) and all significant outcomes were in the expected direction.

The online intervention was limited to those who are computer literate and who have access to the Internet. As might be expected, this was a highly educated group, with a mean education equivalent to nearly 4 years of college. Although this program may not be accessible to everyone, the use of computers allows for participation by people who cannot or will not attend a small-group program. In addition, as computers and the Internet continue to become more accessible, this form of education will be available to a larger population.

Our hypotheses that there would be improvement in health status, health behaviors, and health care utilization at 6 months and 12 months after beginning the online program were partially confirmed. Comparing intervention with usual care control groups, there were significant differences in health statuses. These differences were all found at 6 months and remained significantly improved at 1 year.

Patients with only fibromyalgia appeared to have fewer benefits from the program. It should be noted that these are long-term, nonreinforced changes, and, while significant, are modest. The question arises: do they have clinical meaning? The population is very heterogeneous for disease, age, education, and symptom distribution. In addition, there were no exclusion criteria based on severity of symptoms. Thus, many of the people in the study entered with low levels of 1 or more symptoms and therefore had little room for improvement (floor effect). All of these factors influence effect sizes. Forty-four percent of the treatment group reported ≥0.3 effect size improvements for ≥3 of the 6 health indicators. We know that participants were enthusiastic about the intervention, and reported important personal changes.

In the original ASMP, improvements in self-efficacy measured soon after the completion of the program were associated with improved outcomes at 1 year. The possible mediating effects of self-efficacy within the Internet-based ASMP should also be examined in future studies.

The majority of previous studies have demonstrated positive changes in health status, and sometimes reduction in health care utilization, from randomized trials at 4 or 6 months. The current study adds a randomized comparison at 1 year with continued findings of positive results for health status measures. One recent study failed to replicate these findings, and another study only partially replicated the findings (30, 31). It is unclear whether this was due to the intervention, the fidelity by which the intervention was replicated, the population selection, comorbidity, the sensitivity of the instruments used, or other methodologic problems. If arthritis self-management education is to meet its full potential for improving the lives of people with arthritis, future studies should focus on what factors provide for successful and effective translation from the research setting into practice. There is also a need to further define the population for whom the Internet-based ASMP is most effective.

In summary, the Internet-based ASMP, like its predecessor small-group program, appears to be effective in slowing or reducing the negative effects of arthritis over a 1-year period of time. Both programs should be considered for assisting patients with arthritis. Future studies should focus on the facilitators and barriers to widespread translation as well as to the specific population for whom the programs are effective.


Dr. Ritter had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Lorig, Ritter, Laurent, Plant.

Acquisition of data. Lorig, Laurent, Plant.

Analysis and interpretation of data. Lorig, Ritter.

Manuscript preparation. Lorig, Ritter, Laurent, Plant.

Statistical analysis. Ritter.


We thank Rose Sage Barone and Michael Halaas of the Stanford Information and Technology Office for their technical and programming assistance.