This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Diabetes self-management education has been shown to be effective in controlled trials. However, few programs that meet American Association of Diabetes Educators standards have been translated into widespread practice.
This study examined the translation of the evidence-based Better Choices, Better Health-Diabetes program in both Internet and face-to-face versions.
We administered the Internet program nationally in the United States (n=1010). We conducted face-to-face workshops in Atlanta, Georgia; Indianapolis, Indiana; and St. Louis, Missouri (n=232). Self-report questionnaires collected health indicator, health behavior, and health care utilization measures. Questionnaires were administered on the Web or by mail. We determined hemoglobin A1c (HbA1c) from blood samples collected via mailed kits. Paired
Of the 1242 participants, 884 provided 6-month follow-up questionnaires. There were statistically significant improvements in 6 of 7 health indicators (including HbA1c) and in 7 of 7 behaviors. For each of the 5 conditions, there were significant improvements among those with the condition (effect sizes 0.59–1.1). A total of 662 (75.0%) of study participants improved at least 0.4 effect size in at least one criterion, and 327 (37.1%) improved in 2 or more.
The Diabetes Self-Management Program, offered in two modes, was successfully disseminated to a heterogeneous national population of members of either insured or administered health plans. Participants had small but significant benefits in multiple measures. The program appears effective in improving diabetes management.
Type 2 diabetes is a growing problem. Not only is the prevalence increasing, but also there are increased costs and decreased quality of life. Patients, health care providers, families, employers, and insurers all share this burden. The US government, in Healthy People 2020, set national objectives for improvements in diabetes care and outcomes [
Diabetes self-management education has been shown to be effective in improving health behaviors and in some cases lowering hemoglobin A1c (HbA1c). However, <7% of privately insured persons receive formal education in the year after diagnosis [
The Administration for Community Living (US Department of Health and Human Services) and the Evidence-Based Leadership Council have varied in their definitions of evidence-based programs [
In this paper, we present the national translation of an evidence-based diabetes education program, The Stanford Diabetes Self-Management Program, also known as Better Choices, Better Health-Diabetes (BCBH-D) [
Recent systematic reviews have found that both community and Internet interventions have been effective in reducing HbA1c and improving quality of life for people with type 2 diabetes [
For this translation study, the US National Council on Aging led a collaboration of 5 organizations. It also provided the platform for offering the Internet workshops. Anthem, Inc (Indianapolis, IN, USA) -affiliated health plans recruited participants from their members. The Young Men’s Christian Association of the United States of America in Atlanta, Georgia, and OASIS Institute in St. Louis, Missouri, and Indianapolis, Indiana, offered community programs. The Stanford Patient Education Research Center (Stanford University School of Medicine, Stanford, CA, USA) collected data and provided data analysis. HealthCore (Wilmington, DE, USA), a subsidiary of Anthem, Inc, contributed to the design and data analysis. The team members met for months prior to the beginning of the study to decide on study design, outcome variables, and logistics. They continued to meet at least monthly for the duration of the intervention. The study was approved by the Stanford University Institutional Review Board and New England Independent Review Board.
We hypothesized that over 6 months, people with diabetes participating in the program would demonstrate (1) a reduction in HbA1c, (2) a reduction in symptoms (hypoglycemic symptoms and depression), (3) increases in healthful behaviors (exercise, communication with physicians, and medication adherence), (4) increases in receiving recommended tests (eye, foot, cholesterol, and kidney examinations). We also hypothesized (5) that program effectiveness would be independent of mode of delivery, (6) that participants with baseline HbA1c ≥9.0, 8-item Patient Health Questionnaire (PHQ-8) depression ≥10.0, having 2 or more hypoglycemia symptoms, medication nonadherence, or no aerobic exercise would have clinically significant improvements in the variables of interest, and (7) that most of the participants would have a moderate effect size (0.4) improvement in either reducing HbA1c, reducing depression, increasing medication adherence, reducing hypoglycemic symptoms, or increasing exercise.
In addition, we wanted to explore the potential of BCBH-D to meet some of the Healthy People 2020 objectives: decrease the proportion of people with diabetes with HbA1c >9, and increase the proportion of people with diabetes who (1) have an annual foot examination, (2) have an annual urinary microalbumin measurement, and (3) receive formal diabetes education.
We chose BCBH-D because it was developed for people with type 2 diabetes in real-world settings. We built both programs (face-to-face and Internet) on extensive patient input, as well as assistance from certified diabetes educators. Both programs have been shown to be effective in previous randomized trials [
The BCBH-D Internet program is a password-protected, interactive, Web-based program. The user interface consists of 3 major sections. (1) The Learning Center offers 20–30 pages of didactic and interactive content each week. In addition to reading content within the Learning Center, participants make weekly action plans, give feedback on the plan from the previous week, and answer a question such as “What problems do you have with your diabetes?” The action plan, feedback, and questions populate the Discussion Center. (2) The Discussion Center contains 4 interactive bulletin boards: problem solving, action planning, difficult emotions, and celebrations. Participants can post to any of these boards and respond to posts at any time. (3) My Tools is a series of tools, such as a medication log, food diary, exercise diary, and links to other websites that can be used as participants wish.
The Internet facilitators are trained over the Web by first participating in a workshop, followed by Web-based training and then cofacilitating the workshop with an experienced facilitator. Certified diabetes educators are available to facilitators to answer questions as necessary. All interactions between facilitators and participants take place on the Web. Facilitators assist participants with the program. They model action planning and problem solving, and offer encouragement by posting to the discussion boards. Facilitators monitor the daily posts of all participants and report inappropriate posts to program administrators. They also have access to a certified diabetes educator. Unlike in the community program, in the Internet program facilitators do not deliver content, as this is scripted on the Web.
For the Internet program, 25–30 participants log on at least 3 times a week and participate in weekly activities, including reading content, posting an action plan, and interacting on the discussion boards. Any problem a participant wishes to discuss can be posted in the Discussion Center and responded to by other participants. The Internet program mirrors the original community program, except that it does not require real-time attendance. Participants can log in any time, have no face-to-face interaction, and may return to past weeks’ material [
The community workshops meet for 2.5 hours a week; 10–15 people attend and may bring a friend or family member. This is longer than the 10 hours usually covered by insurance participant’s health plan benefits. The length was determined by the amount of material to be covered, as well as past work that indicated that shorter programs were less effective [
Both the community and Internet BCBH-D content meets the diabetes self-management education recommendations for diabetes self-management education and support [
Because we wanted to replicate real-world settings, we stipulated few inclusion (have type 2 diabetes and be covered by an Anthem plan) or exclusion criteria (currently pregnant or had chemotherapy or radiation treatment for cancer in the past year). There were no inclusion criteria based on level of symptom severity, nor was having had cancer by itself an exclusion criterion.
We recruited Internet participants by email or announcements from their employers or emails directly from an Anthem plan. A small percentage of participants were referred by family, friends, or physicians. Participants in both commercial and Medicare Advantage health insurance programs were eligible. Potential study participants went to the recruitment website, completed a screening form, and, if they met study requirements, completed an informed consent and baseline questionnaire.
We recruited face-to-face participants through mailings, flyers in workplaces, case managers, and automated telephone calls. They were asked to call a local number for details about the workshop, screened for study eligibility, and registered. Programs were available in Atlanta, Indianapolis, and St. Louis. A small percentage of the community participants were not covered by an Anthem plan. All other screening criteria were the same for Internet and community participants.
Participants enrolled between October 2013 and October 2014.
We collected data at baseline, 6 months, and 12 months. This paper reports only on the 6-month results. Internet participants completed consent forms and all questionnaires on the Web. Face-to-face participants completed informed consent forms and baseline questionnaires within a week of their first session. Follow-up questionnaires were sent by mail.
We asked potential participants to supply a capillary sample of blood for HbA1c testing. If willing, they were sent a CoreMedica home test kit (CoreMedica Laboratories, Lees Summit, MO, USA). These were returned to the investigators, bar coded to avoid disclosing private information, and then sent to CoreMedica, a Clinical Laboratory Improvement Amendments-certified laboratory [
We chose measures to be of interest to patients, providers, and the health care system. We gathered all data except HbA1c from validated self-report questionnaires. Demographic variables included age, sex, race, ethnicity (non-Hispanic white), education, and marital status. In addition, we recorded other diseases (asthma, chronic obstructive pulmonary disease, other lung disease, hypertension, heart disease, renal disease, arthritis, cancer, depression or other mental condition). Outcome measures, described below and in full detail elsewhere, fell into two broad categories: health indicators and health behaviors.
All health indicators have been validated in previous studies. A higher score indicates greater symptoms or worse health. Self-rated health consists of a single item from the National Health and Nutrition Examination Survey (range 1–5) [
The health behaviors measures have also been validated elsewhere. A 3-item scale measured communication with physicians. It used a 6-point scale (from never to always) to measure how often the participant prepared a list of questions for the physician related to the illness [
Translation studies, because of study population heterogeneity, present unique methodological challenges. In many diabetes studies, participants are chosen based on having difficulty with the outcome of interest. Thus, participants enter into studies because of high HbA1c, depression, hypoglycemia, or being nonadherent. In this study, no such screening occurred. This results in greater heterogeneity, as not all participants have the same problems and some may have none of the problems. Consequently, we conducted two types of analyses. The first, or classic, analyses determined the changes, significance, and effect sizes for the population as a whole. The second, or subset, analyses examined only that portion of the study population that demonstrated problems in the variable of interest, that is, high HbA1c or low adherence to taking medications. A third analysis sought to reconcile these two by examining the percentage of the total population who achieved a moderate benefit (0.4 effect size) in at least one of the variables of interest.
Univariate statistics described demographic characteristics. Independent sample
For participants who had had no examinations (eye, foot, cholesterol, or kidney) in the year prior to entry, we calculated the percentage who had examinations in the 6 months following baseline (hypothesis 4).
To determine effectiveness by mode of delivery (hypothesis 5), we used independent sample
We conducted subgroup analyses for participants with specific indicators (hypothesis 6). These were HbA1c >9; clinical depression (PHQ-8 of ≥10 [
To determine whether a large portion of the participants benefited on at least one important outcome, we calculated the percentage who had an improvement of at least 0.4 effect size (change score divided by baseline standard deviation) in at least one of the 5 criterial measures (hypothesis 7). We also calculated the mean number of improvements (out of 5) and examined how this varied by the number of the 5 specific indicators each participant had.
There were 4639 potential participants who indicated interest in the study. Of these, 1242 participants met the study criteria, consented, completed baseline questionnaires, and attended at least one session of the workshop, resulting in 1010 Internet participants (
Demographic characteristics of the study sample.
Variable | Community (n=232) | Internet (n=1010) | Entire sample (n=1242) | |
Male, n (%) | 61 (26.3%) | 346 (34.3%) | 407 (32.8%) | |
Education in years, mean (SD, range) | 15.0 (2.86, 8–23) | 15.5 (2.75, 9–23) | 15.4 (2.78, 8–23) | |
Married, n (%) | 117 (50.4%) | 762 (75.5%) | 879 (70.8%) | |
Non-Hispanic white | 148 (63.8%) | 756 (75.0%) | 904 (72.9%) | |
Black | 70 (30.2%) | 98 (9.7%) | 168 (13.5%) | |
Hispanic | 5 (2.2%) | 90 (8.9%) | 95 (7.7%) | |
Age in years, mean (SD, range) | 65.6 (10.1, 28–95) | 55.0 (8.68, 25–91) | 57.0 (9.86, 25–95) | |
Number of other chronic conditions, mean (SD, range) | 1.65 (1.37, 0–8) | 1.40 (1.15, 0–6) | 1.45 (1.19, 0–8) | |
Medicare | 133 (57.3%) | 62 (6.1%) | 195 (15.7%) | |
Private insurance | 166 (71.6%) | 1010 (100%) | 1176 (94.7%) |
Internet group flowchart. HbA1c: hemoglobin Aic.
Community workshop group flowchart. HbA1c: hemoglobin A1c.
Outside of HbA1c, which was optional, no variable had more than 1.7% missing data among those who completed questionnaires. We did not attempt to impute missing data.
Of the 1242 participants, 358 (28.82%) did not complete the 6-month questionnaire, including 28 who withdrew from the study or died before 6 months (
Comparing 6-month noncompleters and completers on demographic variables, those who did not complete the 6-month follow-up tended to be younger (mean 55.4 vs 57.6 years old,
Baseline and 6-month scores for participants with 6-month data in the Internet and workshop groups combined.
Variables | Possible |
Desirable |
No. | Baseline | 6-Month change | Effect |
||||
Mean | SD | Mean | 95% CI | |||||||
Adjusted HbA1ca,b | Lower | 489 | 7.95 | 1.38 | –0.151 | –2.47 to 10.06 | 0.109 | .002 | ||
PHQ-8c depression | 0–24 | Lower | 877 | 5.97 | 5.02 | –1.02 | –1.30 to –0.74 | 0.203 | <.001 | |
General health | 0–5 | Lower | 884 | 2.90 | 0.764 | –0.117 | –0.163 to –0.071 | 0.153 | <.001 | |
Illness intrusiveness | 1–7 | Lower | 877 | 2.79 | 1.20 | –0.082 | –0.153 to –0.005 | 0.068 | .03 | |
Hypoglycemic symptoms | 0–7 | Lower | 865 | 1.36 | 1.40 | –0.188 | –2.78 to –0.108 | 0.134 | <.001 | |
Fatigue | 1–10 | Lower | 884 | 4.84 | 2.34 | –0.508 | –0.657 to –0.341 | 0.217 | <.001 | |
Sleep | 1–10 | Lower | 882 | 3.93 | 2.95 | –0.061 | –0.240 to 0.119 | 0.021 | .51 | |
Aerobic exercise (min/week) | Higher | 876 | 81.6 | 101 | 8.91 | 2.39 to 15.4 | 0.088 | .01 | ||
Medication adherence | 0–4 | Lower | 878 | 1.05 | 1.08 | –0.135 | –0.201 to –0.073 | 0.125 | <.001 | |
Communication with doctor | 0–5 | Higher | 877 | 2.78 | 1.17 | 0.221 | 0.154 to 0.283 | 0.189 | <.001 | |
Eye exam | 0,1 | Higher | 877 | 0.451 | 0.498 | 0.100 | 0.057 to 0.145 | 0.201 | <.001 | |
Foot exam | 0,1 | Higher | 877 | 0.551 | 0.498 | 0.132 | 0.096 to 0.166 | 0.265 | <.001 | |
Cholesterol exam | 0,1 | Higher | 877 | 0.763 | 0.425 | 0.090 | 0.061 to 0.127 | 0.222 | <.001 | |
Kidney exam | 0,1 | Higher | 877 | 0.614 | 0.487 | 0.152 | 0.115 to 0.190 | 0.312 | <.001 |
aHbA1c: hemoglobin A1c.
bBecause of a laboratory recalibration changing HbA1c measurement, HbA1c prior to June 2014 was adjusted by adding 0.4.
cPHQ-8: 8-item Patient Health Questionnaire.
d“Don’t know” set to “no examination.”
Most health status and behavior variables showed statistically significant 6-month improvements. This remained when adjusting for multiple comparisons (using a
For intent-to-treat analyses (not shown), we substituted a value of 0 (no change) for 6-month change scores for those who did not complete 6-month questionnaires. The intent-to-treat analyses of changes resulted in no differences from the
For all subsequent analyses, we included only those who completed 6-month questionnaires. We present these analyses as a means of determining who benefited in this heterogeneous population.
Of those who had not been examined (tested) in any of the 4 areas (eye, foot, cholesterol, kidney) in the year before baseline (n=85), only 12 (14%) had no tests in the next 6 months (not shown in tables). Thus, 73 (86%) of those who had not been examined at all had at least one examination in the 6 months after the program. A total of 46 (54%) of these had 3 or more of the 4 tests.
There was little evidence of differences between participants in the two types of programs (Internet vs community). Only 1 change score differed by program: community participants had greater improvement in sleep (
For each of the 5 outcomes described in hypothesis 6 above, we look at 6-month changes for those who exhibited symptoms and those who did not (not shown in the tables). Then we looked at those who benefited for 1 or more of these 5 problems.
Of those who supplied a blood sample for HbA1c determination at both baseline and 6 months, 20.0% (98/489) had a baseline HbA1c of ≥9. The proportion with HbA1c >9 was reduced to 15.3% at 6 months (n=74,
At baseline, 22.0% (193/877) had symptoms indicating clinical depression (PHQ-8 score of ≥10) [
At baseline 332/865 (38.4%) had 2 or more symptoms of hypoglycemia (mean 2.88, SD 1.04). At 6 months, the number with 2 or more symptoms was reduced to 280 (32.4%). The 6-month change in mean hypoglycemic symptoms was –0.91 for those who had 2 or more symptoms at baseline (
At baseline, 35.0% (n=307) were nonadherent in taking medicine (≥2 on a 0–4 scale), and at 6 months, 29.4% (n=260) were nonadherent (
At baseline, 23.2% (203/876) indicated that they were taking no aerobic exercise. This was reduced to 18.4% (n=163) at 6 months (
In each of the cases above, the improvements for the less well-off group (high HbA1c, indications of depression, etc) were much greater than any negative changes for those who had been doing well at baseline.
At baseline, 70.4% (622/884) of participants fell into 1 or more of the above groups (high HbA1c, high depression, hypoglycemic, nonadherent, or nonexercisers); 40.6% had 2 or more problem scores and 19.5% had 3 or more. The 5 criteria variables were not correlated with each other, with the exception of hypoglycemic symptoms and PHQ depression (
We then looked at what proportion of the total study population completing 6-month questionnaires improved in at least one of the 5 criteria variables (hypothesis 7). Using effect-size improvements of at least 0.4 as an indication of an improvement, 75.0% of the study population (n=662) improved in at least one of the 5 criteria variables, and 37.1% (n=327) improved in 2 or more. The mean number of improvements (of ≥0.4 effect size) was 1.13. When broken down by how many of the 5 of these problems a participant had, the mean number of improvements tended to increase with the mean number of problems. Those with none of the 5 criteria problems had a mean of 0.671 improvements (out of a possible 5, n=268); those with 1 problem or condition had a mean of 1.08 improvements (n=272); those with 2 problems had 1.31 improvements (n=195); those with 3 problems had 1.52 improvements (n=125); and those with 4 or 5 problems had 2.06 improvements (n=52). There was a correlation (Pearson
There were modest but statistically significant improvement in 13 of the 14 outcome measures, and 10 of those were significant at the
BCBH-D may help meet the Healthy People 2020 diabetes objectives. The proportion of study participants with HbA1c >9 was reduced from 20% (n=98) to 15% (n=74), and the proportion of participants tested for foot and microalbumin (kidney function) in the last 6 months increased by 13% and 15% (
Because this is a translation study, participants were not screened for HbA1c or other symptoms. This led to a high degree of heterogeneity among study participants. In addition, many participants were already managing their diabetes and had little room for improvement. It was important, therefore, to determine who might benefit. It appears that those who had an HbA1c of ≥9 decreased their HbA1c by approximately the same amount as one would expect by taking metformin [
Considering the large initial sample (N=1242) and large number of outcomes, there were few statistical differences between 6-month completers and noncompleters. Based on the significant difference in baseline medication adherence, there is some evidence that noncompleters were slightly more likely to be medication noncompliers. They were also slightly younger. There is no evidence that they were more ill or had greater severity of symptoms, which would suggest that attrition would have been unlikely to bias 6-month outcomes.
Despite a few differences at baseline, there were few differences in 6-month changes between those attending community workshops and the Internet group. This suggests that both modes are similarly effective. In this study, very few people had a choice of modes. However, it may be that future studies will find a greater population penetration when people are offered a choice.
Because we lack a control group, we cannot be certain that the improvements observed in this study are not due to other factors. Plausible alternative explanations for the improvements include the introduction of new medications in the marketplace and health plan initiatives that were implemented for the purpose of improving the health of individuals with chronic conditions. It is also possible that there was a beneficial effect from participating in the workshops resulting in a greater likelihood of participants taking advantage of the Anthem initiatives and new medications. While these other factors are important to keep in mind, the consistency of the statistically significant improvements across multiple domains suggests a favorable impact of workshop participation.
By looking at the participants who were worse off on specific measures, we risk that part of the observed improvements might have been the result of regression to the mean. While regression to the mean may have contributed to subgroup improvements, it most likely does not explain all of the improvement.
Most of this study population were covered by a health plan and had a high mean baseline education (15.4 years). This may have contributed to the positive results. However, there were similar results when the small-group program was used with a low-education (mean of 7 years) Spanish-speaking population [
Data on how well improvements are sustained after 6 months are not included in this study. A future study will examine 12-month outcomes and address the sustainability of improvements following the intervention.
In order to reduce participant burden, unfortunately we did not include self-efficacy to manage diabetes in the questionnaires for this study. The program was designed to enhance self-efficacy and has been found to be significantly associated with self-efficacy in earlier randomized trials [
We did not attempt to control for changes in medication usage. Choosing to take medication for diabetes may be a part of self-management, and thus adding medication could be a positive outcome. In other cases, participants may have begun taking medication because of worsening health. A future study might want to look carefully at the relationships between self-management education, medication change, and health outcomes, as well as at other changes in behaviors as mediating variables.
As a community-based public health intervention, BCBH-D offered in two modes demonstrated small but significant benefits. It also showed promise for helping to meet at least some of the Healthy People 2020 diabetes objectives. More important, it demonstrated clinically significant benefits for those with high HbA1c, and important benefits for those with depression and hypoglycemia, as well as nonadherers to medication and nonexercisers. The benefits differed by individual, but a large majority of the population demonstrated meaningful improvements in at least one of the above areas. This study demonstrated that the peer-facilitated BCBH-D in both face-to-face and Internet formats can assist a national sample of health plan members in improving their diabetes management.
Better Choices, Better Health-Diabetes
hemoglobin A1c
8-item Patient Health Questionnaire
The Better Choices, Better Health-Diabetes self-management study received financial support from the Bristol Myers Squibb Foundation. The Anthem Foundation, the National Council on Aging, National Council of Young Men’s Christian Association of the United States of America (Y-USA), and OASIS Health aided in the administration and implementation of the program. Audrey Alonis and Maisoon Ayish at the Stanford Patient Education Research Center were responsible for data collection and management. Maurice Green was laboratory consultant for the HbA1c testing. Kathlyn Plant, then with National Council on Aging (NCOA), helped with recruitment and was liaison between Stanford and NCOA researchers for the Internet program. James Teufel, of OASIS Health, assisted with recruitment and institutional review board submission. Sarah Lovegreen, OASIS Health, and Valerie Lawson, Y-USA, helped with recruitment and administration of community programs. Julie Kosteas, Kristie Kulinski, Simona Eldridge, of NCOA, assisted with administration of the Internet program, and Andrea DeVries, of HealthCore, assisted with analysis.
KL designed and coordinated the study, and wrote and edited the manuscript. PLR researched the data, and wrote and edited the manuscript. JG researched data and edited the manuscript. RMT researched data and edited the manuscript. KE contributed to discussion and edited the manuscript. DDL managed and coordinated the study and edited the manuscript.
KL and DDL have the potential to receive royalties if the program is further disseminated. Other authors have no potential conflicts of interest.