This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on http://mhealth.jmir.org/, as well as this copyright and license information must be included.
Type 2 diabetes is a growing public health problem amenable to prevention and health promotion. As healthy behaviors have an impact on disease outcomes, approaches to support and sustain diabetes self-management are vital.
This study aimed to evaluate the effectiveness of a nurse coaching program using motivational interviewing paired with mobile health (mHealth) technology on diabetes self-efficacy and self-management for persons with type 2 diabetes.
This randomized controlled trial compared usual care with an intervention that entailed nurse health coaching and mHealth technology to track patient-generated health data and integrate these data into an electronic health record. The inclusion criteria were as follows: (1) enrolled at 1 of 3 primary care clinics, (2) aged 18 years or above, (3) living with type 2 diabetes, and (4) English-speaking. We collected outcome measures at baseline, 3 months, and 9 months. The primary outcome was diabetes self-efficacy; secondary outcomes were depressive symptoms, perceived stress, physical functioning, and emotional distress and anxiety. Linear regression mixed modeling estimated the population trends and individual differences in change.
We enrolled 319 participants; 287 participants completed the study (155 control and 132 intervention). The participants in the intervention group had significant improvements in diabetes self-efficacy (Diabetes Empowerment Scale, 0.34; 95% CI –0.15,0.53;
We demonstrated the short-term effectiveness of this intervention; however, by 9 months, although physical activity remained above the baseline, the improvements in self-efficacy were not sustained. Further research should evaluate the minimum dose of coaching required to continue progress after active intervention and the potential of technology to provide effective ongoing automated reinforcement for behavior change.
ClinicalTrials.gov NCT02672176; https://clinicaltrials.gov/ct2/show/NCT02672176
Type 2 diabetes is a growing public health problem amenable to prevention and health promotion [
Having a chronic condition has implications for all aspects of daily life as the individual navigates choices about nutrition, physical activity, sleep, stress management, and medication regimen. Bandura and Adams [
Diabetes education programs and group classes may be effective in the short term but appear to be insufficient to sustain behavioral changes (eg, improvements in physical activity and healthy eating) and self-management skills [
The International Diabetes Federation outlines clinical guidelines for type 2 diabetes management, including educating patients and providers, setting goals for self-management of blood glucose, generating a structured profile, providing feedback to patients about their results, using these data to modify treatments, and engaging in shared decision making [
Mobile technology offers new opportunities to track health behaviors, provide reinforcement through immediate feedback about objective measures of behavior such as steps taken, and improve health outcomes in chronic diseases [
This study examined the impact of a novel intervention using MI-based nurse health coaching combined with wearable activity trackers that integrate patient-generated activity data into the patient’s electronic health record (EHR) to improve health among adults with type 2 diabetes. We hypothesized that individuals randomized to the intervention group would show overall improved self-efficacy compared with individuals in the usual care group.
A detailed description of the study design was previously reported in a clinical trial protocol [
We recruited participants from 2 suburban and 1 urban primary care clinic within an academic health center in Northern California. The inclusion criteria were as follows: (1) aged 18 years or above, (2) receiving care at 1 of the 3 clinics, (3) living with type 2 diabetes and having HbA1c of 6.5% (48 mmol/mol) or higher, and (4) able to speak English. Participants were ineligible if they did not have access to a telephone, were not able to consent because of cognitive impairment, or were pregnant. Our power analysis determined that with 100 participants per arm, we would have 80% power to detect differences in self-efficacy. We queried the health system EHR and diabetes registry to identify eligible participants who subsequently received mailed letters and telephone calls. Potential participants were told that the study focused on using enabling technology to support their health in diabetes. Study data were collected and managed using Research Electronic Data Capture tools hosted at the Clinical Translational Science Center at UC Davis [
Participants in this group received usual care through their primary care clinic. Usual care comprised standard health care visits with providers and access to classes, resources, and services (ie, diabetes management and weight loss education, electronic learning videos, and care coordination). At the orientation meeting, the study team members provided instruction on how to access these resources and services as well as how to use the health system’s patient portal (MyChart).
The intervention group participants received the same care through their primary care clinic and training as those receiving usual care regarding health system services and resources. In addition, the intervention included (1) nurse health coaching and (2) mHealth technology to track PGHD and integrate these data into the EHR (see
The nurse health coaches for the intervention were 3 registered nurses (RNs) with experience in both health coaching and management of chronic disease. To promote fidelity to the intervention and a common approach to coaching participants, the nurses received core training in MI-based coaching and diabetes management. All the RN health coaches delivering the intervention completed the HealthSciences Institute’s Registered Health Coach (RHC) and Chronic Care Professional training programs (www.healthsciences.org). A final performance evaluation using the Motivational Interviewing Treatment Integrity (MITI) 3.1.1 global scale evaluation tool confirmed health coaching competency before the receipt of the RHC certificate [
We paired each participant with a nurse health coach who delivered 6 individual sessions using a counseling style based on the concepts of MI. Sessions were structured to promote mutual goal setting, enhance self-efficacy in health behavior change, and assist individuals to derive meaning from data to reinforce choices and behaviors. Two RN researchers with nurse coaching experience in diabetes audited 8 of the 158 (5%) of the participant sessions and scored the coach using the MITI. They provided timely feedback to the coaches during weekly debriefing sessions, reviewed scores, and discussed optimization strategies by reviewing scenarios.
The participants had an in-person orientation with the nurse coach, followed by telephone sessions every 2 weeks for 3 months (6 contacts total). The initial MI session elicited motivations and set goals with tracking metrics to gauge the progress toward goals at subsequent sessions. Throughout the sessions, the coaches encouraged the participants to identify facilitators and barriers to achieving their health goals.
We provided a wearable tracking device (initially, Basis Peak, then Garmin VivoSmart Heart Rate [HR]) to the intervention group participants. This device generated real-time information about steps taken, distance walked, active minutes, heart rate, and hours of sleep at night and synced the data to either an iPhone operating system mobile phone and/or iPod Touch. We provided the iPod Touch to participants who did not already possess this technology. We preinstalled MyFitnessPal, a mobile app, on the devices to allow participants to log and track nutritional consumption if they chose. We provided in-person or telephonic technical support to all participants—including the usual care group participants—throughout the duration of the study. We encouraged the participants to wear the activity tracker for the entire 9-month duration of the study.
PGHD were integrated into the EHR when participants performed regular synchronization of the activity tracker to their personal device. We used 2 connectors, Apple HealthKit and MyChart, to accomplish the automatic transmission of data to the EHR. We used Synopsis, a feature within the EHR, to design a single screen page of relevant PGHD along with clinically relevant data elements (ie, laboratory values and medications). In the case management module of Epic Electronic Health Record, we designed a summary documentation form for the nurse coaching sessions. We sent a final summary of goals and achievements to the primary care providers. Using these tools, the participants, providers, and nurse health coaches could view trends in activity levels, sleep, and nutritional intake on either their smart device or on a computer.
Early in the intervention period, we experienced an unexpected recall of the Basis Peak activity tracking device because of a safety issue that required identifying and selecting a replacement device. The study team, in collaboration with the advisory boards, worked diligently and promptly to identify, test, and select a replacement (Garmin VivoSmart HR) and then distribute the new device to the participants in the intervention arm of the study, providing technical support to these participants as needed. This recall affected 79 participants; most of these participants received and were oriented to their new devices within 2 weeks of the recall.
The participants completed Web-based surveys at baseline, 3 months (coinciding with the end of the intervention or 3 months from baseline), and 9 months. The baseline survey included demographic information (age, gender, race and ethnicity, education, income, and insurance type), health information (common chronic illness and health status), and technology use and adoption information. Readiness to change was assessed with 2 items measuring intention,
Diabetes self-efficacy (Diabetes Empowerment Scale [DES]–Short Form) [
Depression severity (Patient Health Questionnaire-9 (PHQ-9) [
The surveys also included Patient-Reported Outcomes Measurement Information System (PROMIS) [
Descriptive analysis yielded means and SDs for continuous variables and frequencies for categorical variables. We examined distributions and collinearity to determine whether the data met the assumptions for planned statistical analyses. We compared the demographic and health-related characteristics among the individuals in the intervention group and the usual care group using Student
In the primary analysis, we estimated the difference over time in the effects of the intervention versus usual care in the study participants, controlling for potentially relevant variables such as demographic characteristics, readiness to change, self-reported health, and comorbid disease. This was an intention-to-treat analysis with the assumption that any dropouts were missing at random. In our evaluation, we did not find any significant difference between the participants who dropped out and the participants who continued in the study. We adopted a mixed effects maximum likelihood model that accounts for the missing data from the participants who dropped out from the study. Finally, our sensitivity analysis found no difference with regard to the significance of our findings when we excluded these participants from the analysis. We included all participants, regardless of intervention completion, in the intention-to-treat analysis. We used multivariate regression modeling for all hypotheses testing to estimate within-group and across-group effects of the intervention on the outcomes (significance level:
There were no significant differences in the demographics between the usual care and intervention groups (
Most participants were managing multiple chronic illnesses. The participants tracked metrics related to their health at varying rates as follows: blood glucose (127/319, 39.8%), laboratory results (125/319, 39.2%), physical activity (82/319, 25.7%), nutrition (65/319, 20.4%), and sleep (54/319, 16.9%). A large percentage of participants, 256 out of 319 (80.3%), had no prior experience using mobile apps or sensors. The usual care and intervention groups were similar with regard to their baseline health characteristics and experience with technology.
Characteristics of participants in the study.
Characteristics | Total (N=319) | Control (n=161) | Intervention (n=158) | ||
|
.97 | ||||
|
Female | 148 (47.3) | 75 (47.2) | 73 (47.4) |
|
|
Male | 165 (52.7) | 84 (52.8) | 81 (52.6) |
|
Age (years), mean (SD) | 59.07 (11.4) | 59.18 (11.5) | 58.96 (11.3) | .87 | |
|
.89 | ||||
|
High school or less | 36 (11.4) | 20 (12.6) | 16 (10.3) |
|
|
Some college | 109 (34.6) | 56 (35.2) | 53 (34.0) |
|
|
Associate’s degree | 40 (12.7) | 19 (12.0) | 21 (13.5) |
|
|
Bachelor’s degree | 63 (20.0) | 29 (18.2) | 34 (21.8) |
|
|
Graduate and professional degree | 67 (21.3) | 35 (22.0) | 32 (20.5) |
|
|
.06 | ||||
|
<US $25,000 | 44 (15.9) | 29 (21.3) | 15 (10.7) |
|
|
US $25,000-US $50,000 | 66 (23.9) | 29 (21.3) | 37 (26.4) |
|
|
US $50,001-US $75,000 | 56 (20.3) | 29 (21.3) | 27 (19.3) |
|
|
US $75,001-US $100,000 | 47 (17.0) | 17 (12.5) | 30 (21.4) |
|
|
>US $100,000 | 63 (22.8) | 32 (23.5) | 31 (22.1) |
|
|
.79 | ||||
|
Caucasian | 196 (63.0) | 100 (62.9) | 96 (63.2) |
|
|
African American | 39 (12.5) | 18 (11.3) | 21 (13.8) |
|
|
Asian | 27 (8.7) | 16 (10.1) | 11 (7.2) |
|
|
Other | 30 (9.7) | 14 (8.8) | 16 (10.5) |
|
|
More than 1 race | 19 (6.1) | 11 (6.9) | 8 (5.3) |
|
|
.28 | ||||
|
Hispanic or Latino | 42 (15.2) | 18 (12.9) | 24 (17.5) |
|
|
Not Hispanic or Latino | 235 (84.8) | 122 (87.1) | 113 (82.5) |
|
|
.88 | ||||
|
No other comorbidities | 115 (37.5) | 55 (34.2) | 60 (38.0) |
|
|
1 comorbidity | 95 (30.9) | 47 (29.2) | 48 (30.4) |
|
|
2 comorbidities | 52 (16.9) | 27 (16.8) | 25 (15.8) |
|
|
3 or more comorbidities | 45 (14.7) | 25 (15.5) | 20 (12.7) |
|
|
.74 | ||||
|
Previously used apps or sensors | 63 (19.8) | 33 (20.5) | 30 (19.0) |
|
|
Never used apps or sensors | 256 (80.3) | 128 (79.5) | 128 (81.0) |
|
|
|
||||
|
Blood glucose | 127 (39.8) | 64 (39.8) | 63 (39.9) | .31 |
|
Physical activity | 82 (25.7) | 35 (21.7) | 47 (29.8) | .25 |
|
Nutrition | 65 (20.4) | 33 (20.5) | 32 (20.3) | .46 |
|
Sleep | 54 (16.9) | 28 (17.4) | 26 (16.5) | .42 |
The most common smart goals selected by the intervention group participants were physical activity (103/147, or 70.7%) and nutrition (37 out of 147 or 25.2%), with 7 out of 147 (4.8%) participants selecting other goals such as stress reduction, alcohol cessation, and improving sleep. Across all coaching sessions, participants averaged 172 min of nurse coaching.
Finally, our mixed effect regression models evaluated the effect of the intervention on the primary outcome, diabetes empowerment (DES), over time (3 months and 9 months). We found significant improvement in the DES scores at 3 months (0.50, 95% CI 0.07-1.1;
This study built upon the recognition that chronic disease management is fundamentally a partnership between health care providers and individuals, requiring goal setting, bilateral communication, and motivation. We sought to change the conversation through mHealth technology and nurse coaches’ support that could standardize goal setting and generate relevant patient-generated data for discussion and action. We engaged the stakeholders representing persons with diabetes, clinicians, and health information technology experts to design the intervention and the integration of the technology and coaching into primary care.
This study demonstrated the short-term effectiveness of an innovative diabetes intervention using nurse health coaching and mHealth technology on diabetes self-efficacy and increased physical activity, supporting our hypothesis. However, by 6 months post intervention, although physical activity remained above baseline, the differences in self-efficacy were not sustained.
The change in diabetes self-efficacy score represented a meaningful improvement in the confidence for engaging in self-management behavior in 3 to 4 out of 8 potential areas. The average PHQ-9 score was below the cutoff indicating substantial depressive symptoms, and the score decreased over the course of the intervention. Improved self-efficacy and management of depression may be the most important drivers for positive health behavior change. This finding also suggests a potential for PGHD derived from wearable sensors to play a role in improving self-efficacy. By using tracking devices, the participants became aware of their physical activity and behavior patterns and could visualize and measure their accomplishments, increasing motivation to continue changes to attain their goals. The nurse coaches explaining the data from sensors and supporting the participants as personal difficulties arose were critical in the positive outcomes as the coaching interactions also created accountability, focus, and awareness of how behavior impacted the participants’ health.
The decrease in depression severity scores among the intervention group participants further endorse the importance of personalized support, connection to health care resources to cope with and treat depression, and continuous communication provided during the coaching sessions. Significant differences in these areas did not persist at 9 months. As lifestyle changes take time and require reinforcement, it is possible that a longer intervention period with continuing support (ie, health coaching) could be needed for sustained behavior changes leading to better diabetes management.
Participants in the intervention demonstrated a significant increase in physical activity, as measured by increased steps per week, from 23,700 to 39,167 at 3 months and down to 32,601 at 9 months. Using average stride length, this is an increase from about 11.9 miles to 19.6 miles walked per week. At 9 months, participants were still walking 16 miles. This accomplishment is consistent with the Centers for Disease Control and Prevention’s (CDC’s) recommended guidelines for type 2 diabetes management to get at least 150 min per week of moderate-intensity physical activity [
The lack of long-term sustainability of this intervention echoes other studies [
The high retention rate (89.9% completed the 9-month study), coupled with the qualitative feedback from participants, suggests that the intervention was acceptable and useful despite the fact that over 80% had no previous experience with technology. Both groups engaged with the technology; the usual care group demonstrated interest and use of online diabetes resources, and the intervention group used their wearable activity trackers and online resources. Although the digital divide exists, this study demonstrates that mHealth solutions are viable for novices to technology and individuals with low socioeconomic status. Fu et al [
The passive tracking devices provided data for discussion with the nurse coaches who assisted in making sense of the patterns, generating insights into health habits that can affect outcomes that matter to the patient. These insights form the basis for long-term behavior change necessary for optimal health in diabetes. In a parallel analysis of the qualitative data from this study, participants described new insights about living with type 2 diabetes and defined success in their own terms as changes in awareness, mindset, engagement with health resources, and self-perceptions of emotional or physical health [
This study also demonstrated the feasibility of integrating such PGHD into the EHR for visualization by the patient and the health care team. Although the use of mHealth to improve diabetes self-management is well illustrated in the research literature [
This study had several limitations. First, the sample might have been biased toward those ready to change and those ready to use technology, limiting generalizability. Second, this study has limited generalizability to other settings because of required investments in technology, technology training, and support throughout the intervention. However, we estimated the total cost of the intervention, including staff time and technology, to be less than US $500 per participant. The total cost is a small investment relative to the costs of an office or emergency room visit. Finally, this intervention requires commitment to a model of care to include nurse coaches in a team-based approach.
Although this study demonstrated the short-term effects on self-efficacy, it did not demonstrate sustainability of the effects. Since this study commenced, 2 major trends have accelerated—the use of mHealth technology and the movement toward value-based payment for chronic disease management. These facilitating forces will likely result in approaches akin to those examined in this study. This study answered several logistical questions for scalability and generalizability, demonstrated the feasibility of integrating PGHD into the clinical record, and engaged and supported patients with little previous exposure to technology. As mobile technology is gaining adoption across all age demographics [
Although traditional fee-for-service reimbursement models would not cover such interventions, managed care plans and value-based purchasing will move toward reimbursing interventions that yield quality outcomes, further enhancing the potential for sustaining interventions of this kind. Finally, as PGHD are not highly sensitive, this intervention does not have to be clinic-based. It has the potential for delivery in a variety of settings such as fitness centers, workplaces, and community centers, increasing the potential to scale.
Coupling patient data with other indicators of health provides a more complete picture that could be helpful in a variety of chronic conditions, such as congestive heart failure and chronic obstructive pulmonary disease, which require a partnership between clinicians and patients for effective management. This study combined nurse coaching with mHealth and compared this intervention with usual care; future studies could include more study arms that break out the components to allow greater comparison. Future research could examine this approach in additional conditions and other strategies, such as automated feedback in the form of SMS messaging and online peer support that could enhance the effectiveness of the intervention. Longer study periods with intermittent coach contact could potentially demonstrate how to sustain the effects of the intervention over time. Finally, studies of implementation and dissemination across a variety of settings would improve the translation of research such as this into practice.
Intervention figure.
CONSORT flow diagram.
Outcome measures at baseline, 3 months, and 9 months for control group table.
Outcome measures at baseline, 3 months, and 9 months for intervention group table.
Change in outcomes comparing baseline and 3 months (Difference in Difference).
Change in outcomes comparing baseline and 9 months (Difference in Difference) table.
CONSORT-EHEALTH checklist (V 1.6.1).
Centers for Disease Control and Prevention
Diabetes Empowerment Scale
electronic health record
mobile health
motivational interviewing
Motivational Interviewing Treatment Integrity
Patient and Provider Engagement and Empowerment Through Technology
patient-generated health data
Patient Health Questionnaire-9
Patient-Reported Outcomes Measurement Information System
Perceived Stress Scale
Registered Health Coach
registered nurse
The authors would like to thank the P2E2T2 program staff Rupinder Colby, Sarina Fazio, Daicy Luo, Michael Dang, Nazifa Hamdard, and Sarah Haynes; nurse health coaches Jennifer Edwards, Bridget Levich, Katherine Greene, and Sarina Fazio; IT team members Michael Minear, Kent Anderson, Scott MacDonald, Ryan Peck, and Kristen Augtagne; primary care, health management, and education consultants Thomas Balsbaugh, Victor Baquero, Bridget Levich, Linda Blake, Glee Van Loon, and Deborah Greenwood; and the patient advisory board members Eric Bowser, Diane Goodman, Margaret Hitchcock, Maria Ibarra, Michael Lawson, Joseph McCarthy, and Tarunesh Singh for their contributions to the project.
This study was funded by the Patient-Centered Outcomes Research Institute: IHS-1310-07894. This study was conceived and designed independently of the funders, who did not have any role in the data collection, analysis, or writing of manuscript based on the data.
All authors contributed to the design of the study, interpretation of the results, and discussion of the results. HY wrote the manuscript, with major portions contributed by SM, MD, and YF. MD conducted the statistical analysis. All authors reviewed and edited the manuscript. HY is the guarantor for this manuscript.
None declared.