Using peer mentors to support health-related behavior change may be particularly effective in a VA setting where many patients lack social support on the one hand, but have latent bonds because of shared or similar veteran experiences. The research team has demonstrated that peer mentoring can help African American veterans with hard-to-manage diabetes mellitus improve their glucose control in the short term; however, it is unknown if demonstrated improvements in glucose control persist once mentoring stops and how to best sustain such programs. These are essential questions to address in order to successfully build a generalizable and scalable peer mentor program. This work builds on past and ongoing work with the aim of creating evidence based, low-cost, easy-to-implement peer-mentoring programs that can sustainably support VA PACT efforts.
The current application sought to: 1. Test the effectiveness of a self-sustaining peer-mentoring program that trains former peer mentees to be peer mentors to support health-related behavior change in diabetic veterans with poor diabetes control; 2. Assess the effects of becoming a mentor on those who were originally mentees given a growing literature that being a mentor is good for your health; and 3. Conduct a rigorous qualitative evaluation examining in-depth the mentor-mentee relationship, the transition to becoming a mentor, and exploring factors relevant to broader program implementation.
To test our hypotheses, we conducted a randomized controlled trial (clinicaltrials.gov number NCT01651117). In the first Phase, poorly controlled diabetics were randomized to receive mentoring from well controlled peers or to usual care. In the second Phase, different poorly controlled diabetics were randomized to receive mentoring from peers who had been mentored in the first Phase or to usual care. In addition, to assess whether becoming a mentor in Phase 2 had any benefit for those who were mentored in Phase 1, past mentees from Phase 1 were randomized to either be a mentor in Phase 2 or to usual care.
Potential participants were identified through the electronic medical records (EMR). Diabetic patients were considered in poor control if they had HbA1c > 8% on at least 2 occasions in the 24 months prior to enrollment. Phase 1 potential mentors had to be in good control (with at least one HbA1c 7.5% in the 3 months prior to enrollment) but in previously poor control (at least one HbA1c > 8% in the 3 years prior to enrollment). Additional inclusion criteria included age 30 to 75, type II diabetic, access to telephone for contact with mentor/mentee, ability to understand English.
Mentor training consisted of an hour-long one-on-one training that was informed by motivational interviewing techniques. Mentors were contacted monthly to reinforce training and discuss interactions with their mentees.
All participants received $50 for each in person visit. Mentors also received $20 for each month in which they contacted or attempted to contact their mentee at least weekly.
To achieve 80% power to detect a 0.8 unit change in A1c (standard error 1.6) between Phase 2 mentors and non-mentors, a sample of 64 patients per arm was required. To protect against expected attrition (as seen in the pilot study), we inflated that by 10% to arrive at 72 participants per arm. Working backwards to determine how many poorly controlled diabetics would be needed in Phase 1, we started with 144 (72 for each arm of Phase 2) and inflated by 10% to arrive at 160, thus we intended to recruit 320 poorly controlled diabetics and randomize 1:1 to get 160 mentees and 160 usual care. After mid-study evaluation, with guidance from our Data Safety Monitoring Board (DSMB), it was determined that resources were being wasted on such a large usual care group, so randomization was change to 2:1. Additionally, mid-study evaluation revealed that attrition was higher than expected. As a result, we recruited additional poorly controlled diabetics to Phase 1 to have sufficient Phase 2 mentors. All methodological changes were documented in protocol amendments, and approved by the IRB. All randomization was done using permuted-blocks with varying block size using SAS Proc Plan.
The primary outcome of interest was change in A1c from baseline to 6 months. Change in A1c from baseline to 12 months was a secondary outcome. Additional secondary outcomes included change from baseline of other measurements (LDL, systolic blood pressure, diabetes distress score, and depression score) at 6 and 12 months. All participants had a baseline, 6 and 12 month visits during which A1c, LDL, blood pressure, weight and height were measured and several surveys were administered.
Our basic model for all analyses was an Analysis of Covariance comparing change in outcome (A1c) from baseline to 6 months, adjusting for baseline (A1c). To also assess change from baseline to 12 months, we used a mixed-effects model and included a time fixed effects and a patient random effects. We also looked at dichotomized outcomes for A1c (1% improvement) using logistic regression. Subset analyses including only those who had a baseline HbA1c > 8% (since some patients who were recruited as poorly controlled based on the EMR showed good control at baseline).
Approximately 13% of participants were missing 6-month A1c. To perform intent-to-treat analysis (which included all randomized patients except those who died during the intervention period), we used multiple imputation using Markov-chain Monte Carlo methods with 25 iterations. Analyses were conducted on each iteration and results were combined using Ruben's formula.
Qualitative semi-structured interviews were conducted to understand participants' experiences, their relationship with their partner, and how the intervention impacted their behavior. Purposive sampling was done to ensure adequate representation of mentees who made large strides in improving their A1c, those who made marginal improvements or got worse. All interviews were audio-recorded, transcribed and analyzed for salient themes.
We enrolled 365 poorly controlled veterans (158 into usual care and 207 into the intervention arm), 51% of those who were contacted and eligible into Phase 1. 65 were African American and 96% were male. Baseline biometrics were: A1c 9.5% (SD 1.6), LDL 95.4 mm/dl (SD 35.2), and SBP 137.4 mmHg (SD 18.4). The mean change in A1c was -0.20 (95% CI -0.46, 0.06) in usual care and -0.52 (95% CI -0.76, -0.29) for the intervention arm, (p = 0.06). 67 people had, on enrollment, an A1c 8%. When we limited the analysis to the 298 people with a baseline A1c > 8%, the mean change in A1c was -0.32 (95% CI -0.60, -0.05) for usual care and -0.75 (95% CI -1.01, -0.48) for the intervention arm, (p = 0.03). For the intervention arm compared to usual care the odds of dropping A1c by 1 point was 1.70 (95% CI 1.01, -2.86: p = 0.05). There was no difference in A1c between arms at 12 months. The intervention did not influence LDL control, blood pressure control, diabetes distress or depressive symptoms.
For Phase 2 we enrolled 122 people to receive mentoring from a former mentee (49 to usual care, 73 to the intervention). The mean change in A1c was -0.46 (95% CI -1.02, 0.10) in usual care and -0.08 (95% CI -0.42, 0.57) for the intervention arm, (p = 0.16). Including the additional 158 people who had been randomized to control in Phase 1 did not changes these results. However, when we compared the change in A1c between those who received mentoring from a mentor who had been successful as a mentee (dropped A1c by 1% when a mentee) to those who had not, those who received mentoring from a past successful mentee dropped their A1c by -0.28 (95% CI -0.89, 0.34) compared to those who received mentoring from a past unsuccessful mentee 0.76 (95% CI -0,05, 1.57) (p = 0.05). Mentees in this Phase, at 6 months, showed significantly better improvement in diabetes distress score compared to usual care patients (0.10 usual care v -0.41, p = 0.02). No other outcomes showed significant differences between usual care and mentees. As with the first phase, effects did not persist at 12 months.
Seventy former mentees were randomized to becoming a mentor and 69 were randomized to be a non-mentor. Becoming a mentor did not prove beneficial to former mentees. Both mentors and non-mentors increased their A1c at 6 months (0.14 and 0.32, respectively, p = 0.54).
The intervention was well-received, with most participants describing it as valuable. Participants perceived the intervention to have many benefits including having accessible support, increased self-confidence, increased accountability, increased self-efficacy, improved glucose control, and fulfilling a sense of altruism. Participants did encounter barriers including logistical, interpersonal, and individual obstacles. Mentors struggled with mentees who were dealing with mental health or medical comorbidities and mentees struggled if they perceived their control to be better than their mentors. The more successful mentees tended to be more effusive in their description of their mentors, described a stronger sense of connection to their mentor, described a more structured interaction with their mentor, and tended to be more complimentary of the intervention.
Relying on the inherent strengths of the veteran community, peer mentoring shows potential as a means to improving outcomes in patients with diabetes. Overall this peer support program was well-received but might be optimized by selecting naturally inclined mentors, providing additional training to introduce more structure into mentorship interactions, choosing mentors who have obtained a level of control, and targeting mentees who are not struggling with overwhelming comorbidities.
- Lott BD, Dicks TN, Keddem S, Ganetsky VS, Shea JA, Long JA. Insights Into Veterans' Perspectives on a Peer Support Program for Glycemic Management. The Diabetes educator. 2019 Dec 1; 45(6):607-615.
- Kangovi S, Kellom K, Sha C, Johnson S, Chanton C, Carter T, Long JA, Grande D. Patient and provider perceptions of the patient centered medical home: agreement and tensions. Poster session presented at: Society of General Internal Medicine Annual Meeting; 2014 Apr 23; San Diego, CA.