OBJECTIVE

This cluster (clinic-level) randomized controlled trial (RCT) compared medical assistant (MA) health coaching (MAC) with usual care (UC) among at-risk adults with type 2 diabetes in two diverse real-world primary care environments: a federally qualified health center (FQHC; Neighborhood Healthcare) and a large nonprofit private insurance–based health system (Scripps Health).

RESEARCH DESIGN AND METHODS

A total of 600 adults with type 2 diabetes who met one or more of the following criteria in the last 90 days were enrolled: HbA1c ≥8% and/or LDL cholesterol ≥100 mg/dL and/or systolic blood pressure (SBP) ≥140 mmHg. Participants at MAC clinics received in-person and telephone self-management support from a specially trained MA health coach for 12 months. Electronic medical records were used to examine clinical outcomes in the overall sample. Behavioral and psychosocial outcomes were evaluated in a subsample (n = 300).

RESULTS

All clinical outcomes improved significantly over 1 year in the overall sample (P < 0.001). The reduction in HbA1c was significantly greater in the MAC versus UC group (unstandardized Binteraction = −0.06; P = 0.002). A significant time by group by site interaction also showed that MAC resulted in greater improvements in LDL cholesterol than UC at Neighborhood Healthcare relative to Scripps Health (Binteraction = −1.78 vs. 1.49; P < 0.05). No other statistically significant effects were observed.

CONCLUSIONS

This was the first large-scale pragmatic RCT supporting the real-world effectiveness of MAC for type 2 diabetes in U.S. primary care settings. Findings suggest that this team-based approach may be particularly effective in improving diabetes outcomes in FQHC settings.

In 2019, ∼9% of the U.S. population (28.7 million people) had diagnosed diabetes (1). Although diabetes incidence is increasing, the proportion of individuals with diabetes who meet treatment targets remains low (2). There are also known disparities in diabetes among individuals with low incomes and from minority racial and ethnic groups (3). Relative to non-Hispanic White adults, the percentage of U.S. adults with diagnosed diabetes is higher among Hispanic/Latino/Latinx (hereafter Hispanic) individuals (11.8% vs. 7.4%) (3). Once diagnosed, Hispanic individuals also exhibit poorer diabetes management and outcomes compared with non-Hispanic White individuals (4,5). The role of ongoing self-management support is of unquestioned importance in routine diabetes care, particularly in the presence of complex medical, educational, behavioral, and psychosocial needs (6). However, most diabetes is managed in 15-min primary care visits (7), which are not modeled to provide this level of support (8). This framework may be especially challenging within resource-limited settings, such as federally qualified health centers (FQHCs), that serve patients affected by adverse social determinants of health, which contribute to higher rates of death and diabetes complications. Although disparate diabetes outcomes have been reported for those of low socioeconomic status, fewer evidence-based interventions have been developed for or tested in this high-risk population.

Team-based care is an alternative delivery model that supplements physician care with contact from allied health personnel (e.g., medical assistants [Mas]) who are specially trained to provide ongoing self-management support or health coaching in the primary care setting (9). In practice, health coaching involves ensuring understanding of the care plan, reviewing laboratory results and targets, and engaging in shared decision making and collaborative goal setting and action planning for health behavior change (and may include other elements). The effectiveness of health coaching has been supported by recent reviews in a variety of chronic diseases (10), including diabetes (11–13), as well as in medically underserved groups (14,15). For example, in a racially and ethnically diverse sample of adults with type 2 diabetes who received health coaching from MAs, medication adherence, HbA1c, lipids, and blood pressure improved (16–18), and benefits were maintained >1 year (19).

Evidence supporting the efficacy of MA health coaching (MAC) for type 2 diabetes is strong but heavily based on well-controlled studies with low external validity and questionable generalizability. Pragmatic research is essential to build on prior well-controlled studies (11–13) to determine if and how MAC should be implemented in real-world settings. Although preliminary evidence provides support for its real-world effectiveness (20), and another pragmatic trial is underway (21), relatively little is known about the model’s effectiveness when embedded in real-world primary care clinic workflows.

Moreover, few studies have addressed health coaching in medically underserved groups or in the real-world settings where they typically receive care, such as FQHCs. Our study was designed to address these gaps. The MAC trial compared MAC with usual care (UC) among adults with type 2 diabetes in primary care clinics of two health systems that serve large ethnically, racially, and socioeconomically diverse populations in Southern California: an FQHC and a large nonprofit private insurance–based health system. In a recently published process evaluation, we found the MAC trial was efficient and effective in reaching a representative sample and had high protocol adherence and acceptability (22).

The aim of this article is to report the primary and secondary outcomes of the MAC trial. The primary outcomes were clinical indicators of diabetes management, HbA1c, LDL cholesterol, and systolic blood pressure (SBP), which were collected as part of routine care and extracted from electronic medical records (EMRs) over each participant’s 1-year follow-up period. The secondary outcomes were patient-reported behavioral and psychosocial measures that were administered in a subset of trial participants after enrollment and again at 6 and 12 months. We hypothesized that the MAC intervention would lead to greater improvements in clinical and patient-reported metrics relative to UC in this diverse primary care adult population with type 2 diabetes.

An overview of participants, study settings, procedures, and measures from this cluster randomized controlled trial (RCT) is presented below. A detailed description of the protocol can be found in the report by Fortmann et al. (23).

Participants and Settings

English- and Spanish-speaking adults age ≥18 years with type 2 diabetes and one or more of the following cardiometabolic indicators not at target in the last 90 days were eligible: HbA1c ≥8% and/or LDL cholesterol ≥100 mg/dL and/or SBP ≥140 mmHg (24). To enhance the generalizability of findings and scalability of the intervention, no additional exclusion criteria were imposed, and implementation was conducted in two diverse primary care settings in Southern California. Neighborhood Healthcare, an FQHC and designated patient-centered medical home is an ambulatory clinic system that serves higher proportions of un- and underinsured patients and medically underserved patients (>60% of whom are Hispanic). Scripps Health is a large nonprofit health system comprising five hospitals and 20 primary care clinics that serves a greater portion of non-Hispanic White patients with predominantly private insurance coverage.

Procedures

At each health system, two primary care clinics with similar patient volumes and characteristics were selected for program implementation. One clinic within each health system was randomly allocated to MAC, and the other to UC, using a random-numbers generator created by the study’s statistician. Because allocation took place at the clinic level, participants were assigned to MAC or UC based on their home primary care site. The target sample size was 600 participants, allocated equally between health systems and intervention groups (MAC n = 150 and UC n = 150 at each health system). Automated EMR reports were used to identify English- and Spanish-speaking adults age ≥18 years with type 2 diabetes and with one or more clinical indicators not at target in the last 90 days. Given that randomization occurred at the clinic level and that MAC and UC were implemented as the standard model of care at the respective clinics, along with the precedent set by prior pragmatic trials (25–27), written informed consent was not required. We projected that 50% of individuals enrolled in either arm in the MAC trial would complete the patient-reported outcomes substudy (n = 300), described below. Verbal consent via telephone was obtained for the patient-reported outcomes substudy. All study procedures were approved by the Scripps Health Institutional Review Board (reviewing) and San Diego State University Institutional Review Board (relying). Research assistants who conducted patient-reported outcome assessments, EMR analysts who extracted outcome data, and the study statistician were blinded to study arm. Participants were not informed of their study assignment, which was concealed by the clinic-level assignment and standard model of care design. Recruitment occurred from March 2016 to October 2018, and follow-up data collection ended in April 2020.

MAC

At intervention clinics, the MA health coach conducted a brief EMR review of all eligible patients presenting that day for a routine primary care visit. The MA health coach then joined the primary care visit to understand the care plan and conducted the initial health coaching session in person immediately after the clinical visit (or by telephone within 1 week, if more convenient for the patient). During the initial session, the MA health coach conducted a brief assessment of current self-management behavior and used motivational interviewing communication strategies (28) to conduct collaborative goal setting based on the assessment results. Once the initial health coaching visit was completed, the patient was considered enrolled in the trial. Thereafter, MAC participants received health coaching by telephone over the course of 12 months, which involved review of clinical and behavioral progress, reinforcement of positive changes, discussion and problem solving around barriers impeding progress, and refinement of self-management goals. The target health coaching frequency was weekly for 2–4 weeks after the initial encounter and then tapered to approximately monthly for the rest of the year. Additionally, MAC participants received in-person self-management support from the MA health coach after each of their (approximately quarterly) clinic visits over this same period. The MAC intervention design was guided by the Chronic Care Model (29,30), the Robert Wood Johnson Foundation Initiative Resources and Support for Self-Management Model (31), and the Transtheoretical Model (32). Content was incorporated and adapted from other evidence-based interventions for diabetes, including Project Dulce, that were developed for Spanish-speaking Hispanic adults with low incomes. Consistent with the pragmatic nature of this trial, the MA health coach was afforded the flexibility to tailor intervention frequency and delivery to participants’ unique goals and needs (e.g., incorporating educational visuals when knowledge gap was present). The MA health coach was bilingual, delivered the intervention in the participant’s preferred language (Spanish/English), and served in the role of cultural liaison between patient and provider. Intervention materials were available in English and Spanish and optimized for low literacy and cultural relevance. More detail on intervention development, clinical risk score algorithm, and MAC curriculum are available in the report by Fortmann et al. (23).

UC

For each participant enrolled at a MAC clinic, a research assistant identified a patient with a comparable clinical risk score from the automated daily report who completed a primary care visit that same business week at the respective UC clinic. Because there was no change to care, nor were there any required study visits, UC participants’ enrollment was invisible to these individuals; however, the enrollment process facilitated the inclusion of their clinical data in the primary outcome analyses and qualified them to be contacted for the patient-reported outcomes substudy. Participants at the UC clinics continued to receive evidence-based standard diabetes care without any modifications for the duration of the trial.

Patient-Reported Outcomes (Substudy)

Within 2 weeks of enrollment in the MAC trial, participants in both groups were contacted by telephone by a research assistant inviting them to participate in a survey evaluating their health care experience at their primary care clinic. During the call, research assistants provided a brief overview of the patient-reported outcomes substudy and, if the participant was willing and interested, obtained verbal consent and conducted the baseline survey in participants’ preferred language (Spanish/English). Once the baseline survey was completed, participants were considered enrolled in the patient-reported outcomes substudy and were contacted to complete the month-6 and month-12 telephone follow-up assessments.

Outcomes

Clinical Outcomes

Primary indicators of diabetes management (HbA1c, LDL cholesterol, SBP) were collected as part of routine medical care and extracted from the EMR for outcome analyses.

Patient-Reported Outcomes

Patient-reported outcomes included diabetes self-management behaviors (i.e., healthful eating, physical activity, glucose monitoring, and medication adherence), assessed via seven items from the Summary of Diabetes Self-Care Activities (SDSCA) survey (33,34); perceived physical health and quality of life, assessed using the 10-item Patient Reported Outcomes Measurement Information System (PROMIS) Global-10 Health Scale (35); and patient activation, assessed using the 13-item version of the Patient Activation Measure (PAM) (36–38).

Statistical Analysis

A priori statistical power calculations and assumptions have been published (23). Briefly, we projected that 50% of individuals enrolled in the MAC trial (N = 600) would complete the patient-reported outcomes substudy (n = 300). These sample sizes were deemed sufficient to detect statistical significance for clinically meaningful differences between groups in HbA1c (Δ0.5%; Cohen d = 0.29), LDL cholesterol (10 mg/dL; Cohen d = 0.25), and SBP (5 mmHg; Cohen d = 0.29), as well as a moderate effect size (Cohen d = 0.50) in all patient-reported outcomes, with up to 20% attrition between time points.

Data analysis was conducted in SPSS Statistics 28.0, MPlus version 8.8 (39), and HLM8 (40). Data were screened and cleaned, and distributions were examined for skewness and kurtosis before analysis. There were no significant deviations from normality requiring transformation. Descriptive statistics were calculated to describe the sample characteristics and study variables at baseline, month 6, and month 12. Multilevel modeling was used to examine changes in clinical and self-report outcomes over time using HLM and Mplus, respectively. Mplus and HLM use robust maximum likelihood estimation to produce valid inferences when data are missing at random. For all outcomes, we first estimated change over time across the entire sample, before adding the interaction terms described below sequentially into each multilevel model. The primary effect of interest in all models was a cross-level time by group interaction effect, with time (level 1) represented in months and group (level 2) coded as MAC = 1, UC = 0. Given the uniqueness of the two health care systems, a three-way interaction, time by group by site, was examined to evaluate if the magnitude of the primary effect of interest (time by group) varied significantly by health care system (Scripps = 0, Neighborhood Healthcare = 1). Age, sex, and variables associated with baseline group differences (language) and/or data missingness (site) were included as covariates in all models. Regression coefficients are reported throughout as unstandardized betas (B), and associated P values are reported.

Data and Resource Availability

The data sets generated during and/or analyzed in the current study are available from the corresponding author upon reasonable request and with appropriate data use agreement.

Sample Characteristics

Between March 2016 and October 2018, a total of 600 participants were enrolled in the MAC trial (MAC n = 300 and UC n = 300; Scripps n = 298 and Neighborhood Healthcare n = 302). Supplementary Fig. 1 shows participant flow. Baseline characteristics are listed in Table 1. MAC trial participants ranged in age from 19 to 95 years (mean ± SD 63.5 ± 14.0); a majority were English speaking (68.2%) and non-Hispanic White (60.2%), with a fairly even distribution by sex (56.0% male). Scripps Health participants were more likely to be English speaking and non-Hispanic White than Neighborhood Healthcare participants (P < 0.05).

Table 1

Baseline characteristics of MAC trial participants (N = 600)

OverallIntervention groupHealth care system (site)
MAC
(n = 300)
UC
(n = 300)
Scripps Health
(n = 298)
Neighborhood Healthcare
(n = 302)
Age, years, mean (SD) 63.5 (14.0) 63.9 (14.0) 63.1 (14.0) 70.2 (12.6) 56.9 (12.0) 
Male sex 336 (56.0) 168 (56.0) 168 (56.0) 159 (53.4) 177 (58.6) 
Spanish speaking* 409 (68.2) 193 (64.3) 216 (72.0) 136 (45.0) 273 (91.6) 
Hispanic/Latino* 361 (60.2) 168 (56.0) 193 (64.3) 118 (39.1) 243 (81.5) 
OverallIntervention groupHealth care system (site)
MAC
(n = 300)
UC
(n = 300)
Scripps Health
(n = 298)
Neighborhood Healthcare
(n = 302)
Age, years, mean (SD) 63.5 (14.0) 63.9 (14.0) 63.1 (14.0) 70.2 (12.6) 56.9 (12.0) 
Male sex 336 (56.0) 168 (56.0) 168 (56.0) 159 (53.4) 177 (58.6) 
Spanish speaking* 409 (68.2) 193 (64.3) 216 (72.0) 136 (45.0) 273 (91.6) 
Hispanic/Latino* 361 (60.2) 168 (56.0) 193 (64.3) 118 (39.1) 243 (81.5) 

Data are given as n (%) unless otherwise indicated.

*

Indicates statistically significant baseline difference between health systems: Scripps Health vs.

Neighborhood Healthcare (P < 0.05).

Indicates statistically significant baseline difference between MAC and UC groups (P < 0.05).

Clinical Outcomes

After adjusting for age, sex, language, and site, statistically significant improvements were observed over time in all clinical outcomes in the overall sample: HbA1c (B = −0.09), LDL cholesterol (B = −1.44), and SBP (B = −0.58; all P < 0.001). By examining the time by group interaction effect (Table 2), we found that the rate of HbA1c reduction over time was significantly greater in the MAC versus UC group (time by group interaction effect B = −0.06; P = 0.002), such that MAC participants demonstrated twice the rate of improvement in HbA1c per month compared with UC participants (B = −0.12; P < 0.001 vs. B = −0.06; P < 0.001, respectively). The proportion of participants meeting HbA1c target (<8%) increased by 23.7% (from 30% to 53.7%) in the MAC group and 17.2% (from 43.2% to 60.4%) in the UC group. No statistically significant between-group differences were observed in change over time for LDL cholesterol or SBP (both time by group interaction effect P > 0.05).

Table 2

Change over time in clinical outcomes (overall and between groups)

Clinical outcomesMultilevel modelsDescriptive statistics by time point
B (95% CI)BaselineMonth 6Month 12
HbA1c, %     
 Time* −0.09 (−0.11, −0.07) n = 511 n = 381 n = 214 
 Mean ± SD  8.82 ± 2.20 7.84 ± 1.76 7.89 ± 1.75 
 Change mean (SE)   −0.82 (0.10) −0.72 (0.12) 
n (%) <8%  185 (36.2) 226 (59.3) 122 (57.0) 
 Time × group* −0.06 (−0.09, −0.02)    
  MAC  n = 270 n = 196 n = 108 
   Mean ± SD  9.08 ± 2.23 7.79 ± 1.66 8.00 ± 1.82 
   Change mean (SE)   −1.16 (0.15) −0.95 (0.18) 
   n (%) <8%  81 (30.0) 118 (60.2%) 58 (53.7%) 
  UC  n = 241 n = 185 n = 106 
   Mean ± SD  8.52 ± 2.13 7.89 ± 1.86 7.78 ± 1.66 
   Change mean (SE)   −0.45 (0.13) −0.45 (0.17) 
   n (%) <8%  104 (43.2) 108 (58.4) 64 (60.4) 
LDL cholesterol, mg/dL     
 Time* −1.46 (−2.00, −0.91) n = 329 n = 226 n = 121 
 Mean ± SD  101.81 ± 44.75 87.42 ± 39.10 86.92 ± 34.43 
 Change mean (SE)   −11.83 (3.26) −13.22 (4.48) 
n (%) <100 mg/dL  153 (46.5) 156 (69.0) 79 (65.3) 
 Time × group 0.15 (−0.93, 1.24)    
  MAC  n = 181 n = 111 n = 61 
   Mean ± SD  100.17 ± 45.95 85.45 ± 43.99 87.59 ± 39.15 
   Change mean (SE)   −8.77 (4.79) −12.47 (6.94) 
   n (%) <100 mg/dL  89 (49.2) 79 (71.2) 39 (63.9) 
  UC  n = 148 n = 115 n = 60 
   Mean ± SD  103.92 ± 43.30 89.33 ± 33.82 86.25 ± 29.17 
   Change mean (SE)   −15.28 (4.35) −14.17 (5.20) 
   n (%) <100 mg/dL  64 (43.2) 77 (67.0) 40 (66.7) 
SBP, mmHg     
 Time* −0.59 (−0.75, −0.42) n = 554 n = 475 n = 367 
 Mean ± SD  137.87 ± 19.40 132.44 ± 19.45 131.74 ± 17.03 
 Change mean (SE)   −5.03 (0.84) −5.99 (1.08) 
n (%) <140 mmHg  288 (52.1) 341 (72.7) 259 (71.5) 
 Time × group 0.22 (−0.11, 0.55)    
  MAC  n = 279 n = 238 n = 184 
   Mean ± SD  136.37 ± 19.16 131.00 ± 19.97 131.66 ± 16.79 
   Change mean (SE)   −4.80 (1.52) −5.09 (1.53) 
   n (%) <140 mmHg  159 (57.2) 174 (41.5) 126 (69.2) 
  UC  n = 275 n = 237 n = 183 
   Mean ± SD  139.39 ± 19.55 133.88 ± 18.84 131.82 ± 17.30 
   Change mean (SE)   −5.24 (1.42) −6.89 (1.54) 
   n (%) <140 mmHg  129 (46.9) 167 (72.0) 133 (73.9) 
Clinical outcomesMultilevel modelsDescriptive statistics by time point
B (95% CI)BaselineMonth 6Month 12
HbA1c, %     
 Time* −0.09 (−0.11, −0.07) n = 511 n = 381 n = 214 
 Mean ± SD  8.82 ± 2.20 7.84 ± 1.76 7.89 ± 1.75 
 Change mean (SE)   −0.82 (0.10) −0.72 (0.12) 
n (%) <8%  185 (36.2) 226 (59.3) 122 (57.0) 
 Time × group* −0.06 (−0.09, −0.02)    
  MAC  n = 270 n = 196 n = 108 
   Mean ± SD  9.08 ± 2.23 7.79 ± 1.66 8.00 ± 1.82 
   Change mean (SE)   −1.16 (0.15) −0.95 (0.18) 
   n (%) <8%  81 (30.0) 118 (60.2%) 58 (53.7%) 
  UC  n = 241 n = 185 n = 106 
   Mean ± SD  8.52 ± 2.13 7.89 ± 1.86 7.78 ± 1.66 
   Change mean (SE)   −0.45 (0.13) −0.45 (0.17) 
   n (%) <8%  104 (43.2) 108 (58.4) 64 (60.4) 
LDL cholesterol, mg/dL     
 Time* −1.46 (−2.00, −0.91) n = 329 n = 226 n = 121 
 Mean ± SD  101.81 ± 44.75 87.42 ± 39.10 86.92 ± 34.43 
 Change mean (SE)   −11.83 (3.26) −13.22 (4.48) 
n (%) <100 mg/dL  153 (46.5) 156 (69.0) 79 (65.3) 
 Time × group 0.15 (−0.93, 1.24)    
  MAC  n = 181 n = 111 n = 61 
   Mean ± SD  100.17 ± 45.95 85.45 ± 43.99 87.59 ± 39.15 
   Change mean (SE)   −8.77 (4.79) −12.47 (6.94) 
   n (%) <100 mg/dL  89 (49.2) 79 (71.2) 39 (63.9) 
  UC  n = 148 n = 115 n = 60 
   Mean ± SD  103.92 ± 43.30 89.33 ± 33.82 86.25 ± 29.17 
   Change mean (SE)   −15.28 (4.35) −14.17 (5.20) 
   n (%) <100 mg/dL  64 (43.2) 77 (67.0) 40 (66.7) 
SBP, mmHg     
 Time* −0.59 (−0.75, −0.42) n = 554 n = 475 n = 367 
 Mean ± SD  137.87 ± 19.40 132.44 ± 19.45 131.74 ± 17.03 
 Change mean (SE)   −5.03 (0.84) −5.99 (1.08) 
n (%) <140 mmHg  288 (52.1) 341 (72.7) 259 (71.5) 
 Time × group 0.22 (−0.11, 0.55)    
  MAC  n = 279 n = 238 n = 184 
   Mean ± SD  136.37 ± 19.16 131.00 ± 19.97 131.66 ± 16.79 
   Change mean (SE)   −4.80 (1.52) −5.09 (1.53) 
   n (%) <140 mmHg  159 (57.2) 174 (41.5) 126 (69.2) 
  UC  n = 275 n = 237 n = 183 
   Mean ± SD  139.39 ± 19.55 133.88 ± 18.84 131.82 ± 17.30 
   Change mean (SE)   −5.24 (1.42) −6.89 (1.54) 
   n (%) <140 mmHg  129 (46.9) 167 (72.0) 133 (73.9) 

Multilevel model significance testing was conducted with continuous outcomes. Unstandardized regression coefficients with 95% CIs are shown. Coefficient for time represents marginal main effect for UC. Covariates age, sex, language, and site were included in all models but are not shown for clarity. Unadjusted means and n (%) at target at each time point are presented for descriptive purposes only. Change is shown from baseline for matched pairs and descriptive purposes only.

*

P < 0.05.

A significant time by group by site interaction effect was observed for LDL cholesterol (B = −3.36; P = 0.004), such that MAC participants achieved significantly greater LDL cholesterol improvements relative to UC participants at Neighborhood Healthcare (time by group interaction effect B = −1.78; P = 0.049) compared with Scripps (time by group interaction effect B = 1.49; P = 0.027). As shown in Table 3, at Neighborhood Healthcare, the proportion of participants meeting LDL cholesterol target (<100 mg/dL) increased by 20.6% (from 31.6% to 52.2%) in the MAC group and 13% (from 53.7% to 66.7%) in the UC group. However, at Scripps, the proportion of participants meeting LDL cholesterol target (<100 mg/dL) increased by only 8.4% (from 62.7% to 71.1%) in the MAC group and 32.1% (from 34.6% to 66.7%) in the UC group. No significant time by group by site interaction effects were observed for HbA1c or SBP (P > 0.10; data not shown).

Table 3

Site differences in intervention effect over time in LDL cholesterol

Scripps HealthNeighborhood Healthcare
Multilevel modelsDescriptive statistics by time pointMultilevel modelsDescriptive statistics by time point
B (95% CI)BaselineMonth 6Month 12B (95% CI)BaselineMonth 6Month 12
Time* −1.24 (−1.90, −0.59) n = 183 n = 134 n = 74 −1.91 (−2.80, −1.02) n = 146 n = 92 n = 47 
 Mean ± SD  94.47 ± 41.47 82.43 ± 37.33 82.13 ± 33.56  111.01 ± 47.09 94.69 ± 40.67 94.47 ± 34.78 
 Change mean (SE)   −11.23 (3.95) −6.50 (5.30)   −13.29 (5.82) −25.15 (7.74) 
n (%) <100 mg/dL  92 (50.3) 101 (75.4) 51 (68.9)  61 (41.8) 55 (59.8) 28 (59.6) 
Time × group* 1.49 (0.18, 2.80)    −1.79 (−3.56, −0.04)    
 MAC  n = 102 n = 65 n = 38  n = 79 n = 46 n = 23 
  Mean ± SD  85.82 ± 42.95 79.23 ± 43.15 81.39 ± 37.98  118.70 ± 43.22 94.24 ± 44.12 97.83 ± 39.73 
  Change mean (SE)   −5.81 (5.74) −0.56 (7.60)   −15.47 (8.73) −36.29 (12.25) 
  n (%) <100 mg/dL  64 (62.7) 49 (23.7) 27 (71.1)  25 (31.6) 30 (65.2) 12 (52.2) 
 UC  n = 81 n = 69 n = 36  n = 67 n = 46 n = 24 
  Mean ± SD  105.38 ± 36.98 85.45 ± 30.90 82.92 ± 28.68  101.94 ± 50.11 95.15 ± 37.38 91.25 ± 29.79 
  Change mean (SE)   −17.06 (5.32) −14.83 (6.76)   −10.53 (7.47) −13.15 (8.46) 
  n (%) <100 mg/dL  28 (34.6) 52 (75.4) 24 (66.7)  36 (53.7) 25 (54.3) 16 (66.7) 
Scripps HealthNeighborhood Healthcare
Multilevel modelsDescriptive statistics by time pointMultilevel modelsDescriptive statistics by time point
B (95% CI)BaselineMonth 6Month 12B (95% CI)BaselineMonth 6Month 12
Time* −1.24 (−1.90, −0.59) n = 183 n = 134 n = 74 −1.91 (−2.80, −1.02) n = 146 n = 92 n = 47 
 Mean ± SD  94.47 ± 41.47 82.43 ± 37.33 82.13 ± 33.56  111.01 ± 47.09 94.69 ± 40.67 94.47 ± 34.78 
 Change mean (SE)   −11.23 (3.95) −6.50 (5.30)   −13.29 (5.82) −25.15 (7.74) 
n (%) <100 mg/dL  92 (50.3) 101 (75.4) 51 (68.9)  61 (41.8) 55 (59.8) 28 (59.6) 
Time × group* 1.49 (0.18, 2.80)    −1.79 (−3.56, −0.04)    
 MAC  n = 102 n = 65 n = 38  n = 79 n = 46 n = 23 
  Mean ± SD  85.82 ± 42.95 79.23 ± 43.15 81.39 ± 37.98  118.70 ± 43.22 94.24 ± 44.12 97.83 ± 39.73 
  Change mean (SE)   −5.81 (5.74) −0.56 (7.60)   −15.47 (8.73) −36.29 (12.25) 
  n (%) <100 mg/dL  64 (62.7) 49 (23.7) 27 (71.1)  25 (31.6) 30 (65.2) 12 (52.2) 
 UC  n = 81 n = 69 n = 36  n = 67 n = 46 n = 24 
  Mean ± SD  105.38 ± 36.98 85.45 ± 30.90 82.92 ± 28.68  101.94 ± 50.11 95.15 ± 37.38 91.25 ± 29.79 
  Change mean (SE)   −17.06 (5.32) −14.83 (6.76)   −10.53 (7.47) −13.15 (8.46) 
  n (%) <100 mg/dL  28 (34.6) 52 (75.4) 24 (66.7)  36 (53.7) 25 (54.3) 16 (66.7) 

Multilevel model significance testing was conducted with continuous outcomes. Unstandardized regression coefficients with 95% CIs are shown. Coefficient for time represents marginal main effect for UC. Covariates age, sex, and language were included in all models but are not shown for clarity. Unadjusted means and n (%) at target at each time point are presented for descriptive purposes only. Change is shown from baseline for matched pairs and descriptive purposes only.

*

P < 0.05 (both sites).

Patient-Reported Outcomes

While controlling for age, sex, language, and site, statistically significant improvements were not observed over time in any of the patient-reported outcomes in the overall sample: SDSCA medication adherence (B = 0.02), SDSCA blood glucose testing (B = −0.03), SDSCA physical activity (B = −0.06), SDSCA general diet (B = −0.06), SDSCA specific diet (B = −0.07), PROMIS global mental health (B = −0.07), PROMIS global physical health (B = 0.08), PAM level (B = −0.02), and PAM score (B = −0.30; all P > 0.100). There were also no time by group interaction effects indicating differences in change over time by group in any of the patient-reported outcomes (all P > 0.05 (Table 4). Finally, when the time by group by site interaction effect was added to the patient-reported outcomes models for exploratory purposes, results indicated that the above findings were not moderated significantly by site (all P > 0.10; data not shown).

Table 4

Change over time in patient-reported outcomes (overall and between groups)

Patient-reported outcomesMultilevel modelsDescriptive statistics by time point
B (95% CI)BaselineMonth 6Month 12
SDSCA medication     
 Time 0.03 (−0.06, 0.12) n = 253 n = 206 n = 200 
 Mean ± SD  6.35 ± 1.73 6.45 ± 1.51 6.34 ± 1.77 
 Time × group 0.01 (−0.03, 0.04)    
  MAC  n = 137 n = 114 n = 115 
   Mean ± SD  6.41 ± 1.56 6.57 ± 1.26 6.42 ± 1.64 
  UC  n = 116 n = 92 n = 85 
   Mean ± SD  6.28 ± 1.92 6.32 ± 1.77 6.24 ± 1.94 
SDSCA blood glucose testing     
 Time −0.01 (−0.12, 0.09) n = 274 n = 225 n = 218 
 Mean ± SD  4.73 ± 2.86 4.59 ± 2.89 4.37 ± 3.01 
 Time × group −0.02 (−0.07, 0.03)    
  MAC  n = 149 n = 123 n = 123 
   Mean ± SD  4.93 ± 2.79 4.8 ± 2.81 4.61 ± 2.93 
  UC  n = 125 n = 102 n = 95 
   Mean ± SD  4.50 ± 2.94 4.32 ± 2.99 4.06 ± 3.11 
SDSCA physical activity     
 Time −0.07 [-0.19, 0.05] n = 275 n = 224 n = 218 
 Mean ± SD  4.04 ± 2.61 4.12 ± 2.53 4.42 ± 2.47 
 TimeXGroup 0.02 (−0.04, 0.07)    
  MAC  n = 150 n = 123 n = 123 
   Mean ± SD  4.26 ± 2.53 4.27 ± 2.49 4.75 ± 2.26 
  UC  n = 125 n = 101 n = 95 
   Mean ± SD  3.78 ± 2.68 3.94 ± 2.58 4.00 ± 2.67 
SDSCA general diet     
 Time −0.08 (−0.22, 0.05) n = 274 n = 223 n = 218 
 Mean ± SD  5.34 ± 2.13 5.30 ± 2.04 5.33 ± 2.02 
 Time × group 0.05 (−0.01, 0.10)    
  MAC  n = 149 n = 122 n = 123 
   Mean ± SD  5.39 ± 2.07 5.28 ± 2.18 5.62 ± 1.77 
  UC  n = 125 n = 101 n = 95 
   Mean ± SD  5.28 ± 2.21 5.33 ± 1.87 4.97 ± 2.28 
SDSCA specific diet     
 Time −0.08 (−0.16, 0.01) n = 276 n = 225 n = 218 
 Mean ± SD  4.55 ± 1.68 4.71 ± 1.56 4.86 ± 1.5 
 Time × group 0.01 (−0.03, 0.05)    
  MAC  n = 151 n = 123 n = 123 
   Mean ± SD  4.58 ± 1.6 4.70 ± 1.59 4.95 ± 1.5 
  UC  n = 125 n = 102 n = 95 
   Mean ± SD  4.52 ± 1.78 4.73 ± 1.54 4.74 ± 1.51 
PROMIS global mental health     
 Time −0.06 (−0.20, 0.08) n = 287 n = 229 n = 231 
 Mean ± SD  13.45 ± 3.73 13.43 ± 3.69 13.62 ± 3.68 
 Time × group −0.01 (−0.07, 0.05)    
  MAC  n = 155 n = 124 n = 130 
   Mean ± SD  13.72 ± 3.57 13.49 ± 3.66 13.88 ± 3.52 
  UC  n = 132 n = 105 n = 101 
   Mean ± SD  13.12 ± 3.91 13.35 ± 3.73 13.29 ± 3.87 
PROMIS global physical health     
 Time 0.10 (−0.002, 0.21) n = 287 n = 229 n = 231 
 Mean ± SD  13.11 ± 3.37 13.09 ± 3.50 13.40 ± 3.48 
 Time × group −0.02 (−0.07, 0.03)    
  MAC  n = 155 n = 124 n = 130 
   Mean ± SD  13.35 ± 3.26 13.24 ± 3.48 13.54 ± 3.49 
  UC  n = 132 n = 105 n = 101 
   Mean ± SD  12.83 ± 3.49 12.9 ± 3.53 13.22 ± 3.48 
PAM level     
 Time −0.01 (−0.06, 0.03) n = 298 n = 238 n = 237 
 Mean ± SD  3.20 ± 0.79 3.25 ± 0.82 3.25 ± 0.80 
 Time × group −0.01 (−0.03, 0.01)    
  MAC  n = 161 n = 129 n = 130 
   Mean ± SD  3.29 ± 0.77 3.2 ± 0.86 3.32 ± 0.76 
  UC  n = 137 n = 109 n = 107 
   Mean ± SD  3.10 ± 0.80 3.31 ± 0.78 3.17 ± 0.84 
PAM score     
 Time −0.36 (−1.20, 0.48) n = 298 n = 238 n = 237 
 Mean ± SD  67.94 ± 13.72 68.08 ± 16.28 68.57 ± 17.23 
 Time × group 0.09 (−0.31, 0.48)    
  MAC  n = 161 n = 129 n = 130 
   Mean ± SD  69.15 ± 13.90 67.05 ± 16.94 70.84 ± 14.8 
  UC  n = 137 n = 109 n = 107 
   Mean ± SD  66.52 ± 13.42 69.3 ± 15.46 65.8 ± 19.5 
Patient-reported outcomesMultilevel modelsDescriptive statistics by time point
B (95% CI)BaselineMonth 6Month 12
SDSCA medication     
 Time 0.03 (−0.06, 0.12) n = 253 n = 206 n = 200 
 Mean ± SD  6.35 ± 1.73 6.45 ± 1.51 6.34 ± 1.77 
 Time × group 0.01 (−0.03, 0.04)    
  MAC  n = 137 n = 114 n = 115 
   Mean ± SD  6.41 ± 1.56 6.57 ± 1.26 6.42 ± 1.64 
  UC  n = 116 n = 92 n = 85 
   Mean ± SD  6.28 ± 1.92 6.32 ± 1.77 6.24 ± 1.94 
SDSCA blood glucose testing     
 Time −0.01 (−0.12, 0.09) n = 274 n = 225 n = 218 
 Mean ± SD  4.73 ± 2.86 4.59 ± 2.89 4.37 ± 3.01 
 Time × group −0.02 (−0.07, 0.03)    
  MAC  n = 149 n = 123 n = 123 
   Mean ± SD  4.93 ± 2.79 4.8 ± 2.81 4.61 ± 2.93 
  UC  n = 125 n = 102 n = 95 
   Mean ± SD  4.50 ± 2.94 4.32 ± 2.99 4.06 ± 3.11 
SDSCA physical activity     
 Time −0.07 [-0.19, 0.05] n = 275 n = 224 n = 218 
 Mean ± SD  4.04 ± 2.61 4.12 ± 2.53 4.42 ± 2.47 
 TimeXGroup 0.02 (−0.04, 0.07)    
  MAC  n = 150 n = 123 n = 123 
   Mean ± SD  4.26 ± 2.53 4.27 ± 2.49 4.75 ± 2.26 
  UC  n = 125 n = 101 n = 95 
   Mean ± SD  3.78 ± 2.68 3.94 ± 2.58 4.00 ± 2.67 
SDSCA general diet     
 Time −0.08 (−0.22, 0.05) n = 274 n = 223 n = 218 
 Mean ± SD  5.34 ± 2.13 5.30 ± 2.04 5.33 ± 2.02 
 Time × group 0.05 (−0.01, 0.10)    
  MAC  n = 149 n = 122 n = 123 
   Mean ± SD  5.39 ± 2.07 5.28 ± 2.18 5.62 ± 1.77 
  UC  n = 125 n = 101 n = 95 
   Mean ± SD  5.28 ± 2.21 5.33 ± 1.87 4.97 ± 2.28 
SDSCA specific diet     
 Time −0.08 (−0.16, 0.01) n = 276 n = 225 n = 218 
 Mean ± SD  4.55 ± 1.68 4.71 ± 1.56 4.86 ± 1.5 
 Time × group 0.01 (−0.03, 0.05)    
  MAC  n = 151 n = 123 n = 123 
   Mean ± SD  4.58 ± 1.6 4.70 ± 1.59 4.95 ± 1.5 
  UC  n = 125 n = 102 n = 95 
   Mean ± SD  4.52 ± 1.78 4.73 ± 1.54 4.74 ± 1.51 
PROMIS global mental health     
 Time −0.06 (−0.20, 0.08) n = 287 n = 229 n = 231 
 Mean ± SD  13.45 ± 3.73 13.43 ± 3.69 13.62 ± 3.68 
 Time × group −0.01 (−0.07, 0.05)    
  MAC  n = 155 n = 124 n = 130 
   Mean ± SD  13.72 ± 3.57 13.49 ± 3.66 13.88 ± 3.52 
  UC  n = 132 n = 105 n = 101 
   Mean ± SD  13.12 ± 3.91 13.35 ± 3.73 13.29 ± 3.87 
PROMIS global physical health     
 Time 0.10 (−0.002, 0.21) n = 287 n = 229 n = 231 
 Mean ± SD  13.11 ± 3.37 13.09 ± 3.50 13.40 ± 3.48 
 Time × group −0.02 (−0.07, 0.03)    
  MAC  n = 155 n = 124 n = 130 
   Mean ± SD  13.35 ± 3.26 13.24 ± 3.48 13.54 ± 3.49 
  UC  n = 132 n = 105 n = 101 
   Mean ± SD  12.83 ± 3.49 12.9 ± 3.53 13.22 ± 3.48 
PAM level     
 Time −0.01 (−0.06, 0.03) n = 298 n = 238 n = 237 
 Mean ± SD  3.20 ± 0.79 3.25 ± 0.82 3.25 ± 0.80 
 Time × group −0.01 (−0.03, 0.01)    
  MAC  n = 161 n = 129 n = 130 
   Mean ± SD  3.29 ± 0.77 3.2 ± 0.86 3.32 ± 0.76 
  UC  n = 137 n = 109 n = 107 
   Mean ± SD  3.10 ± 0.80 3.31 ± 0.78 3.17 ± 0.84 
PAM score     
 Time −0.36 (−1.20, 0.48) n = 298 n = 238 n = 237 
 Mean ± SD  67.94 ± 13.72 68.08 ± 16.28 68.57 ± 17.23 
 Time × group 0.09 (−0.31, 0.48)    
  MAC  n = 161 n = 129 n = 130 
   Mean ± SD  69.15 ± 13.90 67.05 ± 16.94 70.84 ± 14.8 
  UC  n = 137 n = 109 n = 107 
   Mean ± SD  66.52 ± 13.42 69.3 ± 15.46 65.8 ± 19.5 

Multilevel model significance testing conducted with continuous outcomes. Unstandardized regression coefficients with 95% CIs shown. Coefficient for time represents marginal main effect for UC. Covariates age, sex, language, and site were included in all models but not shown for clarity. Unadjusted means and n (%) at target at each time point are presented for descriptive purposes only. Change is shown from baseline for matched pairs and descriptive purposes only.

This was the first large-scale implementation and evaluation of an MAC model in a U.S.-based real-world primary care environment. Moreover, by including a majority Hispanic Spanish-speaking sample and evaluating the intervention in a high-need FQHC setting, this study can inform efforts to reduce racial, ethnic, and socioeconomic disparities in diabetes care and outcomes. The pragmatic nature of the trial facilitated a valuable assessment of the effectiveness of this approach under realistic conditions (as opposed to a well-controlled research setting) and across two unique primary care clinic settings and patient populations. The current study evaluated the effectiveness of MAC versus UC for improving diabetes clinical and patient-reported outcomes in adults with poorly managed type 2 diabetes. Results indicate that all clinical metrics (HbA1c, LDL cholesterol, and SBP) significantly improved over 1 year in both MAC and UC groups; however, the MAC group achieved significantly greater improvements than the UC group in HbA1c (both sites) and LDL cholesterol (FQHC only) over this timeframe. In the patient-reported outcomes subsample, results failed to show support for significant change in any of the behavioral or psychosocial measures administered (regardless of intervention group), and there were no significant health system differences in these findings.

Although both groups achieved clinically significant improvements in HbA1c over time, the MAC group demonstrated, on average, twice the rate of improvement (Δ−0.12% vs. Δ−0.06% per month) and a 0.34% greater absolute change in HbA1c over 12 months (Δ−1.08 vs. −0.74) compared with the UC group. Furthermore, this finding was consistent across health systems. Interestingly, the positive effect of the MAC intervention on HbA1c in the absence of similar results for LDL cholesterol (at Scripps Health) and SBP (at both health care systems) mirrors findings reported by earlier well-controlled trials examining similar MAC models (11,18). This pattern of findings may suggest a relatively greater responsiveness of HbA1c to behavioral/health coaching interventions. Although the MAC curriculum was designed to address all cardiometabolic indicators important for diabetes management, it is possible that the focus on behavioral change and monitoring for glucose control was greater and/or emphasized more by the coach, resulting in a greater positive overall impact on HbA1c relative to LDL cholesterol and SBP. Regarding blood pressure, the assessment of SBP under real-world primary care clinic conditions in the current study had many advantages, such as convenience/low burden for participants and increased availability of measurements for outcome evaluation purposes. However, the absence of some processes and controls that are inherent to well-controlled/nonpragmatic research studies could have affected the reliability of this repeated measure over time.

The statistically significant three-way time by group by site interaction effect on LDL cholesterol deserves further investigation to determine if any site-based differences in patient populations or intervention delivery help to explain this difference in LDL cholesterol outcomes across health care systems. A process evaluation of the MAC intervention that was published previously (22) showed health coaching sessions at Neighborhood Healthcare to be more frequent, longer in duration, and more likely to include a focus on goal setting, nutrition, and exercise (relative to glucose and blood pressure monitoring) compared with those delivered at Scripps. Although outside the scope of the present report, a future report will examine these and other system-based differences as potential moderators of outcomes.

In terms of patient-reported outcomes, it is surprising that MAC was not associated with improvements in self-reported self-management behavior, given its explicit focus on dietary, exercise, medication adherence, and glucose monitoring behaviors. Interestingly, few health coaching trials have assessed or reported behavioral outcomes such as diabetes self-management (11). One large-scale randomized trial assessed self-reported medication adherence and found that it significantly improved after health coaching relative to UC (17). Given that we found that MAC led to significant reductions in HbA1c, it is unclear why there was no evidence of diabetes self-management behavior change during the same period. Perceived physical health, quality of life, and patient activation also did not improve over the course of either intervention. Although MAC did not directly address psychosocial well-being, we expected the benefits of self-management support to spill over into other well-being domains. There are many possible explanations for the lack of statistically significant patient-reported effects. First, although the measures used were previously validated in English and Spanish, they have not been extensively validated for telephone administration and remain subject to the known weaknesses inherent to self-report survey methodology. For example, the assessments required participants to recall their behavior over the last 3 months, which may have reduced the validity of these outcomes. Social desirability bias is also an important consideration, particularly because the study was integrated into patients’ health clinics. Second, more frequent, comprehensive, and objective (as applicable) assessment may be needed to identify the self-management and psychosocial underpinnings of clinical benefit from real-world health coaching interventions. Because the MAC intervention was a highly personalized flexible intervention, dosage and timing of delivery varied, but the patient-reported outcome assessment timing was not yoked to these factors and occurred for all participants at baseline, month 6, and month 12. Therefore, it is possible that assessing these behavioral outcomes more frequently (e.g., monthly, daily) and directly (e.g., using objective assessment tools, such as accelerometers) would be a more optimal approach. This would also allow for examination of nonlinear effects of treatment over time. Third, our results are somewhat consistent with those of other studies; a review and meta-analysis of health coaching found only one in 10 studies that assessed quality of life demonstrated significant improvements (relative to UC) (11). It is possible that a more intentional directed focus on psychosocial needs must be incorporated to see benefit in these measures.

The present findings should be considered in the context of several limitations. Although the pragmatic nature of this trial enhanced the ecological validity of findings, it introduced uncontrolled factors that may have affected internal validity. The assessment of clinical metrics as part of routine medical care versus incentivized research assessment visits affected completeness. The evaluation of patient-reported outcomes in just half of MAC trial participants who were willing to complete the surveys may also limit generalizability of findings to the entire sample. In addition, the baseline assessment of patient-reported outcomes was not a true baseline, because it occurred within the 2 weeks after the initial coaching visit. Next, although it is possible that improved clinical control in the overall sample was due to primary care physicians’ initiation of new therapies or modification of existing therapies in response to the same recent elevated laboratory result(s) that qualified patients for the trial, we did not have access to data to formally evaluate this hypothesis. Finally, flexibility in health coaching frequency and focus was allowed and inherent to the pragmatic/real-world delivery; however, this may have resulted in some participants receiving insufficient dosage and/or content coverage to realize the maximum potential benefit of the intervention.

Despite these limitations, this trial builds on knowledge learned from well-controlled trials to provide a valuable understanding of the value and effectiveness of an MAC intervention for type 2 diabetes in real-world primary care settings. The implementation and evaluation of this model in two unique health care environments, an FQHC and a large primarily private insurance–based health care system, and among patients undergoing standard care processes (vs. a consent-requiring research trial) enhance the generalizability of findings to other primary care settings and populations. Finally, the results demonstrate that the MAC intervention was equally effective in improving glycemic control when implemented in a predominantly Hispanic-serving FQHC setting compared with in a large nonprofit health system. These findings are further strengthened by our prior process evaluation of the intervention, which showed that MAC was able to reach members of the intended population, feasible to implement, and viewed as acceptable by the MA health coaches in two diverse real-world clinic workflows (22). Given that the FQHC setting is typically under-resourced, MAC may be one approach to reducing health disparities in diabetes (3). The importance of investigating social determinants of health to address disparities in diabetes outcomes in the U.S. is well established, but evidence-based, bilingual, and culturally tailored interventions designed for high-risk groups and settings are still lacking. Future directions include a cost-effectiveness analysis and an examination of participant characteristics (e.g., race/ethnicity), intervention delivery, and other health system–based differences as potential moderators of the effect of MAC on diabetes outcomes.

Clinical trial reg. no. NCT02643797, clinicaltrials.gov

This article contains supplementary material online at https://doi.org/10.2337/figshare.25651638.

Acknowledgments. The authors thank the study participants and staff, as well as Lauren Scarpelli, research intern, for their valuable contributions to this project.

Funding. This trial was supported by the National Institute of Diabetes and Digestive and Kidney Disease, National Institutes of Health (grant R18DK104250; principal investigator A.P.-T.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.L.F. and E.C.S. conducted the primary analyses and wrote the initial manuscript draft. A.L.F., L.C.G., and A.P.-T. were involved in funding acquisition; conception, design, and conduct of the study; and analysis and interpretation of the results. T.L.C. and H.S. contributed to data collection, research design/methodology, data curation, and project administration. S.R.S.B., S.R., and T.G. contributed to research design/methodology and data analysis. J.A.J. was involved in project administration and funding acquisition. J.S. contributed resources, project administration, and supervision. T.B. was involved in research design/methodology. All authors were involved in writing the manuscript and provided final approval of the version submitted. A.L.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study was presented as part of the ePoster Theater and poster presentation at the 82nd Scientific Sessions of the American Diabetes Association, New Orleans, LA, 3–7 June 2022.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Cheryl A.M. Anderson and Stephanie L. Fitzpatrick.

1.
Centers for Disease Control and Prevention
. National Diabetes Statistics Report. Accessed 23 August 2023. Available from https://www.cdc.gov/diabetes/data/statistics-report/index.html
2.
Kazemian
P
,
Shebl
FM
,
McCann
N
,
Walensky
RP
,
Wexler
DJ
.
Evaluation of the cascade of diabetes care in the United States, 2005–2016
.
JAMA Intern Med
2019
;
179
:
1376
1385
3.
Centers for Disease Control and
Prevention
. By the Numbers: Diabetes in America. Accessed 3 August 2023. Available from https://www.cdc.gov/diabetes/health-equity/diabetes-by-the-numbers.html
4.
Aguayo-Mazzucato
C
,
Diaque
P
,
Hernandez
S
,
Rosas
S
,
Kostic
A
,
Caballero
AE
.
Understanding the growing epidemic of type 2 diabetes in the Hispanic population living in the United States
.
Diabetes Metab Res Rev
2019
;
35
:
e3097
5.
Haw
JS
,
Shah
M
,
Turbow
S
,
Egeolu
M
,
Umpierrez
G
.
Diabetes complications in racial and ethnic minority populations in the USA
.
Curr Diab Rep
2021
;
21
:
2
6.
Powers
MA
,
Bardsley
JK
,
Cypress
M
, et al
.
Diabetes self-management education and support in adults with type 2 diabetes: a consensus report of the American Diabetes Association, the Association of Diabetes Care & Education Specialists, the Academy of Nutrition and Dietetics, the American Academy of Family Physicians, the American Academy of PAs, the American Association of Nurse Practitioners, and the American Pharmacists Association
.
Diabetes Care
2020
;
43
:
1636
1649
7.
Pilla
SJ
,
Segal
JB
,
Maruthur
NM
.
Primary care provides the majority of outpatient care for patients with diabetes in the US: NAMCS 2009–2015
.
J Gen Intern Med
2019
;
34
:
1089
1091
8.
Bodenheimer
T
,
Laing
BY
.
The teamlet model of primary care
.
Ann Fam Med
2007
;
5
:
457
461
9.
Sinsky
CA
,
Bodenheimer
T
.
Powering-up primary care teams: advanced team care with in-room support
.
Ann Fam Med
2019
;
17
:
367
371
10.
Kivelä
K
,
Elo
S
,
Kyngäs
H
,
Kääriäinen
M
.
The effects of health coaching on adult patients with chronic diseases: a systematic review
.
Patient Educ Couns
2014
;
97
:
147
157
11.
Pirbaglou
M
,
Katz
J
,
Motamed
M
,
Pludwinski
S
,
Walker
K
,
Ritvo
P
.
Personal health coaching as a type 2 diabetes mellitus self-management strategy: a systematic review and meta-analysis of randomized controlled trials
.
Am J Health Promot
2018
;
32
:
1613
1626
12.
Racey
M
,
Jovkovic
M
,
Alliston
P
,
Ali
MU
,
Sherifali
D
.
Diabetes health coach in individuals with type 2 diabetes: a systematic review and meta analysis of quadruple aim outcomes
.
Front Endocrinol (Lausanne)
2022
;
13
:
1069401
13.
Almulhim
AN
,
Hartley
H
,
Norman
P
,
Caton
SJ
,
Doğru
OC
,
Goyder
E
.
Behavioural change techniques in health coaching-based interventions for type 2 diabetes: a systematic review and meta-analysis
.
BMC Public Health
2023
;
23
:
95
14.
Ruffin
L
.
Health coaching strategy to improve glycemic control in African-American adults with type 2 diabetes: an integrative review
.
J Natl Black Nurses Assoc
2017
;
28
:
54
59
15.
Dennis
SM
,
Harris
M
,
Lloyd
J
,
Powell Davies
G
,
Faruqi
N
,
Zwar
N
.
Do people with existing chronic conditions benefit from telephone coaching? A rapid review
.
Aust Health Rev
2013
;
37
:
381
388
16.
Thom
DH
,
Ghorob
A
,
Hessler
D
,
De Vore
D
,
Chen
E
,
Bodenheimer
TA
.
Impact of peer health coaching on glycemic control in low-income patients with diabetes: a randomized controlled trial
.
Ann Fam Med
2013
;
11
:
137
144
17.
Thom
DH
,
Willard-Grace
R
,
Hessler
D
, et al
.
The impact of health coaching on medication adherence in patients with poorly controlled diabetes, hypertension, and/or hyperlipidemia: a randomized controlled trial
.
J Am Board Fam Med
2015
;
28
:
38
45
18.
Willard-Grace
R
,
Chen
EH
,
Hessler
D
, et al
.
Health coaching by medical assistants to improve control of diabetes, hypertension, and hyperlipidemia in low-income patients: a randomized controlled trial
.
Ann Fam Med
2015
;
13
:
130
138
19.
Sharma
AE
,
Willard-Grace
R
,
Hessler
D
,
Bodenheimer
T
,
Thom
DH
.
What happens after health coaching? Observational study 1 year following a randomized controlled trial
.
Ann Fam Med
2016
;
14
:
200
207
20.
Wayne
N
,
Perez
DF
,
Kaplan
DM
,
Ritvo
P
.
Health coaching reduces HbA1c in type 2 diabetic patients from a lower-socioeconomic status community: a randomized controlled trial
.
J Med Internet Res
2015
;
17
:
e224
21.
Hadden
KB
,
Arnold
CL
,
Curtis
LM
, et al
.
Rationale and development of a randomized pragmatic trial to improve diabetes outcomes in patient-centered medical homes serving rural patients
.
Contemp Clin Trials
2018
;
73
:
152
157
22.
Clark
TL
,
Fortmann
AL
,
Philis-Tsimikas
A
, et al
.
Process evaluation of a medical assistant health coaching intervention for type 2 diabetes in diverse primary care settings
.
Transl Behav Med
2022
;
12
:
350
361
23.
Fortmann
AL
,
Philis-Tsimikas
A
,
Euyoque
JA
, et al
.
Medical assistant health coaching (“MAC”) for type 2 diabetes in diverse primary care settings: a pragmatic, cluster-randomized controlled trial protocol
.
Contemp Clin Trials
2021
;
100
:
106164
24.
American Diabetes Association
.
8. Cardiovascular disease and risk management
.
Diabetes Care
2016
;
39
(
Suppl. 1
):
S60
S71
25.
Zhang
JZ
,
Nicholls
SG
,
Carroll
K
, et al
.
Informed consent in pragmatic trials: results from a survey of trials published 2014-2019
.
J Med Ethics
2023
;
49
:
34
40
26.
Ferrara
A
,
Hedderson
MM
,
Albright
CL
, et al
.
A pragmatic cluster randomized clinical trial of diabetes prevention strategies for women with gestational diabetes: design and rationale of the Gestational Diabetes’ Effects on Moms (GEM) study
.
BMC Pregnancy Childbirth
2014
;
14
:
21
27.
Krist
AH
,
Glenn
BA
,
Glasgow
RE
, et al.;
MOHR Study Group
.
Designing a valid randomized pragmatic primary care implementation trial: the my own health report (MOHR) project
.
Implement Sci
2013
;
8
:
73
28.
Miller
WR
,
Rollnick
S
Motivational interviewing: preparing people to change addictive behavior. New York, Guilford Press
,
2002
29.
Wagner
EH
,
Austin
BT
,
Davis
C
,
Hindmarsh
M
,
Schaefer
J
,
Bonomi
A
.
Improving chronic illness care: translating evidence into action
.
Health Aff (Millwood)
2001
;
20
:
64
78
30.
Wagner
EH
,
Austin
BT
,
Von Korff
M
.
Improving outcomes in chronic illness
.
Manag Care Q
1996
;
4
:
12
25
31.
Fisher
EB
,
Brownson
CA
,
O’Toole
ML
,
Anwuri
VV
,
Shetty
G
.
Perspectives on self-management from the Diabetes Initiative of the Robert Wood Johnson Foundation
.
Diabetes Educ
2007
;
33
(
Suppl. 6
):
216S
224S
32.
Prochaska
JO
,
Velicer
WF
.
The transtheoretical model of health behavior change
.
Am J Health Promot
1997
;
12
:
38
48
33.
Toobert
DJ
,
Hampson
SE
,
Glasgow
RE
.
The summary of diabetes self-care activities measure: results from 7 studies and a revised scale
.
Diabetes Care
2000
;
23
:
943
950
34.
Vincent
D
,
McEwen
MM
,
Pasvogel
A
.
The validity and reliability of a Spanish version of the summary of diabetes self-care activities questionnaire
.
Nurs Res
2008
;
57
:
101
106
35.
Hays
RD
,
Bjorner
JB
,
Revicki
DA
,
Spritzer
KL
,
Cella
D
.
Development of physical and mental health summary scores from the Patient-Reported Outcomes Measurement Information System (PROMIS) global items
.
Qual Life Res
2009
;
18
:
873
880
36.
Hibbard
JH
,
Mahoney
ER
,
Stockard
J
,
Tusler
M
.
Development and testing of a short form of the patient activation measure
.
Health Serv Res
2005
;
40
:
1918
1930
37.
Hibbard
JH
,
Stockard
J
,
Mahoney
ER
,
Tusler
M
.
Development of the Patient Activation Measure (PAM): conceptualizing and measuring activation in patients and consumers
.
Health Serv Res
2004
;
39
:
1005
1026
38.
Alegría
M
,
Sribney
W
,
Perez
D
,
Laderman
M
,
Keefe
K
.
The role of patient activation on patient-provider communication and quality of care for US and foreign born Latino patients
.
J Gen Intern Med
2009
;
24
(
Suppl. 3
):
534
541
39.
Muthén
LK
,
Muthén
BO
.
Mplus User’s Guide. Eighth Edition. 1998-2017.
Los Angeles, CA
,
Muthén & Muthén
,
2017
40.
Raudenbush
SW
,
Congdon
RT
.
HLM & Hierarchical Linear and Nonlinear Modeling.
Chapel Hill, NC
,
Scientific Software International, Inc
.,
2021
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.