An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Christie A. Befort, PhD, Jeffrey J. VanWormer, PhD, Cyrus DeSouza, MD, Edward F. Ellerbeck, MD, Byron Gajewski, PhD, Kim S. Kimminau, PhD, Allen Greiner, MD, MPH, Michael G. Perri, PhD, Alexandra R. Brown, MS, Ram D. Pathak, MD, Terry T.K. Huang, PhD, MPH, Leslie Eiland, MD, Andjela Drincic, MD, Heather Nelson-Brantley, PhD, RN, NEA-BC, CCRN-K, and Shellie Ellis, MA, PhD.
Author Information and AffiliationsNearly 20% of the US population lives in rural communities. Rural residents experience obesity disproportionally and have less access to effective weight loss programs than do residents of other areas. Primary care has the potential to fill an important need in treating obesity in rural America. Primary care is traditionally provided on a fee-for-service (FFS) model with 15-minute individual office visits, as is reimbursed by Medicare under the intensive behavioral therapy for obesity (IBT) benefit, which was approved in 2011 as the first benefit billed under an obesity diagnosis code. However, alternative models may be better suited for behavioral obesity treatment especially in the rural setting, where having fewer available clinic staff may limit capacity for frequent individual visits.
We conducted a pragmatic cluster randomized trial comparing the Medicare IBT FFS model delivered via in-clinic individual visits with 2 alternative models: (1) in-clinic group visits delivered by practice-employed providers modeled after patient-centered medical home (PCMH) principles, and (2) phone-based group visits delivered by centralized personnel. A total of 36 rural primary care practices were randomized 1:1:1 to the 3 study arms (12 practices per arm). Participants (N = 1407 adults aged 20-75 years who resided in a rural area) were enrolled at participating practices. All participants received a high-intensity lifestyle intervention with the same core components that differed by delivery model: in-clinic individual visits, in-clinic group visits, or phone-based group visits. The primary outcome was 24-month weight change. Secondary outcomes included cardiometabolic risk factors, quality of life (QOL), sleep, and stress. In addition, both quantitative and qualitative measures were used to evaluate RE-AIM measures (ie, reach, adoption, implementation, and maintenance).
Of the 36 included practices, 22 were located in isolated or small rural areas, and 13 were in large rural areas; 10 were rural health clinics, and 12 were Federally Qualified Health Centers. Participants (N = 1407) had a mean (SD) age of 54.7 (11.8) years and mean (SD) body mass index (BMI) of 36.7 (4.0); 76.8% were female, 96.2% were White non-Hispanic, and 46.8% were from an isolated rural area. The trial retained 87% of participants at 24 months. Weight loss at 24 months was −4.43 kg (95% CI, −5.48 to −3.39 kg) in the PCMH arm (in-clinic group visits); −3.92 kg (95% CI, −4.98 to −2.87 kg) in the disease management (DM) arm (phone-based group visits); and −2.56 kg (95% CI, −3.63 to −1.49 kg) in the FFS arm (in-clinic individual visits). Compared with FFS, there was a net difference of −1.87 kg (97.5% CI, −3.51 to −0.23 kg; P = .01) for PCMH and −1.36 kg (97.5% CI, −3.00 to 0.29 kg; P = .06) for DM. The pattern of results across study arms was similar for most subpopulations, except that greater weight loss in DM than in FFS was observed only among participants with an annual income of ≥$50 000. The results for secondary outcomes showed greater improvements in triglyceride levels and physical activity in the PCMH arm than in the FFS arm. All arms showed statistically significant (P <.05) and meaningful improvements through 24 months in physical activity, dietary measures, weight-related QOL, and sleep.
Participant-level implementation measures showed that participants in the FFS arm had the highest attendance, whereas participants in the DM arm had the lowest attendance and worst ratings of experience of care. Participants in the PCMH arm reported the highest overall satisfaction. Analysis of reach revealed a high enrollment rate among eligible participants (86%), with no statistical differences between arms. Mixed-method findings among practice personnel revealed that most perceived the intervention across all arms to be easy to implement, effective, and compatible with what their patients wanted, although for some, the perceived effectiveness declined from midintervention to postintervention. Practice personnel in the DM arm reported lower perceived effectiveness than did those in the FFS and PCMH arms, which some related to lack of care integration. The major barriers to maintaining the intervention across arms were cost and counselor staffing, and these concerns were particularly evident among counselors in the PCMH arm.
The PCMH arm with in-clinic group visits produced greater weight loss than did the FFS arm with in-clinic individual visits. The combined participant and practice personnel implementation measures highlighted downsides to the DM model, which had slightly worse results than those from the PCMH model, including patients' difficulty developing accountability and cohesiveness, lack of support from local peers, and lack of care integration within local practices. Thus, the magnitude of the additional weight loss in the DM arm compared with the FFS arm needs to take into account the less-positive experiences of care. Overall, the PCMH model resulted in the most favorable outcomes; however, to be a sustainable model, implementation strategies are needed to determine financial viability based on reimbursement options across different payers and to address staffing shortages. The findings have implications for policies related to treatment of obesity in primary care, particularly in rural and other medically underserved areas.
The sample was predominantly female and White non-Hispanic, and the rural cultural and geographical context of the trial may affect the generalizability of the findings to urban and more racially/ethnically diverse populations. Because the study was designed to compare current care delivery models under pragmatic conditions, it did not control for the potential impact of different professional backgrounds or training of the counselors. Finally, by design, the study was not limited to older adults covered by Medicare; thus, replication of the IBT provision is needed in a Medicare-insured population.
Rural residents in the United States are disproportionately affected by obesity1,2 and obesity-related illnesses, including diabetes and heart disease.3-5 This higher obesity prevalence is accompanied by a 32% higher age-adjusted death rate compared with that of urban residents outside inner cities.6,7 Nearly 20% of the US population lives in rural communities, as defined by county-based definitions from the Economic Research Service, constituting one of the largest medically underserved populations in the nation.4,8 Decreasing rural health disparities is a high national priority, as outlined by the Department of Health and Human Services,9 US Surgeon General,10 and Centers for Disease Control and Prevention.11
Rural residents encounter multiple barriers to maintaining healthy lifestyles, including an inadequate built environment to facilitate physical activity12,13 and limited access to healthy food outlets14 and evidence-based weight control programs.15-17 However, an estimated 70% to 75% of rural residents have a primary care provider (PCP) within their rural area18 and may look to their PCP to guide them toward healthy lifestyles.15,19
The US Preventive Services Task Force recommends that PCPs screen all adults for obesity and offer (or refer) patients to intensive behavioral interventions as first-line treatment.20,21 However, PCPs often fail to diagnose obesity, and they offer weight loss counseling to only 20% to 40% of patients with obesity.22-24 When brief counseling is provided, it typically falls short of the intensity (frequency and/or duration of sessions) needed for meaningful weight change.25 Low-intensity (eg, quarterly) PCP counseling is insufficient to achieve clinically meaningful weight loss. For example, at 24 months, quarterly PCP visits have resulted in −0.2 to −1.7 kg compared with −4.6 to −5.4 kg with weekly to monthly visits.25,26
Primary care is traditionally provided on a fee-for-service (FFS) model with 15-minute individual visits. In 2011, the Centers for Medicare & Medicaid Services (CMS) approved coverage for intensive behavioral therapy for obesity (IBT) with up to 22 individual 15-minute face-to-face visits over a 12-month period.27 Although this coverage was considered a step in the right direction toward obesity being a medically diagnosed and treated disease, some questioned the limited evidence supporting the FFS delivery model.28,29 Subsequent Medicare claims data revealed that fewer than 1% of eligible beneficiaries with obesity had used the service, and users had on average only 2 visits per year.30 Barriers to uptake may include inadequate visit length, low reimbursement rates, and the requirement that patients attend one-on-one counseling visits during regular clinic hours that are separate from other medical visits.29 The shortcomings of the face-to-face approach have been considered particularly concerning for underresourced rural communities.31,32
Various alternative care delivery models have arisen to challenge the traditional individual office visit. Group visits have been shown to be effective33-36 and to preserve the benefits of a face-to-face encounter and coordination with other health care, consistent with patient-centered medical home (PCMH) principles,37 while also providing unique opportunities for peer support. Group visits for obesity can also be delivered by phone38,39 and integrated into call centers, such as those offered by disease management (DM) programs. Obesity treatment via telehealth may be more convenient for both providers and patients, especially in rural settings.
In February 2014, PCORI issued a request for applications for trials comparing obesity treatment alternatives with the Medicare FFS IBT model in underserved primary care settings.40 We conducted a 5-year cluster randomized trial in primary care practices in the rural Midwestern United States in which we compared FFS obesity treatment with in-clinic group visits delivered by practice-employed providers; this approach was modeled after PCMH principles and phone-based group visits delivered by centralized personnel modeled after DM programs. All study arms included an active 2-year intervention with the same core components. The study was designed as a pragmatic trial that leveraged available primary care staff and facilities across multiple types of rural practices. The primary aim was to compare change in weight (in kg) at 24 months between both the in-clinic group visit arm (PCMH) and the phone-based group visit arm (DM) vs the in-clinic individual visit arm (FFS). Secondary aims included comparisons across the in-clinic group visit arm and phone group visit arm; changes in cardiometabolic risk factors, quality of life (QOL), sleep, and stress; heterogeneity of treatment effects (HTE) across relevant participant characteristics (ie, race/ethnicity, sex, education, income, employment status, and travel time to clinic); and evaluation of the reach, adoption, implementation, and maintenance (RE-AIM) of the intervention.
The trial was guided by patient and stakeholder input from its inception. Stakeholders at local, state, and national levels were engaged, and a patient advisory board (PAB) also was involved. National-level partners at the American Academy of Family Physicians provided input on strategies for comprehensively engaging practices. The Kansas Academy of Family Physicians and other state health organizations assisted with practice recruitment and dissemination of the study findings at annual meetings. Rural PCPs provided essential feedback regarding the feasibility of the trial design and championed the vision of improving obesity treatment. Local practice staff were involved in day-to-day study operations and provided ongoing feedback to the central study team. At the conclusion of the study activities, they offered valuable insight regarding future implementation of the intervention outside the research setting. The PAB included 10 men and women living with obesity in rural communities in the Midwestern United States. They provided input and feedback on multiple aspects of the trial, including intervention materials, recruitment and retention materials and strategies, and selection of patient-reported outcome measures. They were active participants in training sessions and dissemination meetings. Their insight into ways to improve the dialogue between patients and PCPs when communicating about obesity resulted in a publication in Annals of Family Medicine.19 In addition, 2 PAB members copresented with the primary investigator at the 2016 and 2020 PCORI annual meetings. A detailed summary of the engagement activities is included in Appendix A.
Rural Engagement in Primary Care for Optimizing Weight Reduction (RE-POWER) was a cluster randomized pragmatic trial comparing 3 models of providing behavioral weight loss counseling in 36 rural primary care practices. The primary hypotheses were that the in-clinic group visits (PCMH arm) and phone-based group visits (DM arm) would be more effective than would in-clinic individual visits (FFS arm) in reducing weight at 24 months. Secondary outcomes included cardiometabolic risk factors, QOL, sleep, and stress. We also examined potential HTE across relevant participant characteristics (ie, race/ethnicity, sex, education, income, employment status, and travel time to clinic) and evaluated the reach, adoption, implementation, and maintenance of the intervention according to the RE-AIM framework41 and informed by the Consolidated Framework for Implementation Research (CFIR).42
The study was conducted by the University of Kansas Medical Center (KUMC) in collaboration with the Marshfield Clinic Health System in Wisconsin and by the University of Nebraska Medical Center in collaboration with the Veterans Affairs (VA) Nebraska-Western Iowa. The study was approved by the KUMC IRB, which served as the central IRB for the study, under the PCORnet-funded Greater Plains Collaborative, a clinical data research network,43 and the VA Nebraska-Western Iowa IRB.
The study was conducted in Kansas, Iowa, Nebraska, and Wisconsin. The practice types were diverse, consistent with PCORI's emphasis on including heterogeneity in practice and participant recruitment. The characteristics of the participating practices are described in the Results section.
Practices were recruited via outreach and referrals from colleagues. Site visits were conducted to review the study staffing, administrative requirements, and timeline. We evaluated practice eligibility and suitability by whether the practice served predominantly or exclusively rural residents, had a lead PCP to champion the study, was capable of recruiting 40 participants, and was ready to deliver the intervention according to the study timeline. Practices were required to commit to implementing any of the 3 arms before randomization.
Practices were randomly allocated by the study statistician, in equal proportions, to each of the 3 arms using a pseudo-random number generator in the statistical software R. Randomization was stratified by institutional affiliation (each primary care practice was affiliated with 1 of 3 different academic institutions). Practices were recruited and randomized in 3 cohorts. Due to the speed of practice recruitment and interest in starting, 15 practices were randomized in cohort 1 in an allocation ratio of 5:5:5 for FFS:PCMH:DM. One practice randomized to the FFS arm in cohort 1 declined after randomization but before enrolling patients for a reason unrelated to randomized assignment (ie, the study physician became ill). Cohort 2 included 10 practices in a 3:3:4 ratio for FFS:PCMH:DM. Cohort 3 included 12 practices in a 5:4:3 ratio for FFS:PCMH:DM. Two practices in cohort 3 (1 in FFS arm and 1 in PCMH arm) declined after randomization but before enrolling patients for reasons unrelated to randomization (ie, changes in staffing and changes in data security requirements). These 2 practices were replaced with the next recruited practice. A total of 36 practices were randomized and enrolled patients in the trial.
The first participant was enrolled on February 15, 2016, and the final participant was enrolled on October 2, 2017. The final data collection visit was completed on December 30, 2019. Each practice was involved for approximately 3 years and recruited participants within approximately 3 to 9 months for FFS and 3 to 6 months for PCMH and DM (shorter time frame to prevent long delays between recruitment of early participants and the start of groups).
Participants were eligible if they were between the ages of 20.0 and 75.0 years, had a body mass index (BMI) between 30.0 and 45.0, and resided in a rural location as defined by rural-urban commuting area codes,8 Urban Influence Codes,44 amount of agricultural income, or individual commuting patterns. Participants were required to have access to a telephone and to speak English. One individual per household was permitted to enroll in the study. Medical clearance was required for all participants.
As a pragmatic trial, participant exclusion criteria were kept minimal. Exclusions included myocardial infarction, new cancer diagnosis, or stroke in the last 6 months; history of bariatric surgery ever or planned bariatric surgery within 2 years; pregnancy in the last 6 months or planned within 2 years, or currently pregnant or lactating; end-stage renal or liver disease or anticipated transplant within the next 2 years; and plans to leave their primary care clinic in the next 2 years.
Two primary participant recruitment strategies were direct mailings and in-clinic referrals. Each practice was responsible for developing a recruitment list of potentially eligible participants, based on age, BMI, rural ZIP Code, and a clinic visit within the past 18 months. PCPs had the option to review their recruitment lists and remove participants with known medical or behavioral exclusions (eg, current addiction treatment or serious mental illness). Providers referred participants during routine medical visits, and study brochures and opt-in postcards were available at the clinics. Interested participants (referred by mail, providers, or family/friends) initiated contact with the study team via mailed postcard, telephone, or email. Phone screeners asked participants how they heard about the study, confirmed they were an active patient at a participating clinic, and screened them for eligibility. PCP clearance was then obtained, and eligible participants completed the baseline survey online or by mail. Subsequently, local clinic staff contacted the participant to schedule a single consent/baseline visit.
All arms included a 2-year intervention incorporating a 6-month weight loss phase with a goal of 10% weight loss (but with reinforcement of the idea that health benefits could be achieved even with a more modest loss of 5%), followed by an 18-month weight loss maintenance phase. The diet, physical activity, and behavioral recommendations were the same across all study arms and were consistent with the Look AHEAD lifestyle intervention.45 Participants received a calorie goal (1200-1500 kcal/day if <114 kg; 1500-1800 kcal/day if >114 kg) and were instructed to consume a balanced diet with ≥5 fruit and vegetable servings per day. Emphasis was placed on portion control and consumption of low-calorie, high-volume foods.46 To facilitate adherence, preportioned meals and shakes were recommended. Participants were advised to increase physical activity over the first 12 weeks to 225 minutes per week of moderate-intensity activity. During weight loss maintenance, participants were given a new personalized calorie goal based on the Harris-Benedict equation47 and were guided to maintain diet and physical activity behaviors. No food or scales were provided to the participants.
Self-monitoring and goal setting were the core behavioral strategies reinforced throughout the intervention. Participants were instructed to set weekly diet and physical activity goals and to self-monitor daily with the Lose It! commercial app48 or a written log. Visits also addressed strategies for stimulus control, social support, and problem-solving for overcoming barriers,49 including those common to rural environments (eg, limited food outlet choices and physical activity resources). During the maintenance period, key elements of the intervention included continued support, accountability, and help with addressing routine problems. Counselors were instructed to provide feedback on participants' self-monitoring logs throughout the intervention.
The 3 delivery models were designed to represent how they may be typically delivered in clinical practice, including the selection and training of counselors. A summary of the visit, counselor, and training characteristics of the 3 models is shown in Table 1. The characteristics of the study personnel who provided counseling visits are described in the Results section.
Characteristics of Randomized Intervention Delivery Models.
In the FFS arm, practice-employed clinicians (eg, nurses, registered dieticians, advance practice providers, physicians) served as intervention counselors and provided 15-minute face-to-face individual counseling visits according to the frequency reimbursed by Medicare: weekly for 1 month, biweekly for months 2 to 6, and monthly thereafter. Two modifications were made to the Medicare provision: (1) participants were not required to lose ≥3 kg by 6 months in order to receive additional visits, and (2) visits remained monthly during year 2 rather than resuming an annual benefit. These changes were made so that the study arms had the same session frequencies during the final 18 months of the intervention and to prevent higher attrition in the FFS arm. Visits were paid by the study at the rate reimbursed by Medicare (approximately $27 per visit). Practices were encouraged but not required to establish billing for patients insured by Medicare. Each practice selected 1 to 2 counselors, most commonly nurses, who conducted visits incident to PCPs, as permitted by CMS, with coverage allowed if the PCP is physically present in the setting when ancillary providers deliver the counseling.50
Counselors and PCPs received a 1-time, 3-hour, in-person training focused on diet and physical activity guidelines, behavioral strategies, and motivational interviewing. Each practice also received a toolkit to facilitate the delivery of IBT, including example visit objectives based on the 5As model (assess, advise, agree, assist, arrange), case studies, and participant handouts.51 No follow-up training or monitoring was provided.
In the PCMH arm, practice-employed clinicians (eg, nurses, registered dieticians, advance practice providers, physicians) delivered the lifestyle intervention with group visits held at the practice, with a median of 14 participants per group (range, 8-18 participants). Group visits were held during the lunch hour or before and after regular clinic hours, and offered opportunities for interactions with other participants. Group visits were 60 minutes each and were weekly for the first 3 months, every other week for months 4 to 6, and monthly thereafter. The first 14 visits were delivered face to face. Practices had the option to switch to group conference calls for subsequent visits; however, all but 1 practice opted to continue face-to-face visits. One to 3 counselors, predominantly nurses, were chosen locally to implement the intervention.
Training in the PCMH arm included the same 3-hour in-person training as in the FFS arm plus additional training for leading a comprehensive group-based lifestyle intervention. Counselors received a standardized treatment manual and an accompanying participant manual, a 1-day in-person workshop focused on group facilitation skills, and optional biweekly to monthly group telehealth/phone training sessions designed to reinforce core intervention principles and address issues as they arose. Fidelity monitoring was limited using a pragmatic approach; central study personnel observed the program delivery and provided feedback at least once per practice, and counselors used an electronic tool52,53 to capture attendance, participant progress, and a simple checklist documenting the completion of core components (eg, log feedback, weigh-in, check-in, planned topic, checkout).
Participants at the practices randomized to the DM arm received the same group-based lifestyle intervention as the PCMH arm, but the intervention was delivered remotely via conference calls by central study staff. The same treatment manual was used; visits were the same duration and frequency; and groups were of a similar size, with a median of 14 participants per group (range, 10-17 participants). DM groups were composed of patients from multiple practices. Visits were held during the lunch hour or before or after regular clinic hours. Participants were instructed in ground rules for group calls, including being in a place free from distraction, avoiding using mute, and actively participating in the discussion. Counselors encouraged direct patient-to-patient interaction through the use of directed summaries, reflections, and questions. The counselors had graduate degrees in relevant fields (eg, nutrition, exercise science, psychology). Training included the same 3-hour training as in the FFS and PCMH arms plus shadowing an experienced counselor and regular weekly to monthly staff meetings. Fidelity monitoring included a review of recorded sessions; counselors also used the same visit checklist documenting the completion of core intervention components. To enhance integration with primary care, phone counselors sent 5 progress reports to each participant's PCP that included weight loss to date, laboratory test results, diet and physical activity remarks, participant-reported barriers and motivators, and concise recommendations for PCP action (eg, praise weight loss; discuss plans to maintain tracking and exercise).
Local practice personnel were responsible for scheduling and conducting in-person data collection visits at baseline, 6, 18, and 24 months. In-clinic measures included height, weight, blood pressure, and cardiometabolic laboratory tests. Practice personnel were trained and certified in data collection protocols during an in-person training session. Participants were instructed to fast and wear lightweight clothing (shorts and T-shirt); gowns were provided as needed. Practice staff entered data directly in an online study Research Electronic Capture (REDCap)52,53 database or used paper source documents that were faxed to the central study team and quality checked. Practices were paid on a monthly basis for each data collection visit completed once the data were entered in REDCap. Participants completed surveys online or via paper, depending on participant preference. Central study personnel managed survey data completion, and double data entry was used for all paper surveys.
Height was collected at the baseline/enrollment visit only using a stadiometer. Body weight was measured using a calibrated digital study scale accurate to 0.1 kg (MX-115, Befour, Inc). A study scale was provided to each practice.
Blood pressure was measured with participants in a seated position after resting for 5 minutes. Two blood pressure measures were collected consecutively and then averaged. If either the systolic or diastolic measurements were >5 mm Hg apart, a third measurement was taken after a 1-minute break and that value was used. A fasting blood sample was collected at baseline, 6 months, and 24 months. Samples were collected, processed, and analyzed in accordance with the practice's standard operating procedures. Samples were analyzed for fasting blood glucose levels and lipid panel (ie, total cholesterol, triglycerides, low-density lipoprotein [LDL] cholesterol, and high-density lipoprotein [HDL] cholesterol).
All clinic-based and survey measures were collected at baseline, 6 months, and 24 months; diet and physical activity measures were also collected at 12 months; and QOL, sleep, stress, and mood measures were also collected at 18 months.
Patient demographic information and medical history were self-reported. Travel time to the clinic via personal vehicle from the participant's home address was calculated using ESRI ArcGIS Desktop 10.7. Assessment of weight loss history included the lifetime frequency of seeking assistance for weight loss.54
The Modifiable Activity Questionnaire (MAQ) was used to measure the duration and frequency of 38 physical activities over the past 7 days.55 Physical activity was calculated as the metabolic equivalent of task (MET) hours per week derived by multiplying the total time of an activity by the respective MET and summing across activities.56 The MAQ has evidence of convergent validity compared with accelerometer measurement across various intensity ranges.57 Screen time is assessed on the MAQ, with a single item assessing nonoccupational television and other screen time per day over the last 7 days.
The percentage of kilocalories (kcal) from fat was assessed using the National Cancer Institute's Energy Screener. This measure has demonstrated validity among individuals with obesity compared with the Food Frequency Questionnaire58 and 24-hour dietary recall59; however, it should be noted that that the Food Frequency Questionnaire and 24-hour dietary recalls tend to underestimate energy intake compared with objective measures such as doubly labeled water.60 The frequencies of fast food and sugar-sweetened beverage consumption were assessed with 3 items from the Behavioral Risk Factors Surveillance System.61 Fruit and vegetable intake was measured using a 2-item questionnaire assessing the number of cups of fruits and vegetables consumed daily. This low-burden measure has good 2-week test–retest reliability (intraclass correlation coefficient [ICC], 0.63) and moderate validity (r = 0.40) with 24-hour recall values.62
The 12-item Short Form (SF-12) v2 was administered as a widely used measure of physical and mental health–related QOL. The SF-12 differentiates QOL between individuals with obesity and those who are of normal weight.63 The Physical Component Score (PCS) and Mental Component Score (MCS) are based on a t score (mean of 50 and SD of 10).64,65
The Impact of Weight on Quality of Life-Lite (IWQOL-L) includes 31 items and 5 subscales of Physical Function, Self-Esteem, Sexual Life, Public Distress, and Work.66 The IWQOL-L has demonstrated internal consistency reliability and construct validity compared with other QOL measures, detects QOL changes in response to weight loss during an intervention, and distinguishes differences in QOL based on BMI.66-68 Scores (subscales and total) are transformed into a 0 to 100 scale, where 100 represents the best QOL.
The Pittsburgh Sleep Quality Index (PSQI)69 was used to measure overall sleep quality and is based on component scores for subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction. The PSQI distinguishes between sleep quality in individuals with obesity/metabolic syndrome and those who are of normal weight.70 The sum of the 7 component scores yields the global sleep score, with a range from 0 (better) to 21 (worse).
The Perceived Stress Scale (PSS) is a widely used 10-item measure71 with established internal consistency, factorial validity, and construct validity in a variety of samples, including adults with obesity.72 Scores range from 0 to 40, with higher values representing greater levels of stress.73
The Patient-Reported Outcomes Measurement Information System-29 (PROMIS-29) Anxiety and Depression Scales74 have been shown to have good internal reliability and strong convergent validity, with well-established measures of anxiety and depression, including the Generalized Anxiety Disorder-7 (GAD-7) and the Patient Health Questionnaire-9 (PHQ-9).75 Scales are scored using t scores.
Patient experience of care was assessed at 6 and 24 months with 10 Likert items adapted from integrated care performance measure recommendations from the Center for Health Care Strategies.76 Satisfaction with the intervention was assessed with 8 Likert items adapted from similar previous obesity treatment trials.77-79 Participants in the PCMH and DM arms also completed 2 items assessing group cohesion and an 8-item scale assessing overall satisfaction with their group leader.77
The RE-AIM framework was used to evaluate reach (participant participation rate and representativeness of participants), adoption (representativeness of participating practices), implementation (process and extent to which interventions were delivered), and maintenance (at the practice level).80 Both quantitative (survey) and qualitative (interview) evaluation tools were used.
Reach measures included the response rate to the mailing, eligibility rates, participation rates, and representativeness of the enrolled participants. Among participants who were screened, we reported the number and percentage who were eligible and who declined, in total and by study arm. We reported different participation rates, as recommended by Glasgow.81 From least to most conservative, participation rates included the number enrolled divided by (1) the number who were screened and were assumed to be eligible (applying the eligibility rate at each screening step to those who declined before completing screening); (2) the number who contacted the central study team (by phone, email, or mail) and were assumed to be eligible; and (3) the number who were mailed to and were assumed to be eligible. To assess representativeness, we compared participant characteristics (age, sex, race, rurality,82 months since last clinic visit, and BMI) between enrolled participants and nonenrolled participants on mailing lists. Additional comparisons across enrolled and nonenrolled participants were made with the subset of participants from the 10 Marshfield Clinic sites, including comorbid conditions and health care use (ambulatory visits and in-patient days in the past 3 years). These characteristics were extracted from the electronic medical record (EMR) for all participants who met the initial eligibility criteria and were “medically homed” to the Marshfield Clinic Health System, defined by the Marshfield Clinic Institute for Quality, Innovation, and Patient Safety as having reasonably complete capture of all medical care, per standard quality reporting definitions.83
Representativeness of practices was evaluated by comparing practice characteristics (eg, clinic type, ownership, staff size, rurality) of participating and nonparticipating practices that were contacted about the study, within the subset of sites affiliated with KUMC.
Participant-level measures included counseling visit attendance and participant experience of care and satisfaction surveys noted previously. Practice-level measures included assessment of completion of clinician and counselor training and counselor competencies self-assessed at pre- and postintervention. In addition, PCPs, counselors, and office managers completed Likert-scaled survey questions at midintervention (6 months) and postintervention (24 months) assessing barriers and facilitators to implementation. Surveys were completed by the lead study PCP at each site, 1 counselor per site in the FFS and PCMH arms, and the office manager at each site if the site had someone in that role working on the study. Survey questions were based on constructs from the CFIR42 that have been found to be important for implementing weight loss services.84 The questions addressed constructs related to intervention characteristics (eg, relative advantage, complexity, cost), inner setting (eg, climate compatibility, resources, communication), and implementation process (eg, engaging champions, planning, executing). The postintervention questions also focused on the sustainability of the intervention at the practice. The responses formed the basis of subsequent qualitative interviews with a subset of practice personnel (see “Qualitative Measures” section).
Maintenance at the practice level was measured as the number of practices in the study that continue or begin to offer weight management services to their broader patient population with obesity and the number that plan to continue or to begin billing insurance for weight management visits.
Perceptions of barriers and facilitators to implementation and sustainability among practice personnel were assessed using an explanatory sequential mixed-methods design, whereby practice personnel completed quantitative surveys and wherein structured interviews were conducted to help explain the survey responses.85 Interviews were conducted at mid- and postintervention, approximately 2 weeks after the surveys were sent, with personnel purposively sampled at the 21 KUMC-affiliated practices. Interviews were conducted via telephone by a co-investigator (K.S.K.) trained in qualitative methods and lasted 30 to 60 minutes. Detailed notes were taken by the interviewer, and the interviews were recorded and subsequently transcribed verbatim.
In addition, 1 to 2 patients from each practice were randomly selected for interviews after completion of their 24-month visit. Participant sampling for interviews was stratified by level of participation (high vs low visit attendance), magnitude of weight loss (<5%, 5%-10%, >10%), and sex. A random number generator was used to assign rank to the stratified lists, and interviewers contacted participants at each site in assigned order until 2 interviews per practice were completed. The interviews lasted 15 to 30 minutes, were conducted via telephone, and followed a standardized interview guide. Interview questions focused on participants' experiences with counselors, factors influencing visit attendance, impact of rural sociocultural and environmental constraints, and experiences with PCPs regarding weight management during the study. All participant interviews were recorded and transcribed verbatim.
Strategies to improve retention for assessment visits included reminder letters, newsletters, thank-you cards, small incentives, and outreach for participants at risk for attrition (visit not scheduled or prior no-shows). There were no incentives for retention to counseling visits, although counselors were encouraged to reach out to participants who stopped attending. Retention letters for the final assessment visit were sent via certified mail to participants who were nonresponsive to all other forms of outreach. Walk-in visits were offered at clinics where resources allowed. In-home visits were completed on a case-by-case basis for a small number of participants (n = 22). Central study coordinators had regular phone/email contact (at least twice per month) with practice staff and made 1 to 2 visits to each practice per year to maintain collaborative working relationships and help ensure compliance with retention and other study protocols.
To determine sample size, we set the ICC at 0.05 and the type I error at 0.025 (overall type I error rate of 0.05 for 2 primary hypotheses comparing both the PCMH and DM arms with the FFS arm). With 36 practices and 40 participants per practice (N = 1440 participants total), the trial had 80% power to detect a net treatment effect of 2.75 kg (SD = 8 kg). This effect size was supported by the available literature86,87 and our previous work.88
Analyses were conducted using linear mixed-effects multilevel models, which included random cluster (clinic) effects, to examine changes in weight (in kilograms) and percentage weight loss over time. An unstructured covariance matrix was used, and missing weights were treated as missing at random (MAR) and addressed using maximum likelihood methods. The 36 clinics that were randomized and enrolled patients in the trial were included. Generalized linear mixed models were used to compare the percentages of participants achieving the 5% and 10% weight loss thresholds at 6 and 24 months. Models included affiliated academic institution based on stratification in randomization. Potential covariates were evaluated by examining imbalances between arms on baseline characteristics. Characteristics that differed with an effect size of Cohen d >0.20 (sex, race/ethnicity, education, diabetes, cardiovascular disease, and travel time to clinic) were included in the adjusted models in sensitivity analyses.
Three additional post hoc sensitivity analyses were conducted excluding the single VA clinic (from the Nebraska-Western Iowa VA affiliated with the University of Nebraska Medical Center), excluding the 2 nonrandomized practices that were recruited to replace sites that dropped out, and imputing baseline weight for missing 24-month weights. Secondary survey measure outcomes and counseling attendance were also compared across treatment arms using separate hierarchical linear mixed models with adjustment for affiliated academic institution. Generalized linear mixed models were used for categorical secondary outcomes (eg, proportion meeting the guideline of ≥5 servings per day of fruit and vegetable intake). For individual 5-point Likert items assessing participant experience of care and satisfaction across arms, cumulative ordinal logistic mixed models were used. CIs were set at 97.5% for the 2 primary comparisons and at 95% for secondary outcomes. Because of the potential for type I error due to multiple comparisons, the findings for secondary end points should be considered exploratory.
HTE was examined across the following participant characteristics: sex, race/ethnicity (White non-Hispanic vs other), education (Bachelor's degree or higher vs less than a Bachelor's degree), annual household income (≥$50 000 vs <$50 000), employment status (full time vs other), travel time to clinic (≥15 minutes vs <15 minutes), and weight loss history (any history of seeking assistance for weight loss vs no history).54 Interaction terms between each of the above-mentioned participant characteristics, time, and treatment arm were added to hierarchical linear mixed models predicting weight loss over 24 months.
Evaluation of the RE-AIM measures was primarily descriptive and exploratory. For reach measures, analysis of variance was conducted to examine differences across study arms in mailing response rates, eligibility rates, and participation rates. Rates were calculated at the clinic level and then averaged across clinics to account for site-level variation. To evaluate representativeness, we compared the characteristics of study participants vs nonparticipants at the clinic level using paired t tests. For practice-level adoption, practice characteristics across participating and nonparticipating practices were compared using χ2 and t tests, with Cohen d calculation for effect sizes. For practice personnel-level implementation measures and Likert survey questions, simple frequencies were calculated.
Transcripts of practice personnel and patients were coded in Dedoose,89 a cloud-based mixed-methods analysis software. For the practice personnel interviews, transcripts were analyzed using deductive, template analysis,90,91 starting with a priori codes derived from the CFIR. Three members of the study team were trained on the coding template using the interrater reliability testing feature in Dedoose. A Cohen pooled κ value of 0.6 was set as the target level of agreement with the expert coder (the principal investigator [PI]). Coders who did not achieve the target value met with the investigators to resolve coding discrepancies. The coders then coded all interviews, flagging textual units about which they were uncertain, which were reviewed and resolved with the investigators. After all interviews were coded, content analysis92 was used to assess the frequency of coding for each construct and to assess whether each construct was a facilitator or barrier to implementation.
In line with the sequential mixed-methods design,85 we first reviewed the frequencies of quantitative survey responses and noted items where there was discrepancy across arms (defined as a difference of ≥20% who agreed/strongly agreed), as well as the overall degree of favorable to unfavorable responses across domains. Qualitative data were examined for context to help elucidate the reasons for these discrepancies.
To analyze the participant interviews, we used inductive qualitative thematic analysis.92 Interviews were deconstructed into meaning units, condensed meaning units, and codes. Similar codes were grouped into categories through the process of convergence and divergence, leading ultimately to themes.93 Peer debriefing,94 whereby qualitative data coders, analysts, and the PI met weekly to share and challenge initial interpretations and arrive at consensus, was used to enhance the trustworthiness of the findings and methodological rigor.
The study protocol was refined post-award as more detailed implementation plans were developed. Upper limits for inclusion criteria for age and BMI were added, and end-stage renal and liver disease were added as exclusions. The waist circumference outcome was removed, as it proved to be too difficult to perform reliably during initial training of the practice staff. Participant-reported measures were added (ie, IWQOL-L, PROMIS-29 Anxiety and Depression scales, participant experience of care) in part to align with the PROPEL trial funded under the same request for applications.95 The frequency of surveys was reduced to lessen participant burden; all surveys were collected at baseline, 6 months, and 24 months, but some were alternated at interim 12- and 18-month time points. The analysis plan was updated before beginning the analyses. Hierarchical linear mixed models were originally proposed to include baseline and 24-month weights only and were changed to include interim weights at 6 and 18 months as a more efficient approach. The longitudinal model handles missing data by treating values as MAR. As such, the multiple-imputation methods as originally described were no longer needed. All noted changes were discussed with PCORI before implementation and were approved by the KUMC IRB.
Thirty-six rural primary care practices were randomized and enrolled patients in the study (Table 2). Twenty-one (58.3%) were affiliated with KUMC, 10 (27.8%) with Marshfield Clinic Health Systems, and 5 (13.9%) with the University of Nebraska Medical Center. Clinics were located in communities classified as isolated rural to large rural based on rural-urban commuting area codes.8 One exception was the VA practice site, which is located in an urban setting but serves a large proportion of rural residents. The median travel distance from the practices to a tertiary care hospital was 59 miles (interquartile range [IQR], 17-87 miles).
Practice Characteristics.
Ownership of the clinics varied, with 22 (61.1%) being hospital owned, 11 (30.6%) being privately owned, and 3 (8.3%) owned by a nonprofit board. Ten (27.9%) of the participating clinics were designated rural health clinics, and 12 (33.3%) were Federally Qualified Health Centers. The median full-time equivalent (FTE) for physicians was 4.0 (IQR, 2.6-5.0) per clinic, with considerable variability. Advanced practice providers (advanced practice registered nurses, physician assistants) were on staff in all but 2 of the clinics, with a median FTE of 2.9 (IQR, 1.0-3.9) per clinic. Eleven sites (30.6%) had access to an onsite registered dietitian (RD), and 15 (41.7%) had mental health staff (2 sites randomized to FFS and 6 sites randomized to PCMH used 1 of these RDs or behavioral staff to deliver the RE-POWER counseling visits; see below for additional counselor characteristics).
Lead study PCPs at the 36 practices had a median of 15 years of work experience (IQR, 10-24 years), and all were family medicine specialists (31 physicians and 5 advanced practice providers) (Table 3). There was a total of 55 counselors across study arms. Counselors for the FFS arm (n = 19) and PCMH arm (n = 22) were employees of the local practices (median, 2 counselors per practice), whereas counselors for the DM arm (n = 14) were central research personnel employed by KUMC. Counselors were predominantly registered nurses and RDs but included a broad range of health professionals. Only 2 physicians (1 each in the FFS and PCMH arms) and 10 advance practice providers (4 in the FFS arm and 6 in the PCMH arm) served as intervention counselors. Counseling sessions with PCPs were conducted as separate visits distinct from routine medical visits. The median number of years in practice tended to be higher among counselors in the FFS and PCMH arms (both 9 years) than in the DM arm (1 year), whereas the proportion of counselors with prior experience in weight loss counseling tended to be lower in the FFS and PCMH arms (2/19 and 6/22, respectively) than in the DM arm (7/14) (Table 3).
PCP and Counselor Characteristics.
A total of 1931 participants were screened for the study; of these, 255 participants (13.2%) were ineligible, and 204 participants declined or lost contact during the enrollment process (Figure 1). The most common reason for exclusion was BMI being out of range (n = 138), followed by lack of clearance from their PCP (n = 40). A total of 1432 participants were enrolled, with a median of 40 participants per clinic (range, 34-44). Participant mean age was 54.7 years (SD, 11.8); mean BMI was 36.7 (SD, 4.0); 76.8% of participants were female; and 46.8% were from an isolated rural area8 (Table 4). Comorbid conditions included hypertension (45.7%), diabetes (24.0%), arthritis (33.5%), and depression or other mental health conditions (39.2%). The baseline physical functioning score (measured by SF-12 PCS, 45.0 ± 9.8) was 0.5 SD lower than that of national samples (including rural and urban); both the PCS and MCS were 0.2 SD lower than those of a similar primary care–based behavioral weight loss trial.96 Baseline weight-related QOL was poor (IWQOL-L, 66.4 ± 18.1) and on average at the level of severe impairment (defined as ≥3 SD from normative mean among adults who were not seeking treatment).97
Participant Flow Diagram.
Baseline Participant Characteristics.
Participants removed from the study (n = 25; for pregnancy [n = 9], bariatric surgery [n = 4], major medical contraindications [n = 9], or death [n = 3]) were excluded from analyses based on a priori specification in the protocol. Among the remaining 1407 participants, data collection visit retention at 6 months was 91.7% overall (92.2% in the FFS arm, 91.7% in the PCMH arm, and 91.2% in the DM arm). At 24 months, data collection visit retention was 86.7% overall (86.7% in the FFS arm, 87.6% in the PCMH arm, and 85.8% in the DM arm). Survey retention was 92.7% at 6 months and 87.9% at 24 months, with no statistical differences across arms.
At 6 months, estimated weight loss was −8.34 kg (95% CI, −9.24 to −7.45 kg) in the PCMH arm, −7.72 kg (95% CI, −8.62 to −6.81 kg) in the DM arm, and −5.73 kg (95% CI, −6.65 to −4.81 kg) in the FFS arm. The net between-arm change (minus the weight change in the FFS arm) at 6 months was −2.61 kg (95% CI, −3.82 to −1.40 kg; P < .001) for the PCMH arm and −1.99 kg (95% CI, −3.37 to −0.60 kg; P = .002) for the DM arm (Figure 2 and Table 5). At 24 months, weight loss was −4.43 kg (95% CI, −5.48 to −3.39 kg) in the PCMH arm, −3.92 kg (95% CI, −4.98 to −2.87 kg) in the DM arm, and −2.56 kg (95% CI, −3.63 to −1.49 kg) in the FFS arm, corresponding to net between-arm changes of −1.87 kg (97.5% CI, −3.51 to −0.23 kg; P = .01) for the PCMH arm and −1.36 kg (97.5% CI, −3.00 to 0.29 kg; P = .06) for the DM arm.
Estimated Mean Weight Change by Treatment Arm.
Mean Weight Loss (kg and % of Baseline) Over 24 Months, Unadjusted.
There were no statistically significant or meaningful between-arm differences between the PCMH and DM arms (−0.51-kg [95% CI, −1.94 to 0.92 kg] difference at 24 months). The results were similar for percentage weight loss (Table 5). Models with adjustment for participant covariates showed similar results (Appendix B, Supplemental Table S1). Sensitivity analyses after excluding the single VA clinic (the only practice in the PCMH arm that switched to phone-based group visits) and excluding the 2 nonrandomized sites showed similar findings. Likewise, sensitivity analyses addressing the ∼15% missingness at 24 months follow-up by imputing a missing value by substituting the baseline value showed similar findings.
Categorical weight loss results (ie, treating weight loss as a categorical variable instead of a continuous one) are shown in Figure 3. At 6 months, a higher percentage of participants in both the PCMH arm and the DM arm achieved >5% and >10% weight loss than did those in the FFS arm. At 24 months, there were no significant differences across arms in the percentages achieving >5% and >10% weight loss. At 24 months, 44.1% (95% CI, 35.2%-47.8%) in the PCMH arm, 41.4% (95% CI, 37.9%-50.6%) in the DM arm, and 36.0% (95% CI, 30.2%-42.3%) in the FFS arm lost >5% of their weight. For the >10% threshold, the percentages were 22.6% (95% CI, 18.1%-27.9%), 22.3% (95% CI, 17.9%-27.6%), and 17.1% (95% CI, 13.3%-21.8%) in the PCMH, DM, and FFS arms, respectively.
Proportion of Participants Who Met the 5% and 10% Weight Loss Threshold at 6 and 24 Months.
There was a significant treatment × factor interaction for income and a significant treatment by factor × time interaction for employment. There was no effect modification based on participant factors of sex, race/ethnicity, education, travel time to clinic, or history of seeking assistance for weight loss. Mean weight loss values by income and employment subgroups are shown in Table 6. For income, across all time points, participants in the DM arm had greater weight loss than did those in the FFS arm only for those with high annual income (≥$50 000) and not for participants with low annual income (<$50 000). The net between-arm change (weight change in DM arm − the weight change in the FFS arm) at 24 months was −2.46 kg (95% CI, −4.18 to −0.75 kg) among participants with high income, compared with 0.17 kg (95% CI, −1.79 to 2.13 kg) among participants with low income. In addition, among participants with low income (but not high income), participants in the DM arm had less weight loss at 6 months than did those in the PCMH arm (net difference, −1.92 kg [95% CI, −3.48 to −0.36 kg]). A similar pattern was observed for employment status, with higher net weight loss for the DM arm than for the FFS arm observed only for those employed full time; however, this effect modification by employment status was observed at 18 months only.
Mean Weight Loss (kg) Over 24 Months by Income and Employment Subgroups.
Table 7 shows changes in cardiometabolic risk factors over 24 months. Participants in the PCMH arm had statistically greater and clinically meaningful reduction in triglyceride levels at 6 months (−25.4 mg/dL) and 24 months (−19.2 mg/dL) than did participants in the FFS arm (−12.6 mg/dL and −0.5 mg/dL at 6 and 24 months, respectively). Changes in other cardiometabolic risk factors were modest to minimal and not significant across arms.
Mean Change in Cardiometabolic Risk Factors Over 24 Months.
Table 8 shows reported physical activity changes over 24 months. Participants in the PCMH arm had significantly greater and more clinically meaningful increases in physical activity at 6 months (+9.8 MET hours/week) and 24 months (+4.8 MET hours/week) than did participants in the FFS arm (+5.4 MET hours/week at 6 months and +2.6 MET hours/week at 24 months). Physical activity among participants in the DM arm increased significantly by +7.9 MET hours/week at 6 months and +4.3 MET hours/week at 24 months, with no between-arm differences from either the FFS or PCMH arm. Reported screen time was reduced modestly in all arms by approximately 0.5 hours per day at 24 months, with no significant differences across arms.
Mean Change in Reported Physical Activity and Screen Time Over 24 Months.
Changes in diet measures are shown in Table 9. At 6 months, participants in the PCMH arm and the DM arm reported significantly greater reductions in percentage kcal from fat (both −2.5% kcal) than did participants in the FFS arm (−1.7% kcal); however, the between-arm differences were small in magnitude and may not be clinically meaningful. At 6 months, there was a significantly greater and clinically meaningful difference in the proportion of participants in the PCMH arm and DM arm who reported meeting the guideline of ≥5 fruit and vegetable servings per day (31.3% and 30.9%, respectively) than in the FFS arm (23.4%). There were no significant or meaningful between-arm differences at 6 months in sugar-sweetened beverage consumption, number of meals out, or fast food consumption. At 24 months, there were no significant or meaningful differences across arms in diet measures. Participants within all arms reported significant but small improvements from baseline through 24 months in all diet measures (% kcal from fat, sugar-sweetened beverages, meals out, fast food consumption, fruit and vegetable intake), with some attenuation in improved scores from 6 to 24 months.
Mean Change in Reported Diet Measures Over 24 Months.
Changes in QOL and psychosocial measures are shown in Table 10. Overall, participants within all arms had significant improvements in most QOL and psychosocial measures over 24 months, with no significant differences across arms. The 1 between-arm difference was significantly lower improvements in the SF-12 MCS at 6 months in the DM arm than in the PCMH and FFS arms; however, the magnitude of difference was small and below a minimal clinically important difference (MCID).98,99 At 6 and 24 months, participants within all arms had meaningful increases in weight-related QOL (measured by IWQOL-L), including subscale scores for physical function, self-esteem, and sexual life, as well as more-modest but significant improvements in public distress and work. Participants also had improved sleep quality (measured by PSQI; scores decreased) at 6 and 24 months within all arms. Improvements in general QOL (measured by SF-12) and stress (measured by PSS) were modest and not observed across all time points. Participants in all arms had significant increases in anxiety (measured by PROMIS-29); however, changes over time were below an MCID,100 except for small but potentially meaningful increases at 18 months (+3.0 to +3.6 across arms).101
Mean Change in QOL and Psychosocial Measures Over 24 Months.
Visit attendance is shown in Table 11. Attendance was higher among participants in the FFS arm than among participants in the group-based PCMH and DM arms, both during the first 6 months (86.4% vs 71.6% and 66.2%, respectively) and from 6 to 24 months (58.9% vs 40.9% and 35.9%, respectively). The proportions of participants who attended at least 1 visit during the maintenance phase (6-24 months) were 89.2% in FFS, 76.5% in PCMH, and 71.9% in DM.
Mean Number and Percentage of Counseling Visits Attended.
We found several differences in patient experience of care and satisfaction measures across arms at 6 and 24 months. Ratings declined (ie, became less favorable) from 6 to 24 months; however, the decline was similar across arms. Therefore, the ratings at 24 months are presented in Table 12, and the results at 6 months are provided in Appendix B, Supplemental Table S2.
Patient Experience of Care and Satisfaction at 24 Months.
Consistent with their higher attendance levels, participants in the FFS arm were more likely to agree that appointment times were convenient (item 2; for all items, see Table 12) than were participants in both the PCMH and DM arms. Participants in the FFS arm were also more likely to report that the number of visits (item 14) was “just right” than were participants in the PCMH arm, who were more likely to report that the number of visits was “too little.” Participants in the DM arm rated several items related to their experiences at the clinics less favorably than did participants in the FFS and PCMH arms (see items 8-10). They also rated the helpfulness of their PCP lower (27.7% rated “extremely helpful” vs 46.3% in FFS and 47.7% in PCMH; item 16). In addition, participants in the DM arm, compared with those in the FFS arm, were less likely to agree that the clinic staff were sensitive to their unique needs (item 6) and that they felt respected by the clinic staff (item 7). Across the 2 group-based arms, 34.3% of participants in the DM arm vs 50.6% in the PCMH arm were “very comfortable” participating in their group (item 19); similarly, 19.9% vs 54.5% of PCMH group members “bonded very well” with other group members (item 20). However, participants in the DM and PCMH arms reported nearly identical ratings of satisfaction with their group leaders (mean of 34 on a scale from 8-40; item 21). Furthermore, despite lower ratings on several items among participants in the DM arm, their overall satisfaction (eg, 31.2% “very satisfied”; item 17) was similar to that of participants in the PCMH arm (33.4%) and higher than that of participants in the FFS arm (25.2%). Finally, the likelihood of recommending the intervention (item 18) was highest among participants in the PCMH arm, with 80% likely or very likely to recommend, compared with 71% in the FFS arm and 68% in the DM arm.
Mailings were sent to 15 076 participants, with a median of 357 mailings per clinic. Of these, 1990 participants contacted the study telephone line, for a mailing response rate of 13.2%. There was no statistically significant difference in mailing response rates across study arms (P = .42). A total of 2479 potential participants contacted the study line, of whom 66.1% reported being referred by the mailing, 21.5% by a provider during a clinic visit, and 11.0% from media or family/friends. There were no statistically significant differences across arms in the proportion referred by the mailing (P = .94) or by provider (P = .41).
The numbers and percentages of screened participants who were ineligible or declined to participate at various stages are summarized in Table 13. There were no statistical differences in ineligibility or decline rates across study arms nor in participation rates. The participation rates (among those presumed eligible through all stages of enrollment) were as follows: (1) 86.0% of those who screened eligible; (2) 66.3% of those who contacted the study line; and (3) 15.7% of those who received the study mailing.
Number of Participants Who Were Ineligible or Declined to Participate, as Percentage of Participants Screened.
Table 14 shows the representativeness of enrolled participants (n = 1432) vs nonparticipants (n = 17 497) across all 36 clinics, based on available participant characteristics. Compared with nonparticipants, enrolled participants were significantly older (54.1 vs 51.3 years; P < .001) and were more likely to be female (76.9% vs 55.0%; P < .001) and to live in an isolated rural area (46.3% vs 41.7%; P < .001). They also had a statistically significant higher BMI (36.5 vs 35.6; P < .001) and statistically significantly more recent clinic visit (3.9 vs 4.7 months since last visit; P = .02), although the clinical significance of these differences is minimal. In sensitivity analyses to examine the potential influence of the mailing on representativeness, comparisons of these characteristics were also made between enrolled participants and only those nonparticipants who were mailed a study invitation (n = 13 858), as well as for the subsample of participants and nonparticipants from the 10 Marshfield Clinic sites, where uniform population-based EMR data extraction methods were used to create the lists. In both sensitivity analyses, the results were the same, with only 1 exception: there were no differences in rurality between participants and nonparticipants at the Marshfield sites.
Representativeness of Participants: Demographics and BMI of Enrolled Participants vs Nonparticipants Who Were Patients at Participating Clinics.
Table 15 shows representativeness based on comorbid conditions and health care use for the subsample of participants (n = 402) and nonparticipants (n = 6199) from the Marshfield Clinic sites. Compared with nonparticipants, enrolled participants were less likely to have cardiovascular disease (2.5% vs 5.0%; P = .03) and to be current smokers (5.9% vs 16.9%; P < .001) but were more likely to have prior joint replacement surgery (10.9% vs 8.1%; P = .02). They also had fewer inpatient days over the previous 3 years (P = .001). No differences were observed for type 2 diabetes, history of cancer, kidney disease, pulmonary disease, depression, or number of outpatient encounters.
Representativeness of Participants: Medical Comorbidities and Health Care Use of Enrolled Participants vs Nonparticipants Who Were Patients at a Subset of Clinics.
Fifty-three rural practices were contacted by KUMC, and 21 practices participated in the trial. Among practices that declined (n = 32), 14 had no response to a recruitment email/fax; 11 responded but were not interested; and 7 were interested, attended a visit with the study PI and staff to learn more (5 in person at the clinic and 2 by video), and ultimately declined due to competing demands. Table 16 shows the structural characteristics of participating and nonparticipating practices that were approached by KUMC. No meaningful differences were observed for the number of PCPs, presence of RDs or behavioral health staff, or distance to a tertiary care hospital. There was a potential trend (with moderate effect sizes >0.2 SD) for type of practice ownership (P = .20), rural health clinic status (P = .10), and rurality (P = .12). Participating practices were more likely to be private practices (47.6%) than were nonparticipating practices (18.8%) and less likely to have a rural health clinic designation. In addition, participating practices were more likely to be in isolated rural areas (52%) than were nonparticipating practices (28%).
Representativeness of Practices: Characteristics of Participating and Nonparticipating Practices.
All PCPs (n = 36) and counselors (n = 55) across the 3 study arms completed obesity treatment training either in person at KUMC or at a Marshfield Clinic or equivalent make-up training. Additional training completed by the PCMH counselors included the full-day in-person workshop (21 of 22 attended in person) and optional biweekly to monthly telementoring (median attendance, 54.8% [IQR, 31.9%-75.0%]). Additional training completed by all DM counselors (n = 14) included shadowing an experienced counselor, attending mandatory regular team meetings, and receiving individual feedback of recorded sessions (median, 3.5 [IQR, 2.0-5.3] sessions per counselor; varied based on time as a counselor).
Counselors completed a brief 3-item survey assessing their competencies to address diet and physical activity recommendations, provide helpful feedback on participant self-monitoring logs, and effectively lead a group (group counselors only). On a 10-point scale from low to high competence, median scores ranged from 7 to 8 at preintervention and from 8 to 10 at postintervention. Little to no variability was observed across arms at either pre- or postintervention.
Practice personnel's responses to the implementation surveys are shown in Table 17 (midintervention) and Table 18 (postintervention). These responses are placed within the context of the accompanying qualitative interviews and are described below.
Midintervention Practice Personnel Perceptions of Implementation Determinants.
Postintervention Practice Personnel Perceptions of Implementation and Sustainability.
Practices were queried about weight treatment services offered before the study and at study end as part of postimplementation activities. As of this report, some sites are still in the process of planning how to sustain services. Currently, among the 36 practices, 9 (25%) have begun and 5 (14%) have continued (from before RE-POWER) weight loss services; 8 of 14 are billing for these services. Among the 9 practices that have begun services, 6 were in the FFS arm, all of which are continuing to offer individual visits, and 3 were in the PCMH arm, 1 of which continued group visits and 2 of which continued group plus individual visits. Of the remaining 22 clinics, 5 (14%) are currently referring patients to RDs outside the clinic setting, and 17 (47%) are not providing any weight loss services. Qualitative findings with practice personnel described below provide further context regarding factors influencing sustainability.
Interviews were completed with 21 PCPs, 13 counselors from FFS and PCMH sites, and 15 office managers. If someone served in a dual role, they were interviewed only once. Qualitative findings provided rich elaboration on implementation survey questions. Here, we summarize the most prevalent themes and accompanying survey responses according to the CFIR domains (intervention characteristics, inner setting, implementation process) and highlight any observed differences across arms and across time from midintervention (Table 17) to postintervention (Table 18). We also note any observed differences across practice personnel/roles (ie, between PCPs, counselors, and office managers). The survey responses are reported separately by practice personnel role in Appendix B, Supplemental Tables S3 to S8.
At midintervention, across arms, practice personnel largely liked the intervention, thought the materials were well designed and easy to use (93.2% agreed/strongly agreed), made no modifications (74.6%), and would consider using it after the study ended if supported by insurance (74.7%; see Table 17). At postintervention, most practice personnel across arms (71.3%) reported that they were likely or very likely to recommend the intervention to other practices (see Table 18).
Opinions regarding the relative advantage of the intervention were generally positive, with 75.6% at midintervention and 73.8% at postintervention agreeing that it was “more effective than alternative weight loss services available to our patients.” The key criteria most often used to describe the superiority of the RE-POWER intervention were “accountability and support” and the focus on longer-term results. However, perspectives on relative advantage varied across arms at midintervention, when 58.3% of practice personnel in the DM arm agreed it was more effective than alternatives, compared with 75.0% and 90.0% in the FFS and PCMH arms, respectively. This difference may have been because practice personnel in the DM arm were less familiar with the intervention or thought the lack of local support made it less appealing:
Feedback that I heard was … some of them didn't find them productive because they were over the phone. They didn't really know the people … several of them would ask “Who else is in this?”… So I think they were looking for those connections locally to help hold accountable or to see the success of others … I feel like they were looking for … group support, to make that personal connection like someone they might know in that group… . They had people in that group from all different states … I heard that a lot. (Office manager, DM arm)
Not having local peers as part of the intervention influenced the perceived effectiveness of the FFS model:
What I would read in the goals of patients sometimes when they were faltering a little is that they didn't have a support system or buddies to exercise with. And so [in] our program, they [were] kind of isolating on their own. (Office manager, FFS arm)
However, by postintervention, perceptions of relative advantage evened out across arms. Over time, the perception that the intervention's effectiveness was highly contingent on patient motivation became more prevalent across all arms:
It works and it can work long term, but [it's] kind of up to the patient to make it work. And so just like with any diet or any routine, they can all work, but [it] kind of comes down to the patient. (PCP, FFS arm)
Practice personnel provided substantial elaboration on the financial viability of sustaining the intervention. At postintervention, across delivery arms, a minority of practice personnel (27.5%) agreed that RE-POWER was “a financially viable option that we will be considering now that the study is over.” As one PCP in the FFS arm commented, “I love the idea of RE-POWER, but the practical part of ‘How do you get this reimbursed?' I don't think has been figured out enough yet.” Personnel demonstrated variability in their sophistication regarding making billing work. Even with reasonable knowledge, the rules were complex and “unpredictable,” especially considering the different payer mix. As one PCP noted, there are “just not enough Medicare patients in the practice to make it viable.” In addition, practice personnel at rural health clinics commented on their different billing rules that provide added uncertainty, and those who were part of a larger health care system noted that decisions are made at a higher level, as clinic administrators have to “prove that it won't financially hurt them.”
Practice personnel were also concerned about patient co-pays and deductibles and were not confident in patients' ability or willingness to pay. One PCP in the PCMH arm explained, “I have patients complaining of a $9 prescription; they certainly won't pay for weight loss.” Some were concerned about the impact patient co-pays might have on the patient-provider relationship when the patient does not think they should be charged.
Postintervention, PCPs were somewhat more likely to agree that the intervention was financially viable (38.2%) than were counselors (18.2%) and office managers (20.8%), who were more involved in the day-to-day implementation of the intervention. Perceptions of financial viability were especially low in the PCMH arm among counselors and practice staff. One counselor in the PCMH arm noted that a 1-hour group visit took 3 hours of work, and the time outside the visit would not be billable.
At midintervention, the intervention was perceived to be compatible with the practice climate across all arms; 88.0% of practice personnel agreed that the “intervention fits with our practice style and what we think our patients like.” In addition, 78.3% agreed that the intervention would “lead to lasting changes in how we treat obesity in our practice.” Personnel elaborated that the intervention aligned well with their priority in addressing obesity prevention. At postintervention, there was a drop in the proportion that agreed that the intervention “fits with what their patients like” (66.3%). Over time, some thought that the intervention fit was hindered due to patient factors such as depression. In the DM arm, the perception of fit was dampened by a lack of provider involvement:
We don't necessarily know what's going on, or… how the groups work or… participation, or… are able to encourage people, ‘cause we're not interacting. (Office manager, DM arm)
Others commented that they questioned the fit with what their patients want because patients became less engaged over time:
I mean, I felt like it should fit with our practice style, but then we have lots of trouble with retention of people coming, and so that made me wonder if it maybe didn't fit as well with the patient style of what they wanted. (Counselor, PCMH arm)
However, despite the reduction in participant attendance over time and the effect on perceived fit, at postintervention, nearly 77.5% of practice personnel across arms agreed that “patients are interested in us offering weight management counseling.”
The lack of available resources emerged as an important element that impacted readiness for implementation. At midintervention, 58.6% of personnel in the FFS arm and 70.0% in the PCMH arm agreed that they “have had sufficient internal resources.” At postintervention, only 37.9% in the FFS arm and 28.6% in the PCMH arm agreed that they “have staff who may be available and interested in continuing to provide weight management counseling.” Whereas space was not a significant concern, staffing resources were often seen as insufficient, especially when counselors were not able to block off dedicated time and were pulled into other responsibilities:
She's [another counselor] getting told to do another job because we're getting a new computer system … and I'm gonna be doing this RE-POWER by myself now more. And so I'm stressing out a little bit on how I'm gonna get all these people's appointments seen plus do my regular work. (Counselor, FFS arm)
Finding the staff with the skills and interest to deliver the intervention and fitting them into a staffing model was not a trivial task. Both PCPs and counselors also recognized the potential for counselor burnout. At postintervention, many reported that they did not have the staffing “bandwidth” to continue due to retirements, staff leaving, hiring plans being put on hold, competing priorities on staff's time, and finding the right people who were committed to delivering the intervention. As one practice manager in the PCMH arm explained, “I'm trying to find the time to pull somebody away to do another task. It's kind of tough. They're already getting overtime already, and I don't think they want anymore.” One PCP in the PCMH arm shared, “I have no question that it can be done. I think it's just a question of when is it a right time to do it? And catching somebody who's eager, willing, and committed.”
Personnel in the FFS arm were more likely to agree that they regularly communicate internally about the progress of participants in RE-POWER than were personnel in the PCMH arm (79.3% vs 56.7% in FFS and PCMH, respectively). This difference was driven by counselors, as more than 80% of PCPs in both arms agreed that they regularly communicated, compared with 90.9% vs 41.7% of counselors in the FFS and PCMH arms, respectively. Counselors in the PCMH arm shared that they communicated more on an as-needed basis. They reported encouraging patients to talk with their PCP directly about their progress: “I kind of left the [communication], I wanted my patients to be able to show the doctors their success.” Counselors who were in closer proximity to the PCPs were more likely to share that it was easy to communicate informally. Some counselors simply used the EMR to share notes, while others emailed PCPs routinely or reported on RE-POWER as part of monthly staff meetings.
At midintervention, practice personnel uniformly agreed across arms that providers were engaged in encouraging patients to enroll (86.7%), that they themselves were engaged (97.1%), and that they had a clear plan for implementing the intervention (83.1%). At postintervention, despite inadequate coverage, about half of the personnel (53.8%) across all arms reported that their practice had discussed plans for continuing weight management counseling. In terms of executing the intervention, counselors in the PCMH arm were much less likely to agree that “the amount of work required was what I expected” than were counselors in the FFS arm (36.4% vs 72.7% of counselors in the PCMH and FFS arms, respectively). Counselors in the PCMH arm commented on the unanticipated amount of time needed to prepare for group visits and provide individualized feedback: “The monitoring of diet and exercise logs and things like that. That part was very, very tedious and required a lot of more time than I ever anticipated.” However, despite the greater amount of work perceived by PCMH counselors, they did not perceive it to be a difficult process. In fact, all personnel across arms rated the intervention as fairly easy to implement (median of 3 on a 10-point scale from “very easy” to “very difficult”).
Postintervention, 51.7% of personnel in the FFS arm, 42.9% in the PCMH arm, and 26.1% in the DM arm reported that their practice was likely or very likely to offer weight loss counseling after the study ended. These proportions were consistent across PCPs, counselors, and other personnel. Clearly, having gained the experience of implementing the intervention locally contributed to a higher likelihood of continuing services. Across arms, personnel rated their confidence to sustain weight loss services in the moderate range (median of 6 on a 10-point scale). Several thought that if the intervention were provided at a volume lower than that under study conditions, the reduced staffing burden would make it more sustainable. If it was covered by insurance, they envisioned having a staff person dedicated to providing the intervention, and some believed that the insurance coverage would encourage enough patients to seek treatment.
A total of 73 interviews were completed among study participants after they completed their 24-month data collection visit (n = 24 for FFS, 25 for PCMH, and 24 for DM). Analysis of transcripts confirmed 4 major themes: (1) experiences with counselors, (2) factors influencing visit attendance and adherence, (3) impact of rural factors, and (4) experiences with PCPs. Example quotes by subthemes are provided in Table 19.
Participant Quotes.
The vast majority of participants were positive about their counselors and appreciated their encouragement and availability to help. Across arms, patients reported similar positive experiences with counselors, who were local clinic personnel in the FFS and PCMH arms and central study personnel in the DM arm.
Participants used many attributes to describe what made their counselor effective: being pleasant, encouraging, helpful, friendly, firm without being condescending, “always interested,” willing to talk about how life events were affecting weight, “not judgey,” “down to earth,” fun, and accessible. Participants appreciated those who personalized problem-solving to the participant's particular struggles. A common sentiment was relatability or “doing it with them.” Among participants in the DM arm, understanding rural differences was an important characteristic of effective phone counselors. One participant said, “[The counselor] listened to what I had to say and what others had to say and making sure she understood where they were coming from. Because if you don't live in an isolated area, like, some people don't understand.”
A few relationships with counselors “fizzled,” especially when it seemed the counselor did not care, was perceived to blame the participants, did not respond quickly, did not elicit participation equally in the groups, or was “blah,” had “no charm,” or did not bring any excitement. Participants acknowledged that being a good counselor takes time and attention both during the visit and preparing for the visit, and they noticed the impact when their counselor appeared to have too many other competing responsibilities.
Some participants in the FFS and PCMH arms knew their counselors before the study and appreciated those existing relationships. Others enjoyed getting to know their counselors and creating a new relationship, especially because it was a small town. Another benefit to in-clinic visits was feeling comfortable with addressing health issues and going over laboratory test results; according to one participant, “It was kind of like getting a free visit to the clinic.”
Two primary factors influencing attendance and adherence to the weight loss intervention were visit frequency and delivery mode (phone vs in-clinic [or DM vs FFS and PCMH]). Participants unanimously agreed across all study arms that more-frequent visits, either weekly or biweekly, were a facilitator because they kept participants accountable. In addition, participants largely preferred in-clinic visits over phone visits because the face-to-face interaction held them more accountable. Several discussed how looking someone in the eye in person is more likely to “hold your feet to the fire.” A few participants in the DM arm mentioned how when others had so much success to share and they did not, they shut down and did not want to participate. Others noted that the phone format made it more difficult to elicit participation from introverted people.
Although the downsides of the phone approach were raised more frequently, a few participants appreciated not having to travel, especially those who lived long distances from the clinic or who were concerned about winter road conditions. Somewhat surprisingly, travel was not noted as a barrier to participants in the FFS or PCMH arms. They were used to having to travel to get anywhere, so traveling to the clinic was “not a big deal” and provided a reason to go to town and run errands.
Participants spoke of how food choices were deeply ingrained from their upbringing, which in rural areas predominately centered on gathering around food and preferring “meat and potatoes.” Participants also shared how there is less pressure to be thin in rural areas. Several also spoke about the seasonality with farming and how long workdays during harvest affected their ability to adhere to the intervention.
Participants reported having less access to fresh foods and restaurants than people in urban areas, which was both a curse and a blessing. Not having many options for eating out helped, but not having convenient access to fresh fruits and vegetables was a barrier. Lack of sidewalks, poor road conditions, and long harsh winters were the most commonly cited barriers to outdoor exercise.
Most participants did not perceive themselves to be hindered by socioeconomic factors, and they felt that barriers related to cost or education/knowledge could be managed with the right amount of self-discipline: “It's not that hard to comprehend. It's more of if you want to use what you learned.” At the same time, participants expressed empathy for others who may face greater economic struggles. Participants talked about the added costs of eating a healthy diet, but they also realized the trade-offs that could be made: “I know in my heart that it's about what choices you're going to make and what your priorities are going to be.”
Participants reported having unique relationships with their PCPs due to living in a small town, from seeing each other at church, working with family members, or being good friends for many years. However, most participants reported not regularly engaging with their PCPs about their weight. Participants felt that their PCP did not have time for or was uncomfortable talking about weight. When PCPs did engage, participants already knew it was a problem, were not offended, and believed their PCP was trying to help. Many cited interactions with their PCPs as what motivated them to make a change. One participant said, “Having someone else finally say ‘Yes, you need to lose weight,' just kind of triggers in your head, okay, you need to lose it.” Participants especially appreciated when their PCPs tied their weight loss to improvements in laboratory test results and noted that receiving PCP encouragement was important in the long-term, with one saying, “It's the encouragement that keeps you doing it.”
This trial demonstrated that the PCMH arm with in-clinic group visits provided by local practice personnel resulted in greater weight loss through 24 months than did the FFS arm based on the Medicare IBT model with in-clinic individual office visits. The DM arm with phone-based group visits provided by central personnel also resulted in greater initial weight loss relative to the FFS arm, but statistically significant differences were observed only through 18 months. The level of weight loss achieved in the PCMH and DM arms contributes to mounting evidence that behavioral weight loss interventions in primary care can produce 2-year weight loss outcomes that approach the magnitude (4-7 kg) of academic-based interventions.102 Because trials based in academic medical centers are delivered under more-controlled conditions (eg, with lower participant eligibility rates, research personnel delivering the intervention, provision of food or incentives), the higher observed weight losses often do not translate to outcomes observed in practice. To our knowledge, this is the first pragmatic trial comparing the FFS model with the PCMH and DM models with group visits. Other primary care–based interventions with individual phone visits have observed similar weight losses ranging from 4.0 kg to 4.6 kg at 12 to 24 months.86,103,104 To advance the translation of behavioral obesity science, more pragmatic trials are needed.
The findings also demonstrate that a clinically relevant weight loss threshold can be achieved similarly across different delivery models. Wadden et al recently conducted a pilot study of the Medicare IBT provision among 50 adults aged 21 to 70 years, although the study was conducted in an academic setting by clinicians outside of primary care. Their findings showed that the proportion of participants who achieved >5% weight loss at 12 months105 (44%) was similar to that of participants in the FFS arm in this trial, where visits were conducted locally by practice personnel with minimal training.
Maximum weight loss and between-arm differences were observed at 6 months, with steady weight regain on average from 6 to 24 months across all arms, accompanied by reduction in between-arm effects over time. Although ≥5% weight loss, sustained for 1 to 2 years, is likely to have long-term health benefits despite subsequent regain,21 findings indicate that the most effective delivery strategy for weight loss may not translate to the most optimal approach for long-term extended care. Future trials should focus on primary care models for intervening with weight regain prevention.
Participants in the FFS arm showed higher session attendance than did participants in the PCMH and DM arms, although this did not result in greater weight loss. Individual visits are easier to schedule/reschedule around personal schedules, and consistent with this, participants in the FFS arm were more likely to report that appointment times were convenient than were those in both group-based approaches. Moreover, qualitative interviews indicated that travel to the clinics during regular office hours was not a major barrier. In fact, most participants reported benefits from the FFS model, including the opportunity to meet one-on-one with a counselor, to “look them in the eye” and be held accountable, to discuss their other health conditions, and for some, to interact with a counselor they already knew from living in a small town. Reasons why the higher attendance in the FFS arm did not lead to greater weight loss are likely multifactorial and may be related to the short duration (15 minutes) of the visits, the lower frequency of visits from months 2 to 3 (biweekly compared with weekly for the group-based arms), the lack of ongoing training for the counselors, and/or the lack of group support and accountability.
Participants in the DM arm had the lowest attendance and the lowest ratings on several experience-of-care domains. They were the least likely to agree that they had received the resources they needed at their clinics and that their PCPs were helpful. They were also less likely to feel respected by clinic staff than were participants in the FFS arm, and compared with participants in the PCMH arm, they were less likely to feel comfortable participating in their group. Many commented that the phone delivery lacked sufficient relationship building to hold them accountable, because they did not have to “actually look someone in the eyes,” and it was easier to “shut down” or “tune out” if things were not going well. Participants linked the lower group cohesiveness in the DM arm directly to the phone delivery (“you don't know who you are really talking to”) and not to lower satisfaction with their counselors.
Overall satisfaction and likelihood of recommending the intervention was highest among participants in the PCMH arm. Peer support/accountability was perceived to be a key factor for participants in the PCMH arm and is likely a key component contributing to the greater weight loss.33,35,106 Of note, both participants and practice personnel in the PCMH arm preferred to keep meeting in person rather than switch to conference calls after the first 14 visits. This further highlights the experienced benefits of in-clinic visits.
The observed treatment effects were consistent across most participant characteristics. However, the DM model was less effective among participants with lower income, and the greater weight loss in the DM arm than in the FFS arm was seen only among participants with higher income (≥$50 000). Among different indicators of socioeconomic status, income has emerged as one of the most predictive of health outcomes,107 although the impact of income may be modified by education level.108 To the extent that participants with lower income were dealing with more systematic barriers to health behavior change, the lower accountability and peer support from the phone delivery may have contributed to its worse effectiveness in this subgroup.109 However, these findings are exploratory, and further research is needed to confirm and disentangle why income level may moderate response to in-person vs phone-based delivery.
There were few differences across arms in secondary outcomes, and statistical differences should be interpreted with caution due to multiple-hypothesis testing. Significant modest changes in triglyceride and HDL cholesterol levels were observed through 24 months in both the PCMH and DM arms. In contrast, blood pressure was essentially unchanged in all arms. The modest to minimal reductions in lipid parameters and blood pressure are similar to those reported in prior primary care–based weight loss trials with similar weight loss at 15 to 24 months.86,110,111 With baseline values near normal in these trials, the capacity to detect change is limited with the observed magnitude of weight loss.
The PCMH arm reported greater increases in physical activity at 6 and 24 months than did the FFS arm. In addition, both the PCMH and DM arms reported greater improvements at 6 months in percentage kcal from fat and the proportion meeting ≥5 fruit and vegetable servings/day; however, with attenuation over time, these differences were no longer observed by 24 months. The peer support and modeling in the PCMH and DM arms may facilitate initial changes in physical activity and diet behavior, or the structure of the group sessions may have better reinforced the importance of goal setting and self-monitoring. The magnitude of increases in physical activity in both the PCMH and DM arms are comparable to those from other primary care–based trials.103,112,113 The improvements we observed for percentage kcal from fat compare favorably at 6 months and similarly at 24 months to the POWER-UP primary care trial that included 32 in-clinic individual visits.113 The improvements in reported fruit and vegetable intake at 24 months also compare favorably with findings from the POWER-UP trial, perhaps due to lower baseline scores for measures of fruit and vegetable consumption in our sample and hence greater room for improvement.113 The findings support the ability to successfully intervene in diet and physical activity behaviors in a primary care setting. However, with peak changes observed at 6 months, the findings also confirm that the major challenge is delivering extended care for supporting long-term maintenance of behavior changes in the months and years after weight loss.
Changes in most QOL and psychosocial measures were modest to minimal over time with no statistically significant differences across arms. However, substantial and clinically meaningful changes were observed in weight-related QOL (IWQOL-L), especially for physical function, self-esteem, and sexual function domains. IWQOL-L changes at 6 months were fully sustained through 24 months. These improvements are important given the severe impairment in mean IWQOL observed at baseline. A systematic review found that 4 out of 11 weight loss trials with >5% mean weight loss showed significant improvements in IWQOL-L score.114 Prior weight loss trials have likewise found greater improvements in weight-related QOL than in general QOL,115 and the small to minimal improvements we observed in general QOL (SF-12) are consistent with findings from prior systematic reviews and meta-analyses.114,116
The reach of the intervention, including eligibility rates and participation rates, was consistent across the 3 study arms, suggesting that the different delivery models were about equally “attractive” to participants. This further tempers presumed concerns about travel burden, as the in-clinic delivery did not appear to present a systematically greater enrollment barrier. Factors such as provider encouragement and patient readiness may be more critical than travel requirements for participants' decisions to engage in a health promotion program.117
As expected due to the pragmatic design, we observed a high eligibility rate (87%) in comparison with previous primary care–based weight loss trials, which noted eligibility rates ranging from 21% to 66%.86,103,104,111,118-121 Among those presumed eligible, we observed participation rates of 86% of those who were screened and 16% of those who received the mailing. Participation rates are difficult to compare with other studies due to sparse literature and inconsistency in how the rates are calculated. An internet-based weight loss trial that used targeted mailings to health plan members observed only a 5% participation rate.122 To enable comparison with a larger group of studies, we estimated participation rates from published participant flow diagrams where possible. These studies recruited through proactive phone calls, and using as the denominator the number of participants for whom contact was attempted multiplied by the eligibility rate, participation rates ranged from 26% to 61% of participants called.81,104,111,118,123 Although proactive phone calls may result in higher participation rates than do mailings, due to added cost and clinical burden, this strategy may be best reserved for subgroups that may be less likely to respond to mailings.81,120 Our findings support the combined strategy of targeted mailings and provider referrals as a cost-efficient means for engaging a sufficient volume of participants in weight loss services.124
Our analysis of representativeness is one of the first to empirically compare characteristics of participants with the broader primary care clinical population. Overall, we found that participants were more likely to be female and were slightly older. Within our subsample of Marshfield Clinic sites, participants were also less likely to have cardiovascular disease or to smoke, and they had fewer inpatient hospital days. These findings are consistent with those from a previous trial in 3 health maintenance organization (HMO) settings where enrollees were more likely to be female and nonsmokers, and had a lower disease risk score compared with other HMO patients with obesity.122 There is a long history of men being underrepresented in behavioral weight loss trials,125 and surface-level tailoring of recruitment materials (eg, including pictures of men as we did in this study) is likely insufficient. More proactive strategies, including explicit training for providers to raise awareness and address any communication concerns in referring men and younger patients, are needed.125,126
Practice personnel overall perceived that they had a clear plan for implementing the intervention and that it was easy to implement, was effective, and fit with what their patients wanted. However, among some, the perceived effectiveness and fit declined from mid- to postintervention; as attendance dropped, they perceived patient motivation to be more critical than the intervention per se. Personnel in the DM arm reported lower relative advantage at midintervention, and interviews revealed a sense of dissatisfaction from the lack of care integration, not being fully aware of what was going on, and hearing from patients that they wanted to connect with someone local. One of the surprising lessons learned during practice recruitment was that many PCPs preferred the FFS and PCMH arms because they wanted to offer weight loss services in house, even if ensuring staff resources was challenging.
The cost and financial viability of the intervention combined with counselor staffing were the major barriers to sustainability. Without uniform coverage across payers, billing was perceived to be complex and coverage insufficient. Medicare IBT covers 22 counseling visits per year (14 visits in the first 6 months and monthly visits thereafter), and local PCPs in the FFS arm overall felt it was feasible to provide this type of care with the right staff person in place to provide the counseling and champion the service. However, not all private payers offer the same benefit, and often, the number of covered visits falls below guideline-based care. Local clinicians in the trial found it difficult to determine coverage options across different payers, provider types (eg, RD vs nurse), and comorbid conditions (eg, diabetes) and to identify staff who could fill the counselor role with enough billable services. Anecdotally, PCPs and other clinicians wanted to provide weight loss services for their entire patient panel, not just patients with Medicare coverage. Thus, even with the Medicare IBT benefit in place, financial viability for the practice at large remained a major barrier to implementation.
Staffing and billing barriers were especially high among counselors in the PCMH arm, who were more likely to be surprised by the amount of extra time to prepare for group visits and provide individual patient feedback. They were also less likely to routinely communicate with the treating PCPs than were counselors in the FFS arm. We observed that PCMH counselors were more likely than were FFS counselors to be delivering the intervention in isolation from other care, because the group visits did not align as much with routine workflows and were typically offered during the lunch hour or before and after regular clinic hours.
Despite the barriers related to billing and staffing, more than half of personnel across all arms reported that their practice had discussed continuing weight loss services, and nearly 80% across arms agreed that their patients wanted them to offer these services. However, personnel in the DM arm reported that they were less likely to offer services, and at the time of this report, no practices in the DM arm had started offering weight loss services since participating in RE-POWER, compared with 6 practices (50%) in the FFS arm and 3 practices (25%) in the PCMH arm.
This trial has several major strengths, including its scope and pragmatic design. To our knowledge, this is the largest primary care behavioral obesity trial to date in the United States, and it is the largest to address the needs of rural residents. We eliminated a no-treatment control arm and designed the trial to be highly pragmatic, with comparator arms that represent viable chronic care management delivery models. It is also the first study to directly test the Medicare IBT reimbursement model in a primary care setting. Other pragmatic features of the trial enhance the generalizability to underresourced rural practices, including the diverse mix of practice types across a wide geographic region, the high patient eligibility rate, having practice-employed staff deliver the intervention, and designing the training to represent a real-world scenario of the amount of time a practitioner might take off from clinical care. The RE-AIM evaluation provides essential information related to patient uptake and factors affecting the ability of local primary care practices to adopt, implement, and maintain the intervention; it also includes extensive qualitative findings among both practice personnel and patients.
There are several limitations to the trial. First, the sample was predominantly female and White non-Hispanic; however, the race/ethnicity of the sample represents the population within the participating rural practices. Second, the rural cultural and geographical context of the trial may impact the generalizability of the findings, and further research is needed to extend the findings to other settings. Third, 2 practices in the third cohort dropped out and were replaced with the next recruited practice, which represents a break in randomization for 2 of 36 sites; however, sensitivity analysis excluding these 2 sites revealed similar findings. Fourth, the inclusion of patients across practices in the phone-based groups within the DM arm presents a potential group effect across clusters, and the effect of this on statistical inference is unclear. The DM arm was intentionally designed to have the intervention groups span multiple practices, as this is a practical benefit of scaling up centralized remote interventions. However, new methodological advances are needed to explore the ramifications of multiple levels of clustering. Fifth, by design, the FFS arm was modified from the Medicare IBT benefit in 2 ways: (1) it did not require >3 kg of weight loss for continued sessions after 6 months, and (2) it did not restart weekly sessions in year 2 as with an annual benefit. Thus, replication of the Medicare IBT benefit is needed. Requiring >3 kg of weight loss for continued sessions may result in less favorable 2-year outcomes; however, allowing a higher frequency of sessions at the beginning of the second year may help reengage some patients. Sixth, physical activity was measured by self-report and did not include accelerometry, and diet was assessed with brief measures that do not capture energy intake or overall diet quality. Seventh, numerous secondary outcomes were tested without lowering the P value that is considered significant; thus, there is risk for type I error (concluding a difference across arms when there is no difference in reality). Last, the study was designed to compare current care delivery models under pragmatic conditions and thus did not control for the potential impact of different professional backgrounds or training of the counselors.
To promote sustainable models for obesity treatment in rural and other primary care settings, future research is needed on implementation strategies to address gaps in billing, staffing, and internal care coordination. Implementation strategies across delivery models may include resources to facilitate shared staffing across clinics, guidance and consultation pertaining to billing options, and financial viability analysis before implementation. In addition, with the rapid expansion of computer- and phone-based medical visits during the COVID-19 pandemic, telehealth has ever-increasing potential to address the downsides to phone-based care. Prior trials in other preventive services such as smoking cessation127 and genetic testing128 have pointed to enhanced patient experience of care with telehealth visits conducted physically within primary care clinics vs phone visits conducted from home. In particular, patients have noted the advantage of telehealth for reading body language.128 Group-based behavioral weight loss interventions delivered via telehealth to patients in their homes represent a delivery strategy that is now widely accessible and in need of further investigation.
The findings from this study support adequate patient uptake in a primary care setting using multiple referral approaches, although future research is needed on strategies to recruit men. The findings also suggest that the travel burden among rural residents in this Midwestern region may not be as substantial as in other regions or as may be assumed. Thus, further research is warranted on the actual experienced travel burden, particularly in light of the potential benefits of attending visits in a health care setting.
Greater mean weight loss was observed through 24 months in the PCMH arm and through 18 months in the DM arm than in the FFS arm. Participants in all arms experienced meaningful improvements in triglyceride levels, physical activity, diet, and weight-related QOL, with triglycerides and physical activity improving significantly more in the PCMH arm than in the FFS arm.
As a whole, the combined participant and practice personnel implementation measures revealed favorable perceptions of the intervention but highlighted downsides to the DM delivery model. These downsides included patients' difficulty in developing accountability and cohesiveness, lack of support from local peers, and providers' dissatisfaction with the lack of care integration. Thus, the magnitude of the additional weight loss in the DM arm needs to be weighed against less-positive experiences of care. Satisfaction with the intervention overall was highest among participants in the PCMH arm, yet PCMH counselors anticipated challenges with sustainability due to unbillable time outside the visits.
Taken together, findings speak to the benefits of offering behavioral weight loss within a primary care integrated model. Although building strong linkages to community-based resources may be an effective approach,129 the reality is that evidence-based community programs are lacking in many rural areas.17 The potential positive impact of capitalizing on long-term patient-provider relationships, in addition to building partnerships with different commercial or remote programs, is particularly important in light of the need for ongoing extended care and the frequent comorbid medical conditions in this population.130 The Medicare IBT benefit could be readily adapted to group treatment. Moreover, findings support the assertion that intensive behavioral treatment can be effectively delivered by a range of health care providers with various professional backgrounds.131 However, the mismatch between the current reimbursement landscape across a mix of payers and the resources needed to implement IBT continues to limit its sustainability. The findings from this trial have critical implications for policies related to the treatment of obesity in primary care, particularly in rural and other medically underserved areas where the burden and needs of obesity continue to increase.
Research reported in this report was funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (OB-1402-09413). Further information available at: https://www.pcori.org/research-results/2014/comparing-three-ways-offer-weight-management-program-patients-living-rural
Details of Patient and Stakeholder Engagement (PDF, 134K)
Supplementary Tables (PDF, 254K)
Befort CA, VanWormer JJ, DeSouza C, et al. (2021). Comparing Three Ways to Offer a Weight Management Program to Patients Living in Rural Areas—The RE-POWER Study. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/07/2021.OB.140209413
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
