It is unclear whether training physicians to counsel obese patients leads to weight loss. This study assessed whether a 5-h multimodal longitudinal obesity curriculum for residents on the basis of the 5As (assess, advise, agree, assist, and arrange) was associated with weight loss in their obese patients.
Design and Methods:
Twenty-three primary care internal medicine residents were assigned by rotation schedule to intervention (curriculum) or control groups. We then conducted follow-up chart reviews to determine weight change at up to 12 months following the index visit. 158 obese patients (76 in the intervention group and 82 in the control group) completed exit interviews; 22 patients who presented for acute care at the index visit were excluded. Chart reviews were conducted on the 46 patients in the intervention group and 41 patients in the control group who were seen again within 12 months of the index visit and had follow-up weight measurements.
The main outcome of interest was mean change in weight at 12 months compared between the intervention and control groups. Patients of residents in the intervention group had a mean weight loss of −1.53 kg (s.d. = 3.72) although the patients of those in the control group had a mean weight gain of 0.30 kg (s.d. = 3.60), P = 0.03. Six (15.8%) patients in the intervention group and 2 (5.4%) patients in the control group lost >5% body weight (P = 0.14).
Although the magnitude of weight loss was small, this study shows that training physicians to counsel patients can produce measurable patient outcomes.
The majority of Americans are diagnosed as obese or overweight (1), with obesity being associated with higher mortality (2) and increased risk of several co-morbidities (2,3,4,5,6,7). Intensive healthcare provider counseling has been proven to help clients reduce weight (8), but evidence for brief, focused, physician-delivered counseling has been inconclusive (9). Some studies show that counseling by physicians and other providers is associated with positive behavioral and weight-loss outcomes (10,11,12,13,14,15) and that patients want physicians to deliver lifestyle and weight-loss counseling (16) and may be more likely to change their lifestyle behavior when counseled by a provider (15). Given that there are 902 million annual ambulatory visits in the United States alone (17), most of which are made to primary care providers, weight management counseling by these providers could have a significant public health impact. Therefore, the United States Preventive Services Task Force recommends that primary care providers screen patients for obesity and counsel patients about their weight (18,19).
Despite this, providers frequently fail to effectively counsel obese patients to lose weight (14,15,20,21-,22,23). Lack of reimbursement for weight management counseling has traditionally been a barrier to providing weight management counseling (24). Recently, the Center for Medicare and Medicaid services (CMS) announced that Medicare will now cover intensive behavioral counseling for obese patients (25); this should increase physician motivation for providing weight management counseling. However, the CMS decision will not overcome other barriers to promoting high-quality physician counseling including perceived lack of effectiveness (26), competing demands on time during the medical visit (27), lack of training (15,24,28), and lack of skills (29). To address the last two issues, we developed a 5-h, multimodal curriculum for resident physicians designed to improve their obesity counseling skills (30). In 2010, we published the results of a non-randomized, wait-list-control design study to explore the impact of this obesity counseling curriculum for residents on the quality of obesity counseling they delivered to patients (30). We found that training was associated with higher quality of counseling as perceived by 158 obese patients when patient, provider, and visit characteristics were taken into account (standardized beta = 0.18; R2 change = −2.9%, P < 0.05) (30). Encouraged by this, we sought to explore whether the curriculum was associated with greater weight loss in patients who visited the clinic within 12 months after the index visit. To do this, we conducted a retrospective chart review to determine weight change differences in patients of residents who received the curriculum (intervention) and those who did not (control).
Methods and Procedures
Information about participants, participant recruitment, and study design has been published with the results of our 2010 study (30).
All 23 residents in the New York University School of Medicine (NYUSOM) primary care internal medicine residency program (academic year 2008-2009) were included in the study. Residents are considered to be a vulnerable population as they are employees of NYUSOM. Therefore, we could not collect identifying variables. However, between 2006 and 2011, 56 residents were part of the 3-year residency program. The average age of the residents was 28 upon entering residency, with 66% being female and 57% identifying themselves as non-Hispanic White.
The residents in the study were stratified by training level and assigned, on the basis of their clinical rotation schedule to either an intervention (5-hour, 5As obesity counseling curriculum, n = 12) or a wait-list control group (standard residency training, n = 11). Clinical rotation schedule was determined independently and a priori by an outside administrator blinded to the study. Residents in the control group received the curriculum 6-8 months later. All residents had consented to be part of a medical education research registry approved by the IRB at New York University.
The obesity curriculum intervention consisted of a 5-hour training, using multiple interactive teaching modalities (didactics, role-playing, standardized patients, and videotape review of a patient encounter) given over three weekly sessions. The main objective of the curriculum was to train physicians to counsel obese patients on the basis of the 5As model (assess, advise, agree, assist, and arrange) (29,31).
Patients were recruited from Gouverneur Healthcare Services, a publicly-funded outpatient clinic located in the Lower East Side of Manhattan. Gouverneur has a small, clinic-based weight management program that offers individual assessments by a nutritionist and/or physician with expertise in weight management. Patients are usually referred by their primary care provider.
Over a 7-month post-intervention period, we recruited 158 obese patients (body mass index ≥ 30 kg/m2) immediately after their medical visit; this was considered the index visit.
We were interested in seeing the trajectory of patients' weights by viewing multiple visits in real clinical time, and therefore chose to collect data via chart abstraction. One to two years post-index visit, two trained research staff conducted chart reviews of the electronic medical record for those patients seen again in the clinic within the 12-month period after the index visit, whether it was with the same provider or not. The reviewers were blinded to which residents (and therefore patients) were in the intervention or control groups. They abstracted weight and determined the number of return visits to the same resident and other providers, chief complaint, number of medications, and number of co-morbidities. Visits were classified by a research clinician blinded to intervention group as “acute” if the chief complaint was centered around an acute pain syndrome (new headache, back injury, limb injury, etc.), a new/acute infectious disease (upper respiratory infection), or a new psychiatric complaint (e.g., depression or anxiety). We decided a priori to exclude patients with an acute care visit at the index visit because we felt that these patients were less likely to both receive and respond to obesity-related counseling. We only included patients who had a follow-up visit within 12 months of the index visit in our analysis, though not necessarily with same provider. We also excluded patients who did not have their weight measured at any follow-up visit. Finally, research assistants extracted the number of referrals to the weight management program.
Patient exit interview questionnaire. This bilingual (English and Spanish) questionnaire was developed to assess patient characteristics, including sociodemographic data and “Stage of Change” for weight management using questions adapted from Prochaska and DiClemente's transtheoretical stages of change model of behavior change (32). In addition, the questionnaire asked patients whether they were referred to a dietician or weight management program.
Initial weight and height measurements. At the start of the study, clinical nursing staff members were trained to weigh patients using one of four calibrated, analog scales, each equipped with a stadiometer to take height measurements with the patients fully clothed except for shoes and jackets.
Statistical analysis. Our main outcome was the mean patient weight change at 12 months compared between the intervention and control groups. Secondary outcomes included mean weight change at 1, 3, and 6 months, number of patients maintaining, losing, and gaining weight in each intervention group, number of patients with clinically significant weight loss (>5%), and number of referrals to a weight management program (patient-reported at exit interview and/or extracted from the medical record). Weight change was calculated by subtracting the index visit weight from the last weight recorded in the electronic medical record up to the 12-month time period. Patients were characterized as having: (i) lost weight if the last weight minus the index visit weight was < −1 kg, (ii) maintained weight if the weight change was between −1 and +1 kg, and (iii) gained weight if the weight change was >+1 kg. We compared mean weight change between patients of intervention and control group residents using independent t-tests and used Fisher's exact χ2 tests to assess the significance of differences between the curriculum groups in the proportion of patients with weight change. Given that a substantial number of patients did not return for a follow-up visit during the 12-month study period, we also conducted a sensitivity analysis that included patients who did not have any follow-up using independent sample t-tests to compare weight loss in the intervention and control groups. For this analysis, patients without follow-up weight data were presumed to have no weight change because more precise imputation methods would not be feasible with a small sample size and large loss of follow-up.
Figure 1 shows a flow chart of the residents and patients included in our analysis. Twenty-two patients (13.9%) with an acute visit at the index visit were excluded from analysis. Thirty-nine patients who did not have a visit within the 12-month follow-up and an additional 10 patients for whom weight data was unavailable at the follow-up visits were excluded as well. There were no differences in gender distribution between the residents in the intervention (58% female) and control (54% female) groups. As compared with those who did not, patients who followed up within 12 months were more likely to have a body mass index ≥35 (23 vs. 30%, P = 0.02) and had a higher mean number of medications (0.84 (1.05) vs. 0.44 (0.84), P = 0.02). Further, patients who followed up within 12 months were less likely to have the index visit classified as “acute” (0 vs. 23% of index visits were acute, P < 0.001). However, they did not differ significantly with respect to mean body mass index, gender, race/ethnicity, number of co-morbidities, language concordance between patient and provider, or stage of change. In the patients with follow-up within 12 months, the mean number of visits was 3.06 (s.d. = 3.08, range 1-9). Fifty-three percent of follow-up visits were with the same resident.
Table 1 depicts demographic data of patients included in the main analyses. The demographics of the full sample of 158 patients have been described elsewhere (30). There were no statistically significant differences between patient and visit characteristics in the control and intervention groups, but the difference in mean number of visits appears to be clinically significant. However, post-hoc analyses demonstrated that the difference seen in the number of visits between the intervention and control groups (3.47 vs. 2.56 visits) was because of an outlier in the intervention group who had 17 follow-up visits. When this outlier was removed from analysis, the differences between the two groups became smaller (3.00 (s.d. = 2.31) vs. 2.75(s.d. = 2.05) visits, P = 0.58).
Table 1. Characteristics of sample at initial (index) visit and nature of subsequent visits
Table 2 shows 12-month outcomes. Over a 12-month period following the index visit, the mean weight change in all patients was −0.38 kg (s.d. = 3.6). Weight change at up to 12 months was not correlated substantially nor significantly with either number of visits (Pearson's r = −0.17, P = 0.10) or number of visits with the same resident (Pearson's r = −0.10, P = 0.36). Figure 2 shows mean weight change in the intervention and control groups in patients seen within 1, 3, 6, and 12 months after their index visit. Before 12 months, there were no statistically significant weight changes in the intervention group when compared with the control group, although by 6 months we found a trend for greater weight loss for the intervention group (P = 0.06). At 12 months, on average, patients initially seen by a resident who completed the obesity counseling curriculum had lost weight (mean weight change = −1.53 kg) although, on average, patients initially seen by control group residents had gained weight (mean weight change = +0.30 kg, P = 0.03). Figure 3 shows the weight trajectories of the intervention and control groups for patients seen each month. With the exception of months 2 and 10, the weight loss trajectories for the intervention and control groups are similar to the overall results. In addition, a greater proportion of patients lost weight or stayed the same in the intervention group (76.3 vs. 48.6%, P = 0.01). Few patients in either group lost more than 5% of their body weight (15.8 vs. 5.4%, P = 0.14).
Table 2. Weight status for patients of intervention vs. control residents at 12 months
An analysis was also conducted comparing the weight change of patients in the intervention and control patients that included 39 patients with no follow-up visit and assumed no weight change in these patients (n = 126). Mean weight change up to 12 months was −0.76 kg (s.d. = 3.06) for the intervention group and +0.20 (s.d. = 2.89) for the control group, P = 0.11).
Impact of referrals
Table 2 shows that the referral rates to weight management clinic were higher in the intervention group. Post-hoc analyses explored whether we would still see greater weight loss in the intervention group if we controlled for receiving a weight management referral. Focusing only on those patients who did receive a referral, we found a similar but not significant pattern of weight loss: control group patients who received at least one referral, n = 11, had an average weight change of 0.03 kg (s.d. = 3.03) and intervention group patients who received at least one referral, n = 21, had an average weight change of −1.24 kg (s.d. = 4.29), P = 0.39.
Patients of resident physicians who completed a 5As-based obesity counseling curriculum lost more weight than patients of residents who had not completed the multimodal curriculum. Although the magnitude of weight loss was small, this study shows that training physicians to counsel patients can produce measurable patient outcomes. Further, larger numbers of patients in the intervention group either lost or maintained weight. This is encouraging because obese patients often progressively gain weight over time and training physicians to counsel patients might dampen this trend. It also supports the CMS decision to pay physicians for weight management counseling services, provided that they receive adequate training. It has been controversial whether or not physician counseling promotes appreciable weight loss in patients. In observational studies, patient report of physicians' weight loss counseling has been associated with weight loss (15,30,33-,34,35). But a systematic review conducted by Tsai and Wadden (9) did not show an impact with brief provider counseling. We believe the lack of impact may be because of physicians' lack of obesity counseling skills (29) and negative attitudes toward obesity (30) leading to delivery of poor quality counseling. In a study of over 300 providers from three different specialties at New York University School of Medicine, 65% did not feel competent obtaining a diet history, 40% did not feel competent to help patients set realistic weight loss goals, and 39% did not feel competent providing brief counseling to help patients lose weight (36). The present study suggests that when physicians are properly trained, they can promote weight loss in their obese patients.
Admittedly, the magnitude of weight loss was small and may not be clinically significant (37-,38,39). Clearly, physician counseling alone is not adequate to treat obesity in primary care. The literature suggests that system changes and office support tools (i.e., clinical reminders, health coaches) may increase the impact of physician counseling. In a study of 45 physicians and 1,162 patients randomized to physician counseling only, physician counseling plus office support tools, and a control condition, patients in the second group lost a mean of 2.3 kg over 1 year as compared with 0 kg in the control group (P < 0.001), but a statistically significant change was not seen in the group who received solely physician counseling (−1.0kg, P value not reported) (40). In contrast, our study showed a slightly higher magnitude of weight loss that was statistically significant in the intervention group who received physician counseling only. The reasons for this difference need to be explored further but may be because of referrals to a weight management program at our institution. Future intervention studies in primary care should explore the combination of provider training, office support, and other systems-level and community-based interventions.
The mechanism of weight loss is unclear. The curriculum for the residents was based on the 5As and focused on teaching and practicing skills needed to promote long-term, healthy lifestyle changes and weight loss in patients (29). Residents in the intervention group were encouraged to properly assess the patients' medical history, stage of change, and current behaviors (assess), provide patient-centered advice (advise), engage in collaborative goal setting to promote behavior change (agree), address barriers to change using brief motivational interviewing techniques (assist), and help patients obtain adequate support through frequent follow-up and/or ancillary and community services such as hospital-based weight management programs or weight watchers (arrange).
Our data suggests that the curriculum may have increased referral rates to a weight management clinic, either through increased awareness about the program by the physicians or increased patient willingness to attend more intensive programs. Regardless of reason, this increased referral rate may be an important reason why patients lost weight. However, referrals alone did not fully account for study findings. Presumably, the interaction with the providers also encouraged patients to make lasting changes in their dietary and physical activity behaviors leading to the increased weight loss, but our study was not designed to assess these intermediate outcomes. Future studies should further elucidate the mechanisms by which physician counseling may promote weight loss.
Our study has several limitations. Almost half of the patients did not follow-up during the 12-month period, and we did not find significant changes in weight when we included them in our analysis. It might be argued that highly activated patients are likely to both follow-up with medical visits and lose weight, entirely explaining our findings. However we found no significant differences in demographics, mean body mass index, comorbidities, and stage of change between those who followed up and those who did not.
There are also limitations in our study design. Residents were not randomized to each intervention group because of having fixed rotation schedules and they could not be blinded to intervention or lack thereof. Although we believe assignment by rotation schedule was likely to lead to statistical equivalence in age and race/ethnicity between the two groups, we did not have access to data that could confirm this. The residents and patients come from a small primary-care oriented program and this limits the generalizability of our findings and also increases the likelihood of contamination as residents might even teach techniques learned to one another. However, if such contamination occurred, it would likely have minimized inter-group differences.
Despite these limitations, our study shows that there is value in evaluating the impact of provider training interventions on physician behavior and patient outcomes. To date, few studies have sought to determine the impact of training providers in caring for obese patients on actual patient level outcomes including the gold-standard, weight loss (40). Given the high prevalence of obesity, the negative impact on morbidity and mortality, and the increasing public investment toward physician-delivered obesity treatment by CMS, determining the relative contribution of provider training to achieving positive patient outcomes is valuable to medical educators, clinical leaders, and policy makers. Larger scale, multi-center studies are needed to explore the incremental impact of training physicians as part of a system-based intervention using a multidisciplinary team and office-based tools that facilitate obesity management.
We thank Dan Erck and Alfredo Axtmayer for their diligence and help with the study. We also thank David Stevens for facilitating patient recruitment. This research was supported by HRSA grant # 12-191-1077, Academic Administrative Units in Primary Care and by a New York State Empire Clinical Research Investigator Program Award (ECRIP).