Ockene IS, Hebert JR, Ockene JK, Saperia GM, Stanek E, Nicolosi R, Merriam PA, Hurley TG. Effect of Physician-Delivered Nutrition Counseling Training and an Office-Support Program on Saturated Fat Intake, Weight, and Serum Lipid Measurements in a Hyperlipidemic PopulationWorcester Area Trial for Counseling in Hyperlipidemia (WATCH). Arch Intern Med. 1999;159(7):725-731. doi:10.1001/archinte.159.7.725
To evaluate the effectiveness of a training program for physician-delivered nutrition counseling, alone and in combination with an office-support program, on dietary fat intake, weight, and blood low-density lipoprotein cholesterol levels in patients with hyperlipidemia.
Participants and Methods
Forty-five primary care internists at the Fallon Community Health Plan, a central Massachusetts health maintenance organization, were randomized by site into 3 groups: (1) usual care; (2) physician nutrition counseling training; and (3) physician nutrition counseling training plus an office-support program. Eleven hundred sixty-two of their patients with blood total cholesterol levels in the highest 25th percentile, having previously scheduled physician visits, were recruited. Physicians in groups 2 and 3 attended a 3-hour training program on the use of brief patient-centered interactive counseling and the use of an office-support program that included in-office prompts, algorithms, and simple dietary assessment tools. Primary outcome measures included change at 1-year of follow-up in percentage of energy intake from saturated fat; weight; and blood low-density lipoprotein cholesterol levels.
Improvement was seen in all 3 primary outcome measures, but was limited to patients in group 3. Compared with group 1, patients in group 3 had average reductions of 1.1 percentage points in percent of energy from saturated fat (a 10.3% decrease) (P = .01); a reduction in weight of 2.3 kg (P<.001); and a decrease of 0.10 mmol/L (3.8 mg/dL) in low-density lipoprotein cholesterol level (P = .10). Average time for the initial counseling intervention in group 3 was 8.2 minutes, 5.5 minutes more than in the control group.
Brief supported physician nutrition counseling can produce beneficial changes in diet, weight, and blood lipids.
AN ELEVATED blood cholesterol level is now clearly established as a major risk factor for coronary heart disease, and lowering such levels reduces the subsequent risk of clinical coronary events. Guidelines published by the National Cholesterol Education Program suggest that dietary intervention should be the first step in dealing with this problem.1,2
Physicians who regard themselves as competent to offer dietary counseling are more likely to provide it to their patients with hypercholesterolemia.3 We have developed and tested a training program for teaching nutrition counseling skills to physicians that was adapted from our previously described patient-centered smoking intervention training program,4 for which we have demonstrated its efficacy on physicians' counseling skills5 and patients' smoking behavior.6 We also have developed an easily used office-support program that includes in-office prompts and algorithms to remind the physician to counsel for dietary change, as well as a simple dietary assessment form that is filled out by the patient while in the waiting room. The training and office-support system are part of a randomized clinical trial, the Worcester Area Trial for Counseling in Hyperlipidemia (WATCH). We previously reported that only physicians receiving both counseling training and the office-support system carried out the counseling sequence appropriately.7 These physicians, as assessed by patient exit inter views, carried out a significantly greater number of steps in the counseling sequence than did either the counseling training-only group or a control group. The latter 2 groups did not differ significantly from each other in the number of counseling steps used. Herein we report on the effect of this intervention on the clinical end points of the study: changes in patients' dietary saturated fat intake, weight and body mass index (BMI, a measure of weight in kilograms divided by the square of the height in meters), and blood lipid levels.
Forty-five primary care internists at a central Massachusetts health maintenance organization, the Fallon Community Health Plan (FCHP), took part in the program. Forty-six physicians were available; 1 declined to participate. The FCHP has some 180,000 enrollees, and has been repetitively rated as one of the best health maintenance organizations in the country.8 A recent review (July 1997) by the National Committee for Quality Assurance gave FCHP its highest rating for preventive health services. Physician characteristics were previously reported: 64% were men, their average age was 38.1 years, and they saw an average of 76 patients per week.7 All but 1 were board certified in internal medicine. Physicians were randomized by site (12 separate patient care units) to 1 of 3 groups (4 sites in each group): 1, usual care; 2, physician nutrition counseling training; and 3, physician nutrition counseling training plus office support. Dietary counseling is universally available at FCHP and physicians are encouraged to refer patients when indicated.
The physician nutrition counseling training program was developed to teach physicians the counseling intervention methodology and the use of the dietary assessment tools. A study-specific algorithm recommended that patients in the highest 10th percentile for low-density lipoprotein (LDL) receive both physician counseling and immediate referral to a dietitian, whereas the remaining patients in the upper quartile first should be given a trial of physician-based nutrition counseling alone. The algorithm suggested reevaluation at 4 to 6 weeks and at 3 months, with referral to nutrition services if physician counseling alone did not meet the goal as defined by the National Cholesterol Education Program criteria.1 All physicians (n = 31) in groups 2 and 3 were trained to deliver nutrition counseling. The 14 internists in group 1 were not trained. The 12 clinical sites are each self-contained, and physicians did not transfer between sites.
The training consisted of 2 sessions: a 2.5-hour small group session (3-10 physicians) and a 30-minute individualized tutorial. The content of these sessions has been previously described.4 The initial session included didactic instruction, videotape observation, and role-playing. The physicians were instructed in the use of the slightly modified Dietary Risk Assessment (DRA) form originally developed by Ammerman and colleagues9 (an abbreviated food frequency questionnaire to be filled out by the patient while in the waiting room). Within 2 weeks following the group training, physicians participate in a 30-minute individual tutorial session in their offices, which involves role-playing the counseling-intervention approach with a patient simulator, who provides immediate feedback.
The patient-centered counseling model includes the following elements: (1) advise nutrition change and use personalized information to reinforce the need for such change; (2) assess experience with dietary change to determine the patient's resources for change (strengths) and factors that inhibit it (barriers); (3) review current diet using the DRA; (4) prioritize areas of high-fat intake; (5) develop a plan for change; and (6) arrange for follow-up. The entire dietary intervention is designed to take at most 8 to 10 minutes, and can be usefully carried out in less time.
The office-support program used in group 3 was designed to assist the physician in carrying out the counseling sequence and to be incorporated easily into usual clinic procedures. For each study-patient encounter:
The patient was given the DRA while in the waiting room, and brought it in to the physician filled out.
The physician was provided with a copy of the patient's lipid profile results, flagged for his/her attention.
The physician was provided with a copy of the counseling algorithm.
The physician was provided with a series of handouts to be given to the patient, including the DRA goal sheets containing dietary change recommendations, a brief cooking and recipe guide, tips for eating out, motivational material, and suggestions for further reading.
All these supportive elements also were available at the group 2 physician sites, but were not provided in any systematic fashion. Thus, the group 2 physician would need to find the lipid profile results among the usual laboratory data, obtain and administer the DRA forms, and provide himself/herself with, and use without outside prompting, a copy of the WATCH algorithm and handouts.
Patients who had appointments to be seen by the study physicians during the 24-month recruitment period (January 2, 1992-December 30, 1993) had not been screened previously, and were between the ages of 20 and 65 years, were contacted by telephone by the FCHP telemarketing services and invited to participate in the trial. Initial eligibility was assessed (no prior lipid-lowering drug treatment, no dietitian referral within 1 year), and the patient was invited for an initial screening fingerstick cholesterol determination performed at the clinic sites (L.D.X. analyzer, Cholestech Corporation, Hayward, Calif).10 Because cholesterol level varies significantly by age and sex, and because we had previously noted and described seasonal variation of cholesterol levels in this population,11 we used percentile cutoff points that were adjusted for age, sex, and season. If the initial measurement was in the upper 25th percentile, the patient was asked to return 1 week later in the fasting state, and a second fingerstick cholesterol determination was obtained. If the mean of the 2 successive cholesterol determinations continued to place the patient in the upper 25th percentile, full-informed consent was obtained, and 2 successive fasting blood samples for cholesterol, triglycerides, high-density lipoprotein (HDL), and LDL by calculation were drawn 1 week apart and analyzed in a Centers for Disease Control and Prevention standardized reference laboratory to establish baseline measurements (University of Massachusetts at Lowell). Patients were excluded if total cholesterol level was more than 7.76 mmol/L (300 mg/dL), LDL level more than 5.17 mmol/L (200 mg/dL), or triglyceride levels more than 4.52 mmol/L (400 mg/dL). The first 2 exclusions were used because it was thought that such individuals were likely to require pharmacological therapy, and the last because the primary lipid end point was LDL cholesterol level, and calculation of LDL was precluded by triglyceride levels of more than 4.52 mmol/L (400 mg/dL). Patients also were excluded if they had a history of coronary heart disease, were thought to have a secondary form of hyperlipidemia (eg, hypothyroidism), or if they had an illness thought to limit life expectancy. Eleven hundred sixty-two patients were recruited.
Study participants were followed up for a period of 1 year. In addition to fasting lipid profiles (2 at baseline, 1 at 1 year, all analyzed in the study laboratory), baseline and 1-year measures included the collection of demographic information, height, weight, blood pressure, and dietary intake data. Psychosocial factors relating to self-efficacy, social support, attitudes toward nutrition, and experience with nutrition change also were assessed and will be reported separately. The primary end point measurement of dietary intake was by 24-hour recall using version 2.3 of the Nutrition Data System of the Nutrition Coordinating Center at the University of Minnesota12 carried out once at baseline and repeated at the 1-year point. Given the considerable variability in day-to-day dietary intake, 24-hour recall data in this study were not used for analyses at the level of the individual, but rather as grouped data.
The number of physician visits and physician-ordered lipid profile measurements also were tracked for each participant, as was the usage of lipid-lowering medication and referrals to dietitians.
As we thought it inappropriate to withhold such data, the results of the paired study baseline lipid profiles, measured outside of the FCHP system, were provided to the patients' physicians in all 3 study groups as standard laboratory reports. Only a line identifying the report as having originated at the University of Massachusetts at Lowell marked this as a research study result. There was some concern that such identification of patients in the control group as having elevated lipid levels would in itself constitute an intervention and thereby obscure a possibly modest counseling intervention effect. However, given the busy clinical practice of these physicians and the minimal identification of these results as study blood tests, we concluded that it was unlikely that providing this information would produce a materially relevant effect in the control group.
This study was reviewed and approved by the institutional review boards of the University of Massachusetts Medical Center, Worcester, and the FCHP.
The WATCH study recruited a total of 1162 subjects with total cholesterol measurements in the highest 25th percentile. The data set for this report includes only individuals with paired baseline and 12-month LDL cholesterol data, the primary study end point. Included patients comprise 78.8%, 82.6%, and 82.7%, respectively, of groups 1 to 3. Exclusions included 7 dead individuals, 15 individuals with triglyceride levels of more than 4.52 mmol/L (400 mg/dL) at follow-up (that prevented the calculation of LDL), 211 individuals without follow-up LDL measures, and the 2 patients of the 2 physicians who each had only 1 patient in the study (dropped because of the heavy weighting they would receive in the analysis), leaving a total sample of 927 (80% follow-up). Comparison of included and excluded patients across all of the study variables revealed significant differences only for cigarette smoking (17.0% vs 29.4%), age (49.4 vs 47.9 years), and BMI (28.9 vs 29.8), respectively. There were no significant differences by study group.
Descriptive statistics were used to characterize the overall study population. Intervention-related measures including cholesterol medication use, nutrition class participation, the number of physician visits, and the number of extra lipid measurements were compared across treatment groups using the χ2 test for homogeneity. A 1-way analysis of variance was used to test differences among the 3 groups for baseline measures of lipoproteins, weight and dietary intake, and the intervention-related measures of perceived length of the clinic visit and perceived length of the dietary advice discussion (from the patient exit interviews).7
Multivariable analyses for the test of the treatment effect were conducted using mixed-model analyses of covariance that accounted for the nested design of the study, ie, physician was nested within site (the unit of randomization) that was nested within treatment group. For the main outcome variables—LDL cholesterol, weight, and dietary intake of saturated fatty acids—we calculated a change score by subtracting the baseline value from the 1-year value. To select control variables for the final models, associations were examined between outcome change and factors that were potential predictors of change, including patient characteristics, intervention-related factors, and the interaction between these variables and treatment group. For all multivariable models we fit age and sex as covariates. In the lipid models, an interaction term for treatment group was included. Although 2 intervention-related variables, cholesterol-lowering medication use and nutrition counseling participation were associated with change in several of the outcome measures and varied between treatment group in the degree of implementation, we opted not to include these variables in the final models, to maintain an intention-to-treat analytic approach. Separate analyses that included these 2 variables in the model produced results that were not materially different from the intention-to-treat model.
Results from the multivariable models are presented as the adjusted least squares means for each treatment group. The statistical test is a t test for the difference between the least squares means for group 1 (the control group) and group 3 (the full-intervention group). All analyses were conducted using SAS statistical software.13
The mean age of the participants was 49.3 years, 44% were men, and the group as a whole was overweight, with a mean BMI of 28.8 kg/m2. Table 1 and Table 2 provide other baseline data for the study population. There were significantly more women in group 2 (65.5%). The mean baseline blood cholesterol level was 6.42 mmol/L (248 mg/dL); mean LDL level, 4.34 mmol/L (168 mg/dL).
Table 3 presents intervention-related measures. The participants receiving cholesterol-lowering medication during the study was highest in the control group (12.2%). The number of physician-ordered lipid profiles was significantly different in the 3 groups (means, 1.5, 1.6, and 2.1 for groups 1, 2, and 3, respectively), as was the number of physician visits (visits for any purpose: means, 3.4, 3.1, and 3.6 for groups 1, 2, and 3, respectively).
The WATCH algorithm encouraged physicians to defer referral to dietitians for patients in the 75th to 90th LDL percentile range until their own counseling had been demonstrated to be ineffective. Thus, we anticipated a lower level of referrals to dietitians in the 2 counseling groups (2 and 3). However, a remarkably low level of referral was seen in group 2 (mean percentage of participants referred, 20.8, 4.8, and 11.8 for groups 1, 2, and 3, respectively; P<.001).
In an unadjusted analysis of variance, greater change was seen at the end of 1-year follow-up in patients in group 3 compared with those in group 1: total cholesterol level, −0.065 mmol/L (−2.51 mg/dL); LDL cholesterol level, −0.039 mmol/L (1.5 mg/dL); percentage of energy from saturated fat, −1.6%; and weight, −2.1 kg. The definitive analyses, as follow, adjust for study design and covariates.
In a model controlling for age, sex, and education, there was a decrease of 0.10 mmol/L (3.8 mg/dL) in LDL cholesterol level in group 3 compared with group 1 (P = .10) (Table 4). Education (high school graduate or less compared with some college or more), medication use, and nutrition class participation were independently associated with a reduction in LDL levels.
A similar effect was seen for total cholesterol level, with an average difference of −0.12 mmol/L (4.8 mg/dL) comparing groups 3 and 1 (P = .07). The HDL increased somewhat, resulting in a favorable decrease in the total cholesterol-HDL ratio (P = .004).
Of 927 participants with complete lipid data, 43 were missing dietary data at both times; 59, baseline data; and 183, 12-month data. One other 24-hour recall was judged unreliable. Thus, data from 700 participants were used in this analysis.
For saturated fat intake, group 3 was significantly lower than the other 2 groups, which were not significantly different from one another (Table 4). Compared with the controls (group 1), group 3 reported a mean percentage of energy from saturated fat that was 1.1 percentage points lower (P = .01), a relative difference of −10.3% in percent of energy from saturated fat.
Paired weight and height data were available for 550 members of this data set (59%), distributed as 161, 192, and 197 (57.7%, 57.7%, and 62.5%) members of groups 1, 2, and 3, respectively. Follow-up weights were obtained at the time of the 1-year blood studies by laboratory technicians blinded to participant group. Early in the study the technicians failed to obtain weights at times when they were busy; this problem was later corrected. This was a random error, and there were no significant differences in any of the measured baseline factors between those individuals weighed and not weighed. A markedly favorable effect was seen in group 3, with a mean reduction in weight of 2.3 kg and in BMI of 0.81 kg/m2 compared with group 1 (P<.001 for weight and P<.001 for BMI) (Table 4).
The time spent by physicians in delivering the intervention was estimated using patient exit interviews.7 Physicians in group 3 were reported to have spent an average of 8.2 minutes discussing diet, some 5.5 minutes more than did physicians in group 1 (average, 2.7 minutes). This corresponds closely to the 8- to 10-minute period described in the training sessions. Since these data are not derived from direct observation, but rather reflect patient perception, it is possible that patients perceived the time spent as longer than it was. This is suggested by the perception of patients in group 3 that the entire visit was some 40 minutes in length, compared with an average of 28.1 minutes for patients in group 1. If true, this time would have had to have been made up elsewhere in the physician's schedule, as these were 30-minute visits.
Patients who attended nutrition classes had greater decreases in LDL levels than those who did not, and there appeared to be an interaction with group (LDL decreases of 0.22, 0.35, and 0.53 mmol/L [8.5, 13.6, and 20.5 mg/dL] for groups 1-3, respectively) (P = .001 for overall effect) but the numbers attending the classes were small (Table 3).
Recurring costs for the program are estimated in Table 5. Physician training (3 hours) took place only once. Training in patient-centered counseling is not specific to nutrition, and is useful in many areas of patient counseling.6,14 The overall cost of the intervention was $1.86 per patient per year, with much of the cost of the intervention offset by the lower use of lipid-lowering medication in group 3 compared with the control condition.
We previously reported that only physicians in group 3, receiving both training in counseling for nutrition change and a supportive office environment, effectively delivered the counseling intervention.7 In this report we extend this finding to the diet, weight, and lipid outcomes of the WATCH study. The strength of our findings lie in their consistency: it is only patients in group 3 who demonstrate benefit, and such benefit is seen in all the expected areas, with reduced dietary intake of saturated fat, significant weight loss, and lower LDL levels. Furthermore, these effects were seen with a modest intervention effort. The initial intervention was fit into an already scheduled visit in a busy health maintenance organization practice, and compared with the control group, physicians in group 3 spent an average of only 5.5 extra minutes discussing diet. Physicians are still perceived by the public as the most reliable and credible sources of health information and advice, and often underestimate the potency and value of even brief counseling efforts.15,16
The borderline significant lipid outcomes are of modest degree, especially when compared with the results of recent studies involving pharmacological intervention.17,18 The WATCH, however, was a primary care study involving a low-cost, minimal effort counseling intervention. A 10% reduction in energy from saturated fat, associated with a weight loss of 2.3 kg, a modest reduction in LDL cholesterol levels, and a favorable shift in the cholesterol-HDL ratio in patients not receiving pharmacological therapy, maintained 1 year out from the intervention, has important public health implications.19
The magnitude of the weight loss seen in patients in group 3 was unexpected. Obesity is a causal factor in many chronic diseases, and the accelerating rate of increase in the prevalence of obesity seen in the United States over the last decade presents a formidable challenge.20 Even modest weight loss has a beneficial effect on cardiovascular risk factors,21 and can be brought about by small sustained changes in energy intake, such as those generally associated with reduction in dietary fat and saturated fat intake in free-living individuals.22 Although we do not have data on physician counseling for physical activity, it is also possible that physicians did recommend increased physical activity in conjunction with their dietary counseling efforts. Any such increase in activity would reduce the concomitant dietary change needed to effect weight loss.
Physician participation in the training program was close to 100%, and we have shown that the counseling behavior seen in group 3 physicians persisted for as long as 3 years after training.7 It is clear that when dealing with a low-acuity preventive intervention in a busy practice setting, appropriate office support is necessary to cue the physician. Training without support may be counterproductive, as physicians in group 2 referred remarkably few patients to nutrition services, yet failed to carry out the intervention themselves. This has implications for continuing medical education programs and other forms of physician training, which rarely consider the entire intervention delivery system.
The American College of Physicians has recommended cholesterol screening guidelines that limit screening in young, elderly, and low-risk patients, and has raised concerns that increased screening will lead to increased and inappropriate use of cholesterol-lowering medication.23,24 We saw no evidence of such an effect; rather, use of medications was highest in the control group.
Even physicians in group 3 did not follow the recommendations of the National Cholesterol Education Program for follow-up, scheduling few extra visits and referring few patients to nutrition classes. This suggests that further improvements in nutrition intervention need to come at a systems level, with improved methods for patient tracking and prompting for repeated testing and appropriate dietitian referral.
Our study should be generalizable to the large numbers of physicians and patients now providing and receiving care in managed care environments in the United States. They may also be applicable to any environment where physicians practice under time pressure, and deliver interventions that are important but of low acuity.
The combination of a primary care–based program of physician training in patient-centered counseling intervention for nutrition change and a low-cost office-support system has beneficial effects on patients' dietary fat intake, weight, and blood lipid levels.
Accepted for publication August 15, 1998.
This work was supported by grant RO1-HL44492 from the National Heart, Lung, and Blood Institute, Bethesda, Md.
Reprints: Ira S. Ockene, MD, Division of Cardiovascular Medicine, University of Massachusetts Medical School, 55 Lake Ave North, Worcester, MA 01655 (e-mail: email@example.com).