Nancy M. Allen LaPointe, PharmD; Elizabeth R. DeLong, PhD; Anita Chen, MS; Bradley G. Hammill, MA; Lawrence H. Muhlbaier, PhD; Robert M. Califf, MD; Judith M. Kramer, MD, MS 

Am Heart J.  2006;151(5):992-998.  ?2006 Mosby, Inc.

Abstract and Introduction

Abstract

Background: Despite a survival benefit and guideline recommendation for β-blockers in left ventricular systolic dysfunction, β-blockers are underused in clinical practice.
Methods: Medical practices with ≥15 patients with heart failure (HF) in the Duke Databank for Cardiovascular Disease (DDCD) were identified for a prospective, randomized study using a multifaceted intervention to improve β-blocker use. Intervention practices received provider education, patient education materials, feedback on β-blocker use of their patients with HF, and access to telephone consultation with an HF expert. The primary outcome was a comparison between intervention and control practices of the proportion of patients with HF self-reporting β-blocker use on their first routine DDCD follow-up in the postintervention year. A random effects model was used for the analysis.
Results: Post intervention, 2631 patients (1701 in 23 intervention practices and 930 in 22 control practices) completed DDCD follow-up. No significant difference in the proportion of patients with HF reporting β-blocker use was found in the intervention versus control groups (OR 1.16, 95% CI 0.94-1.43, P = .2), although more patients in the intervention group started a β-blocker than stopped a β-blocker during the study period (P = .02).
Conclusions This multifaceted intervention did not significantly increase the mean proportion of patients taking β-blockers within practices exposed to the intervention, although favorable trends were observed. Further studies are needed to identify and evaluate strategies for translating evidence into clinical practice to reduce the global health burden associated with HF.

Introduction

With the publication of several pivotal studies of β-blocker therapy for left ventricular systolic dysfunction between 1996 and 1999, the evidence of a survival benefit became clear.^[1-5] These data were then incorporated into clinical practice guidelines for the management of chronic heart failure (HF). The first was published in 1999.^[6] This was soon followed in 2001 by the American College of Cardiology/American Heart Association Guidelines for the Evaluation and Management of Chronic Heart Failure in the Adult.^[7]

This recommendation for β-blocker therapy in patients with HF represented a complete reversal of previously held beliefs that β-blockers were harmful in patients with left ventricular systolic dysfunction.^[8,9] Given the radical change in recommendations for β-blockers in HF, coupled with the typical slow and incomplete translation of evidence and clinical practice guideline recommendations into clinical practice, the extent and speed of translation of this recommendation into clinical practice was uncertain.^[10-12]

Of all patients with coronary artery disease and an ejection fraction (EF) ≤40% or clinical HF who had been referred to our institution for a cardiovascular procedure, only 48% reported outpatient use of a β-blocker in 1999. Therefore, a multifaceted intervention using Internet-based academic detailing, physician feedback, and patient education to improve the outpatient use of β-blockers in patients with HF was developed and tested.

Methods

Practice Identification, Recruitment, and Randomization

Because of concerns regarding the risk of contamination across patients within practices, the unit of randomization for this study was the medical practice. Medical practices were identified based upon the number of patients with HF that each practice referred to Duke University Medical Center for a cardiovascular procedure in 1999. Using the Duke Databank for Cardiovascular Disease (DDCD), patients with an EF ≤40% and/or documented clinical HF at any time were identified. The DDCD includes an annual clinical follow-up for all patients with significant coronary artery disease in which patients list outpatient medications and their physicians' names. For this study, the physicians listed by patients with HF in 1999 were grouped by medical practice site, and those medical practices with ≥15 patients with HF were identified.

Physicians within identified medical practices were approached by fax, e-mail, and/or telephone to participate in the study. For medical practices with no initial physician response, personalized telephone recruitment calls were made by one of the investigators. Each physician who agreed to participate was asked to recruit the other physicians, if any, within his/her practice. Medical practices in which at least one physician agreed to participate in the study were then randomized to the control or intervention arm after stratification by the number of patients with HF per medical practice referred to Duke. A stratified randomization scheme was used to attempt to balance the number of patients with HF followed up in DDCD between intervention and control arms. The entire recruitment and randomization process took place between July 2000 and January 2001. Neither physicians nor medical practices were financially compensated for participation in this study. The study protocol was reviewed and approved by the Duke Medical Center Institutional Review Board.

Control and Intervention Groups

Physicians were notified in January 2001 of their medical practice's randomization status. Physicians in both groups were given a 1-page summary of the evidence for β-blockers in patients with HF and a patient-oriented brochure on β-blocker use in HF that the physician could distribute to his/her patients. Physicians in the intervention group were also given patient-focused educational videotapes for distribution to their patients. Both groups received feedback on β-blocker use in patients with HF. For control group practices, baseline data was provided on the overall proportion of patients with HF within the DDCD who were receiving a β-blocker. Intervention group physicians were given patient-specific information on β-blocker use from each of the patients who listed him/her as their physician. The β-blocker use status for each patient was determined by patient survey conducted between October 2000 and April 2001, and the report was sent to each physician in June 2001.

Physicians in the intervention group were also invited to participate in a live Internet educational program focused on β-blockers in HF management. This interactive academic detailing program occurred in February 2001 and provided continuing medical education credit. The program was archived and available through the end of the intervention period for physicians unable to participate in the live program. In August 2001, a text copy of the program was sent to each physician within each intervention group practice. Physicians in the intervention group were also given a toll-free number they could use to discuss issues and questions regarding β-blocker use with a cardiologist with expertise in HF management. All components of the intervention were completed between February 2001 and September 2001.

Medical Practice Characteristics

The following baseline physician characteristics were obtained from the American Medical Association: medical specialty, boarded specialty, age, years since medical school graduation, and race. Medical practices in which all physicians listed cardiology as their specialty were considered cardiology practices, medical practices in which all physicians listed either internal medicine or family medicine were considered primary care practices, and medical practice in which there was a mix of specialties or a mix between primary care and specialties were considered mixed practices.

Each practice was identified as a rural, urban, or urban cluster practice as determined by the zip code of the medical practice using US Census data. Through communication with the medical practice staff and with data from the DDCD, the following practice characteristics were determined: the number of physicians within each practice, number of practices affiliated with Duke, number of patients with HF in the DDCD in the year 2000 within each practice, and the proportion of patients with HF reporting β-blocker use in the year 2000 within each practice.

Patient Identification and Patient Characteristics for Analyses

Patients with heart failure who completed their annual DDCD clinical follow-up questionnaire between October 2001 through September 2002 (postintervention period) and also listed one of the physicians participating in a control or intervention group practice as their physician were included in the analyses. Patient characteristics were obtained from the patient's history and physical examination that occurred just before their entry into the DDCD clinical follow-up. The collected patient characteristics included age, sex, race, ejection fraction, New York Heart Association class, documentation of a clinical diagnosis of HF, number of recorded myocardial infarctions, and smoking history.

Primary End Point

Patient self-reported β-blocker use was obtained from the completed DDCD clinical follow-up questionnaires obtained during the postintervention period. Report of any β-blocker was considered confirmation of β-blocker use. The proportion of DDCD patients reporting β-blocker use in the postintervention period in each practice was determined, and the mean proportion of β-blocker users for the intervention and control groups was calculated. The primary end point was a comparison of mean proportion of β-blocker users between the intervention and control arms.

Secondary End Points

Several prespecified secondary analyses were undertaken. First, because the postintervention data collection period spanned a period of 12 months, we explored changes in reported β-blocker use over that period and compared any observed changes between the intervention and control groups. Secondly, we looked at patients who were within a control or intervention group practice in the postintervention period and who also had a DDCD clinical follow-up during the preintervention period. From these patients, we determined and compared the proportion of intervention group patients and control group patients who reported β-blockers both before and after the intervention. Of patients with inconsistent β-blocker use before and after the intervention, the proportion of patients within each arm that started a β-blocker during the intervention period was compared with the proportion that stopped a β-blocker during the intervention period.

In a post hoc analysis, we also summarized specific β-blocker medications reported by patients during the postintervention period. The number and proportion of patients reporting either carvedilol or metoprolol succinate was determined and compared between intervention and control groups. In addition, the proportion of patients reporting ≥50 mg/d of carvedilol or ≥100 mg/d of metoprolol succinate within each group was also determined and compared.

Physician participation in the Internet educational program was tracked. The number of physicians and the number of practices represented by those physicians who participated in the program were determined.

Statistical Methods

Patient and practice characteristics were assessed for the intervention and control groups. Means with SD or medians with 25th and 75th percentiles were reported for continuous variables and frequencies for categorical variables. Because practice characteristics did not appear to be balanced, we compared patient characteristics using the Wilcoxon rank sum test for continuous variables and Fisher exact test for categorical variables.

For the primary analysis, a patient-level random effects model was used, in which the practice was considered the random effect.^[13] This model allows for β-blocker use in each practice to vary and accounts for the inherent correlation of outcomes for patients within a practice. In our model, we compared β-blocker use between the control and intervention group in the postintervention period. Because we randomized practices to a control or intervention arm, we would expect most patient and practice characteristics of the 2 groups to be similar. However, any imbalances that were identified were also entered into the model. In addition, a post hoc analysis was conducted that excluded patients without a documented EF ≤40%.

For the trend analysis, we explored β-blocker use as a function of time with respect to intervention and control groups. A random slopes, random intercepts model with a time by study arm interaction was used to test whether the time trend in β-blocker use differed between intervention and control.

For the subset of patients who were in a control or intervention practice in the postintervention period and also had DDCD clinical follow-up in the preintervention period but who did not have consistent reporting of β-blocker use, we further explored the change of β-blocker use between the periods to evaluate if the intervention had any effect on these changes. McNemar test was used to compare the proportions of patients who started a β-blocker after the intervention to those who stopped a β-blocker after the intervention.

The proportion of patients reporting use of either carvedilol or metoprolol succinate was compared among the control and intervention group patients. In addition, reported use of targeted doses of carvedilol or metoprolol succinate was compared between the groups. χ² Test was used for all comparisons among the control and intervention groups.

A P value of <.05 was established as the level of statistical significance for all tests. All analyses were performed using SAS software (version 8.2, SAS Institute, Cary, NC).

Results

Sixty-six medical practices consisting of 319 physicians were identified. A total of 66 physicians representing 45 practices personally agreed to participate in the study. Twenty-three practices were randomized to the intervention arm and 22 practices were randomized to the control arm (Figure 1). After randomization, 2 practices in the intervention arm withdrew from the study. One practice withdrew because of the death of the lead physician and the other practice withdrew over concerns about using educational materials not developed within their own health system. Data from both of these practices remained in the analyses.

Figure 1. 

Physician/practice recruitment.

     

Practice and Patient Characteristics

Characteristics of the intervention and control group practices are presented in Table I . A total of 2717 patients with HF listed a control or intervention group physician on their DDCD clinical follow-up questionnaire during the postintervention period: 1760 patients in intervention group practices and 957 patients in control group practices. Patient characteristics are listed in Table II . Differences between the groups in race and proportion of patients with documented clinical HF were identified ( Table II ).

Primary Analysis

Of the 2717 patients identified in control and intervention group practices, 86 (3%) had missing medication data on the postintervention clinical follow-up questionnaire, leaving a total of 2631 patients for inclusion in the primary outcome analysis. After the intervention, the mean use of β-blockers in control group practices was 63% and the mean use of β-blockers in intervention group practices was 66%. There was no statistically significant difference in β-blocker use between intervention group practices and control group practices (OR 1.16, 95% CI 0.94-1.43, P = .2). This difference remained statistically nonsignificant after adjustments were made for the imbalances between the groups in race and documented clinical HF. This difference also remained statistically nonsignificant with an additional model adjusting for preintervention β-blocker use. In addition, results remained unchanged when the 2 medical practices that withdrew from the study were excluded from the model (OR 1.17, 95% CI 0.94-1.45, P = .2) and when patients without a documented EF ≤40% at any point were excluded (OR 1.15, 95% CI 0.89-1.49, P = .3).

Secondary Analyses

Time trend analysis of the reported use of β-blockers during the postintervention data collection period indicated a trend toward higher β-blocker use reported in the later months of the data collection period than in the earlier months. However, there was no statistically significant difference in these trends between the intervention and control group patients (P = .2) (Figure 2).

Figure 2. 

Mean bimonthly β-blocker use reported during the postintervention period (October 2001 through September 2002).

     

A total of 1680 patients were in a control or intervention practice in the postintervention period and also had a DDCD clinical follow-up in the preintervention period (601 in a control group practice and 1079 in an intervention group practice). Of the 601 from the control group, 359 (60%) reported β-blocker use before the intervention period and 369 (61%) reported β-blocker use after the intervention period. Of control patients who did not have consistent reporting of β-blocker use before and after the intervention, 66 patients started a β-blocker and 56 discontinued a β-blocker representing no overall statistically significant change in β-blocker use during the study period (P = .4). Of the 1079 from the intervention group, 689 (64%) reported β-blocker use before the intervention and 722 (67%) reported β-blocker use after the intervention. Of intervention patients who did not have consistent reporting of β-blockers before and after the intervention, 113 patients started a β-blocker during the intervention period and 80 stopped a β-blocker representing a statistically significant change during the intervention period (P = .02).

The most frequently reported β-blocker during the postintervention period in both groups was atenolol with 252 (42%) of the 607 control group patients with medication data and 425 (36%) of the 1181 intervention group patients with medication data reporting its use. There was no statistically significant difference between the groups in the proportion of patients reporting use of carvedilol or metoprolol succinate (219 [37%] in control group and 465 [40%] in intervention group, P = .2). Of these 684 patients reporting carvedilol or metoprolol succinate, 569 (83%) also reported dosage. A total of 61 (34%) of 179 control group patients and 153 (39%) of 390 intervention group patients reported dosages of carvedilol and metoprolol succinate at or above the targeted dose (P = .2). However, of the 115 patients reporting carvedilol dosage, there was a significantly higher proportion of patients at or above target dose in the intervention group (25 [36%] of 70 intervention group patients versus 7 [16%] of 45 control group patients, P = .02).

The only measure of physician participation in the study was their participation in the live interactive academic detailing program. Only 11 (9%) of the 120 eligible physicians participated, representing 9 of the 21 eligible medical practices.

Discussion

This multifaceted intervention aimed at practicing physicians did not produce a significant increase in β-blocker use. However, the trend toward greater use of β-blockers in the intervention group, coupled with results of prespecified secondary analyses demonstrating significantly more patients starting a β-blocker than stopping a β-blocker in intervention practices, puts the findings in a more encouraging light.

The failure to achieve a significant difference in β-blocker use in this trial could be attributed to several reasons. First, busy clinicians may not have had time or interest in fully participating in the multifaceted intervention. Secondly, β-blocker use in the study population was greater than expected during the study period and may have approached a ?ceiling? of use that for legitimate clinical reasons cannot be exceeded in practice. Third, the number of recruited practices was smaller than anticipated and coupled with the higher-than-expected baseline use of β-blockers; there was less power to be able to detect a difference between the groups. Finally, the strength of the intervention may have been limited by the dynamic nature of clinical practice.

Previous studies have suggested that multifaceted interventions are necessary to improve physicians' adherence to guidelines.^[14-21] However, most studies have met with minimal success, and investigators continue to explore new types of interventions, physician incentives, and technological advances to stimulate adherence. Regardless of the methods used, improvement in adherence requires the allocation of time and resources to implement, sustain, and evaluate an intervention. We designed a multifaceted intervention and outcome measure that would minimize intrusions on the physicians' time. By using a clinical database, we were able to identify patients and practices for the study, collect patient demographics and medical history, and track patients' self-reported β-blocker use without requiring additional work from the medical practice, physicians, or patients. We hoped that this decreased burden on the medical practices and physicians would encourage study participation. Although there was no overall measure of physician participation, poor participation in the Internet educational program may have indicated lack of awareness of the study and/or poor overall study participation.

Not all patients with HF will be eligible for β-blockers because of comorbid conditions or documented adverse events from a β-blocker. Therefore, achieving β-blocker use in 100% of patients with HF cannot be the goal of any intervention. Some studies have estimated that ≥12% of patients with HF may have relative contraindications to β-blocker therapy.^[22] Gupta et al^[23] retrospectively evaluated 500 patients enrolled in an outpatient HF clinic to determine a realistic target for β-blocker use in quality improvement initiatives. The authors concluded that about 70% of patients with HF can be successfully treated with a β-blocker in a specialized HF clinic. In another study evaluating the early impact of an HF management program, about 76% of patients with HF were receiving a β-blocker.^[24] Thus, the ?ceiling? of use for β-blockers in HF treatment may be far below 100%, and as the ?ceiling? level is approached, it may become increasingly difficult to demonstrate any significant improvements. The mean proportion of β-blocker use of 66% in our intervention group very nearly approached this suggested ?ceiling.? In this situation, demonstrating additional significant increases in β-blocker use may not be possible with any intervention.

Interventions designed to change outpatient practice behavior and patterns may require months or years to complete. During this time, unpredictable events may occur that can either enhance or detract from the intervention. In our population, β-blocker use in HF was steadily increasing during the intervention period. In 2000, 2538 (51%) of all 4964 patients with HF within the DDCD reported β-blocker use. In 2002, this proportion increased with 3265 (62%) of 5269 patients with HF reporting β-blocker use.^[25] Assuming 60 of 66 practices agreed to participate, the study had 85% power at the 5% significance level to detect a difference of 7 percentage points between the postintervention average percentage of 65 assumed for the control group and an average percentage of 72 (SD 9 percentage points) for the intervention group. Because the study was only able to recruit 3 quarters of the anticipated sample size and the difference in average β-blocker use (3 percentage points) was much lower than anticipated, the power was substantially reduced to 20%. In addition, patients were not required to remain within one medical practice throughout the study period and there were no requirements that the patient visit their physician during the intervention period. Thus, the ?assignment? of a patient to one medical practice to analyze a practice-based intervention presents numerous challenges.

Despite these challenges, even modest increases in β-blocker use in patients with HF would be expected to carry substantial benefits. Based upon the results of the MERIT HF study, for every 27 patients with HF treated with a β-blocker, 1 life would be saved in 1 year.^[2] According to the American Heart Association, there are about 5 million Americans with HF. Thus if an intervention could increase the use of β-blockers by only 1%, an additional 1850 lives would be saved in 1 year.

Specific methods to improve outpatient use of β-blockers in patients with HF have not been extensively studied. Heart failure disease management programs have developed around the country and have demonstrated some favorable results.^[23,24,26] However, these programs are resource intensive. In a recent study comparing use of β-blockers in patients with HF whose physicians were randomly assigned to a control group, a nurse facilitator group or prompt/feedback system demonstrated that only the nurse facilitator group was effective in increasing the use of β-blockers and titrating the β-blocker dose up to target.^[27] The authors concluded that adherence to complex guidelines such as those for β-blockers in HF require an active intervention rather than a passive intervention and preferably one involving a system change rather than an attempt to change individual behavior.

There are several limitations of this study. Basic educational materials were provided for physicians and patients in both the intervention and control arms. Although the materials for the intervention arm were considered more comprehensive than those for the control arm, this may have contributed to improvements seen in both groups and compromised our ability to detect a difference between the groups. In addition, physicians in both the intervention and control arms were given educational materials to distribute to their patients with HF. We were unable to assess the extent or uniformity of use of these materials among practices within a study arm or between the study arms.

Relying on patient self-report could be considered a strength or a limitation of our study design, depending upon the relative accuracy of physician's notes in a medical record and the patients' accuracy in recording their medications. In addition, this study was not designed to assess patients' daily compliance with the intended regimen. Finally, there was an unequal number of patients in the intervention and control group practices despite stratification in the randomization scheme by number of patients followed by DDCD within each practice.

Conclusions

The use of β-blockers in medical practices exposed to the multifaceted intervention was higher than in those practices not exposed to the intervention, but the difference was not statistically significant. The survival benefit associated with β-blockers in patients with HF is large and, thus, interventions with even modest increases in β-blocker use could result in substantial benefits. Future interventions need to identify creative ways to integrate system changes with personalized and intensive patient activation.