Adherence to Evidence-Based Guidelines for Heart Failure in Physicians and Their Patients: Lessons From the Heart Failure Adherence Retention Trial (HART)


James E. Calvin, MD, Rush University Medical Center, Department of Medicine/Section of Cardiology, 1653 W. Congress Parkway, Suite 1021 Jelke, Chicago, IL 60612-3833


The Heart Failure Adherence and Retention Trial (HART) provided an opportunity to determine adherence to evidence-based guidelines (EBG) in patients with heart failure (HF). Ten hospitals were the source of 692 patients with HF (EF < 40%). Physicians of patients with HF were classified as adherent to EBG if the patient chart audit showed they were on a beta-blocker, ACE-inhibitor (ACE-I), or angiotensin receptor blocker (ARB). Patients were classified as adherent to EBG if MEMS pill caps were used appropriately more than 80% of the time. Sixty-three percent of physicians prescribed evidence-based medications that were adherent to clinical practice guidelines. New York Heart Association (NYHA) III patients were less likely to be adherent (P < 0.001), as were those with renal disease (P < 0.001) and asthmatics (P < 0.001). Nonadherent physicians were less likely to treat patients with beta-blockers (39% vs 98%, P < 0.001) and ACE-I or ARBs (71% vs 98%P < 0.001). Thirty-seven percent of patients prescribed evidence-based therapy failed to use the MEMS pill cap bottles appropriately and were more likely a minority or higher NYHA class. Adherence to evidence-based therapy is less than optimal in HF patients based on a combination of both physician and patient nonadherence. ©2011 Wiley Periodicals, Inc.

Heart failure (HF) continues to be a major public health problem, affecting an estimated 5 million Americans, with an incidence of 550,000 new cases diagnosed annually.1 Its prevalence rises with age, affecting 1.5% to 2% of the population between the ages of 40 to 59 years and more than 10% of persons older than 60.1 HF is the most common diagnosis-related group discharge in persons older than 65 years.2 This has two important public health concerns. First, HF carries substantial mortality when clinically evident. In 2004, the overall mortality for HF was 19.1%,1 with a sudden cardiac death rate 6 to 9 times that of the normal population.1 Second, HF also accounts for a large burden in rising health care expenditures, translating into 3.4 million office visits and 1 million hospital days per year.3 The indirect and direct costs of HF treatment in the United States are now $33.2 billion annually, siphoning off a significant portion of the cardiovascular budget.1

Because of the high mortality and tremendous costs associated with HF treatment, adherence to evidence-based therapy is critical. Evidence-based literature supports the value of medical therapy and lifestyle modification in delaying progression and improving survival in patients with HF.4 However, patient adherence to these treatment regimens show high variability, with rates ranging from 10% to as high as 85%.5–8 The most common factors associated with patient nonadherence to HF treatment recommendations include complicated medical regimens, poor discharge instructions, lack of patient understanding, lower socioeconomic status, minority status, psychosocial variables, and younger age.6,7,9–11

The Heart Failure Adherence and Retention Trial (HART) was a National Institutes of Health–sponsored clinical trial designed to determine the efficacy of self-management training to improve clinical outcomes in patients with HF through improved patient adherence to evidence-based therapy.12 It featured a subgroup of 692 patients with mild to moderate systolic dysfunction (ejection fraction <40%) and included a significant proportion of both women and minorities.

Baseline data from HART provide the opportunity to study variations in quality of care in patients with systolic HF where practice guidelines can be applied. Specifically, we determined the rates of both physician and patient nonadherence to prescribing and correctly taking evidence-based medical therapy and the roles of socioeconomic status, level of education, ethnicity, and illness acuity in affecting adherence patterns.

Materials and Methods

A complete description of the design and methods of HART has been provided.12 Briefly, HART was based at a single site, Rush University Medical Center in Chicago, but recruited patients from 10 collaborating medical centers within the Chicago metropolitan area. Medical centers were within the city as well as the surrounding southern, western, and northern suburban areas.

Eligibility.  For this report, only patients with systolic dysfunction were eligible, defined as meeting the following: left ventricular ejection fraction ≤40% by echocardiography, radiographic ventriculography, or radionuclide ventriculography (systolic dysfunction) and taking diuretic therapy for at least 3 months.

Exclusion criteria included patients:

  •  With an uncertain 12-month prognosis (New York Heart Association [NYHA] class IV, likelihood of cardiac transplant over the next year, sustained ventricular tachycardia not controlled with therapy within the past 3 months, and other illnesses that would limit 12-month survival).
  •  Who had NYHA class I HF who were not likely to reach the primary end point during the duration of the trial.
  •  Who were not likely to gain benefit from behavioral intervention (ie, underlying cognitive dysfunction or psychosocial comorbidity such as substance abuse, psychosis, suicidal ideation).
  •  Whose symptoms may be alleviated with surgical intervention (ie, severe aortic stenosis).
  •  Whose participation was limited due to logistic issues (eg, patients who were already enrolled in a conflicting research protocol or patients who did not speak English).
  •  Whose physicians refused participation.
  •  Who were not agreeable to making lifestyle changes now or prospectively.
  •  Who had unstable angina, myocardial infarction, coronary artery bypass grafting, or percutaneous transluminal coronary angioplasty PTCA within the past month (temporary exclusion).

Recruitment.  Patients were recruited on a voluntary basis from 10 hospitals throughout the Chicago area, selected because they were able to offer a large number of patients and could provide diversity in sex and ethnicity. Each of the 10 hospitals had HF specialists, nurse practitioners on staff, and HF clinics as well as cardiac catheterization laboratory facilities. Each participating hospital had a designated cardiologist who served as the local principal investigator at the site and co-investigator for the trial. Three recruiting strategies were utilized: inpatient screening, outpatient screening in ongoing clinics, and referrals from local cardiologists and internists. The screening to enrollment ratio was approximately 25% enrolled of those screened.

Once a potential patient was identified, a HART nurse coordinator obtained permission from the physician of record to inspect the medical records for exclusion criteria. If the patient was eligible, he/she was approached and informed about the trial. After a second assessment of eligibility, informed consent was obtained and a baseline examination and chart audit was performed at the local recruiting hospital.

Assessment of Nonadherence.  The American College of Cardiology/American Heart Association (ACC/AHA) guidelines for the management of HF that were current during HART were used to determine physician nonadherence.4 In the case of patients with coronary artery disease, the ACC/AHA 2002 guideline for stable angina was also used.13 Based on the baseline chart audit, physicians were deemed to be nonadherent in this study if they failed to prescribe, in the absence of contraindications, any of the following: (1) angiotensin-converting enzyme (ACE) inhibitor or angiotensin receptor blocker (ARB); or (2) β-blocker. The physician was also deemed to be nonadherent if he/she prescribed a medication in the presence of a known contraindication. In the case of β-blockers, a heart rate of ≤45 beats per minute, a systolic blood pressure <80 mm Hg, and the presence of asthma severe enough to require the use of one asthma medication were contraindications. In the case of an ACE inhibitor or ARB, a systolic blood pressure <80 mm Hg or serum creatinine ≥2.5 mg/dL was used as a contraindication.

Patient adherence to prescribed medication was measured by MEMS electronic pill caps, which record each time a pill cap bottle is opened. The MEMS bottle cap contains a microelectronic circuit that registers times when the pill bottle is opened and closed. The medication events are transferred to a Windows-based computer system and analyzed by AARDEX software (AARDEX Group, Ltd, Sion, Switzerland). These measurements were then compared with the actual regimen that was prescribed by the physician for accuracy. Patient adherence to prescribed medication was defined as taking assigned medication ≥80% of the time.

Statistics.  Univariate comparisons between adherent and nonadherent patients and physicians were performed using t tests for continuous variables and chi-square tests for categorical variables. Multivariate modeling was performed using a stepwise logistic regression strategy. At each iteration of the stepwise process, the least significant covariate was removed from the model until all remaining variables had a nominal P value ≤0.20. Baseline covariates included in the initial model were age, sex, education, minority status, family income, NYHA class and number of comorbidities.


Physician Adherence.  A total of 692 of the 902 patients enrolled in HART had systolic dysfunction, defined as an ejection fraction <40%. Table I compares the baseline characteristics of patients whose physicians were fully adherent (63%) to evidence-based medical therapy as defined above and those whose physicians were not (37%). The patients of nonadherent physicians were less likely to be taking a β-blocker or ACE inhibitor or ARB. They were more likely to be in NYHA class III, to have asthma, or renal disease; less likely to be able to walk 620 feet in 6 minutes; more likely to be a minority; less likely to have at least a high school education; were older; and less likely to have some college education.

Table I.   Univariate Comparison of Baseline Factors Associated With Physician Adherence to Prescribing Evidence-Based Therapy
VariablePhysician Adherent to Evidence-Based MedicineP Value
All, No. (%)Yes, No. (%)No, No. (%)
  1. Abbreviations: ACE, angiotensin-converting enzyme; ARBs, angiotensin receptor blockers; BP, blood pressure; HS, high school; MI, myocardial infarction; NYHA, New York Heart Association; SD, standard deviation.

No.692433 (62.6)259 (37.4) 
 Age, mean (SD)62.4 (13.4)61.6 (13.5)63.8 (13.3).043
 Female290 (41.9)176 (40.6)114 (44.0).385
 Minority282 (40.8)160 (37.0)122 (47.1).009
 ≤HS education311 (44.9)181 (41.8)130 (50.2).032
 Family income <$30,000320 (50.3)195 (48.8)125 (53.0).304
Medical characteristics
 NYHA III220 (31.8)116 (26.8)104 (40.2)<.001
 Hypertension507 (73.6)311 (72.3)196 (75.7).334
 BP >130/80 mm Hg206 (30.0)119 (27.5)87 (34.3).064
 Diabetes272 (39.4)162 (37.4)110 (42.6).174
 Previous MI326 (47.7)204 (47.7)122 (47.8).964
 Asthma83 (13.6)35 (9.2)48 (21.1)<.001
 Renal disease104 (15.2)45 (10.5)59 (23.0)<.001
 Coronary artery disease386 (58.0)232 (55.9)154 (61.6).149
 Comorbidities, mean (SD)3.1 (1.7)3.0 (1.7)3.4 (1.8).081
 6-Min walk distance <620 ft195 (30.7)106 (26.4)89 (38.2).002
 Drug adherence ≥80%368 (63.3)239 (65.5)129 (59.7).164
 Current smoker65 (9.4)37 (8.6)28 (10.8).328
Medical treatments
 Diuretics606 (87.6)383 (88.5)223 (86.1).364
 Statins346 (50.0)225 (52.0)121 (46.7).182
 Aspirin317 (45.8)212 (49.0)105 (40.5).031
 Nonspecific vasodilators171 (24.7)96 (22.2)75 (29.0).045
 ACE inhibitors or ARBs609 (88.0)425 (98.2)184 (71.0)<.001
 β-Blockers524 (75.7)422 (97.5)102 (39.4)<.001

Table II presents the actual reasons for nonadherence to evidence-based therapy. Of 641 patients who should be taking a β-blocker, 152 (23.7%) were not. Of 51 who should not be taking a β-blocker, 35 (68.6%) were prescribed one. Of 638 patients who should be taking ACE/ARB therapy, 69 (10.8%) were not. Of 54 patients who should not be taking ACE/ARB therapy, 74.1% were.

Table II.   Number of Patients Who Were or Were Not Prescribed a Medication From the Indicated Class, According to Whether the Medication Was Contraindicated
MedicationRecommended (No Contraindicators)Not Recommended or Contraindicated
Total, No.On Rx, No, (%)Not on Rx, No. (%)Total, No.On Rx, No. (%)Not on Rx, No. (%)
  1. Abbreviations: ACE, angiotensin-converting enzyme; ARBs, angiotensin receptor blockers, NA, not available; Rx, therapy.

β-Blockers641489 (76.3)152 (23.7)5135 (68.6)16 (31.4)
ACE/ARBs638569 (89.2)69 (10.8)5440 (74.1)14 (25.9)

We identified 4 significant independent predictors of physician nonadherence on multivariate analysis: (1) the number of comorbidities (odds ratio [OR], 1.11; 95% confidence interval [CI], 1.01–1.22; P=.030); (2) age (OR, 1.02; 95% CI, 1.00–1.03; P=.016); minority status (OR, 1.81; 95% CI, 1.28–2.56; P<.001); and NYHA class III HF (OR, 1.64; 95% CI, 1.17–2.30; P=.004) (Table III).

Table III.   Multivariate Predictors of Physicians’ Nonadherence to Prescribing Evidence-Based Therapy in Patients With Systolic Dysfunction
VariablesOdds Ratio95% CLP Value
  1. Abbreviations: CL, confidence limit; NYHA, New York Heart Association.

NYHA class III1.641.17–2.30.004

Patient Adherence. Table IV compares participant nonadherence with prescribed medication based on MEMS pill cap records. Of 581 patients for whom data were obtained, 213 (37%) were classified as nonadherent. Patients who were nonadherent were more likely to be a minority and in NYHA class III. They were less likely to be taking a statin and an ACE inhibitor or ARB.

Table IV.   Univariate Comparison of Baseline Factors Associated With Participant Adherence to Prescribed Medicine Based on Electronic Pill Caps
VariableParticipant’s Pill Cap Adherence ≥80%P Value
All, No. (%)Yes, No. (%)No, No. (%)
  1. Abbreviations: ACE, angiotensin-converting enzyme; ARBs, angiotensin receptor blockers; BP, blood pressure; HS, high school; MI, myocardial infarction; NYHA, New York Heart Association; SD, standard deviation.

No.581368 (63.3)213 (36.7) 
 Age, mean (SD)62.9 (13.5)63.6 (12.5)61.6 (15.0).085
 Female241 (41.5)151 (41.0)90 (42.3).774
 Minority217 (37.3)108 (29.3)109 (51.2)<.001
 ≤HS education258 (44.4)165 (44.8)93 (43.7).784
 Family income <$30,000260 (49.1)165 (47.8)95 (51.6).405
Medical characteristics
 NYHA III187 (32.2)106 (28.8)81 (38.0).022
 Hypertension425 (73.5)261 (71.3)164 (77.4).112
 BP >130/80 mm Hg170 (29.6)98 (27.0)72 (34.0).077
 Diabetes228 (39.3)134 (36.4)94 (44.3).060
 Previous MI272 (47.6)169 (46.6)103 (49.3).530
 Asthma66 (12.6)39 (12.3)27 (13.2).771
 Renal disease85 (14.8)53 (14.5)32 (15.2).815
 Coronary artery disease324 (57.9)208 (58.4)116 (56.9).718
 Comorbidities, mean (SD)3.1 (1.7)3.1 (1.7)3.2 (1.8).230
 6-Min walk distance 20 ft171 (32.5)103 (29.9)68 (37.4).084
 Current smoker46 (7.9)26 (7.1)20 (9.4).322
Medical treatments
 Diuretics508 (87.4)325 (88.3)183 (85.9).400
 Statins300 (51.6)204 (55.4)96 (45.1).016
 Aspirin274 (47.2)176 (47.8)98 (46.0).673
 Nonspecific vasodilators140 (24.1)89 (24.2)51 (23.9).948
 ACE inhibitors or ARBs508 (87.4)332 (90.2)176 (82.6).008
 β-Blockers445 (76.6)289 (78.5)156 (73.2).147

Table V presents the results of the multivariate model. Minority status was the only independent predictor of nonadherence using pill caps (OR, 3.18; 95% CI, 2.11–4.78; P<.0001).

Table V.   Multivariate Predictors of Patients’ Nonadherence (<80% Adherent) to Prescribed Medication Regimen
VariablesOdds Ratio95% CLP Value
  1. Abbreviations: CL, confidence limit; NYHA, New York Heart Association.

Income <$30,0000.730.49–1.10.1363
NYHA class III1.320.89–1.96.1661

Overall Adherence. Figure 1 depicts the proportions of the 4 possible combinations of physician and patient adherence in the 581 patients for whom data were available. In 41% of cases, both physician and patient were adherent to both prescribing and taking evidence-based therapies, respectively. In 15% of cases, both physician and patient were nonadherent. When nonadherence to evidence-based guidelines disregards whether the source was physician or patient, 59% of patients were not receiving life-saving therapy.

Figure 1.

 Breakdown of both physician and patient adherence.


The baseline data of HART, collected across 10 institutions within the greater Chicago metropolitan area, provided an opportunity to focus on various prevalence and predictors of participant and physician nonadherence to evidence-based therapy in HF patients with systolic dysfunction. This cohort was notable for its high proportion of both minorities and women. In this study, both physician and participant nonadherence to evidence-based therapy was relatively high and, when combined, even worse. Indeed, nonadherence stemming from both patients and physicians was 59%. That is, a large majority of patients are not benefitting from life-saving drug therapy. Physician nonadherence was adversely and independently associated with a high number of comorbidities, older age, more advanced HF, and minority status. Patient nonadherence was primarily associated with minority status.

The algorithm used to judge physician nonadherence was simple and straightforward. Joint Commission core measures during the time of the study included use of ACE inhibitors or ARBs for all patients admitted with HF. β-Blockers were not included in core measures although their beneficial effect on early mortality has been shown.5 We simplified our test for physician adherence to ACE inhibitors or ARBs and added β-blockers because of their proven impact on mortality. Data were collected on heart rate, blood pressure, and asthma medications to make some judgment about contraindications to β-blockers. Serum creatinine was collected to assess whether a contraindication existed for ACE inhibitors or ARBs. Creatine kinase and liver enzyme measurements were not performed to evaluate contraindications more objectively. Nonetheless, we believe that our approach provides an assessment of the degree of physician nonadherence that is reasonably robust, clinically reasonable, and easily reproduced.

In 2005, Fonarow and colleagues8 examined physician adherence to 4 Joint Commission on Accreditation of Healthcare Organizations core performance measures and found that median rates of conformity varied from 24% to 86%, with the lowest rate found for discharge instructions, the highest for assessment of left ventricular dysfunction, and a rate of 72% for prescription of ACE inhibitors. Yancy and associates14 extended analysis of physician adherence to other metrics including providing pneumococcal vaccination, treatment of low-density lipoprotein cholesterol levels to <100 mg/dL, and treatment of systolic blood pressure to <140 mm Hg. Physicians did not achieve satisfactory adherence on any of these goals. This suggests a need for improved process of care, better documentation, and increased measures to promote physician adherence of HF therapies. Our baseline findings of physician adherence rates to ACE/ARBs were higher (89%) than the 72% observed by Fonarow. It is possible that this is because patients were enrolled in a trial where health tends to be greater than in the general population. Our results are unique in that they provide insights on the interaction between physician and patient adherence.

The Institute of Medicine (IOM) has concluded that variations in quality of health care are to a great extent multifactorial, but could be categorized under 3 sources of variation: provider-level, patient-level, or system-level variables.15 In the IOM report, provider-level variations in quality of care were attributed to potential health care provider prejudice, bias, or stereotypes held against ethnic minorities; medical decisions made under time pressure with limited information; and increased clinical uncertainty when interviewing minority patients.15 Patient-level variables were attributed to overall mistrust and refusal of recommended treatment, poor adherence to treatment regimens, and delay in seeking medical care. However, these patient-level variations were felt to be small and not significant enough to explain health care disparities.15 Our data suggest that patient-level influences may be greater. A total of 37% of our patients were not adhering to prescribed therapy using medication event monitoring systems (MEMS) pill cap analysis.16 Others, using patient self-report, have determined a much higher patient adherence rate to taking medications (up to 90% of patients). Pill caps analysis assesses patient accuracy of adherence objectively rather than self-perception measured by self-report.

Reasons for patient nonadherence included depression, cost, cultural factors, attitudes regarding medications, effects on sexual function, comorbidity, poor health care literacy, and being unconvinced of the utility of the medications.17,18 Provider-level variables, such as provider prejudice, bias, clinical uncertainty or time pressure were also not specifically evaluated. More study is obviously necessary with emphasis on specific reasons for both patient and physician nonadherence. Our results do suggest, however, that nonadherence is a problem, particularly in minorities.

Limitations and Strengths

Reasons for physician nonadherence could be a combination of provider-level and system-level variables. A limitation of this study is that detailed data were not collected on potential reasons for nonadherence. Our data are clear, however, in pointing to greater physician nonadherence in the presence of more complicated HF and ethnic minority patients. It is likely that factors such as patient-provider interaction, individual system access, and insurance status could be implicated based on these clues.

Strengths of the study were that it was a cross-institutional analysis of 10 hospitals with varying health care delivery systems, including private community, university, and public hospital models included within the mix. All of the institutions offered specialized care with HF specialists on staff, nurse practitioners on staff, and HF clinics, and had access to cardiac catheterization laboratory facilities. Therefore, results could be generalized to clinical practice.

Our results highlight the paradox of diminished care to patients who need it the most, such as those with comorbidities, more severe HF, and cultural barriers. This paradox has been identified in other cardiovascular disorders.19 The reasons for paradoxical care are not clear. Comorbidity undoubtedly puts a huge financial burden on the patient. This is in part related to the large number of medications that HF patients are required to take. Could patients be making choices about what drugs to take based on finances? Poor functional capacity from either comorbidity or the functional impairment from HF itself limits access to health institutions and physicians offices. If so, expansion of home health care options could benefit patients with HF.


Adherence to life-saving drug therapy in patients with systolic HF is unacceptably low. It is urgent to conduct more research on patient-level, provider-level, and system-level factors that account for this. Such data can inform new multilevel trials that aim to improve adherence using interventions that go beyond simply targeting patients. Patient-level factors may simply be the tip of the nonadherence iceberg.