Understanding and Modifying Physician Behavior for Prevention and Management of Cardiovascular Disease

One of the more vexing problems in health services research is understanding why, despite evidence from randomized trials, physicians do not follow published clinical guidelines. Four articles in this issue of the Journal of General Internal Medicine shed light on various aspects of this problem as it relates to the prevention and management of cardiovascular disease.

To understand the cause of poor physician compliance, we must first determine the factors affecting physician behavior. These can be classified as knowledge (of guidelines), attitude (physician's and patient's view of the guideline), and external incentives including disincentives. Physician knowledge has been consistently high when examined and is unlikely to be a major contributor to noncompliance.

On the other hand, attitudes may be important in explaining poor physician compliance with guidelines. Most practitioners agree that randomized trials have shown that β-blockers improve survival following myocardial infarction, but their own (or the patient's) concerns about side effects may deter a busy clinician from prescribing a β-blocker. Two studies in this issue support this view. The report by Ubel et al. examines primary care physicians' attitudes toward the use of β-blockers and diuretics for the treatment of hypertension, the treatments recommended by the Joint National Commission on High Blood Pressure at the time of the survey (1997).¹ They found that physicians believe diuretics are less effective than β-blockers, calcium antagonists, or angiotensin converting enzyme (ACE) inhibitors. Physicians in their survey also believed that β-blockers are not tolerated as well as drugs in the other three classes. Both of these views were associated with physicians' unwillingness to prescribe diuretics and β-blockers. Ubel et al. note that multiple randomized trials have shown no clear differences in effectiveness or tolerability between the four classes of medications, implying that these negative attitudes toward diuretics and β-blockers do not appear to be justified.

The article by Foley et al. examines physicians' attitudes toward treatment of hyperlipidemia.² Foley et al. find that attitudes, as measured by a newly developed survey instrument, are associated with physicians' intention to treat hyperlipidemia to appropriate thresholds. Physicians who were less willing to treat to recommended low-density lipoprotein (LDL) cholesterol levels were more likely to view high doses of statins to be risky, to believe levels near threshold were sufficient, to feel less time pressure in reaching threshold, to experience time and resource constraints, and to be pessimistic about their ability to treat the patient to the LDL goal.

Do incentives exist today that affect provider behavior? For decades, pharmaceutical companies have provided incentives for physicians. In the Ubel study, the availability of free samples of medications was independently associated with using ACE inhibitors or calcium antagonists instead of β-blockers or diuretics for treatment of uncomplicated hypertension.¹ Although industry interventions clearly have had an effect in choice of drugs, the overall effect is difficult to judge. Improved use of statin and ACE inhibitors in appropriate patients is in the interest of many pharmaceutical companies, while treatment with generic diuretics and β-blockers is not. Do nonindustry incentives exist? Peer review of provider care is required by the Joint Commission on Accreditation of Health Care Organizations (JCAHO). The impact of these reviews on physician behavior is unclear, but may be significant if the reviews evaluate guideline compliance and are performed by physicians known to the reviewee.

Many interventions have been developed to educate physicians regarding clinical practice guidelines. Guidelines for LDL cholesterol are particularly difficult to memorize because treatment depends on incorporating multiple risk factors into a global coronary heart disease risk. In this issue, Sheridan et al. review various risk calculation tools that have been developed to make global risk calculation easier for the physician.³ They find that these tools, varying from paper charts to electronic calculators, provide comparable risk estimation to the full equations from the Framingham Heart Study (from which they were developed). Sheridan et al. note that only a few studies have examined the effect of risk calculators on clinical practice and these studies did not demonstrate a discernable effect on treatment.

Computer-generated reminders may be an attractive intervention given the low cost and wide applicability. Tierney et al. examine computer-generated evidence-based cardiac care suggestions that target primary care physicians and pharmacists (who then counsel physicians).⁴ Cardiac care suggestions for physicians were printed at the end of the medication list on the encounter form and displayed as suggested orders on physicians' workstations. The investigators observed a trend toward an effect for pneumococcal vaccination (P = .09), but saw no effect on initiation or increased dosing of any cardiac drug (e.g., ACE inhibitors, β-blockers, or diuretics). Why were reminders ineffective in this study? With any reminder intervention, one could argue that contamination occurred if somehow the intervention affected the control patients. However, the meticulous study design including randomization at the provider level should have limited if not eliminated this problem. A more likely reason is that it takes a high-impact intervention to get an already reluctant physician to prescribe drugs that may have significant side effects. This explains why in this study and a prior study⁵ reminders influenced use of vaccinations, but not treatment with cardiac medications. We should not act on these negative findings by limiting further research into computer reminders. Such interventions are so low cost that even a tiny effect may be cost effective.

So what does it take to improve quality of care? Simple notifications and reminder systems have at best a small effect. One attractive option may be paying for quality. As Tierney et al. point out, the few studies that have examined this topic have had mixed results. Although one could argue that higher payments would have increased physician compliance with guidelines, it is unlikely that cash bonuses to physicians can be supported by the current health care payment system. Alternatively, a medical group could choose to hire or retain only “quality” physicians by paying a higher salary. Identifying “quality of care” is expensive and requires detailed chart reviews. Plans could start by insisting on proxy measures such as recent board certification. Once the cost of measuring quality care is reduced through the use of detailed electronic medical records, practices or individual physicians could be certified for having met a quality standard. Rating individual physicians will be challenging because many physicians have too few patients with any given disease (e.g., diabetes mellitus) to accurately measure disease-specific quality of care.⁶ More high-quality studies, like those presented in this issue, are needed to help us understand and modify physician behavior so that we can more effectively reduce morbidity and mortality from cardiovascular disease.