Do as the doctor says, not as he or she does

Doctors are full of advice: Eat healthy. Exercise. Don’t miss your medications. Don’t get that unnecessary test.

The premise behind this advice is simple: most people have little medical knowledge and doctors, by giving their patients more and better information about their health and available treatments, can help them make better decisions. But health education isn’t just the cornerstone of the patient-doctor relationship, it’s also the backbone of an enormous number of policies designed to improve public health.

It turns out that doctors don’t always follow their own advice, at least not as well as we might expect them to. That may have implications for how much we can rely on information alone to shape health behaviors. Educational campaigns to get people to practice healthier lifestyles, shared decision-making tools to inform patients about the risks and benefits of specific health care decisions, and public efforts to report the costs and quality of various types of medical care are all based on the notion that providing more information to patients is an essential step toward improving health care choices and outcomes.

In a new study published by the National Bureau of Economic Research, we analyzed the health care decisions that doctors make when they are patients. We assumed that because doctors have so much more information and knowledge about health and health care than non-doctors, they would make better decisions for themselves.

To do this, we used data on more than 35,000 doctors in the military health system about 70% of whom were men. In this unique database, we could link information on health care treatments to job occupation, allowing us to better understand the decisions that doctors — in this case military doctors — make as patients.

When it came to so-called low-value health care services — services that the medical profession says offer little clinical benefit to patients, such as imaging or blood testing before low-risk surgeries like cataract surgery — doctors didn’t receive much lower rates of them. There’s no reason to have a chest x-ray or a comprehensive metabolic panel before a low-risk surgery, yet almost 10% of non-physicians undergoing low-risk surgeries had these tests. Among physicians, 10% had metabolic panels and 7.5% had chest x-rays.

C-section is widely thought to be excessively used and often unnecessary for childbirth, yet 29% of deliveries by physician mothers in our study were by C-section, compared to about 31% among non-physician mothers. That’s a much smaller difference than we expected for a group of individuals with so much medical knowledge.

The fact that doctors receive similar rates of low-value care as non-doctors is surprising because the enormously popular Choosing Wisely campaign aims to promote conversations between doctors and patients about reducing the use of low-value care.

Doctors also didn’t perform much better than non-doctors when it came to high-value care — health care services that the medical profession thinks should be performed, like individuals with heart disease taking a statin or routine immunizations.

In our study, among individuals with high cholesterol who should have been taking a statin, about 70% of non-doctor patients filled prescriptions for these drugs, compared to 74% among doctors. While that’s a smidge higher than 70%, it is still well below the recommended level of 100%. For doctor and non-doctor patients with diabetes, there was no difference in rates of testing for hemoglobin A1C, a recommended measure of glucose control. We found similar patterns for a number of different high-value services.

Studying the health care behaviors of doctors as patients can tell us something about the potential for clinical and policy efforts to improve health and reduce wasteful care that are based primarily on providing patients with more information and education.

There are important ongoing efforts, for example, to inform patients about what end-of-life medical care actually looks like, with the direct goal of improving individuals’ experiences at the end of life and the indirect goal of reducing intensive and expensive end-of-life care. (In one study of Medicare beneficiaries, during the six months before their deaths, 30% were hospitalized in an intensive care unit, 28% had undergone surgery, and Medicare expenditures averaged $22,000.)

Although information-based efforts have demonstrated some success in this realm, it’s also been shown that the intensity of end-of-life care received by doctors and lawyers — who should be aware of the low odds of benefiting from this care — is actually quite similar to that received by the general population.

The potentially limited ability of information-based efforts to improve health care behaviors also plays an important role in considering the impact of policy efforts to increase price transparency in health care, including President Trump’s recent executive order on price and quality transparency, and the shift of individuals into high-deductible health insurance plans where they have more “skin in the game.” These and other efforts assume that consumers who are well informed about the costs and benefits of specific health care choices will vote with their feet and choose the health care services that are best for them.

But recent work suggests that when patients face higher out-of-pocket costs for health care, they not only cut back on low-value care (a good choice) but also on high-value care (a bad choice). The reduction in use of high-value care presumably stems from not being aware that the foregone care is truly beneficial. But our findings suggest that more information alone may not significantly improve the decision-making.

Doctors have far more information about health, and the benefits and risks of tests and treatments, than most of their patients. Yet when doctors need health care they act like much like their patients. This suggests that only providing patients with more information is unlikely to substantively improve their health care decisions. Other approaches, such as nudges targeted towards patients or providers, and financial incentives, also targeted towards both patients and providers — are likely to have greater effects.

Michael D. Frakes, J.D., is a professor of law and economics at Duke Law School. Jonathan Gruber, Ph.D., is a professor of economics at Massachusetts Institute of Technology. Anupam B. Jena, M.D., is an associate professor of health care policy and medicine at Harvard Medical School.