When Doctors Disagree

The American public has been warned that medical care may soon have to be rationed. In an article in Newsweek (November 12, 1984), Dr. William Schwartz writes that “either we will accept the continued rise of hospital costs that result from full exploitation of technological advances, or we will start to ration hospital care. And if it’s the latter, we will then have to say to some people, ‘Yes, a new liver would be good for you, but as a society we can’t afford it.’ ” The economist Henry Aaron, in collaboration with Dr. Schwartz, has elaborated on this theme in a recent, widely publicized book, The Painful Prescription: Rationing Hospital Care,¹ as well as in articles in The New York Times and elsewhere.

The truth is that medical care is already rationed and it always has been. One of the disgraces of national policy is that the poor and unemployed who cannot afford to pay for medical care or have no medical insurance must often accept inferior treatment, if they can get it at all. What is new is that the costs of care, whether it is paid for directly or through insurance, are rapidly exceeding the ability or willingness of even the middle class to pay. The main reason for this is that medical scientists are inventing more new diagnostic and therapeutic “medical technologies”² than the economy can pay for. The artificial heart is a famous example. If a workable model were developed and made available to the estimated fifty thousand terminally ill cardiac patients each year who could benefit from it, that would cost between three and five billion dollars. Since it is unlikely that the government will pay the bill—as it has for kidney transplants and hemodialysis for patients dying from kidney failure—we face the prospect that this enormously expensive medical technology, if it succeeds, will be made available primarily to those willing to pay for it.

The case of the artificial heart is by no means unusual. The potential costs of new medical technologies are enormous. They include the exchange of blood plasma (plasmapheresis) in the treatment of rheumatoid arthritis and related diseases, and diagnosis by means of nuclear magnetic resonance imaging, and liver transplants. We can assume that new medical technologies will continually appear, and we can also assume that major controversies will arise over who will benefit from them and who will pay for them.

A number of measures could be taken short of rationing. These include preventive medicine that would decrease the need for medical care, such as encouraging people to stop smoking, improving prenatal care, and increasing programs for immunization. In addition, the large US investment in basic research, primarily at the National Institutes of Health, will continue to produce health-improving technology and all of it may not increase costs. As Lewis Thomas has pointed out, truly effective medical technology, such as immunization and antibiotics for specific diseases, will be less expensive than the complex but inefficient “halfway technologies,” including the artificial kidneys and hearts that make up such a dramatic part of current medical practice.³ Finally, substantial savings could be achieved by reorganizing the delivery of health care. As an example of how large these savings might be, Dr. John Wennberg of Dartmouth, appearing before a recent Senate Appropriations Committee hearing, estimated that 40 percent of hospital costs could be saved if hospitals gave “out-patient care” to many patients who are now required to occupy a hospital bed if they are to be treated.⁴

But a quite different approach to reducing the need for rationing has, I think, been insufficiently stressed. Doctors tend to believe that they should recommend whatever they think will be beneficial, and they tend to be optimistic about new medical techniques. But their optimism is often not based on sound evidence. Not every new treatment is successful—indeed, much of current medical practice may be ineffective. Discovering what works and what does not is something the medical profession is not very good at, and the amounts now invested in studies of the effectiveness of US medical technology have been inadequate. Furthermore, patients often fail to take part in decisions affecting their own health, or uncritically accept the optimism of their doctors. These factors, I believe, have led to widespread overtreatment of patients. What we would save from a successful national program to assess medical technologies might greatly help to avoid rationing medical care.

How strong is the evidence for overtreatment? Without comprehensive studies of the effectiveness of medical care, evidence for widespread overuse of medical technology is somewhat indirect but it is nonetheless convincing. In 1970, I reported that there were twice as many surgeons in the United States for each million of population as there were in Great Britain, and that they performed twice as many operations.⁵ That is to say, in a given year, an American was twice as likely to undergo surgery as a person in Great Britain. This disproportion in the probability of surgery has not changed since 1970. Yet the general health and life expectancy of the British population compares favorably with that of our own.⁶ Henry Aaron and William Schwartz argue convincingly that the British have failed to provide enough of some new medical technologies, such as hemodialysis and open-heart surgery, to meet the needs of their population. While their use of some procedures may be too low, ours are certainly too high.

Dr. Wennberg has produced the strongest evidence suggesting that medical treatment can be excessive in the US. Wennberg made surveys of the frequency of surgery and hospitalization in such states as Vermont, Iowa, and Massachusetts. He divided individual communities of a county or state into what he called “hospital markets” or “hospital service areas” in which most of the population used one local hospital or group of hospitals. He then gathered information about the rates of use of medical treatments, diagnostic tests, and surgical procedures in various hospital markets by examining health insurance records (such as Medicare and Medicaid) and hospital discharge records. He found, he says, that many “hospital markets” had “highly variable rates of use for most specific medical treatments, diagnostic tests, and surgical procedures, and by widely different resource use rates.” For example, he observed that

in Maine, by the time women reach seventy years of age in one hospital market the likelihood they have undergone a hysterectomy is 20 percent while in another market it is 70 percent.

In Iowa, the chances that male residents who reach age eighty-five have undergone prostatectomy range from a low of 15 percent to a high of more than 60 percent in different hospital markets.

In Vermont the probability that resident children will undergo a tonsillectomy has ranged from a low of 8 percent in one hospital market to a high of nearly 70 percent in another.⁷

Similarly, Wennberg has shown that Americans living in some cities or towns in New England and in Iowa are twice as likely to undergo surgery or to be hospitalized for nonsurgical conditions as those living in other cities or towns of those states. Moreover, contrary to what one might expect would result from such wide variations in treatment, no sharp differences have been observed in the reported incidence of disease or in the mortality rate in the areas compared by Wennberg.

Why not? No one has convincingly explained why treatment varies so much. Wennberg’s studies failed to show any significant correlations between high rates of use of medical treatments and rates of illness, or insurance coverage, or access to medical service. The variations, he wrote, “also persist after adjustment for age, which tends to account for most illness-related differences in populations.” There are large variations among patients even in a state such as Vermont, where the population is relatively homogeneous in economic and social status. Wennberg considered the possibility that some variation might be explained by the varying supply in different hospital markets of physicians and facilities. But he concluded that “most” of the variations remained “unexplained” even if such differences were taken into account. He cited one study on rates of hysterectomy which showed that “per capita number of beds and gynecologists are virtually the same in hospital markets with low and with high per capita hysterectomy rates.”

Moreover, the variations in the use of medical treatments could not be explained by the incentives given to physicians by “fee-for-service medicine,” since the pattern of variation is similar not only in “fee-for-service” markets in the United States but also in “the health care regions in Canada, England, and Norway, even though obvious differences exist in the supply of surgeons, the organization and financing of services, and in the cultural and demographic characteristics of hospital market area residents.”

Wennberg’s own explanation of the variations is that physicians differ in what he calls “practice style”—in the “subjective factors” that help to determine whether a physician recommends that a patient receive a specific service and whether the treatment is offered in a hospital or an outpatient facility.

But this seems only a partial explanation. Why do physicians with similar “practice styles” seem to concentrate in certain geographic areas? If their “practice styles” were truly “subjective,” and were not intended to match the behavior of their colleagues, then variations in practice patterns would most likely not be observed. A student of these variations, Dr. David Eddy, has argued that “a physician who follows the practices of his or her colleagues is safe from criticism, free from having to explain his or her actions,” so that a “tendency to follow the pack is the most important single explanation of regional variations in medical practice.” According to this view,

differences between regions are observed because individual physicians tend to follow what is considered standard and accepted in the community. A community standard evolves from statements published in national journals and textbooks, from the opinions of established physicians, and from new ideas brought to the community by new physicians.⁸

But how could such different patterns take shape without marked effects on the health of local populations? The most convincing explanation, I believe, derives from the fact that for perhaps a third to a half of the care doctors provide, the benefits and risks are evenly balanced, and the net benefits are exceedingly small, if they exist at all.⁹ In this situation doctors who are in positions of authority may, without seeming to incur excessive risks, help to establish a pattern of greater or lesser surgical intervention or hospitalization. Neither they nor their patients may be aware that the pattern is quite different in another part of the state, or that the effectiveness of their treatments may be uncertain. The more closely we look at specific operations and treatments, moreover, the more likely it seems that some are being used excessively.

Removal of the gallbladder, for example, has for many years been routinely prescribed for many patients with gallstones, even in the absence of a previous attack of gallbladder pain. The reason for doing so is simple and has intuitive appeal. Elective surgery to remove the gallbladder now carries a relatively small risk of death. If an emergency should later arise and the gallbladder must be removed, however, the risk of death will certainly be much greater, depending on how old and sick the patient has become.

This intuitive clinical rule is called into question, however, when one examines more closely the probability of risks and benefits. Emergency surgery for gallbladder removal does indeed carry a higher risk than elective surgery, but it may never be necessary at all; and if emergency surgery eventually becomes necessary, a good many years may already have passed. Using available data, and taking account of the age and general health of the patients, my former associates at Harvard and a separate group of health policy analysts at Case Western Reserve have calculated that removal of the gallbladder for gallstones that are “asymptomatic”—i.e., cause neither pain nor severe indigestion—will increase life expectancy by no more than a week or two, and perhaps even decrease it. The relatively few people who will later die during emergency surgery gain years of life if they have an operation right away; among those who take the risks of elective surgery now, some will die, losing an approximately equal number of years. Thus, on average, life expectancy is minimally affected.¹⁰

This example is a relatively simple one since both the benefits and risks of surgery are measured by the standard of survival or death. However, for much medical care the analysis is more difficult, particularly when benefits are measured in relief of symptoms, such as pain or disability, and risks are measured by loss of life or by the grave consequences for which doctors use the word “complications.” Elective removal of the prostate in men suffering from benign prostatic hypertrophy—a benign enlargement in the prostate gland—will prevent complete obstruction to the flow of urine; such obstruction will threaten the patient’s life and require emergency surgery. But complete obstruction is uncommon, and elective prostatectomy is carried out usually to relieve such symptoms as frequent and slow urination. Its success in relieving symptoms must be weighed against the risks of operating, which include not only death but also possible loss of sexual potency. There is no really satisfactory way to measure relief of symptoms quantitatively, or to weigh the value of such benefits against the risk of death.

That risks and benefits are fairly evenly balanced in the case of elective prostatectomy, however, is strongly suggested by the fact that the surgeons who perform this operation are sharply divided over when it should be done. This difference of opinion is reflected in Wennberg’s findings. In some parts of Vermont and other states, for example, the operation is two to three times more frequent than it is in other places. Forty percent of the men living in some hospital service areas will have undergone prostate removal by age eighty. In other places as few as 20 percent will have done so. There is no evidence showing that men who live where fewer operations are performed are in any way worse off, although we lack specific data on how tolerable many men found their symptoms if they did not have an operation. Still, it seems reasonable to assume that there is some overuse of this operation, perhaps a great deal.

Hysterectomy, or removal of the uterus, is perhaps the single operation most widely believed by the public to be performed too frequently. It is obviously justified for the removal of cancer of the cervix or endometrium, or for prolapse of the uterus (popularly known as “falling of the womb”). Elective hysterectomy for sterilization, for relief of the symptoms of menopause, or for preventing cancer is, however, highly controversial among physicians, health policy experts, and women generally. As with prostatectomy, the difference in professional opinion is reflected in a two- to threefold variation in the frequency of surgery in otherwise comparable geographic areas. Wennberg calculates that in some hospital service areas, between 45 and 50 percent of women will have had their uterus removed by age sixty-five; in others only 10 to 15 percent will have done so.

Again it is difficult to say unequivocally that the high frequency of elective hysterectomy in some hospital service areas shows that the operation is widely overused or unnecessary. By removal of the uterus and thus the chance of future uterine cancer, the life expectancy of women who have this operation is increased by two to three months. In fact, the unidentifiable 1.3 percent of women who would have contracted cancer at a later time receive all the benefit, each one being spared several years of lost life. This considerable benefit will likely be accompanied by an improvement in the quality of life through the relief of menopausal symptoms. For a healthy middle-aged woman, moreover, the risk of surgical or anesthetic death from the operation is relatively low.

All this seems to suggest that elective hysterectomy may be a good risk to take. However, so far as preventing cancer is concerned, most of the benefits of elective hysterectomy can be achieved equally well by diagnostic techniques such as regular Pap smear exams for cancer of the cervix and careful observation for the abnormal bleeding associated with cancer of the endometrium. After early diagnosis of cancer, hysterectomy or other appropriate treatment can be expected to have a high likelihood of success.

In the absence of strong medical reasons for performing the operation, therefore; it is by no means clear that elective or discretionary hysterectomy should be performed as frequently as it is by many doctors. The same could be said of many other operations, including back surgery, cataract removal, knee operations, and dental extractions performed in hospitals. It is especially true of those operations that are often performed for conditions that are part of the aging process. These include hysterectomy and prostatectomy, but also tonsillectomy and coronary bypass surgery for mild angina. “Well-defined scientific norms,” Wennberg writes, “simply do not exist to limit the practice options physicians select to treat these maladies.”

A major reason for overtreatment of all kinds is the natural optimism that physicians bring to their work and that patients themselves no doubt encourage. Who wants a pessimist for a doctor? But medical optimism takes peculiar forms that have long been recognized by social scientists who have studied the medical profession. In his study of doctors’ behavior, the sociologist Eliot Freidson writes that since the physician “believes that the work he does is all for the good of the client, [he] typically assumes that it is better to impute disease than to deny it and risk overlooking or missing it.”¹¹ In the surgical locker room, one often hears quips reflecting this view: “A chance to cut is a chance to cure”; “Every patient has at least three surgical diseases—all you have to do is find them”; “When in doubt, take it out.”

Such professional optimism was the subject of quantitative surveys made by Professors Frederick Mosteller of Harvard and Thomas Chalmers of Mount Sinai Hospital in New York and their colleagues. They showed that the enthusiasm with which a new surgical procedure is reported in the medical literature is often inversely related to the quality of the experimental design of the study by which it is evaluated. For example, Chalmers traced the history of the “portacaval shunt” operation, which is designed to decompress the portal vein and thereby to prevent life-threatening bleeding from the esophagus in patients suffering from cirrhosis of the liver. Chalmers and his associates reviewed thirty-two evaluation of portacaval shunt operations in which there was no control group, and six evaluations in which the results of the operation were compared, on a random basis, to those of conventional, nonsurgical treatment.

A review of the thirty-two reports of cases in which there was no control group showed that the operation was considered a marked success by the investigators in twenty-four reports, a moderate success in seven, and unsuccessful in one. By contrast, in the six studies in which the results of the operation were randomly compared to those of conventional nonsurgical treatment, the operation was judged unsuccessful in three cases, a moderate success in three, and a marked success in none.¹² Most medical procedures, in fact, have not been subjected to rigorous, well-controlled clinical trials. From the research of Chalmers and Mosteller, we can anticipate that many new procedures, perhaps most, may be used excessively.

Of course the enthusiasm of doctors for medical treatment is by no means limited to medical innovations. In 1974, Byron Brown, a professor of biostatistics at Stanford University, and I conducted a survey of how physicians and their wives behaved when they were themselves surgical patients. Their surgical histories were compared to those of lawyers, ministers, and graduates of Stanford Business School, along with their wives. The study was carried out in parts of California with high per capita income and high ratios of doctors and hospital beds to the local population. What, we had originally asked ourselves, would be the effects on demand for medical care if consumers were extremely well informed about medicine and were under no financial constraint? Of all patients, we reasoned, doctors would be the best informed; and if doctors or their families are not insured, doctors usually do not charge them fees. We therefore wondered whether doctors would undergo fewer elective operations than the other three professional groups. We found just the opposite. Physicians and their wives had operations just as frequently as other groups, in some cases more frequently.

Indeed hysterectomies were performed on physicians’ wives at such high rates that half of the women in our sample group will have lost their uteruses by age sixty-five. We concluded that the physician, as a patient, places a high value on surgical care, and that doctors generally believe in the procedures that surgeons provide.¹³

The overconfidence of doctors is only part of the problem. If medical costs are to be controlled and the enormous waste in overtreatment eliminated, much will depend on changing the attitudes of patients, who at present are remarkably passive in their relations with doctors. The patient’s legal right to all the information necessary to make decisions about his health is now well established,¹⁴ but in fact relatively few ask for such information, even though the patient should have it if he is to give “informed consent” before undergoing a recommended medical procedure. More important than getting information, however, is the way patients use this information, and whether they are willing to become actively involved in the medical decisions that affect their lives.

We do not know how the volume of medical care would be affected if patients were better informed about the risks of the procedures their physicians recommend, and particularly if they insisted on having a larger part in the decisions affecting them. But we have reason to believe the values patients place on the benefits of treatment may be quite different from those of doctors. According to limited experimental evidence, when it comes to taking risks in medical care, patients behave much as they do in other situations; and psychological research on decision making strongly suggests that most people are willing to take serious risks when they expect losses but not when they expect gains. Given an offer of a 50-50 chance of winning or losing a modest sum of money, say $500, most of us will decline. We will, however, accept the same odds when faced with the prospect of losing that amount: when told we are about to lose $500 we will probably risk $1000 for a 50-50 chance of losing nothing.¹⁵

Similarly, although the available evidence is far from conclusive,¹⁶ it is reasonable to assume that most people are glad to accept risks involving surgery in order to avoid the large losses of imminent life-threatening disease. For elective procedures involving risks and benefits that appear to be evenly balanced, on the other hand, we can conjecture that they will, if carefully informed, tend to avoid risks.

Thus almost all of us will accept the substantial risks of emergency surgery to stop internal bleeding. The anesthetic or operation could kill us, but the bleeding is more likely to do so. Most of us will also without hesitation accept surgery to remove a cancerous organ, even if the chances of cure may be small. However, if we are faced with the prospect of gallbladder removal for gall-stones that have produced no symptoms in the past, and if we are fully informed about the probabilities of success or failure, many, perhaps most, of us would choose instead to wait to see whether an operation would eventually become necessary.

Most surgery, perhaps as much as 85 to 90 percent, is for such purposes as relief of discomfort, disability, or disfigurement, and does not deal with conditions that threaten life. If, as I have suggested, much of contemporary medicine consists of elective interventions involving evenly balanced risks and benefits, most patients; if they were fully informed about the risks of each treatment, would decline some interventions. Perhaps many patients would decline many of them. The volume of surgical procedures might therefore decline substantially. And so might mortality, since all surgery carries some risk of death and thus may shorten the life expectancy of the local population from which the patients are drawn. In 1976, for example, Los Angeles doctors took part in a slowdown for five weeks, during which emergency surgery, but not elective surgery, was performed. The total population mortality for Los Angeles County fell by approximately one-third and rose again soon after elective surgery was resumed.

Overtreatment is only one of many medical issues with which citizens generally should be concerned. So far, however, few groups have emerged to represent the public interest in state and local decisions concerning health policy. Among the groups that have tried to do so, however, several have been impressive. The Oregon Health Decisions Group, for example, which has several thousand members, has taken strong positions challenging current medical standards and practices.

In a report released in January, the group recommends that doctors who grant the wishes of patients to forgo lifesustaining care be protected from prosecution. It also recommends programs for health education, increased taxes to penalize unhealthy behavior such as smoking, and the broadening of medical insurance coverage to improve medical services for the needy. I was particularly interested to see that the group urges that their state medical and hospital associations, together with the insurance industry, establish explicit “guidelines for the prudent introduction of, and reimbursement for, new procedures or technology.” To do this such agencies will certainly need better information on the effectiveness of such new technologies than that currently available.

Organizations representing doctors have in the past generally opposed the gathering of information that might be used by government agencies to regulate medical practice. If rationing of medical care is to be avoided, physicians will have to recognize the need to document the effectiveness of the procedures they recommend. They must also overcome their reluctance to give up the almost complete authority over the patient they have traditionally assumed. No one doubts that doctors should be able to give orders in emergencies when decisions involving life or death must be made. The problem is that authoritarian relations between doctor and patient were taken for granted when medicine typically consisted either of such emergency procedures or of modest efforts to relieve symptoms of disease.

Now that so much current medical care is discretionary, as well as of unmeasured effectiveness, physicians should reconsider their way of treating their patients and the extent to which patients should participate in medical decisions. For the most part, doctors now try to provide whatever care they consider possibly beneficial and to refrain only from using procedures that will do certain harm. My own view, as a doctor and a sometime patient, is that at least for most conditions that do not threaten life, a more appropriate general rule would be to provide care only when careful studies have shown that it leads, with high probability, to relief from symptoms of disease or to cure. It is of course an impossible requirement of a medical treatment that it never be accompanied by risks of any kind, whether they be pain or inconvenience or disability—or, in some cases, death. But if we act on the best information available to us, and if patients are encouraged to share in medical decisions, we might be able to provide all, or almost all, care that is likely to be effective in this qualified sense. As a result, Americans would receive far better care at less cost.

But what of the widely advertised threat of malpractice suits that are brought when a physician allegedly fails to do everything that might help a sick patient? Such suits may be valid in lifethreatening situations, when patients expect everything to be done to save them and are willing to accept the accompanying risks. For purely discretionary surgery, on the other hand, when the potential benefits may be relatively small, patients are less apt to accept risks, even small ones. Indeed, it is in exactly these situations, especially if the physician allows the patient to harbor exaggerated expectations, that a disappointing result, or even a minor complication, is more likely to lead to a lawsuit. Thus, a patient undergoing surgery for cancer may suffer a wound infection requiring an extra week in the hospital but, glad to be alive, will not file suit. A thirty-nine-year-old woman undergoing elective hysterectomy for minor symptoms who suffers the same complication will probably be much less forgiving, and often will sue the responsible doctor.

There is, then, a clear need for basing medical decisions upon much more reliable evidence. That need will not be met, however, unless the government provides greater support for assessing medical procedures and technologies, much as it now so effectively promotes basic medical research. A National Center for Health Care Technology was established in 1978 for this purpose but was allowed to expire for lack of funding. Congress has recently shown renewed interest in evaluating medical treatment and in 1984 passed Public Law 98-551 which expanded the responsibilities of the National Center for Health Services Research to include research on health care technology. This was signed into law by the President last December. It is in the enlightened self-interest of the government, as well as of the public and the medical profession, that this time the effort to ascertain the true benefits of medicine not fail.