Shocking News

Much has been written lately about over-treatment of older patients. Only rarely does anyone suggest that older patients are getting too little treatment, but a new study in JAMA does just that. The reality isn't quite so clear.

The treatment is the implantable cardioverter defibrillator (ICD) and the patients are people over the age of 65 who have had a heart attack and are found afterwards to have a weak heart (defined as an ejection fraction less or equal to 35%). These patients are at risk of sudden death, of an irregular heart rhythm such as ventricular tachycardia, and the ICD is designed to deliver an electric shock if that happens, effectively bringing the patients back from death. By looking at the National Cardiovascular Data Registry, which keeps track of heart attack patients, the authors of the article found that only 8.1% of “eligible” patients actually received an ICD. As a result, they claim, the 92% of patients who didn’t get an ICD were more likely to die than their counterparts who did.

This is a surprising finding in light of the persuasive and cogent argument made by Sharon Kaufman in her recent book, Ordinary Medicine: Extraordinary Treatments, Longer Lives, and Where To Draw The Line. Kaufman makes the case that many high tech treatments come to be seen by physicians and patients as normal and necessary once Medicare agrees to pay for them. The end result for many marginally beneficial, burdensome, and expensive treatments, including the ICD, is that patients just can’t say no. If that's true, why are so few older people getting an ICD? 

Now it wouldn't be the first time that ageism or misinformation prevented older people from getting beneficial treatment. Many years ago, patients who were over a certain age were precluded from receiving clot-busting drugs (thrombolytic therapy) because it was widely assumed that in older age groups, the risks outweighed the benefits. It turned out that clot-busting drugs were actually more beneficial in older patients, basically because their heart disease tended to be severe which meant they stood to gain a great deal from treatment. Elevated systolic blood pressure was likewise once assumed to be normal in the geriatric population, or even desirable in order to improve blood flow to the brain. Studies eventually showed that elevated systolic blood pressure, even in older patients, predisposed to stroke and other unfortunate outcomes, and warranted treatment—though the recommendations about just how much blood pressure should be lowered have evolved over time. Is the ICD implantation rate just another case of bias or ignorance at work?

Dr. Robert Hauser of the Minnesota Heart Institute, writing an editorial published alongside this article, blames our fragmented health care system. He speculates that primary care physicians may not realize that their patients were supposed to get an ICD. The fact that there's supposed to be a 40-day waiting period between the onset of the heart attack and implantation of the ICD contributes to the problem. Hauser suggests that the primary care physician is so frazzled and overburdened that he is apt to neglect to send his patient to a cardiologist. Is this the explanation?

It can’t be the whole story. While patients who saw a cardiologist after hospital discharge were more likely to wind up with an ICD than patients who didn’t, only 30% of the patients who saw a cardiologist had an ICD implanted. Recall that 100% of them were, technically speaking, “candidates” for an ICD. So what else is going on?

Hauser hints at another explanation: “It is possible that some older patients may refuse ICD treatment for personal reasons or because comorbidities such as endstage kidney disease or advanced frailty were considered in the decision regarding ICD implantation.” He doesn't accept this explanation as sufficient, rightly recognizing that patients are very likely to accept whatever technological intervention their physician recommends and that shared decision-making, if it takes place at all, is apt to reflect the physician’s preferences as well as the patient’s. So the problem, if it is a problem, must lie with doctors, too. Physicians are not systematically and emphatically recommending ICD implantation to their older patients. Even the most technologically sophisticated academic medical centers only implanted ICDs in 16% of their eligible older patients. But is this a problem that needs fixing, like under-treatment of heart attacks with clot busters and inadequate treatment of high blood pressure in the past?

Dr.  Hauser believes it is, saying “even though the use of ICD for primary prevention may not seem to make as much sense for an 80 year old patient as it does for a patient in his 50s or 60s, an older patient at risk for sudden cardiac death should have the same opportunity to choose potentially lifesaving therapy.” But the benefits of ICD in those over 80 are far from clear. The studies include very few people in this age group. What data there is indicates that there is little if any survival benefit. Moreover, ICDs implanted in older people fire erroneously half the time. That means they deliver a very unpleasant electric shock to the hapless patient. In addition, if the ICD does work as intended, what that means is the abolition of sudden death. 

Maybe, just maybe, the low rate of ICD implantation in older people is a refreshing instance of massive civil disobedience—of both patients and doctors refusing to abide by prevailing clinical guidelines. We all have to die of something. An ICD virtually guarantees that the something will involve a protracted period of decline and suffering. If you had to choose between cancer, Alzheimer’s disease, and sudden death, which would you pick?

Putting the Genie Back in the Bottle

You can’t put the toothpaste back in the tube. You can’t put the genie back in the bottle. You can’t turn back the clock. We have lots of expressions denoting the irreversibility of certain actions. Getting patients to stop taking a particular medication or to abandon a procedure—and persuading doctors to stop prescribing it—seems to be in the same camp. Unless something comes along that’s much better, or at least that the pharmaceutical or device manufacturing company says is much better, we are extremely reluctant to change. A series of articles in JAMA Internal Medicine

discuss this phenomenon, calling it “the undiffusion of established practices.”

One article reports that physicians gradually adopted the practice of tight control of blood sugars in the ICU after a single study showed that patients did better when their blood sugars were as close to normal as possible, but then abruptly gave up the practice after a later study drew the opposite conclusion. A second article describes how physicians jumped on the bandwagon of ordering noninvasive  screening for coronary disease in patients scheduled for non-cardiac surgery, only to discover later that the screening was not beneficial—and the routine use of such testing decreased. A third article discloses that professional medical societies tend to continue to support practices that have been challenged or even debunked if their members believe the practices to be useful.

What does all this tell us about “undiffusion?”  How do we change behavior? One answer is that we should move slowly to adopt something new, whether a medication, a procedure, or a process. We should exercise caution, given the reality that the best new thing often turns out not to be the best. A second answer is to design studies that are better at looking for subgroups that may benefit from a new approach—even if the approach doesn’t make sense for everyone. But I think there’s a lot more to be said about putting the genie back in the bottle, particularly as we begin to recognize that less may be more.

It’s instructive to look at the cases where the genie did go back in the bottle and compare them to the cases in which he didn’t. Consider, for example, the process of tightly controlling blood sugars in the ICU, a practice that physicians readily abandoned. What’s striking about this is that controlling blood sugars very tightly is a nuisance—it means checking the blood sugar four times a day and adjusting insulin doses every time. That’s labor intensive—someone has to do the tests and someone else has to respond to the results. And there’s nothing sexy about pricking a patient’s finger to get a drop of blood. It’s not nearly as exciting as seeing the patient’s organs in color on a CT-scan image—a test that was not so readily abandoned. Finally, there’s no drug company or device manufacturer that stands to gain from widespread use of fingersticks, though there is from diffusion of 128-slice CT scanners, not to mention 256-slice machines.

What about the recent example in which use of antidepressants in young people decreased after the FDA issued a black box warning about potential side effects? Some critics argue that in this instance, putting the toothpaste back in the tube was wrong-headed: many more people suffered from not being treated than from the side-effects of treatment. Whether or not it was a good idea to cut back on antidepressant use, the major driver of the behavioral change appears to have been the media. Newspapers reported that antidepressants were bad; they were dangerous; they killed people. They presented a far less nuanced picture than the original article. But the combination of publicity and fear-mongering was effective and antidepressant prescribing rates fell significantly.

Finally, the most dramatic, if obvious, examples of turning back the clock occur when the FDA orders a drug company to recall its drug. When Vioxx was taken off the market, after revelations that it significantly increased the risk of heart attacks, use of Vioxx plummeted. In a similar vein, when health insurance companies decide not to pay for something, use generally falls. Bone marrow transplantation for the treatment of metastatic breast cancer ceased when, based on revelations that an earlier study had been fraudulent, coverage was halted.

Going back is hard. Commenting on the articles in JAMA Internal Medicine, Frank Davidoff quotes David Hume: “Men generally fix their affections more on what they are possessed of, than on what they never enjoyed. For this reason, it would be greater cruelty to dispossess a man of any thing, than not to give it to him.” Likewise, research in cognitive psychology confirms that people are far more distressed by what they perceive to be losses than they are ecstatic about what they see as gains. So reversing course is likely to require a multi-pronged approach. We need to have more solid evidence to adopt change in the first place and we should be looking more carefully at subgroups of the population when we draw our initial inferences about efficacy and we need more measured and judicious reporting by the news media and we need to use the nuclear option—removing ineffective or harmful interventions from the market or from third party coverage. Only then will we gain control over the phenomenon of persistent use of established but suboptimal treatment.

In his 2005 book, On the Take, former New England Journal of Medicine editor Jerome Kassirer inveighed against the influence that drug and device companies exerted over physicians. Through a variety of strategies, ranging from subsidizing elaborate meals at medical conferences to lucrative consulting arrangements to all-expenses-paid travel to exotic resorts in exchange for product endorsement, corporate America affected and sometimes determined physician behavior. In response, many academic medical centers have banned pharmaceutical sales reps, the “detail men” of yesteryear and put sometimes stringent conflict of interest policies into effect. The Sunshine provision of the Affordable Care Act requires drug companies and device manufacturers to reveal how much money they gave to doctors and to hospitals and to whom they gave it. As of September, 2014 this information is publicly available on a special website maintained by the Centers for Medicare and Medicaid Services. Is this provision having an effect? Will it?

An article written by a health economist and a health lawyer predicted the law would be unlikely to have a direct effect on either patients (by dissuading them from seeing particular physicians) or on physicians (by shaming them into severing their relationship with drug or device manufacturers). It suggested there might nonetheless be a beneficial effect if what the authors called “learned intermediaries,” such as health insurance companies, begin discriminating against people or institutions that are the recipients of what they regard as excessive largesse.

Six states had already introduced legislation requiring disclosure of payments in the form of gifts, food, travel and fees long before the Sunshine Act was passed, and the results in those states give us a preview of what is to come. One study of the Massachusetts law, which went into effect in 2009, concluded that the disclosure requirements were associated with a fall in the total volume of prescriptions, a decrease in the use of brand name drugs, and a rise in the use of generic drugs. But it’s hard to tease out how much of the change was due to the law and how much due to other simultaneous changes in the culture of medicine.

It’s too soon to draw conclusions about the federal law, but some preliminary observations are raising red flags. According to an investigative journalism piece just published in the New York Times, the information available on OpenPayments, the CMS database, is incomplete and replete with errors. Some companies, whether intentionally or accidentally, misspell the names of their drugs and devices or use different names for the same drug in different sections of the database. All told, 40% of the records in the database are missing the names of the doctor or the hospital that received the payment. And indeed, when I checked this out by looking up a local academic hospital, I found that if I entered "Beth Israel Deaconess Hospital," the report said there were no conflicts entered for this facility, but when I entered "Beth Israel Deaconess Medical Center," I found dozens of them.

Even if the reporting were accurate, the real question is whether transparency alone will have a significant effect. I’m doubtful. For some time, medical journals have had disclosure requirements for articles. The result is that the reader knows about consulting relationships, advisory board memberships, and other financial ties authors have with their sponsors. Perhaps the reader will take more seriously those papers in which the authors have no disclosures to make than those where they have a litany of them. But in either case, the article has passed muster: it has made it past the gauntlet of peer review, it has the journal’s stamp of approval. Does anyone seriously discredit a paper because of the author’s pedigree?

Does any of this matter? In particular, what does it have to do with medical care for older people? A great deal. Older patients comprise a huge part of the market for most drugs and most devices.  The average 65-year-old takes 4 prescription medications a day. The average nursing home resident takes closer to 9. In 2015, Medicare spending on Part D, the prescription drug benefit, is expected to account for 14% of total Medicare expenditures in 2015, or $76 billion.

And while the top ten prescription drugs dispensed to older people include such inexpensive generics as hydrochlorothiazine and atorvastatin, many ultra-expensive specialty drugs are used principally in older people. Some of these drugs are one-of-a-kind; many are life-saving. This is where the concern about undue influence of the manufacturer comes in. Precisely because older people take so many medications (and use so many devices) and precisely because they amount to such a big chunk of health care resources, we need to be concerned with what drugs doctors prescribe. Public disclosure of gifts from PhRMA is unlikely to make much of a difference in the all important power of the pen. 

If we want to assure the integrity of the prescribing process--and this is critical because of the number and cost of the medicines taken by older people--we need to prohibit certain kinds of corporate/physician interactions, and to give such bans teeth.

A Matter of Opinion

In an op-ed in the New York Times this week, my fellow Harvard physicians Pamela Hartzband and Jerome Groopman argue that it’s pernicious to reward physicians for meeting quality standards. These measures of performance, they say, are “population-based and generic, and do not take into account the individual characteristics and preferences of the patient or differing expert opinions on optimal practice.” Really? Are there no practices that physicians should routinely follow if they are to practice good medicine?

The answer revolves around the concept of “preference sensitive care,” articulated by Jack Wennberg and Elliott Fisher of the Dartmouth Center for the Evaluative Clinical Sciences. Preference-sensitive care involves treatments that involve significant tradeoffs affecting the quality and or length of life. It is contrasted with effective care, which involves essentially uncontroversial treatments that are unambiguously optimal, or for which there are no viable alternatives. What Hartzband and Groopman are arguing is that essentially all care is preference-sensitive and there is no such thing as effective care. Therefore, they conclude, there is no way that we can legitimately grade treatment as inherently good or bad—it either reflects patients’ wishes—or physicians’ recommendations—or it doesn’t.

The view that there are no medical practices that constitute “effective care” and that all medical care is preference-sensitive strikes me as profoundly misguided. In 2003, RAND evaluated the quality of medical care in the US and found that only 55% of adult patients surveyed by telephone had received recommended medical treatment. These treatments included counseling on smoking cessation to patients who had had heart attacks and long-acting inhalers for people with chronic asthma. Did the 45% of patients who did not get standard treatment actually receive care that, while not consistent with expert recommendations, conformed well to their preferences? Could 45% of physicians have deliberately and thoughtfully chosen a course that deviated from expert recommendations because they had unique insight about what was best for their patients? Is there a plausible reason why physicians should not advise cardiac patients to stop smoking or patients with at least moderately severe asthma to use inhalers? I doubt it.

What about older people? Perhaps this is one population where most treatment is preference-sensitive. After all, as I have often argued in this blog, what makes sense for people who are frail, dying, or have dementia depends heavily on their goals of care. Even a seemingly clear-cut decision such as whether to have surgery for a broken hip might have a values component. A patient who is dying, for example, and whose only goal of care is his comfort, might choose to have bed rest and pain medication rather than the rigors of surgery. So are there quality indicators, those “metrics” that Hartzband and Groopman vilify, that apply to the geriatric population?

I looked at the most recent HEDIS (Healthcare Effectiveness Data and Information Set) measures, the set of indicators used by health insurance companies throughout the US to evaluate physician performance. These were designed by the NCQA (National Committee for Quality Assurance), a non-profit, non-governmental, independent body that relies on expert judgment and public feedback to determine quality indicators. It turns out that there are 5 HEDIS measures for older adults, addressing physical activity, pneumococcal vaccination, osteoporosis testing, medication management, and fall risk management.

The 5 domains were chosen for good reason. Physical activity plays a role in maintaining physical function as well as in staving off depression, cardiovascular disease, and other chronic illnesses common in old age.

Pneumonia is very common in the elderly and it’s a killer. When it doesn’t kill, it causes shortness of breath and confusion, often resulting in hospitalization, and may trigger other conditions such as a heart attack. Osteoporosis is responsible for tremendous suffering, mainly in the form of pain and impaired mobility, and predisposes to fractures. Falls, like osteoporosis, are the cause of an enormous amount of misery in old age. Nobody, and I mean nobody, wants to fall, just as nobody wants to have osteoporosis. Medications are commonly used by older people—on average, they take about 5 prescription drugs—and the more medications a person takes, the greater the likelihood of side effects.

If the targets are reasonable, are the specific quality indicators themselves? In general, the answer is yes. The pneumonia metric, for example, looks at the percentage of adults 65 and older who report ever having received a pneumococcal vaccination. The only people for whom getting a pneumonia shot doesn’t make much sense are those who are dying. And even patients who are dying presumably weren’t dying a year or two years ago, at which point they should have gotten immunized against pneumonia. Since the measure asks if a person has ever been vaccinated, the only older patients for whom this is inappropriate are those who develop a lethal illness at age 65 or those who regard life as worse than death (usually also people with a fatal illness) and who hope they will get pneumonia. There will always be people who refuse the shot out of a misguided belief that it will cause them harm. Some people will have religious objections—though I have yet to find an established religion that forbids vaccination (the exception is Christian Scientists but they aren’t likely to go to doctors in the first place). So physicians will rarely achieve a score of 100% on the vaccination metric. But the question is not whether vaccination can always be achieved; it’s whether vaccinating against pneumonia constitutes good medical practice.

It would be nice to have separate measures for patients  who are near the end of life, for those who meet the criteria for enrollment in hospice (whether they have chosen to enroll or not). At the very least, for patients with a limited life expectancy, the physician should be able to indicate that the quality indicators are “not applicable.” But apart from this concern, the quality measures strike me, on balance, as entirely reasonable.

According to a report just released by the NCQA, 54% of the US population (171 million people) now have health insurance from a plan that uses HEDIS measures. Over the last 3-5 years alone, 46% of the HEDIS measures have shown consistent, statistically significant performance gains. Surely this is progress. Some treatment decisions are preference-sensitive and for these cases, we should find a way to determine if physicians are eliciting their patients’ preferences and treating them accordingly. For those treatments that constitute “effective care,” the quality indicators make good sense.