Yes, There Is Low-Value Care in the Very Ill, Despite MIT Study

John M. Mandrola, MD; Vinay Prasad, MD, MPH


July 13, 2018

It's the worst: performing procedures or delivering chemotherapy to people in their last months of life. We hate doing this, because it's neither kind nor wise.

Hematologist-oncologist Vinay Prasad

Scrutiny of low-value care at the end of life went mainstream when Atul Gawande, MD, writing in the New Yorker,[1] made popular the finding that 25% of Medicare expenditures goes to beneficiaries in their last year of life.[2]

The notion of wasteful spending at end of life is easy to accept. Anyone who rounds in hospitals in the United States bears witness to a crisis of common sense and humanity. Beneficent intent, death denial, and fee-for-service incentives combine to induce a proliferation of folly in sick people nearing end of life: high-risk procedures, hail-Mary chemotherapy, dialysis in those with multiorgan failure, and perhaps the silliest of them all—sending frail elders to nursing homes to "get stronger," with the promise that they can then return to receive aggressive care.

Cardiologist John Mandrola

Everyone wants to improve this crisis. The problem comes when people say that doing so could save money. That raises the disquieting topic of rationing care and thinking that doctors can predict who will live or die. Many people cite anecdotes of an elder saved from death by an aggressive—and often expensive—treatment. The message to those who seek to lessen low-value aggressive care at end of life is clear: tread lightly, beware of hubris.

Now, provocative findings[3] from researchers at MIT have further complicated the idea of "right care" at end of life. Using millions of patients in a Medicare database and machine-learning methods, the authors report that although spending on patients who die is high, it is nearly impossible to predict which patients will die before treatment begins.

That observation is provocative because it tempts some to say: See...since you can't know who will die before therapy begins, you can't say anyone is wasting resources on low-value care.

We disagree. This is a nice paper, but it does not refute the idea that low-value care is commonplace in the care of medically complex people at end of life. Furthermore, we worry that misunderstanding this study could foster even more low-value aggressive care in patients at end of life.

How They Did It

The researchers begin by pointing out that those who end up dying are not the same as those who were sure to die. They note that previous studies chronicling high spending in the last year of life look backward (ex post). The high spending on those who died could simply have been because sicker people consume more healthcare.

The MIT group set out to model spending before death (ex ante). Their machine-learning techniques generated a prediction of each individual's chance of death over the next 12 months. They inputted baseline variables available in the Medicare data set of almost 6 million people in their prediction model, including healthcare use in the months before the study began.

They then assessed spending at each level of predicted mortality. For instance, what was the spending on individuals with a 10% predicted rate of death versus those with a 50% predicted rate of death?

Key Findings

The authors found no sizable group of people for whom imminent death was certain. Most patients who died were lower-risk. In fact, 95% of patients had 1-year predicted mortality rates below 25%, and less than 10% of those who died, had a predicted mortality above 50%.

Patients with the highest predicted mortality (the sickest) accounted for only a small share of total spending. The sickest patients, those with predicted mortality above 46%, account for under 5% of total spending, and crucially, almost one half of these patients survived the year. The authors describe multiple methods of checking their prediction model, including use of data from the time of hospital admission; the results held up.

Spending on the sick explained a great deal of spending at end of life. Almost one half of spending on patients who died could be explained by the observation that spending is higher in those who have higher predicted mortality.


This paper does not show whether spending on those who live or die is wise, helpful, or harmful. It just says that we can't identify patients in whom we could stop delivering futile care because death is near. But as Farzad Mostashari, MD, wrote on Twitter: "Who is asking that question though? Has anyone proposed death panels that would cut off services to people likely to die? Seriously, anyone?" The answer is, no one. A few years ago, the BMJ clinical evidence team picked 3000 treatments used in the UK's National Health Service and found that about one half were of unknown effectiveness, and only 11% were clearly beneficial.[4] Much of the care of the very ill is based in eminence, not evidence. Scant few studies include elderly persons or those with cognitive decline, advanced kidney disease, or frailty.

The reason we think that healthcare spending at the end of life is often wasteful is not because of any retrospective data on the costs of those who die. It's because much of what is done to people, especially sick people, lacks empirical evidence or is based only on pathophysiology.

As doctors, we also feel the intangible pressures of today: Profit-driven healthcare incentivizes aggressive care. Mass marketing of healthcare feeds on the natural fear of death. Society has come to expect miracles. These forces align against common sense and compassion at end of life. Patients accept aggressive care; everyone in the system makes money delivering aggressive care. "Do everything," becomes the default.

The findings of this paper, the uncertainty of prediction, the spending on the sick, have reinforced our view that some patients live because of our care, some live despite it. Sadly, too, some die because of our care.

What this paper should do, and perhaps it will, is to stimulate further research to identify the drugs, procedures, and strategies that deliver benefit in medically complex patients.

For the practicing physician, the paper should change nothing. Every intervention that we use for patients with advanced or terminal illnesses should be applied judiciously. It should ideally be based on high-quality evidence and aligned with patient's goals and desires as part of a discussion of shared decision-making. And for patients in the intensive care unit, we should ask ourselves whether every single pill, procedure, or device is necessary.

What we fear most, though, is that this paper by the MIT researchers will be misinterpreted and misapplied. Some may use it to justify even more aggressive care in the elderly and sick. This would be tragic, because the core problem in end-of-life care isn't wasteful spending; it's inhumane care.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.