For years, it’s been generally accepted that more than $100 billion dollars is spent every year on futile treatments for people nearing death. The number frequently cited to make this point certainly makes that inference plausible: 25 percent of Medicare spending is on people in their last year of life. That percentage has held steady for decades. However, the oft-cited notion that resources are perceived as “wasted” on patients who are soon to die turns out to be much more complicated than previously thought. And the data shows it.

An analysis of nearly 6 million patients led a team of researchers to these surprising conclusions, which they describe in a recent paper published in Science. They found that the overwhelming majority (95 percent) of Medicare spending is actually on patients who were expected to survive a year or longer. In other words, the notion that most of the spending is on people who were identified as unlikely to benefit turns out to be wildly wrong.

Another, more remarkable finding also came from the research that may make some people uncomfortable. However, it is precisely the kind of information that we need to seriously consider if we are ever going to get to the root of the health care spending debate. Based on the data, the research team concluded that physicians are just not very adept at predicting death. Even among the patients who were considered likely to die within one year, nearly half lived longer.

These findings should provide an important dose of humility and cause us to shift focus to what we actually know. If we are unable to accurately predict which patients are going to die soon, then its unwise to make health care spending decisions based on the tenuous distinction between patients who are “sick” those who are “near-death.” Instead, we should limit our focus to figuring how to optimize care for sick patients, irrespective of their duration of life.

Another important consideration that the study didn’t account for was quality of life. For those patients who were predicted to die soon but lived longer than expected, what was the quality of that “extra” time? Would a person want to live longer if it meant being hooked up to machines and confined to a nursing home? Could treatment slow the spread of cancer long enough for a parent to see a child’s graduation or marriage?

These are difficult questions and hard to assess, and the answers may differ from one person to another. But the only way that we can begin to reduce overspending on health care is if we base decisions on whatever the evidence indicates – even if it goes against conventional wisdom or upends long-held beliefs. And this research does not just have implications for care decisions; it is representative of the bold approach that is necessary in confronting the very practices and misconceptions that have caused our health care spending issues in the first place.