The following originally appeared on The Upshot (copyright 2016, The New York Times Company).
When I started out as a doctor in 1999, the Institute of Medicine published a blockbuster report that declared that up to 98,000 people were dying in United States hospitals each year as a result of preventable medical errors. Just a few months ago, a study in the BMJ declared that number has now risen to more than 250,000, making preventable medical errors in hospitals the third-largest cause of death in the country in 2013.
Those numbers warrant some further reflection. Although medical errors should concern us all, these statistics are more controversial than you might think.
After the original report, some future colleagues of mine at Indiana University wrote an article in The Journal of the American Medical Association rebutting the numbers being thrown around. First, they pointed out that many of those 98,000 were very sick; they couldn’t be compared with the general public, but needed to be compared with other sick, hospitalized patients.
They focused on the study from which the 98,000 was extrapolated. It involved an observational analysis of 7,743 “high-severity” patients in a New York hospital admissions database, which found that 13.6 percent had died, at least in part, because of an adverse event.
But this didn’t account for the baseline rate of death. Using New York State Health Department data, and applying the calculated death rate for in-hospital acute care admissions, they found that about 13.8 percent of patients in the “high-severity” group should have been expected to die over all.
This means that the death rate in the group with medical errors was probably similar to the death rate in a group without medical errors, casting doubt on those errors as being the cause of death. The authors of the original study disagreed.
In May of this year, researchers looked at studies published since the 1999 report and extrapolated to 2013. Their conclusion: 251,454 hospitalized patients died from medical errors each year. They further stated that they believed this was an underestimate. Things apparently have only gotten worse.
Once again, others disagree. Some of these disagreements are compelling. For instance, there are about 2.5 million deaths each year in the United States, about 700,000 of which are hospitalized patients. This means that medical errors — in hospitals — would have to account for up to 10 percent of all deaths, or up to more than a third of hospitalized patients. That’s hard to fathom.
Doing research in this area is very difficult, for a number of reasons. When someone dies in a car accident, it’s clear what caused the death. Same for a drug overdose, homicide or suicide. But when an 86-year-old with dementia and cancer dies and also had been given a drug in a slightly-too-high dose a few weeks earlier, is it the error that killed her or the underlying disease and age?
The researchers who engage in this type of work do their absolute best to tease apart these factors. I’m not questioning their intent or their acumen. But even with the best methods available, it’s very, very hard to prove what events cause death and which are correlated with it.
A study published in 2001 asked physicians to review cases of deaths and then rate not only whether they thought a preventable error might have contributed to the end result, but also how likely death might have been in the absence of an error. As with other studies, they found that almost a quarter of hospital deaths might have been at least “partially preventable.” Only 6 percent of deaths were “probably preventable.” But the “inter-rater reliability,” or the amount that reviewers agreed on each case, was low. It’s hard to parse this determination, and even experts disagree on each case.
They also noted that after considering the three-month prognosis and adjusting for the variability of ratings, only 0.5 percent of patients who died would probably have lived at least three months more in good health if care had been optimal. That’s far, far fewer than the numbers cited.
A similar study (in Britain) from 2012 found that while 5 percent of deaths in hospitals may have a more than 50 percent chance of being preventable (lower than these recent studies), more than half occurred with older, sicker patients who were thought to have less than one year of life left to live.
This isn’t to say that these error-associated deaths aren’t sad or meaningful. They are. I can’t say that enough. But the potential harms of hospitals have to be weighed against the potential benefits. People who think that going to a hospital, or even a doctor’s office, carries no risk are fooling themselves. Beyond mistakes that occur, you’re exposing yourself to risk just by being around many sick patients. Further, the sickest patients are likely to have more medical interventions, and therefore more opportunities to have a preventable error occur.
But it’s somewhat sensationalistic to keep coming up with increasing numbers. I’m not sure it’s doing much good. After the publication of the initial report, defenders of the 98,000 number argued that even if the numbers were wrong, bringing attention to this problem would be good in itself.
Unfortunately, research doesn’t necessarily back that up. A 2010 study in The New England Journal of Medicine followed 10 North Carolina hospitals in the 10 years after the Institute of Medicine report. They found that the overall rate of harms, and the rate of preventable harms, did not significantly improve over that period.
Was this because we didn’t pay enough attention in 1999? Will the now much larger numbers force us to pay attention? Maybe. But it’s also possible that this tight focus is a distraction. After all, the vast majority of preventable medical errors don’t result in death. They result in other harms, which are probably more pervasive and might have a bigger overall impact. They also occur outside the hospital setting.
It makes headlines to say that medical errors in hospitals kill more people than guns or cars. But that’s debatable. Moreover, our continued focus on this number — and the hospital setting — may be draining resources and attention from more effective harm reduction.