How Many Die From Medical Mistakes in U.S. Hospitals?


MEDICAL ERRORS (1)Note: This is being republished, with permission, under a Creative Commons license. See the original ProPublica page and comments.

by Marshall Allen ProPublica, Sep. 19, 2013, 10:03 Charles Ornstein and Lena Groeger, ProPublicaby Dan Nguyen, Charles Ornstein and Tracy Weber, ProPublica by Robin Fields, Al Shaw and Jennifer LaFleur, ProPublica

It seems that every time researchers estimate how often a medical mistake contributes to a hospital patient's death, the numbers come out worse.

In 1999, the Institute of Medicine published the famous “To Err Is Human” report, which dropped a bombshell on the medical community by reporting that up to 98,000 people a year die because of mistakes in hospitals. The number was initially disputed, but is now widely accepted by doctors and hospital officials — and quoted ubiquitously in the media.

In 2010, the Office of Inspector General for Health and Human Services said that bad hospital care contributed to the deaths of 180,000 patients in Medicare alone in a given year.

Now comes a study in the current issue of the Journal of Patient Safety that says the numbers may be much higher — between 210,000 and 440,000 patients each year who go to the hospital for care suffer some type of preventable harm that contributes to their death, the study says.

That would make medical errors the third-leading cause of death in America, behind heart disease, which is the first, and cancer, which is second.

The new estimates were developed by John T. James, a toxicologist at NASA‘s space center in Houston who runs an advocacy organization called Patient Safety America. James has also written a book about the death of his 19-year-old son after what James maintains was negligent hospital care.

Asked about the higher estimates, a spokesman for the American Hospital Association said the group has more confidence in the IOM's estimate of 98,000 deaths. ProPublica asked three prominent patient safety researchers to review James' study, however, and all said his methods and findings were credible.

What's the right number? Nobody knows for sure. There's never been an actual count of how many patients experience preventable harm. So we're left with approximations, which are imperfect in part because of inaccuracies in medical records and the reluctance of some providers to report mistakes.

Patient safety experts say measuring the problem is nonetheless important because estimates bring awareness and research dollars to a major public health problem that persists despite decades of improvement efforts.

“We need to get a sense of the magnitude of this,” James said in an interview.

James based his estimates on the findings of four recent studies that identified preventable harm suffered by patients – known as “adverse events” in the medical vernacular – using use a screening method called the Global Trigger Tool, which guides reviewers through medical records, searching for signs of infection, injury or error. Medical records flagged during the initial screening are reviewed by a doctor, who determines the extent of the harm.

In the four studies, which examined records of more than 4,200 patients hospitalized between 2002 and 2008, researchers found serious adverse events in as many as 21 percent of cases reviewed and rates of lethal adverse events as high as 1.4 percent of cases.

By combining the findings and extrapolating across 34 million hospitalizations in 2007, James concluded that preventable errors contribute to the deaths of 210,000 hospital patients annually.

That is the baseline. The actual number more than doubles, James reasoned, because the trigger tool doesn't catch errors in which treatment should have been provided but wasn't, because it's known that medical records are missing some evidence of harm, and because diagnostic errors aren't captured.

An estimate of 440,000 deaths from care in hospitals “is roughly one-sixth of all deaths that occur in the United States each year,” James wrote in his study. He also cited other research that's shown hospital reporting systems and peer-review capture only a fraction of patient harm or negligent care.

“Perhaps it is time for a national patient bill of rights for hospitalized patients,” James wrote. “All evidence points to the need for much more patient involvement in identifying harmful events and participating in rigorous follow-up investigations to identify root causes.”

Dr. Lucian Leape, a Harvard pediatrician who is referred to the “father of patient safety,” was on the committee that wrote the “To Err Is Human” report. He told ProPublica that he has confidence in the four studies and the estimate by James.

Members of the Institute of Medicine committee knew at the time that their estimate of medical errors was low, he said. “It was based on a rather crude method compared to what we do now,” Leape said. Plus, medicine has become much more complex in recent decades, which leads to more mistakes, he said.

Dr. David Classen, one of the leading developers of the Global Trigger Tool, said the James study is a sound use of the tool and a “great contribution.” He said it's important to update the numbers from the “To Err Is Human” report because in addition to the obvious suffering, preventable harm leads to enormous financial costs.

Dr. Marty Makary, a surgeon at The Johns Hopkins Hospital whose book “Unaccountable” calls for greater transparency in health care, said the James estimate shows that eliminating medical errors must become a national priority. He said it's also important to increase the awareness of the potential of unintended consequences when doctors perform procedure and tests. The risk of harm needs to be factored into conversations with patients, he said.

Leape, Classen and Makary all said it's time to stop citing the 98,000 number.

Still, hospital association spokesman Akin Demehin said the group is sticking with the Institute of Medicine's estimate. Demehin said the IOM figure is based on a larger sampling of medical charts and that there's no consensus the Global Trigger Tool can be used to make a nationwide estimate. He said the tool is better suited for use in individual hospitals.

The AHA is not attempting to come up with its own estimate, Demehin said.

Dr. David Mayer, the vice president of quality and safety at Maryland-based MedStar Health, said people can make arguments about how many patient deaths are hastened by poor hospital care, but that's not really the point. All the estimates, even on the low end, expose a crisis, he said.

“Way too many people are being harmed by unintentional medical error,” Mayer said, “and it needs to be corrected.”

See how you can help ProPublica investigate patient safety and join our Facebook group on the topic.

daily newsletter

ProPublica is a Pulitzer Prize-winning investigative newsroom. Sign up for their newsletter.

What do you think? Please scroll down (or click) to post a comment. Or please share the post with your thoughts on LinkedIn – and follow me or connect with me there.

Did you like this post? Make sure you don't miss a post or podcast — Subscribe to get notified about posts via email daily or weekly.

Check out my latest book, The Mistakes That Make Us: Cultivating a Culture of Learning and Innovation:

Get New Posts Sent To You

Select list(s):
Previous articleNot the Safest Way to Win “Toast Kaizen?”
Next articleFree Webinar Tomorrow – #Lean & Daily Management by @KaiNexus


  1. This conversation always reminds me of the Aviation Safety Reporting System–a system sponsored by the FAA to allow voluntary, confidential reporting of near miss and close-call incidents in civil and commercial aviation. They go so far as to have NASA run the system, simply because NASA has no enforcement powers, which guarantees that the aircrew involved will not be punished for reporting incidents.
    Modelling this type of system in healthcare would go a long-way toward reducing these tragic numbers.


  2. This high number is hog wash. When people are very ill and are using more medical care than ever to stay alive, for instance dyalisis, blood transfusions, ventilators. When something goes wrong with this complex care or not quite as planned, the death is still a result of there underlying illness.

  3. Frank: I don’t believe it’s hog wash at all. These reviews take into account comorbidities and other conditions that make patient survivability a challenge to begin with. Having worked in hospitals for over 20 years and having seen cases where healthy individuals come in and end up dying during the course of their stay or are “treated and released” and die shortly thereafter, I can tell you that the numbers presented are probably fairly close to accurate (RCAs of those particular cases pointed to medical errors–not the patients’ conditions). You seem to overlook the fact that not all of those deaths are people who are coming in near the end of their life and require considerable medical care to stay alive (even in those instances, it doesn’t justify medical error that results in an unexpected or unanticipated death). Many are relatively healthy individuals, who are coming in for treatment that many thousands of others receive–without error–and live to talk about it.

  4. Having worked at two different hospitals with two different perspectives I think Frank and Robert could both be right. I would rather my family go to the hospital that doesn’t excuse errors away as a known side-effect, within reasonable care, or as complications of a life-saving effort. Hospitals have a long way to go towards transparency well beyond current reported measures.

  5. I wish these reports would provide some detail on what these really mean. Is it 250,000 people that were completely healthy coming in for a physical and they die when they would have been healthy if the medical system didn’t exist? I doubt it. Is it 2,000 completely healthy people and 248,000 people that were under intensive medical care for years keeping them alive and now we slipped up and they died? Probably not, again, but my guess is it is closer to the second.

    Preventing errors is obviously important. And in health care it is very important, of course. But just because you use data doesn’t mean it isn’t misleading. Medical errors leading to death is just too big an operational definition to be very meaningful in my opinion. For these numbers to provide much insight I really think they need to be segmented more:

    – perfectly health person that was going to be perfectly healthy for decades but were killed by medical error.
    – person that needed life saving care of they were going to die that month and we routinely should be able to provide the very easy care to make them perfectly healthy again but they were killed.


    – person that was extremely sick with many problems for years and was saved with medical care over and over again. Complex care was needed and much of it was done well but in the very challenging situation there was a mistake and that mistake is the proximate cause of death.

    We should be working on making everything better and eliminating medical errors that cause damage and death. But there are huge differences between a medical error conditions that caused death and to me those differences are so huge lumping them together is hardly useful.

    • Great question. Some of these numbers are very unknowable numbers. But we must try to improve anyway.

      I guess there is a difference between a preventable medical error that kills a 50-ish woman and robs 20 to 40 years of her life (the Louise Batz story) or an error that steals the life of a young child versus an error that kills a 90-year old who is ill already.

      I know you’re not suggesting that errors are acceptable or more acceptable in the ill and elderly.

      The broken processes and systems that potentially harm us all can be the improved processes that better protect us all.

      A recent email from a personal friend from elementary school:

      “My mom is [at a hospital in my hometown], and there have been several instances where they would have done things wrong if she didn’t tell them.

      The latest is that transport came to take her for an x-ray. She asked why and said she didn’t want to go unless she knew why.

      They asked the nurse, and she deleted the order. My mom asked why there was an order for an x-ray. They said it was an error.

      They also did a CT Scan for a possible blood clot in her lung. My mom said to the tech, “Oh, I didn’t know a CT scan could show a blood clot.” The tech said, “Oh, this is for a blood clot? The kind I was going to do wouldn’t show that. A different kind of CT is needed. She then confirmed with the doctor and did the kind that can detect a clot.”

      Those might not have been errors that would have killed her, but it doesn’t give one much confidence in our chaotic, overburdened, screwed up hospitals.

  6. Right, l think the thing I am getting at is there is a big difference between making a mistake while you are in a complex situation where any of 30 bad decisions in a pressure situation could result in death (and where doing nothing results in death) versus a situation where there is no risk of death until you do an absolutely idiotic thing that turns my visit to get a physical into death.

    Everything should be constantly improved and made safer with mistake-proofing thinking… And healthcare needs this more than most everything due to the dangers and consequences involved.

    When there are headlines like 100,000 deaths due to medical error every year that reads to me like John was walking along the street and boom a medical error/piano dropped on his head. But I don’t believe that is true. I bet it is true that are lots of deaths due to just unforgivable errors – someone is given a drug which was indicated in numerous sensible ways would kill them due an allergy but they were given it anyway and died. But I don’t trust how much of the deaths attributed to error are really 20% error, 19% cancer, 18% diabetes, 17% cardio-vascular disease, 16% long term high level use of powerful drugs ravaging the body, 10% car accident (which also someone else might say is 5% error, 30% cancer, 25%…). The error is still bad, and the system needs to be improved to reduce the frequency and consequences of errors. But I just don’t know how to take the error to death data without much more explanation.


Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.