The Computer Crashed. Changes Not Saved. Patient Dies. | [node:field-byline] | Healthcare Blogs Skip to content Skip to navigation

The Computer Crashed. Changes Not Saved. Patient Dies.

January 25, 2010
by daphne
| Reprints

I just read a harrowing account of system errors that effectively killed patients being treated with radiation from linear accelerators. Basically, the patients mistakenly got obscenely high levels of radiation and died horrifying deaths —one of them at my old hospital’s cancer center, St. Vincent’s in Manhattan.

As I started to read the blow-by-blow of how the failure occurred, my blood turned cold. Because most of us have been in the same situation at one time or another: “Did I save or not?”

In this case, the medical physicist was using a computer program to set the exact calibrations for the size of the beam for treatment. “As Ms. Kalach was trying to save her work, the computer began seizing up, displaying an error message.” The error message asked if she wanted to save her changes before the program aborted. The physicist answered yes, and the physician in charge then approved the new plan.

The changes were not saved.

When the patient presented for the radiation, the multileaf collimator on the linear accelerator, which was supposed to pinpoint focus the radiation beam precisely on the tumor, was wide open. The patient received three such treatments before the error was found. The patient died, after much suffering, as a result of acute radiation toxicity.

But the story doesn’t end there.

State health officials issued an alert, reminding operators “of the absolute necessity to verify that the radiation field is of the appropriate size and shape prior to the patient’s first treatment.” That same day, another patient in New York, this time at Downstate Medical Center in Brooklyn, received radiation treatment for breast cancer. She received TWENTY EIGHT treatments, though she kept complaining of the sore on her chest that was peeling and wouldn’t heal. Two weeks after her treatment ended, the hospital looked into the possible causes of her wound.

Root cause analysis revealed that a therapist had programmed “wedge out” rather than “wedge in” on the machine. Another therapist failed to catch the error. On 27 occassions, during treatment, the computer screen clearly read “wedge out.” 27 times! And no one noticed that on the computer screen during treatment.

The wound on her chest got worse, and the patient endured four operations to repair the damage and reconstruct her chest wall. She was hospitalized for a year and eventually died, again, after much suffering.

I put the paper down slowly. We are entrusting more and more of our care to computers. I started thinking how important fail safe measures are when it comes to systems—and I hope every vendor in the space is reading this article as a wake up call.

Our caregivers can join them: in the second case, a computer screen had the wrong setting clearly displayed 28 times, and not one person noticed it—indeed, there was no protocol to check.

This investigative report was very thought-provoking on many levels. I urge you all to read it, whether you’re a vendor, caregiver—or patient.



The follow up article. "New technology has outpaced its oversight."

Thanks for the link Joe. I'm glad someone is focusing on this topic. Koppel is including errors due to EMRs and barcoding, and agreed, that's what most people think of with informatics errors, myself included. Impotant to realize they can come from places we don't often consider.


Last month at Johns Hopkins, I had the honor of attending Ross Koppel, PhD's Informatics grand rounds titled "From the front: Adventures in Healthcare Information Technology-enhanced clinical errors"

Here's the blurb and URL:

December 18, 2009 12:15 - "From the front: Adventures in Healthcare Information Technology-enhanced clinical errors"

Video archive (available here: )

Ross Koppel, PhD Professor Sociology Department, the School of Medicine Graduate School of Education University of Pennsylvania

Objectives: 1. To describe errors due to electronic medical records 2. To describe errors due to barcode technologies 3. To describe vendor-clinician relationships that may impede remediation of problems 4. To identify potential system fixes that deal with these errors.

Koppel has been publishing on this topic for years and his sophistication as a presenter as well as his reasoned elaboration of options are a must-see for anyone involved in HCIT policy. (Requires RealPlayer.)

Thanks again to Harold Lehmann and his faculty and team for continually putting out this great educational resource for some many years. I know of only one other institution that is similarly committed to the community as Hopkins Division of Health Sciences Informatics and it's Bloomberg school.

Are you tracking on Jim Walker's approach? Geissinger, recent 90 minute presentation at AMIA 2009 in November. I bring this up because it broad and systematic, and informed by other industries.

...informed from other industries...just like the checklist...