The order was written correctly in the electronic medical record (EMR) for phenytoin, 800 mg IV. The drug-dispensing machines stocked phenytoin in 250 mg/1 mL vials. The correct dose therefore would require 4 vials and be equal to 3.2 mL to be added to a small IV bag. The nurse misread the order as 8000 mg (8 g) and proceeded to administer that dose to the patient, which was a 10-fold overdose and 2 to 3 times the lethal dose. The patient died several minutes after the infusion.
This nurse had to work hard to make the error:
An audit of the pharmacy system revealed that the nurse had taken 32 vials out of 3 different pharmacy dispensing machines to accumulate 8 g of IV phenytoin. Moreover, the nurse had to use two IV bags and a piggyback line to give that large a dose.
And all this had to go unnoticed by people nearby:
Within 100 feet of the ED nurses' station were several ED doctors, a number of nurses, and a pharmacy with a PhD pharmacist on duty. The nurse did not ask anyone to check her calculations, nor did anyone notice or comment when she was moving around the unit amassing the vials needed for the dose.
What do we conclude? Elizabeth Manias writes:
In this case, the nurse made a series of cognitive errors that contributed to a 10-fold overdose of phenytoin. The nurse did not recognize that it was unusual to use 32 vials of phenytoin to obtain the required dose. She did not acknowledge that it was uncommon to need two intravenous (IV) fluid bags to administer the single dose of phenytoin. The nurse also did not double-check the IV medication with another clinician. Most important, she appeared not to know the toxic dose of the medication she was administering.
Incompetent? No necessarily:
Every day, well-intentioned clinicians carry out their medication activities in environments that are set up to fail them. Mistakes with medications occur not because a clinician has been incompetent by making an error, but rather because this single act is the final link in a chain of failures.
Indeed, some of the worse mistakes come from good intentions:
In this case, one can imagine a well-meaning nurse trying to do everything she could to collect the medication for her allocated patient. Although her persistence is laudable, it is probably also an example of anchoring bias. When the order is so difficult to complete and so unusual, it is far more likely to be in error than to reflect an idiosyncrasy of the prescribing physician or the patient. While the nurse was undoubtedly trying to be helpful, the instinct of all clinicians has to change from one of "this is unusual, but I'll just get it done" to "this is unusual, I wonder whether it is correct."
So back to the nurse in Brazil. I don't know if she was incompetent. I do know that variations of the kind of error she made happen thousands of times, even by highly trained folks. Manias concludes with these take-home points:
- Good communication between clinicians is a key factor to minimizing the risk of producing a medication error.
- Clinicians can train themselves to recognize warnings associated with medication errors.
- Medication errors generally occur as a result of system failures rather than faults produced by particular people.
As the article notes, there are limits to error--preventing safety design:
ReplyDelete"It is impossible to think of all the ways in which a clinician completes a well-intentioned but fatally flawed task."
There is no question that crashingly incomprehensible errors are rendered more likely in our current chaotic medical 'system' (not) where improvisation and workarounds are common. However, as Robert Wachter and others have noted, somewhere there is a line between personal accountability and system failure. It is where to draw that line that is the difficulty. Must one, or can one, design a system that assumes the worker has no baseline level of intelligence and competence?
nonlocal MD
I'd have to agree that a system focus is more productive than the "bad worker" method of error prevention, although both are possible.
ReplyDeleteIn this case, the normal team process, accepted by everyone in the ER, was for the nurse to read a number on screen and give that dose to the patient in isolation. Simply mis-counting the zeroes in "800" thus results in death.
Such errors aren't a matter of intelligence. To illustrate, I'd like to point out that the case report in the blog says that the phenytoin concentration in the vial is "250 mg/ml". This is wrong, as this concentration would be too high to be stable in solution. "250 mg/5ml" is a standard vial. Such errors don't correlate with educational attainment, as far as I know. In my humble experience, such errors correlate negatively with hours of sleep, and positively with convulsing patients.
An old-time data-entry card-punch rule-of-thumb was that you could expect one typo per 600 characters of arbitrary machine code. If better accuracy is needed, you must double-check. Clearly redundancy is more effective than stress at reducing errors to zero.
In the twenty-first century, we accept the statement "The order was written correctly in the electronic medical record.." This illustrates the difference between recording and communicating.
The MD and RN clearly did not communicate correctly. Verbal and written communication customs have evolved to prevent or detect such mistakes. It seems unlikely to me that the RN would have mistaken the phrase "eight hundred milligrams" for "eight grams" in normal speech. Nor would an MD normally write "8000 mg" on paper. Electronic medical records miss some of the built-in redundancy of English. For instance, we would switch from "mg" to "gm" units for a big dose. We use commas to help show the order of magnitude of numbers. We adjust the numeric precision to show the significant digits. We can use two-way communication to ensure understanding and reasonableness, rather than "entering" an order from another room. Hand-written paper orders allow the RN multiple opportunities to read back, then re-read the number in the process of drawing up the medicine, since you can carry a piece of paper or a chart more easily than a computer.
In other words, the EMR was an unreliable messenger. Reducing communication to database records ignores a hundred thousand years of social evolution. We can and will adapt to these EMR mandates. The process, however, will require identifying unacknowledged dangers.