Trapping errors

Mary PW Chin 钱碧慧博士
PhD (Wales), MSc (Surrey)

Continuing education & training

Trap errors before they ever get a chance to reach the client, patient or product. Errors do happen — but all errors can be trapped. If the staff can confidently step out of the premise without worrying about bird droppings landing on their heads, clients or patient should be able to enter the premise with equal confidence that the visit will incur no incidents.

Trap errors before they ever get a chance to reach the patient. Everyone makes mistakes; human errors, we say. Yes, it is impossible to avoid mistakes completely. I make mistakes too. But mistakes can always be trapped before reaching the patient; there is no excuse. Letting mistakes reach the patient is not being human, but is just a manifestation of weak protocols.

If the staff can step confidently out of the clinic, into the open, without bird droppings falling on his or her head, patients should be able to enter the clinic with equal confidence that the visit will incur no incidents. I used to ask hospital staff, "Has anyone been so unfortunate that bird droppings dropped on your head when you were out in the open (not under a tree)?" The answer was, "Never. The probability is way too tiny." The probability is tiny because the cross section of the droppings and the cross section of our head are way too tiny compared to the area of the earth.

So, how do we minimise the probability of medical incidents to a level way to tiny?

  1. implement multiple rounds of independent cross-checking;
  2. cross-check, don't double-check;
  3. drop superstitions about bad luck causing mistakes, take up the responsibility.

Say, my error rate is 2% (that is 1 error in 50 cases). An independent cross-check would cut the effective error rate to 0.04%. Subsequent rounds of independent cross-check would bring the effective error rates down to 0.0008%, 0.000016%, 0.00000032% and so on. Multiplied by the number of cases I handle throughout my career, I should increase n till I can be sufficiently confident that the probability of an incident becomes minuscule. It is therefore entirely within our control to press the probability below that of catching a bird's dropping, and below that of winning a lottery.

Let E be my error rate and N be the rounds of independent cross-check, the effective error rate would be E^N. Independent cross-checks reduce the error rate by the power law. In the preceding example, I used E=2%=0.02, N=1,2,3,4,5.

Error rate E can be lowered with a bit of discipline:

Table 3. Double-checking versus cross-checking.
Initial taskDouble-checkCross-check
Example 12 + 5 = 72 + 5 = 77 - 5 = 2
Example 2digits typed out
left to right,
the usual way
check from
left to right,
the usual way
move backwards,
from right to left,
Example 3digits typed out
row after row,
the usual way
check row by row,
top to bottom
start from the last row,
move upwards,

So what is the difference between double-checking and cross-checking? In cross-checking, the mind goes over the same contents via a fresh route. In double-checking, a lingering shadow remembering the previous, familiar route tends not to notice mistakes. Cross-checking is more robust and independent than double-checking.

Checking by a second person is by nature an independent cross-check. Further checks by the third and the fourth person would reduce the effective error rate dramatically. Many protocols require multiple signatures for this reason. That notwithstanding, individual staff should be well equipped with error-trapping skills, apply the skills diligently and confidently before passing the check to the next colleague in line. This is particularly crucial in the absence of a complete team e.g. during on-call after hours. Even at hours of reduced workforce it is entirely possible to practice rigorous cross-checking to prevent any errors from reaching the patient.