0:01

The next topic I wanna talk about in with regard to

0:05

reducing errors is to detect those errors early.

0:11

And this is something

0:13

that we had have established at Johns Hopkins

0:15

and have promoted in the literature.

0:17

And it was started by, um, actually a conversation I had

0:22

with my late brother Sam,

0:24

who was a pathologist at UPMC in Pittsburgh.

0:28

He turned me onto the concept that in, in pathology at UPMC,

0:33

they would not release pathology cases

0:37

until they had been peer reviewed.

0:39

So they had a system of peer reviewing 10% of their cases,

0:44

and, um, it would get reviewed by a colleague

0:50

before it was finalized to the medical record.

0:53

Now, just imagine being able to detect the error

0:59

before it's made in public as opposed to

1:03

six months later, re-looking at the slides

1:06

and saying, no, you know, there is a little spot

1:08

of cancer on the, on this pathology slide

1:11

of the prostate biopsy.

1:14

So he, he said, you know, Dave, the, the key is

1:18

to catch the errors early, not

1:21

after the tumor has grown for six months.

1:25

And so we implemented that at Hopkins

1:27

and have, uh, promoted that, um, commercially as well,

1:31

that prospective peer review

1:34

where you are not looking at the prior studies

1:37

to peer review or your colleagues,

1:39

but you are reading cases

1:40

that they have read within 24 hours.

1:43

So you pick up that missed cancer, that missed stroke,

1:47

that missed aneurysm immediately, rather than

1:51

five months later when the patient presents

1:54

with subarachnoid hemorrhage

1:55

because of an aneurysm that was missed.

1:57

So I highly recommend this.

1:59

You might wanna think about

2:01

whether you're doing the patient a,

2:03

a service when you're finding the errors,

2:06

when a prior study from a year ago, two years ago,

2:08

six months ago, as opposed to finding the errors

2:12

within 24 hours of the error having been occurred.

2:15

I think it's brilliant

2:17

and I credit my brother for recommending that to me.

2:20

So find the errors early.

2:22

Um, find the errors in studies that are high risk.

2:26

Remember that peer reviewing with AI is also a

2:31

great application to double read the cases

2:34

with artificial intelligence

2:37

and, um, then have that discussion.

2:40

All right, we, it seems like Dave Yousome is making most

2:43

of his peer review errors on aneurysms.

2:46

Maybe we should have Dave take another course,

2:49

or, you know, be given the feedback

2:52

that your errors are revolving around missing aneurysms

2:56

as opposed to missing cancers or wrong

2:59

Interpretation of white matter regions.

3:01

So I think that, um, peer review is important for

3:06

in the group, reducing the errors

3:10

and that peer review may be from a computer rather than

3:13

among, uh, colleagues, human colleagues.

3:17

So this is, uh, my brother's, uh, paper on, uh,

3:21

pre-sign out quality assurance.

3:23

Um, and that was written in what, 2010

3:27

and in 2019, not Sam Newso,

3:30

but Dave Yousome wrote a paper about neuroradiology peer

3:33

review in the impact of, uh,

3:36

dinging someone with a peer review.

3:38

So this is, um, um,

3:43

a technique of peer review that is important.

3:45

We also ratchet up the number of cases that

3:50

we peer review in new faculty

3:54

during a startup period of their first year, as opposed

3:58

to the individuals who have been on faculty for 20 years.

4:01

You could do that, you can do that.

4:03

You can adjust the, the numbers for peer reviewing someone

4:06

who has a higher rate of errors, for example,

4:10

and I'm the, the shark.

4:12

I tend to report a lot of, um, errors,

4:16

even if they're minor.

4:17

And that's just to make sure

4:19

that the peer review process is, you know, honest

4:24

in trying to identify, you know, blind spots

4:29

or errors among the colleagues.