WEBVTT 00:00:00.000 --> 00:00:02.120 Okay, so I want to start today by throwing a 00:00:02.120 --> 00:00:05.120 term at you. And I guarantee, unless you're, 00:00:05.120 --> 00:00:07.280 I don't know, a very specific type of doctor 00:00:07.280 --> 00:00:09.699 or maybe a lawyer who's deep in industrial lawsuits, 00:00:10.000 --> 00:00:13.119 you have never heard this before. And honestly, 00:00:13.359 --> 00:00:16.699 it sounds, it sounds underwhelming. Oh, yeah. 00:00:16.940 --> 00:00:21.559 What's the term? Bereader. Ah, yes, it does sound 00:00:21.559 --> 00:00:24.120 a bit. Second rate, doesn't it? Right. I mean, 00:00:24.120 --> 00:00:25.679 if I told you I was a B reader, you'd assume 00:00:25.679 --> 00:00:28.339 I was the understudy. The backup plan, the person 00:00:28.339 --> 00:00:30.100 who reads the script when the A reader is out 00:00:30.100 --> 00:00:32.320 sick. Or maybe someone who only reads, like, 00:00:32.340 --> 00:00:35.460 B -list celebrity autobiographies. Exactly. It 00:00:35.460 --> 00:00:37.759 just implies lesser than. But what if I told 00:00:37.759 --> 00:00:39.560 you that being a B reader is actually one of 00:00:39.560 --> 00:00:43.740 the most elite, exclusive, and high -pressure 00:00:43.740 --> 00:00:46.359 designations in the entire American medical system? 00:00:46.600 --> 00:00:48.740 That is a much, much more accurate way to put 00:00:48.740 --> 00:00:51.399 it. It's a tiny club of people with a very, very 00:00:51.399 --> 00:00:54.560 specific skill set. And what if I said these 00:00:54.560 --> 00:00:57.719 people hold a power that is, frankly, kind of 00:00:57.719 --> 00:01:00.329 terrifying? Imagine a job where you're sitting 00:01:00.329 --> 00:01:03.130 in a dark room staring at this blurry black and 00:01:03.130 --> 00:01:06.069 white image of a lung. And your opinion, your, 00:01:06.230 --> 00:01:09.290 you know, subjective take on a shadow, can decide 00:01:09.290 --> 00:01:12.390 the fate of massive lawsuits, millions of dollars, 00:01:12.489 --> 00:01:14.609 and the medical truth for thousands of workers. 00:01:14.870 --> 00:01:17.250 It's the ultimate high -stakes game. And the 00:01:17.250 --> 00:01:20.849 B definitely does not stand for backup. So today 00:01:20.849 --> 00:01:23.590 on The Deep Dive, we are unpacking the very strange 00:01:23.590 --> 00:01:26.370 world of the B -reader. We're going to look at 00:01:26.370 --> 00:01:29.489 this intersection of, um... medicine, human error, 00:01:29.629 --> 00:01:32.409 and the law. We've got NIOSH protocols, some 00:01:32.409 --> 00:01:34.590 pretty dense medical journals, and a few legal 00:01:34.590 --> 00:01:36.189 studies that I think are really going to make 00:01:36.189 --> 00:01:38.170 your jaw drop. And it's a perfect topic because 00:01:38.170 --> 00:01:40.109 it really challenges something we all just assume, 00:01:40.269 --> 00:01:42.310 you know, that medicine is subjective. We think 00:01:42.310 --> 00:01:44.650 an x -ray is a picture of the truth. But as we're 00:01:44.650 --> 00:01:47.129 about to find out, an x -ray is just a shadow. 00:01:47.409 --> 00:01:49.709 And shadows can change depending on who's looking. 00:01:49.870 --> 00:01:52.890 Okay, so let's start at ground zero. What technically 00:01:52.890 --> 00:01:56.659 is a bee reader? Okay, so a B -reader is a physician. 00:01:56.920 --> 00:01:59.319 They're usually already a specialist, like a 00:01:59.319 --> 00:02:02.260 radiologist or a pulmonologist, but they've gone 00:02:02.260 --> 00:02:06.200 a step further and gotten certified by NIOH. 00:02:06.480 --> 00:02:09.259 That's the National Institute for Occupational 00:02:09.259 --> 00:02:12.360 Safety and Health. And their whole job, their 00:02:12.360 --> 00:02:15.879 specific certification, is to classify chest 00:02:15.879 --> 00:02:20.259 x -rays for signs of pneumoconiosis. Pneumoconiosis. 00:02:20.280 --> 00:02:22.960 That's the big umbrella term for like dust diseases. 00:02:23.000 --> 00:02:25.219 That's it. Yeah. It comes from the Greek. Pneumo 00:02:25.219 --> 00:02:28.120 for lung, conus for dust. So you're talking about 00:02:28.120 --> 00:02:30.360 things like asbestosis, silicosis, and the one 00:02:30.360 --> 00:02:32.900 that really kicked this whole thing off. Coal 00:02:32.900 --> 00:02:36.500 workers, pneumoconiosis. black lung. So these 00:02:36.500 --> 00:02:38.180 are the doctors who look at a long x -ray and 00:02:38.180 --> 00:02:40.180 say, yep, there's dust damage here, or nope, 00:02:40.259 --> 00:02:41.939 these lungs are clear. Well, see, that's where 00:02:41.939 --> 00:02:43.599 you have to be really careful with the wording. 00:02:43.639 --> 00:02:45.759 You said diagnose, which implies you're treating 00:02:45.759 --> 00:02:48.860 a patient. A B reader isn't necessarily diagnosing 00:02:48.860 --> 00:02:51.520 someone in that clinical sense. They are classifying 00:02:51.520 --> 00:02:53.979 an image. That feels like a subtle distinction. 00:02:54.219 --> 00:02:55.939 What's the real difference there? A clinical 00:02:55.939 --> 00:02:58.379 diagnosis is holistic. A doctor looks at you, 00:02:58.419 --> 00:03:00.580 listens to your breathing, asks about your history. 00:03:01.199 --> 00:03:04.000 A B reader is doing something much colder. It's 00:03:04.000 --> 00:03:07.060 epidemiological classification. They are looking 00:03:07.060 --> 00:03:09.900 at an image and mapping it against a really rigid 00:03:09.900 --> 00:03:12.539 international standard to generate a data point. 00:03:12.699 --> 00:03:14.639 So they're turning a picture into a spreadsheet 00:03:14.639 --> 00:03:17.139 entry. Precisely. And to do that, they use a 00:03:17.139 --> 00:03:19.439 system called the ILO classification. That's 00:03:19.439 --> 00:03:21.840 from the International Labor Office. And this 00:03:21.840 --> 00:03:23.719 is where the science part gets really, really 00:03:23.719 --> 00:03:26.319 intense. I was looking at the notes on this ILO 00:03:26.319 --> 00:03:29.960 system and it is just. It's wildly complex. It's 00:03:29.960 --> 00:03:32.599 not just, is there dust? Yes or no? Not at all. 00:03:32.699 --> 00:03:35.080 It's a whole language. When dust gets in the 00:03:35.080 --> 00:03:37.240 lungs, the tissue reacts by forming these little 00:03:37.240 --> 00:03:40.240 scars, fibrosis. On an x -ray, they look like 00:03:40.240 --> 00:03:43.479 small white spots or opacities. The ILO system 00:03:43.479 --> 00:03:46.800 makes the doctor describe those spots with incredible 00:03:46.800 --> 00:03:49.639 precision. First, the shape and size. And they 00:03:49.639 --> 00:03:52.060 use letters for this, right? Like P, Q, R, R. 00:03:52.120 --> 00:03:54.120 They do. So if the spots are round, you have 00:03:54.120 --> 00:03:56.949 to measure them. Are they type P? That's up to 00:03:56.949 --> 00:03:58.750 one and a half millimeters. Are they type Q? 00:03:59.370 --> 00:04:01.310 That's one and a half to three millimeters. Yeah. 00:04:01.389 --> 00:04:04.229 Or are they type R3 to 10? So you're distinguishing 00:04:04.229 --> 00:04:07.009 between a dot the size of a pinhead and one the 00:04:07.009 --> 00:04:10.189 size of a lentil on a blurry film. Exactly. And 00:04:10.189 --> 00:04:12.050 that's just for the round ones. If the scars 00:04:12.050 --> 00:04:15.250 are irregular like little lines or squiggles, 00:04:15.330 --> 00:04:18.670 you have to classify them as type S, T, or U 00:04:18.670 --> 00:04:20.709 based on their width. And then you judge how 00:04:20.709 --> 00:04:23.959 many there are. Sort of. You judge the profusion, 00:04:23.980 --> 00:04:26.959 the density, how crowded are these spots. And 00:04:26.959 --> 00:04:28.839 this is the part that just blows my mind. They 00:04:28.839 --> 00:04:31.519 don't just use a simple 1, 2, 3 scale. They use 00:04:31.519 --> 00:04:33.439 a 12 -point scale that actually incorporates 00:04:33.439 --> 00:04:36.459 uncertainty. I saw this in the notes. One slash 00:04:36.459 --> 00:04:39.519 zero or one slash two. Explain that because it 00:04:39.519 --> 00:04:41.639 seems so weird. It's not a fraction. It's like 00:04:41.639 --> 00:04:44.639 a mental toggle. So let's say I look at an x 00:04:44.639 --> 00:04:46.740 -ray. I think it's a category one definite, but 00:04:46.740 --> 00:04:49.660 light scarring. But I hesitated for a second. 00:04:49.699 --> 00:04:52.199 I briefly thought maybe this is a category zero, 00:04:52.420 --> 00:04:54.779 which is normal. So I coded as 10. So the code 00:04:54.779 --> 00:04:57.439 literally says I landed on one, but I thought 00:04:57.439 --> 00:05:01.360 about zero. That's it. Or 21 means I decided 00:05:01.360 --> 00:05:03.839 it's a two, but I considered calling it a one. 00:05:04.329 --> 00:05:06.730 The system is built to capture the doctor's own 00:05:06.730 --> 00:05:09.430 internal debate. It acknowledges that the line 00:05:09.430 --> 00:05:12.009 between healthy and sick is, you know, fuzzy. 00:05:12.519 --> 00:05:14.660 That is incredibly nuanced. It's like they're 00:05:14.660 --> 00:05:17.000 trying to build a mathematical model for human 00:05:17.000 --> 00:05:19.720 hesitation. They are. And that is why you need 00:05:19.720 --> 00:05:22.240 a special certification. You can't just grab 00:05:22.240 --> 00:05:24.980 any doctor and ask them to distinguish a type 00:05:24.980 --> 00:05:27.920 Q opacity from a type R while deciding between 00:05:27.920 --> 00:05:30.720 11 and a 12. Okay, so why did this whole system 00:05:30.720 --> 00:05:33.379 get created? Why couldn't regular radiologists 00:05:33.379 --> 00:05:36.500 just handle it? Because of the absolute chaos 00:05:36.500 --> 00:05:39.180 of the 1970s, you have to go back to the origin 00:05:39.180 --> 00:05:43.240 story here. 1974. The era of bell bottoms, the 00:05:43.240 --> 00:05:46.860 energy crisis. And a massive focus on coal miners. 00:05:47.060 --> 00:05:49.720 Right. The federal government had this huge surveillance 00:05:49.720 --> 00:05:52.120 program for coal miners. They were taking thousands 00:05:52.120 --> 00:05:54.879 of chest X -rays to track black lung. But when 00:05:54.879 --> 00:05:57.459 the data came back, they found a huge problem. 00:05:57.519 --> 00:05:59.990 Which was? Reader variability. Which is just 00:05:59.990 --> 00:06:02.149 a polite scientific way of saying that doctors 00:06:02.149 --> 00:06:04.310 couldn't agree on anything. It was a mess. I 00:06:04.310 --> 00:06:06.350 mean, you'd send a miner's x -ray to a doctor 00:06:06.350 --> 00:06:09.250 in Kentucky, and he'd say, this man has stage 00:06:09.250 --> 00:06:12.589 2 fibrosis. Get him out of the mine. You send 00:06:12.589 --> 00:06:15.290 the exact same film to a doctor in West Virginia, 00:06:15.389 --> 00:06:18.389 and he says, totally clear, send him back to 00:06:18.389 --> 00:06:21.709 work. That is a nightmare. For the miner, I mean. 00:06:21.810 --> 00:06:23.870 Yeah. Your whole life, your health, your compensation, 00:06:24.110 --> 00:06:26.569 your job, it all depends on which doctor happens 00:06:26.569 --> 00:06:29.079 to pick up your file. It was scientifically unacceptable. 00:06:29.360 --> 00:06:32.079 And legally, it was a disaster. It just proved 00:06:32.079 --> 00:06:34.480 that truth in these images was totally subjective. 00:06:35.120 --> 00:06:37.639 So NIOSH stepped in. They said, we need to calibrate 00:06:37.639 --> 00:06:40.060 the human eye. And they launched the B -Reader 00:06:40.060 --> 00:06:42.279 program. And just to close the loop on the name, 00:06:42.439 --> 00:06:45.339 I saw a note that the B probably stands for the 00:06:45.339 --> 00:06:47.600 Black Lung Program. That's the prevailing theory, 00:06:47.720 --> 00:06:50.459 yeah. The Black Lung Connection. But honestly, 00:06:50.620 --> 00:06:53.060 given how hard the exam is, the B might as well 00:06:53.060 --> 00:06:55.120 stand for brutal. Let's talk about that exam. 00:06:55.529 --> 00:06:57.269 Because this isn't some weekend seminar with 00:06:57.269 --> 00:06:59.709 free bagels. Oh, not at all. Becoming a bee reader 00:06:59.709 --> 00:07:02.430 is famously difficult. Just to give you a sense 00:07:02.430 --> 00:07:06.589 of the scale, as of October 2017, how many certified 00:07:06.589 --> 00:07:09.009 bee readers do you think there were in the entire 00:07:09.009 --> 00:07:12.550 U .S.? I don't know. There are thousands of radiologists. 00:07:12.730 --> 00:07:17.970 Maybe a thousand. 181. 181. That's it. For the 00:07:17.970 --> 00:07:20.250 whole country? That's it. Plus about 58 international 00:07:20.250 --> 00:07:23.610 ones. It is a microscopic club. Why so few? Does 00:07:23.610 --> 00:07:25.470 nobody want to do it? Oh, people want to do it. 00:07:25.490 --> 00:07:29.550 It can be very lucrative. But the exam is a massive 00:07:29.550 --> 00:07:32.310 filter. It takes place at one specific lab alert, 00:07:32.490 --> 00:07:34.970 the Appalachian Laboratory for Occupational Safety 00:07:34.970 --> 00:07:37.750 and Health in Morgantown, West Virginia. Even 00:07:37.750 --> 00:07:40.689 the location sounds intense. Right in the heart 00:07:40.689 --> 00:07:43.269 of coal country. It is. And the test involves 00:07:43.269 --> 00:07:47.009 a timed classification of 125 different radiographs. 00:07:47.009 --> 00:07:49.689 You have to sit there under the clock and correctly 00:07:49.689 --> 00:07:52.810 classify 125 x -rays using that super strict 00:07:52.810 --> 00:07:55.329 ILO standard. You have to get the P's and Q's 00:07:55.329 --> 00:07:57.189 right. You have to nail the perfusion. And you're 00:07:57.189 --> 00:07:59.350 graded against a master key, right? A consensus 00:07:59.350 --> 00:08:02.050 from a panel of experts. Yes. And if you deviate 00:08:02.050 --> 00:08:04.370 too much, you fail. The pass rates are historically 00:08:04.370 --> 00:08:08.069 low. But here is the real kicker. It's not like 00:08:08.069 --> 00:08:10.129 a degree where once you have it, you have it 00:08:10.129 --> 00:08:13.149 for life. Every four years. If you don't pass 00:08:13.149 --> 00:08:15.470 the recertification exam, you lose your status. 00:08:15.589 --> 00:08:17.819 You're out of the club. That's incredible pressure. 00:08:18.120 --> 00:08:20.660 It's like having to retake your final exams every 00:08:20.660 --> 00:08:23.079 four years for your entire career. It's to make 00:08:23.079 --> 00:08:25.220 sure their skills don't drift. Because remember, 00:08:25.339 --> 00:08:27.139 the original point was for research and compensation 00:08:27.139 --> 00:08:30.040 data. They needed the human eye to act like a 00:08:30.040 --> 00:08:33.360 machine. So we have this elite, super trained, 00:08:33.519 --> 00:08:38.179 constantly tested group of 181 experts. Theoretically, 00:08:38.240 --> 00:08:40.980 this should solve the variability problem. We 00:08:40.980 --> 00:08:43.669 should have perfect objectivity. Theoretically. 00:08:43.809 --> 00:08:46.389 But this is where the story shifts from, you 00:08:46.389 --> 00:08:49.029 know, a medical science story to more of a legal 00:08:49.029 --> 00:08:51.529 thriller because the context started to change. 00:08:51.730 --> 00:08:54.269 We move beyond just monitoring coal miners. Right. 00:08:54.570 --> 00:08:56.870 Bee readers became critical for all the dust 00:08:56.870 --> 00:09:00.230 diseases, asbestos, silica. And suddenly we weren't 00:09:00.230 --> 00:09:01.549 just talking about government surveillance. We 00:09:01.549 --> 00:09:04.929 were talking about toxic torts, massive lawsuits. 00:09:05.110 --> 00:09:06.450 And where there's litigation, there's money. 00:09:06.570 --> 00:09:09.210 Huge money. Billions. And where there is money, 00:09:09.370 --> 00:09:11.750 there is pressure. So let's unpack variability 00:09:11.750 --> 00:09:14.629 again. We talked about random variation, just 00:09:14.629 --> 00:09:18.330 doctors disagreeing. But the sources highlight 00:09:18.330 --> 00:09:22.190 a different kind. Systematic variation. Which 00:09:22.190 --> 00:09:27.049 is a polite way of saying bias. Well, is it corruption? 00:09:27.409 --> 00:09:29.169 Well, let's stick with the science of bias for 00:09:29.169 --> 00:09:32.490 a second. It can be cognitive. It happens when 00:09:32.490 --> 00:09:34.909 a reader has outside information that primes 00:09:34.909 --> 00:09:38.070 their brain. So if I hand you an x -ray and I 00:09:38.070 --> 00:09:41.429 say... This is Mr. Smith. He's 65, and he worked 00:09:41.429 --> 00:09:43.710 in an asbestos factory for 30 years. What is 00:09:43.710 --> 00:09:45.590 your brain going to do? My brain is going to 00:09:45.590 --> 00:09:47.509 go hunting. I'm going to expect to see damage. 00:09:47.629 --> 00:09:49.750 I'll squint at every shadow until it looks like 00:09:49.750 --> 00:09:51.730 a problem. Exactly. You lower your threshold 00:09:51.730 --> 00:09:54.009 for calling something positive. It might not 00:09:54.009 --> 00:09:55.929 even be conscious. You're just primed to see 00:09:55.929 --> 00:09:58.129 the pattern. But then you add the money. If I'm 00:09:58.129 --> 00:10:00.070 a lawyer for Mr. Smith, I need a positive read 00:10:00.070 --> 00:10:02.370 to win my case. If you, the doctor, give me a 00:10:02.370 --> 00:10:03.909 negative read, well, I don't pay you for your 00:10:03.909 --> 00:10:05.919 testimony. I just go find another doctor. And 00:10:05.919 --> 00:10:07.820 that creates a marketplace for positive readings. 00:10:08.240 --> 00:10:11.759 And this leads us to the most explosive piece 00:10:11.759 --> 00:10:14.240 of evidence we found, the Gitlin study. I read 00:10:14.240 --> 00:10:15.759 this three times. I had to make sure I wasn't 00:10:15.759 --> 00:10:18.100 misinterpreting the numbers. This study basically 00:10:18.100 --> 00:10:21.039 took a flamethrower to the whole idea of B -reader 00:10:21.039 --> 00:10:23.860 objectivity. So walk us through the setup. Okay. 00:10:24.259 --> 00:10:28.879 So they had a set of 492 chest x -rays. Group 00:10:28.879 --> 00:10:31.460 A was B readers who were retained by lawyers 00:10:31.460 --> 00:10:33.799 in asbestos litigation. These were the hired 00:10:33.799 --> 00:10:37.600 guns for the plaintiffs. Group B was an independent 00:10:37.600 --> 00:10:40.279 panel of B readers who were blinded. So they 00:10:40.279 --> 00:10:42.559 didn't know who the patients were, no history, 00:10:42.740 --> 00:10:44.500 and they weren't being paid by a lawyer to find 00:10:44.500 --> 00:10:47.639 a specific result. So Group A, the plaintiff's 00:10:47.639 --> 00:10:50.039 experts, versus Group B, the neutral scientists, 00:10:50.299 --> 00:10:53.309 same films. What happened? The B readers hired 00:10:53.309 --> 00:10:55.950 by the lawyers found abnormalities, signs of 00:10:55.950 --> 00:11:00.429 disease in 95 .9 % of the cases. 96%. Okay, so 00:11:00.429 --> 00:11:02.309 basically almost everybody's sick. Almost everyone. 00:11:02.389 --> 00:11:04.769 Now, the independent blinded panel looking at 00:11:04.769 --> 00:11:07.370 the exact same images, they found abnormalities 00:11:07.370 --> 00:11:10.690 in only 4 .5 % of the cases. Wait, stop. Let 00:11:10.690 --> 00:11:13.570 that sink in. 96 % versus 4 .5%. That's the gap. 00:11:13.769 --> 00:11:16.789 That is not a gap. That's a canyon. I mean, if 00:11:16.789 --> 00:11:18.909 it was 60 percent, it's 50 percent. I'd say, 00:11:18.970 --> 00:11:22.889 OK, subjective. But 96 to 4. That means one group 00:11:22.889 --> 00:11:25.409 is seeing a plague and the other group is seeing 00:11:25.409 --> 00:11:27.809 a room full of healthy lungs. It highlights the 00:11:27.809 --> 00:11:30.690 massive, massive impact of context and financial 00:11:30.690 --> 00:11:33.029 incentive. The study suggests that when you are 00:11:33.029 --> 00:11:35.450 hired to find something, you find it. And remember, 00:11:35.590 --> 00:11:38.309 these aren't just random doctors. These are all 00:11:38.309 --> 00:11:42.580 NISH certified bee readers, the elite 181. This 00:11:42.580 --> 00:11:45.240 is just, I mean, the implications are huge. If 00:11:45.240 --> 00:11:47.919 a lawyer can hire a B reader who will say positive 00:11:47.919 --> 00:11:51.100 96 % of the time, that's almost a guaranteed 00:11:51.100 --> 00:11:52.980 ticket to a settlement, isn't it? The certification 00:11:52.980 --> 00:11:55.639 itself was being weaponized. That's exactly the 00:11:55.639 --> 00:11:57.399 argument that started to be made, that the B 00:11:57.399 --> 00:11:59.620 reader stamp of approval had become a commodity. 00:12:00.159 --> 00:12:02.299 Lawyers weren't hiring the most accurate doctors. 00:12:02.360 --> 00:12:04.679 They were hiring the most positive doctors. But 00:12:04.679 --> 00:12:07.100 just to play devil's advocate for a second, is 00:12:07.100 --> 00:12:09.019 it possible the independent panel was wrong? 00:12:09.460 --> 00:12:11.740 Maybe they were being too conservative. It's 00:12:11.740 --> 00:12:14.860 possible. Radiology is subjective. But a 90 point 00:12:14.860 --> 00:12:17.559 spread. That suggests something systemic is going 00:12:17.559 --> 00:12:20.500 on. It suggests that objectivity dissolves when 00:12:20.500 --> 00:12:23.220 you know which side you're playing for. So surely 00:12:23.220 --> 00:12:25.480 NIOSH didn't just let this go. I mean, their 00:12:25.480 --> 00:12:28.759 entire mission is accuracy. If their gold standard 00:12:28.759 --> 00:12:30.820 certification is being used to game the legal 00:12:30.820 --> 00:12:33.840 system, that's a really bad look. It was a crisis 00:12:33.840 --> 00:12:36.340 of credibility. Absolutely. And this brings us 00:12:36.340 --> 00:12:39.899 to the cleanup phase. NIOISH responded by establishing 00:12:39.899 --> 00:12:43.100 a formal code of ethics for all B readers. We 00:12:43.100 --> 00:12:45.419 have the code here. It's 10 points. And honestly, 00:12:45.539 --> 00:12:46.980 reading through them, some of these feel like 00:12:46.980 --> 00:12:48.519 things you shouldn't have to write down for a 00:12:48.519 --> 00:12:50.720 doctor. Usually if a rule exists, it's because 00:12:50.720 --> 00:12:53.460 someone somewhere did the opposite. Right. OK, 00:12:53.519 --> 00:12:55.960 look at rule eight. B readers shall not accept 00:12:55.960 --> 00:12:58.379 compensation that is... contingent upon the findings 00:12:58.379 --> 00:13:01.159 of their chest radiograph classifications. That's 00:13:01.159 --> 00:13:03.120 the bounty hunter rule. It tackles that problem 00:13:03.120 --> 00:13:05.539 head on. You can't have a fee that says, I get 00:13:05.539 --> 00:13:08.480 $500 for a negative reeb or $5 ,000 if I find 00:13:08.480 --> 00:13:11.899 disease. That is explicitly forbidden. You get 00:13:11.899 --> 00:13:14.220 paid for your time, not for the result. That 00:13:14.220 --> 00:13:17.299 seems like ethics 101. It is. But you can imagine 00:13:17.299 --> 00:13:20.179 how creative the incentives could get. This shuts 00:13:20.179 --> 00:13:23.440 that down. Then there's rule three. Bee readers 00:13:23.440 --> 00:13:26.919 shall not make clinical diagnoses. Based on chest 00:13:26.919 --> 00:13:29.080 radiograph classification alone. And this circles 00:13:29.080 --> 00:13:31.419 back to what we said at the very beginning. A 00:13:31.419 --> 00:13:34.100 B reading is a classification. It's a data point. 00:13:34.220 --> 00:13:37.480 It is not a diagnosis. A doctor can't use this 00:13:37.480 --> 00:13:39.960 form to tell a patient, you have asbestosis. 00:13:40.080 --> 00:13:42.820 You need more evidence than that. But then look 00:13:42.820 --> 00:13:45.679 at rule four. This one feels like the moral safety 00:13:45.679 --> 00:13:48.519 net. It says if a B reader sees something clinically 00:13:48.519 --> 00:13:50.820 important, they have to make sure the individual 00:13:50.820 --> 00:13:53.580 gets notified. Right. Because imagine you're 00:13:53.580 --> 00:13:56.340 a B -reader hired by a company. You're blinded. 00:13:56.340 --> 00:13:58.500 You get an x -ray and you see a massive tumor. 00:13:58.879 --> 00:14:02.240 But your only job is to look for dust. Technically, 00:14:02.399 --> 00:14:04.259 you could just check the no dust box and move 00:14:04.259 --> 00:14:06.200 on. And the patient could die a year later because 00:14:06.200 --> 00:14:08.899 nobody told them. Exactly. Rule 4 says you have 00:14:08.899 --> 00:14:11.240 a duty to the human being, not just to the form. 00:14:11.379 --> 00:14:13.279 You have to make sure someone alerts the patient. 00:14:13.460 --> 00:14:15.639 You can't just hide behind the process. And the 00:14:15.639 --> 00:14:17.740 sources also mention a big push for blinding. 00:14:18.009 --> 00:14:20.090 as a recommended practice. It's the ultimate 00:14:20.090 --> 00:14:23.610 fix. The Gitlin study proved it works. If you 00:14:23.610 --> 00:14:25.429 want the truth, you have to hide the context. 00:14:25.710 --> 00:14:27.669 If the reader doesn't know who the patient is, 00:14:27.830 --> 00:14:31.149 who the lawyer is, or the job history, that bias 00:14:31.149 --> 00:14:34.870 just evaporates. The 96 % drops down to 4%. It's 00:14:34.870 --> 00:14:37.289 so ironic, isn't it? We usually think that to 00:14:37.289 --> 00:14:39.289 make a good decision, you need more information. 00:14:39.570 --> 00:14:42.330 But here, to get to the truth, the doctor actually 00:14:42.330 --> 00:14:45.169 needs less information. That is the whole paradox 00:14:45.169 --> 00:14:48.000 of the B -reader. Context creates a narrative 00:14:48.000 --> 00:14:50.860 and narrative creates bias. To get to the raw 00:14:50.860 --> 00:14:53.519 data, the truth of the photons on the film, you 00:14:53.519 --> 00:14:55.799 have to strip the story away. So where does this 00:14:55.799 --> 00:14:57.639 leave us today? Is the program still running? 00:14:57.799 --> 00:15:00.240 It is. It's still the gold standard for tracking 00:15:00.240 --> 00:15:03.139 occupational lung disease. The exam in Morgantown 00:15:03.139 --> 00:15:05.379 is still brutal and the certification is still 00:15:05.379 --> 00:15:08.039 incredibly rare. But are they still fighting 00:15:08.039 --> 00:15:10.159 this variability battle? I think variability 00:15:10.159 --> 00:15:12.340 will always be a factor as long as humans are 00:15:12.340 --> 00:15:14.559 doing the reading. The whole B -Reader program 00:15:14.559 --> 00:15:16.720 is an attempt to standardize human perception. 00:15:17.259 --> 00:15:19.519 It's an attempt to turn a subjective opinion, 00:15:19.879 --> 00:15:23.000 that shadow looks weird, into objective data. 00:15:23.259 --> 00:15:25.000 And maybe that's the real takeaway here. We think 00:15:25.000 --> 00:15:28.039 of medical tests as binary, positive or negative. 00:15:28.480 --> 00:15:31.039 You have it or you don't. But an x -ray is just 00:15:31.039 --> 00:15:33.720 a shadow on a plate. It requires interpretation, 00:15:34.000 --> 00:15:37.080 and interpretation requires a human brain. And 00:15:37.080 --> 00:15:40.159 brains are messy, pattern -seeking, bias -prone 00:15:40.159 --> 00:15:43.059 things. It really makes you think, we spent this 00:15:43.059 --> 00:15:46.460 whole time on chest x -rays, but if these trained 00:15:46.460 --> 00:15:50.799 experts, the elite 181, can disagree 96 % versus 00:15:50.799 --> 00:15:52.980 4 % of the time, depending on who's signing the 00:15:52.980 --> 00:15:56.120 check, where else in medicine is objectivity 00:15:56.120 --> 00:15:59.259 just an illusion? What's happening with MRI interpretations 00:15:59.259 --> 00:16:02.399 for back surgery? or pathology slides for cancer? 00:16:02.559 --> 00:16:04.360 That's a really uncomfortable question, isn't 00:16:04.360 --> 00:16:06.860 it? The bee reader controversy just made it visible 00:16:06.860 --> 00:16:09.240 because we had the data from that one study. 00:16:09.360 --> 00:16:10.899 It makes you wonder what's happening in all the 00:16:10.899 --> 00:16:12.820 areas where we haven't done the study yet. On 00:16:12.820 --> 00:16:14.580 that slightly terrifying note, I think we'll 00:16:14.580 --> 00:16:16.480 wrap up this deep dive. Always a pleasure to 00:16:16.480 --> 00:16:18.940 undermine your faith in absolute truth. Next 00:16:18.940 --> 00:16:21.600 time you see a medical report, remember the bee 00:16:21.600 --> 00:16:25.320 reader. Remember the P's and Q's. And maybe think 00:16:25.320 --> 00:16:27.659 about getting a blinded second opinion. We will 00:16:27.659 --> 00:16:28.519 catch you on the next one.