Subscribe to The Podcast by KevinMD. Catch up on old episodes!

L. Joseph Parker, a research physician, sheds light on how AI has been employed to target physicians and reshape the landscape of health care, especially in the context of the opioid crisis. Discover the ethical and legal implications, the impact on patient care, and the challenges faced by doctors in this AI-driven health care environment.

L. Joseph Parker is a research physician.

He discusses the KevinMD article, “AI enforcement in health care: Unpacking the DEA’s approach to the opioid epidemic.”

Our presenting sponsor is Nuance, a Microsoft company.

Together, Microsoft and Nuance are leveraging their rich digital technology and advanced AI capabilities to tackle some of health care’s biggest challenges. AI-driven technology promises to revolutionize patient and provider experiences with clinical documentation that writes itself.

The Nuance Dragon Ambient eXperience, or DAX for short, is a voice-enabled solution that automatically captures patient encounters securely and accurately at the point of care. DAX Copilot combines proven conversational and ambient AI with the most advanced generative AI in a mobile application that integrates directly with your existing workflows.

Physicians who use DAX have reported a 50 percent decrease in documentation time and a 70 percent reduction in feelings of burnout, and 85 percent of patients say their physician is more personable and conversational.

Discover AI-powered clinical documentation that writes itself. Visit https://nuance.com/daxinaction to see a 12-minute DAX Copilot demo.

VISIT SPONSOR → https://nuance.com/daxinaction

SUBSCRIBE TO THE PODCAST → https://kevinmd.com/podcast

RECOMMENDED BY KEVINMD → https://kevinmd.com/recommended

GET CME FOR THIS EPISODE → https://earnc.me/UEf2cU

Powered by CMEfy.

Transcript

Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast, and get CME for this episode by clicking on the CME link in the show notes. We welcome back today L. Joseph Parker. He’s a research physician. Today’s KevinMD article is “AI enforcement in health care: Unpacking the DEA’s approach to the opioid epidemic.” Joseph, welcome back.

L. Joseph Parker: Thank you, Dr. Pho.

Kevin Pho: So go to KevinMD.com/podcast, you can hear Joseph’s past episodes to hear his story. But today let’s jump right into his most recent article about AI enforcement in health care, unpacking the DEA’s approach to the opioid epidemic. So what’s this one about?

L. Joseph Parker: A lot of physicians have been wondering what criteria the government is using to target different physicians. Physicians treating pain and addiction have been targeted across the United States, and we’ve had a lot of trouble determining what is it that they’re using. We had identified 12 criteria that they thought made a pill mill, but how are they deciding who to look at to see whether or not they’re a pill mill? And it turns out that under Attorney General Jeff Sessions an AI was created in 2017 for this purpose.

Kevin Pho: All right, so tell us exactly what that entails. And when you say an AI was created for this purpose, walk us through what’s perceived that the government is doing with this AI.

L. Joseph Parker: With the transition to cloud-based medical records and to prescription monitoring programs, law enforcement has access to databases of some of the most personal information you could possibly get. Now, back in the Nixon era, the FBI and CIA had to try to break into a psychiatrist’s office, or I should say two people, one working for the FBI and one for the CIA, tried to break into a psychiatrist’s office to get dirt on Ellsberg, to find out what he was being treated for. Well, they don’t have to do that now. They just type in a search query and they can know.

And they wanted to. In 2017 Mr. Sessions apparently decided that the opioid epidemic in America was not due to fentanyl being produced by cartels, but to physicians coddling those suffering from pain and addiction and prescribing too much medication for them, which made it possible for patients to divert those. So he decided to train an AI, and they hired a company that had worked for 45 years for the government creating databases and algorithms. And they had just started getting to where they could make neural net AIs that are kind of heuristic. They learn almost like a child, so you feed them information and they learn from that.

Kevin Pho: So tell us what kind of criteria you believe these AIs are using to identify whether physician practices are pill mills or not.

L. Joseph Parker: Well, thanks to a Freedom of Information Act lawsuit we have the actual criteria that was used. Instead of going to the American Society of Addiction Medicine to learn about addiction, or the American Academy of Pain Medicine to learn about the proper treatment of pain, the Department of Justice went to health insurance companies and said, what is a bad doctor? And the health insurance companies said, bad doctors make us spend money on our beneficiaries, bad doctors order genetic tests, ancillary labs, stuff like that. So that was programmed into the AI. And then they went to the DEA and said, what’s a bad doctor? And the DEA basically said, a bad doctor prescribes a lot of medicine, a lot of pills, a lot of morphine dose equivalents and all that. And so they programmed the AI with that.

So AI, we have to understand, artificial intelligence is not artificial sentience. These are glorified calculators, extremely advanced, but all they can do is cogitate upon the information you give them, and they know nothing else. There’s nothing else in the universe except what you taught them. So if you tell them a bad doctor makes us spend money and a bad doctor prescribes lots of pills, they believe it with 100 percent accuracy. Belief is probably a bad term, they adhere to it with 100 percent accuracy.

And so this AI combed through tens of thousands of physicians, hundreds of thousands of patient interactions, and started kicking up doctors it labeled as criminal. It also kicked up patients, and they created target lists of patients prior to raiding a doctor’s office, so they could specifically access all the records of those patients.

So this is almost a pre-crime, sort of like the Tom Cruise movie, where they think they can identify someone who’s going to be a criminal even if they’re not a criminal. Or once they have your name, they’re going to prove that you’re a criminal, because they believe that if the AI says it, it’s real. And it’s no more real than any virtual reality system you put out there. It’s just an opinion, and a biased opinion.

Kevin Pho: So are there any specific cases that you can cite, or stories that you can tell, where AI mistakenly identified physicians, groups of physicians, or a physician?

L. Joseph Parker: Absolutely. The AI was trained that if a physician is associated with a person with a felony conviction, then they are more likely to be criminal and are therefore tagged as criminal by the AI. One-third of all African-American men have a felony. Any black physician who has more than three male friends or family members has almost a 100 percent chance of someone having a felony conviction. And these AIs, they comb through property records, all kinds of ancillary records that are publicly available that can be fed into this thing.

So I’ll give you the example of Dr. Henry Lewis. Dr. Lewis was African-American, and if we search, there’s probably somebody he knows or someone in his family who has a criminal conviction. African-Americans are more highly observed, highly policed, highly prosecuted, everything. So a 5 percent difference at every step ends up with a greater than 50 percent difference. So for white doctors the odds of them being associated with someone with a felony conviction is less than half of what it is for black doctors. So Dr. Henry Lewis can be identified as criminal.

Now then it also looks at whether you own a nice car or a boat or lakeside or seaside property. Well, Dr. Lewis unfortunately owned a 1963 Rolls-Royce. Now, it cost less than a modern Ford F-150, but he liked old cars. I used to restore old cars too, I had an old Porsche at one time, a ’73. So he was restoring this car. So now you’ve got a doctor driving a Rolls-Royce, and he’s associated most likely with someone with a felony conviction. Now he’s brought up on charges.

So the government brought him and several other physicians up on a half billion dollar claim of fraud, and they lost against all the doctors. Which makes you question, what drove them to be so wrong? Why were they so convinced he was criminal? Because once this AI spits out your name, they won’t let it go. They will find some excuse to prosecute you. And an excuse instead of a reason. A reason to prosecute is, a crime has been committed, let us solve that crime and identify the person. An excuse to prosecute is, we’ve decided that this person is a criminal and we’re going to find something to charge him with and convict him of.

Kevin Pho: So why hasn’t there been more pushback against this, other than what we’re doing here on this podcast? Why haven’t we heard more about these AI algorithms to indiscriminately target physician practices?

L. Joseph Parker: The algorithms were kept secret. A lawsuit had to be filed by a very tenacious doctor, Dr. Neil Anand. He’s very good at data analysis, and he and his group were able to get this information. They kept it basically classified or confidential, not like when I was in the military as top secret, but they would say, by using a private entity to create the AI, the private entity could claim that this is proprietary information and therefore not subject to release.

Now, there are federal rules of evidence that say that any scientific metric that you use to identify and charge someone and try to convict someone must be made available to the defense. But they’ve been able to get around that rule by claiming this is proprietary. So they just tell the jury this doctor is determined to be criminal, and they throw out criteria for which there is no law, but it has a big impact on the jury, because you’ve got a data scientist up there saying, hey, we put this into the search engine and it spit out Dr. So-and-so.

And sometimes they don’t tell the truth about why they targeted that doctor. They will say, we got a call. But if you look back through the record, you will see they had already targeted and identified that doctor long before they had someone from their task force make a call for them. So just like the police used to make anonymous calls to themselves to get a search warrant, that’s exactly what they’re doing. They are having members of their task force call the DEA itself. So members of the DEA task force call the task force and make a complaint, to give them justification to investigate a physician that they’ve already targeted based on AI.

Kevin Pho: Now, what’s the situation today? I’m sure that the government is still using these algorithm-based processes to potentially target physicians who indiscriminately or maliciously prescribe. What’s the situation today in terms of using these algorithms?

L. Joseph Parker: It is still going full tilt. I just saw published, I believe yesterday, an article from the DEA that was justifying their use of this. And they will see it as an innovative tool, and it will sometimes help them detect a pill mill. But we have probable cause in this country, the Fourth Amendment protects us. Otherwise everything would be open to the authorities. Why can’t they just process all your bank records? Why can’t they just access, I would say access all your emails, but they do that for intelligence purposes, they’re not supposed to for criminal purposes. Why aren’t we just open books? Why do we have a Fourth Amendment?

If the only goal in this country is to stop crime, then we should live in Orwell’s 1984, where everything we do is under strict scrutiny. We don’t, because the founders of this nation felt that to be able to pursue happiness and live a good life you needed to have some privacy in your papers and things. So the Fourth Amendment is supposed to require probable cause before they can even look at you, and that’s what they’re circumventing with these AIs.

Kevin Pho: Now, is there a role for some of these algorithms or AI, especially as AI improves with these large language models and neural networks? Is there a role for these AIs, do you think, going forward?

L. Joseph Parker: Absolutely, if they had just told the AI to read these two textbooks and apply their definitions of the usual practice of medicine. Because these are written by hundreds of doctors who are specialists in these fields, and not a single expert’s opinion, right? Because you can find an expert who believes anything. There was an ER doctor during the COVID thing talking about how spike proteins cause sterility. Now, having practiced emergency medicine, I think you need a little more time in the immunology lab before you make those claims. But she has a right to her opinion and to voice it. But her opinion cannot be given the force of law, and that’s what’s happening in American courtrooms.

Doctors are going to court and they’re being tried not by the textbook but by a single physician’s opinion, who’s paid hundreds of thousands and sometimes millions of dollars to agree with the government. And the jury knows nothing about medicine, it’s very complicated. The DEA hasn’t gone through their four years of medical school, so they think they’re right, they think they’ve targeted a bad doctor. The jury thinks you wouldn’t be targeted unless there was something, where there’s smoke there’s fire. And then they get a quote-unquote expert to get up there and say that everything you did was wrong.

It’s amazing that Dr. Henry Lewis was acquitted. It’s amazing when any doctor is acquitted. But I’ll give you another example. A physician was convicted and pled guilty to fraud in sober homes. He turned in, quote, four other doctors, said they were corrupt and evil. They were all prosecuted, they were all acquitted. Yet the government gave that doctor, they cut his sentence in half, from 20 years to 10 years. So if they were acquitted in court, that means they are innocent, shouldn’t it? So why would you cut a man’s sentence in half for lying about innocent people? And that’s because conviction has become everything in our system. It’s not about justice. In a just criminal system, in a just court system, you would not need a defense attorney and you would still receive justice.

Kevin Pho: We’re talking to L. Joseph Parker. He’s a research physician. Today’s KevinMD article is “AI enforcement in health care: Unpacking the DEA’s approach to the opioid epidemic.” Joseph, once again let’s end off with your take-home messages to the KevinMD audience.

L. Joseph Parker: These AIs must be made public, their training parameters must be published, and they must be vetted and run through to see that they are accurate before they’re unleashed and determined to be probable cause for targeting charges. Otherwise we will have the destruction of American health care. The doctors treating the sickest patients are in the most jeopardy.

Kevin Pho: Joseph, once again, thank you so much for coming on the show, sharing your time and insight.

L. Joseph Parker: Thank you so much.