Tuesday, February 28, 2017

The Radiologist Protection Act of 2024

Medical X-Ray imaging ALP02 nevitI think the single most telling political fight likely to happen in the near future is how radiologists and pathologists will respond to AI. This is a fight that will help define the future of health care, class, politics, and the nature of employment.

Radiologists are some of the highest paid and most educated workers in the country who are performing a task ripe for replacement by AI. At the most basic level, radiology is about using high-level pattern recognition to diagnose diseases, and pattern recognition is something computers are getting dramatically better at every year. Already, programs can perform as well or better than top doctors in these types of tasks.

According to one study, "The retrospective error rate among radiologic examinations is approximately 30%, with real-time errors in daily radiology practice averaging 3–5%." Given the current error rate among human radiologists and the speed at which technology is improving, it is easy to see how using AI could soon not just be dramatically cheaper and faster but significantly safer.

It might be five years, ten years, or twenty years, but there will be a point where having a human read your test results instead of an AI-powered program will seem even stupider than hiring a person to multiply several large numbers with pen and paper (which was once a common job) instead of using a calculator. It would not just be more expensive but extremely dangerous to use a human.

It is not a question of if most radiologists can be replaced by machines, but a question of what they will do when it happens.

So far, groups in the health care industry have proven to be very adept at lobbying for laws that protect them from competition or grant them special privileges. That is partly why it is such a rapidly growing segment of the economy with the greatest number of highly paid workers. That is why I have no doubt that when the time comes, radiologists will fight to keep their highly paid jobs either by stopping these programs or requiring a highly paid radiologist to needlessly rubber stamp results from a computer.

I suspect radiologists are probably the single group best positioned to win a fight against being replaced by AI given their wealth, prestige, and lobbying power. Additionally, it is hard to dismiss them with the advice offered to most workers being displaced by technology:  "Just get more education."

If radiologists do get their Protection Act, it will serve as a model for other occupations. Jobs and salaries aren't based on usefulness but on political clout and defined privileges. We could see a future that more resembles a medieval guild-based economy than a free market. If, on the other hand, radiologists lose, I don't see any group stopping the robots.

Wednesday, February 22, 2017

The danger of the billionaire strawman argument

I see a very concerning strain of thought among some Republicans, Democrats and even liberals that everything should be means-tested as opposed to universally available. They often trot out what I call the "billionaire strawman argument," which I consider very destructive for long term progressive ideals.

For example, here is Hillary Clinton using it against Bernie Sanders' plan for universal higher education. From TPM:

“I believe that we should make community college free. We should have debt-free college if you go to a public college or university. You should not have to borrow a dime to pay tuition," Clinton said. "I disagree with free college for everybody. I don’t think taxpayers should be paying to send Donald Trump’s kids to college."

So rather than supporting tuition-free higher education for everyone, Clinton endorsed debt-free tuition. While it's theoretically true that anyone could apply to a public university, it is worth pointing out most billionaires' children are probably going to a private university, which weren't included in Sanders' plan. Ivanka Trump went to UPenn, and Eric Trump went to Georgetown.

Here is the billionaire strawman argument again, recently used against House Republicans' new ACA replacement proposal in the New York Times.

"That means that the biggest financial benefits would go to older Americans, like, say, Secretary of State Rex Tillerson. If he didn’t have a job in the Trump cabinet and access to government coverage, a 64-year-old multimillionaire like him would get the same amount of financial assistance as someone his age, living in poverty, and he would get substantially more money than a poor, young person."

Again, this is theoretically true -- Sanger-Katz is simply pointing out how the program could work. However, I doubt you could find even a dozen working-age Americans of Tillerson's net worth who don't have employer-provided insurance.

There are many good reasons to oppose this House Republican proposal; for instance, it wouldn't provide enough funding or enough consumer protections. But the fact that Republicans are calling for a law that would make everyone entitled to the same government assistance to afford health insurance is a good thing. It would be a paradigm shift in the role of government in health care that not even Obamacare made, which is why some conservatives are so opposed to it. Progressives should not oppose a universal tax credit; they should only focus on making it large enough.

Worrying that universal programs might benefit a few billionaires will cause major technical and political problems. 

First of all, there simply aren't that many truly rich people who would actually take part. Only 1.5 percent of the country makes more than $250,000 a year, so excluding the truly rich for any universal program would save very little.

Any attempt to exclude people means you need a bureaucracy to verify eligibility that could cost more than what you even save. A means-testing bureaucracy can also find it very difficult to provide everyone with what they are supposed to receive.

To actually save any money by excluding the rich, you need to also exclude many middle class families. This means some will fall through the cracks. That is what the Affordable Care Act did by denying tax credits to those only making more than 400% of the federal poverty level. This is why under our "universal health care plan," a 62-year-old making $48,000 would need to pay 20% of their income just to get basic coverage.

The political problem with the billionaire strawman

There is a reason our most popular government programs are universal: Medicare, Social Security, public highways, public schools, national parks, etc... When everyone is technically part of it, everyone has a political stake in its success. It is a much simpler argument to make that if we think something is essential, we should provide access to everyone.

When a program is universal, you talk about making it better for everyone. It is easier to rally a large number people for expanding Social Security or having Medicare cover prescription drugs. It is much harder to rally popular support for, say, increasing ACA subsidies for those in the 300-380% FPL. When you means test, you also leave a program open to a death by a thousand cuts and restrictions.

If you support single payer health care, you support everyone from the poor to the super rich getting access to the same benefit. That should never be seen as a bug in any policy, that is a feature. 

Thursday, February 16, 2017

Why I'm a technology optimist: malaria vaccine

This is possibly the best news for humanity as a whole in the past 10 years. We are now closer to a viable malaria vaccine. From the NIH:
In 2015, 212 million cases of malaria occurred worldwide, and 429,000 people with malaria died, largely African children under five years old, according to the World Health Organization. Although only 1,500 to 2,000 cases of malaria are diagnosed in the United States each year, the disease is a concern for international travelers, aid workers and military personnel worldwide.[...]

The investigators report that the vaccine candidate was well-tolerated and safe with no serious adverse events. Among the 40 participants who received five placebo doses, 93 percent (37 participants) developed P. falciparum malaria infections; by comparison, 66 percent (27 participants) of the participants who received five doses of the PfSPZ Vaccine (41 participants) developed malaria infection. Based on the primary study analysis, PfSPZ Vaccine demonstrated a 48 percent protective efficacy by time-to-first positive malaria blood smear and 29 percent efficacy by proportion of participants with at least one positive malaria blood smear during a full 20-week malaria transmission season. By both measures of protective efficacy, there was statistically significant protection in the vaccine group as compared with the placebo group.

I spend much of my day studying political news, policy news, and science news. The political and policy news can often be very depressing, especially lately. But the latest scientific development are an endless source of hope for me and should be for everyone.

Being able to more effectively deal with malaria, one of the oldest and possibly deadliest diseases to ever plague our species, would help millions. It would be one of the greatest ways to improve human welfare ever.

I believe the future can be better because I can see how scientist successfully working on ways to make it better.

Thursday, February 9, 2017

A sign Trump may not be deadset on destroying Obamacare (at least not yet)

The Trump administration is thinking about letting insurers on the Obamacare exchanges charge older people more and younger people less -- the move would require a very twisted reading of the Affordable Care Act. According to the Huffington Post, “Today, premiums for the old can be only three times as high as premiums for the young, which is what the Affordable Care Act stipulates. According to sources privy to HHS discussions with insurers, officials would argue that since 3.49 'rounds down' to three, the change would still comply with the statute.”

Nicholas Bagley argues this move would be illegal, which it probably is -- but it could actually increase the number of people using the exchanges next year. Given the makeup of the ACA exchange population and the convoluted way affordability tax credits (APTC) are applied, this move could have a modest net gain in exchange enrollment. Using data from the Kaiser Family Foundation Health Insurance Marketplace Calculator makes it easier to understand why.

Subsidies under Obamacare are designed so that you pay no more than a set percentage of your income for the second cheapest silver plan, as long as you make less than 400 percent of the federal poverty level (roughly $47,000). For example, as a person making $46,000 a year, you will get as large of a tax credit as necessary so that you are spending only 9.69% of your income ($371 per month) on premiums to get a silver plan. Whether your premiums are technically $400 a month or $3,000 a month has little impact on how much you pay.  Since average silver plan premiums for adults 50 and above are well over $371 per month, all older people who currently qualify for tax credits on the exchanges are effectively protected from any premium increase that would arise from this Trump policy change. It would cost the government more -- not these consumers. 

This new "age band" of 3.49-to-1 would matter to older people on the exchanges who are making over 400% of FPL because they get no subsidies. However, almost no one making over 400% FPL actually uses the exchanges. According to CMS, last year just 2% of people who bought coverage on the exchanges made over 400% of FPL. On net, this means relatively few older people using the exchanges would be impacted significantly. (Note: There are many who do buy individual coverage off the exchanges who could be affected, but this group is not well studied. In general, the higher an individual's income, the more likely they are to have employer-provided insurance.)

On the other hand, this change in the age band would reduce premiums for younger people using the exchange -- and the change could impact a large number of them. Since premiums for younger people are already lower, they are less likely to qualify for subsidies. For example, an average 25 year-old making only $36,000 would receive no tax credits. In addition, since actuarially speaking, young people use about one-fifth as much health care as older people, this change could make insurance seem like a better deal, causing more young people to sign up next year.

A Rand analysis from two years ago projected that a much larger increase in the age band (making it 5-to-1) would on net increase the number of people using the exchange -- more young people would get insurance, and there'd be only a relatively small drop in coverage among older wealthier individuals.

Democrats initially chose the 3-to-1 age band as a compromise, back when they envisioned the ACA exchanges being a much larger, more popular, and transformative policy. They envisioned the exchanges as a tool that would become popular with people of all income levels and businesses. The exchanges, though, haven’t turned out to be nearly as popular as once predicted, and the program is now primarily for people of limited means. With that in mind, increasing “premiums” on those on the exchange who aren’t paying the full price anyway and lowering premiums for those who are doesn’t seem like a terrible move.

The actual math is a very wonky point of debate. However, it's worth noting that the Trump administration is considering policy changes that might actually end up growing the exchange population, improving the risk pool modestly, and/or reducing the chance that insurers will drop out of the exchanges next year. Based on earlier actions, there was legitimate concern the Trump team was going to try to completely sabotage the exchanges from the inside. Now the picture is more complex. Perhaps it is a sign the Trump team is starting to realize the political quagmire health care policy can become.

Monday, February 6, 2017

Eliminating the filibuster would be a gift to red state Democratic senators

Capitol-SenateDemocrats are planning on filibustering Neil Gorsuch's nomination to the Supreme Court, but there are some Democrats from states won by Trump who fear this move will result in Senate Republicans getting rid of the filibuster entirely. This concern is misplaced, since if Republicans actually do use the nuclear option, it would be a real political gift to these endangered Democrats.

Currently, Senate Democrats' ability to block the GOP agenda is akin to Schrodinger's cat: It both exists and doesn't exist, and we won't know which it is until Democrats actually try to use it.

From a political perspective, trying to maintain this illusion of possible power is worse than being exposed as having no power at all. The base will get very angry at Democratic senators who don't take part in a filibuster, even if these Democratic senators are convinced it won't work. Under this current dynamic, voters will put some of the blame on red state Democrats for everything that does or doesn't happen in the Senate, even if these senators actually don't have any real power.

On the other hand, if the filibuster is eliminated, Senate Democrats are freed from blame or responsibility. They can clearly and easily say everything is the Republicans' fault. They can honestly tell their base they did everything they could. This is the reason you don't see any liberal base pressure being put on House Democrats, because it is known they can't do anything.

Because of the way Merrick Garland was treated, Senate Democrats have the best moral and political case for a filibuster they will ever have. If they do filibuster Gorsuch and it works, they score a big win. If Republicans eliminate the filibuster, they at least free themselves from blame. Not filibustering, though, is the worst of all possible outcomes. It achieves nothing and makes them responsible.

Wednesday, February 1, 2017

Can anyone beat the gatekeeper?

Samsung Galaxy S II in hand
Google Maps now lets you request Uber or Lyft directly via their app. Since people are likely to first look up an address via Google Maps, this makes it more convenient to order a car via Google Maps than through either company’s own app. It takes fewer steps on the phone, and in the app game, that is very significant.

At the moment, this isn’t a problem for Uber or Lyft because Google doesn’t offer rideshare services, but Google’s parent company is investing heavily in self-driving cars, so it is likely at some point they will. That raises the question: why would anyone bother using Uber or Lyft after Google's AI-powered car fleet comes online? To beat Google, which will be the more convenient option for customers, they are going to either need to offer significantly cheaper services or significantly better services. Given how much money Uber has already burned through, I don’t think they could ever beat Google in a straight up price war if Google wanted that. Since people also seem to only care about getting from point A to B  as quickly as possible, I don’t see how either company really competes on service.

This dynamic highlights what I think could be the biggest economic issue that will only grow as we rely more and more on technology, which is that of beating the gatekeepers. People spend a huge amount of their time and money online, but almost all of that is directed through a very small number of companies. Almost all phone/tablet operating systems are run by Google, Amazon, or Apple. Almost all links people click on are via Facebook or a limited number of other social media.

To succeed, digital companies are going to either need to be involved in a business these companies don’t want to bother with or do a dramatically better job, since they will almost always be the less convenient choice. For now, ride sharing is something the gatekeepers don’t want to bother with, but it seems like way too lucrative a market once self-driving cars are good enough.