Monday, December 21, 2009

Mammograms Revisited

Richard Thaler (coauthor of Nudge; I'll reference him often) wrote a piece for the NY Times about the current mammogram controversy. It's worth a read. He doesn't talk about "saving lives", he brings up the problems with prostate cancer treatment, and he provides some interesting numbers about the likelihood that breast cancer is not fatal and mammograms don't detect the types that are fatal, though not the expected life spans of women who are treated for breast cancer. I have a little commentary to add.

The stated false positive rate is 10% per test (which is huge for something with such serious consequences). For the math impaired, here's how that converts into the probability that women are likely to get a false positive over a decade. The simplest way to do this is to focus on the 90% chance that each test does not give a false positive. We use what is called the Multiplication Rule for Independent Events. For ten years, we multiply .9 (90%) by itself nine times: .9 * .9 * .9 * .9 * .9 * .9 * .9 * .9 * .9 * .9. This is more simply written as .9 ^ 10, and the product is about .35. So, there is a 35% over ten mammograms that a woman will not get a false positive, and a 65% chance that she will get a false positive. Of course, this assumes that there are no characteristics of a breast that would make it prone to false positives, which I imagine is unlikely. The 10% comes from aggregate data. I think it is more probable that women who produce false positives at age 40 are continuously more likely to do so each year than women who did not have false positives at age 40. Without the longitudinal data in front of me, I remain skeptical of the appropriateness of multiplying the false positive probabilities as though they are always independent events.

Thaler writes a little about the costs of all these mammograms in terms of unnecessary treatments that cause side effects, and the stress that women feel when diagnosed with breast cancer, but he avoids using dollar amounts to describe all the waste. What is missing is mention of opportunity costs. The cost of all the unnecessary procedures is not just the dollars exchanged for them, but also the other things we can not buy because our finite dollars were poorly allocated. I covered in my last post the idea that we could extend several times as many life years for Americans by taking the money spent on mammograms for women under 50 and spending on other endeavors, such as nutrition programs for children, or having nurses visit patients after operations to reduce treatment noncompliance complications. It is vital to consider opportunity costs. Mammograms for women under 50 don't just allocate $2 billion per year in exchange for a few thousand life years. They also cost us many thousands of other people's potential life years, and the unnecessary treatments and stress result in much lost work productivity. The practical net result, the way I see it, is that our system is effectively just killing other people early by insisting on giving mammograms to women under 50 who lack high risk indicators. Are women in their 40s more important than everyone else?

Many people have lost or almost lost loved ones to breast cancer. It is okay to feel empathy for them. If you are one of them, I feel empathy for you. That is not justification to ignore the data we have and allocate our limited resources on inferior practices. Emotions limit our perspectives and distract us from the ultimate goal of doing the most good we can with the resources we have. Please, let the cooler minds prevail and establish a system that gives America the greatest benefits it possibly can.

Autism Prevalence

A CDC report about autism has lead to some dramatic reactions. Here are some of my thoughts about this "1 in 100" "epidemic".

Summary, in case you want to skip the rest: Diagnosis got "better" for a while, but now I think we're overdiagnosing. There is really poor standardization in the medical field, and practically no oversight. Doctors refuse to let "bureaucrats" (epidemiologists, statisticians, and scientists, really) get in between them and their right to do whatever they feel like to their patients. New Jersey and Missouri have relatively high rates of ASD diagnoses because of nonstandard diagnosis practices. I expect the diagnosis rate to continue to increase, and that this is inefficient overall.

Firstly, remember that autism is not a disorder. There is a spectrum of severity of different features. To get an autism spectrum disorder (ASD) diagnosis, you don't need all the features, and they don't have to be severe. The primary feature involves impaired social interaction. Repetitive behaviors and sensory sensitivity are also common.

One big reason for the rise in the rate of autism diagnoses is that kids who would have otherwise been identified as mentally retarded are now being correctly classified. That is good because more helpful interventions can be assigned when diagnosis is accurate.

Unfortunately, another big reason for the rise is overdiagnosis. ASD is popular and famous. Neurotic parents whose kids aren't meeting expectations, who aren't the superstars the parents wanted, are looking for reasons and targets of blame and sources of hope. A lot of doctors just don't understand what they're doing, want to appease neurotic parents, practice defensive medicine, and benefit from making referrals to friends or businesses they have stakes in. The same nonsense that we see with ADHD and Bipolar diagnoses is happening with ASD. Little Billy likes to play by himself? ASD! Off to the occupational therapist! Little Billy has a slightly smaller vocabulary than his peers? ASD! Off to the speech therapist! Now, there are often cases in which some intervention would legitimately help a child, but interventions are expensive and different third-party payers have different diagnosis requirements for covering them. Well-meaning doctors commit a lot of insurance fraud, handing out inaccurate diagnoses in order to get a payer to pay for a useful intervention. This can be good for a client, but it is bad for our system. I generally see an enormous amount of money spent on small improvements.

The CDC report pertains to 8-year olds. There is generally a spike in diagnoses around age 8 because that is the transition to a developmental stage in which certain skills (reading and writing) are expected, as well as levels of self-regulation and interpersonal behavior. This is when school puts on more responsibility. Many learning disorders are suddenly noticed around this age. It makes sense for the CDC to use this age as it is likely to include cases caught late, though most autism should be recognized by age 3, when language is rapidly developing. A few teachers over the years increase the odds of recognizing problems.

What would be very interesting to look at is how many of these kids meet diagnostic criteria at age 12 or 18. There are a lot of reasons that kids develop skills and behaviors differently from each other, but a lot evens out the older they get. Many people carry around diagnoses like tattoos long after their natural developmental processes or interventions brought them into the normal range. All it takes is one doctor to use a label and a person who doesn't know better believes he's disabled for life, no matter what improvements are made. Anecdotally, I am familiar with a little girl who is almost three. She is quite a chatterbox now, but a doctor flagged her a year ago as having a low vocabulary. He offered the mother a referral to a therapist who would help the girl develop an average vocabulary for her age. Quite ridiculous and unnecessary and expensive. This was a doctor in New Jersey, by the way, which has the highest ASD rate in the country. Perhaps there is a medical culture in NJ that tends towards overdiagnosis. Perhaps their medical schools need to teach statistics better.

There is a lot of variance out there. People are different from each other. Every skill or characteristic of a person lies within a distribution. Half of people are below the average, and half above. There is a statistical term called "standard deviation" that describes how far from the average a person is relative to how widely people vary from each other in general. Most people are within 1 standard deviation of the average. There is precedent for saying that a person has a disorder when a characteristic of health or function is more than two standard deviations worse than average (a bit more than 2% of the population). Mental retardation is considered when a child's IQ is two standard deviations or more from the average. Many psychological tests flag problems when a person scores more than two standard deviations from the average on scales related to depression, anxiety, etc... But that alone should not be used to demark disorders or assign diagnoses. There has to be a convergence of information showing meaningful dysfunction.

Based on this precedent, we should not be surprised if the ASD diagnosis rate approaches 1 in 40. Kids who are socially awkward enough to rate as two standard deviations away from average on whatever metric someone makes up (there are a bunch of ASD symptom checklists out there, and they are not used well) would qualify for ASD. Asperger's and "high-functioning autism" are ballooning. We've gone past the point where formerly-mentally-retarded kids are being correctly identified with ASD, up to the point where formerly normal kids are being dragged into disorder land. They were probably called "weird" before, but is that something we need to fix? Is it really a problem?

I can't help but compare Asperger's to homosexuality. It was not long ago that homosexuality was considered a mental disorder. Now the legitimate medical community accepts homosexuals as healthy people who are just in a minority on a characteristic. We are surrounded by people who meet criteria for Asperger's, but they do well in academics, science, and technology. They are computer programmers and engineers and scientists and professors. Do we really need to say they are disordered because they're socially awkward and repetitive? Criteria C is redundant because criteria A necessarily involves social impairment. Are we really helping by imposing labels and interventions on these children?

Maybe. I don't know. I hope someone is recording a lot of data, and that there are enough "natural experiments" going on to eventually give us the answers. We definitely do need better standardization of diagnosis practices.

Reblog this post [with Zemanta]

Thursday, December 17, 2009

Prescription Information Availability

I got a little excited when I read that Congress might limit Big Pharma's ability to use doctors' prescription behaviors for their marketing. Of course, the proposed idea was quickly eliminated.

Pharmaceutical companies spend incredible amounts of money on marketing. Besides outright bribes to doctors with free food and paid speaking opportunities (averaging $3900 per targeted doctor), the marketers use information on which doctors are prescribing what and how often to custom-tailor their pitches. They can tell which of their sales techniques are most effective by looking at the spikes in prescriptions after each sales pitch. They know if a doctor was a waste of bribes, and they can focus their efforts on the ones who respond well to incentives.

The companies will claim that their behavior is to help keep doctors up to date on what works, but that is an outright lie. The marketers push the new, still under patent, expensive drugs over the cheaper drugs regardless of effects. Expensive drugs that do not work better are touted. There is no good reason that doctors should listen to the obviously biased salesmen of a drug instead of looking at peer-reviewed research articles in selective journals. There are problems even with those articles, but they are better.

I have attended drug rep presentations at medical facilities. They use anecdotes about outlier cases to hype up the drug, and lay out some lunch and branded office supplies. The MDs in the room didn't look at the fine print, which revealed to me the variance in the drugs effects (40% of participants for one drug got worse, and those who got "better" were still severely ill, so why use this expensive drug with lots of side effects?). A medical degree is no guarantee that a person is going to be careful or attentive, or even understands statistics, or keeps up to date with research. There is practically zero oversight of doctors in most settings because they run the show and they only listen to each other. Doctors are fallible and subject to manipulation. Smart hospitals have banned drug reps.

I am all for accurate and complete information about drugs being disseminated to prescribers. It should be done by unbiased parties, and be presented in a way that doctors can understand, comparing the risks and benefits of the drugs, and I would also include the costs. The PDR is obviously inadequate, and tends to just collect dust on a shelf. No drug marketing should be allowed, ever. Until that happens, let's keep prescription information from the marketers so they are less able to target unethical and mentally weak doctors.

Consumer Reports, an independent organization, may be a good resource. They have a free website about drugs. For example, recent 60-study analysis showed that $10/month Doxazosin is as effective as $246/month Flomax. Flomax is heavily advertised because it is new and expensive, not because it is more helpful.

Sunday, December 13, 2009

Rationing is Ethical

This is what I'm talking about. Allocate our finite resources in such a way that they bring our whole country the greatest benefits. It is completely detrimental and unsustainable to keep spending huge amounts of our resources at the end of life. I want my tax dollars and insurance premiums to result in the greatest possible increase in quality-adjusted life-years, though I understand the inherent subjectivity in determining how to measure or weigh "quality". Even ignoring quality and focusing on life-years would be a vast improvement over our current system. $80,000 of Avastin for a few more months of life? That money could lead to dozens or scores of extra life-years if spent elsewhere. If we, as a country, invested more in childhood nutrition, health education, and smoking cessation instead of pharmaceutical and biotech companies, we would have greater national health and longer, happier lives.

Saturday, December 12, 2009

Small Businesses

I want to start a small business. I'll start out with a few employees, and try to grow. How is my business affected by the various plans proposed for health insurance in America?

Status Quo (most states): Since I'm small, I don't have to give my employees an insurance plan. They pay for their insurance out of pocket, if at all. Some will likely not get insurance so they can spend their money on other things, like bigger houses. Those with insurance will overpay for their coverage because they can't negotiate, and will probably be underinsured. I may have to pay slightly higher salaries to make up for the lack of benefits in order to attract good workers, but since people generally don't understand the values of employer insurance packages, I won't have to pay much more. My costs are kept low, and I am better able to function as a small business. Republicans and Libertarians like this arrangement because it helps small business development and entrepreneurship, even though it exploits workers a bit and sets up people for hardship and bankruptcy if they develop health problems. Half of US bankruptcies, largely the cause of our economic collapse, are due to inability to pay medical bills.

Individual Mandate: My employees will be required by law to buy their own insurance. They are likely to buy catastrophic coverage with the lowest premium and least coverage. They will probably be underinsured. The minimum wage is the same, and my competition could consider ending employer-provided insurance, so I won't have to worry about paying much more in salaries, if any more at all, to make up for my employees' health insurance costs. Insurance companies love this arrangement because the government is forcing people to give these companies money, and the government will probably pitch in for people who need help paying. Free money for insurance companies, cheaper labor for businesses. Real Republicans and Libertarians don't like the government interference (most Rs are bought off by insurance lobbyists and have no values), and liberals don't like the poor quality of insurance and the burden on the working poor who may not qualify for enough subsidies.

Employer Mandate: This is my nightmare. The government would force me to pay for my employees' insurance. Since I have a small business, I don't have leverage to negotiate a good price. There are some small business collectives, though, that can negotiate together, depending on my state. I sure wish there was a national insurance market so the collectives could negotiate more freely. I have to deal with the local insurance monopoly. Combined with minimum wage laws and competition with larger businesses that can negotiate better, paying for health insurance plans for my employees may inhibit my small business's growth, or even make it unsustainable. My business may not last long, or even get started. It's a shame because consumers would really benefit from my business's existence, but not enough to justify paying the prices I would have to charge to stay in business if I have to pay for health plans. This plan also leaves unemployed people in the lurch, and when my business fails my employees will have nothing. Maybe my business can survive if I fire some people and work the others harder. Big businesses like the employer mandate because they already give their employees well-negotiated health plans, and this would make it harder for new competition to sprout up.

Public Plan: Ah, I could relax. Everyone has basic health coverage. They can buy fancier private plans if they want, but the basics are covered, and are better than the old private catastrophic plans. I am more likely to be able to pay competitive salaries or reinvest profits. My employees are secure. Real Liberals (not the bought-off ones), unions, health economists, and the poor favor this idea because it provides security and needed health care to all Americans, and improves the stability of our nation's economy while setting a stage that fosters small business innovation. Republicans and Libertarians hate the plan because it is government control (they blindly hold on to a belief and a value despite all the harm their decisions cause), and insurance companies hate it because they love the profits they get from our current horrible system.