Doctors, Pharmaceuticals, and Statisticians

The results are hardly a surprise; they mirror results of a similar study of coronary bypass surgery (CABG) vs. drug therapy, exercise, diet and lifestyle changes. The CABG lost in that study. Sadly both patients and certainly hospitals prefer the quick, if expensive fix of a CABG or angioplasty. It always surprises me how many patients prefer a surgical quick fix to that of lifestyle change. I don’t know what the actual statistics show but anecdotally (my 30 years in healthcare, multiple hospitals) there are a large percentage of repeat CABG and angioplasties.

I do not understand your continual disparagement of physicians. Even before COURAGE was published, many physicians (including myself) would NOT have recommended angioplasty to patients with 70-85% coronary artery narrowing, especially if the patients were likely to comply with medical and diet therapies. The fact that some physicians do advocate angioplasty is due to recent changes: medical therapy has improved to the point where it almost always does as well or better than angioplasty. That was not true 10 years ago.

Some physicians believe that drug companies make too much money. So what. Almost all the general public believes that.

I see no collisions between doctors and statisticians. Every clinical trial published in first and second tier medical journals contains statistics, and most of those journals employ at least one biostatistician to verify clinical trial statistics. Medical students and residents learn about statistics, probability, Bayes theorem, sample size effects, etc. We conduct journal clubs to critically analyze medical papers and their statistics. The vast majority of physicians accept clinical trial statistics, as long as the trial methodology was appropriate.

I'm not a doctor (or a statistician!), but it seems to me that longevity isn't the entire story. Quality of life is also important. The story mentions that after three years, 67% of those in the angioplasty group were free of angina, compared to 62% in the medication-only group. That's not a huge difference, but I'd be curious about those numbers in other time frames, as well as other quality of life measures (e.g. ability to hold a job, walk some distance, etc.)

Jim

Dr. T,
It is interesting that you mention Bayes' theorem. Several years ago, I cited Bayes' Theorem to a Harvard-trained doctor in order to question his recommendation, and he went ballistic. He clearly had never heard of it, and he called my wife to tell her that her that he thought I was daft.

My experience is that some doctors know statistics--my doctor's oncologist, for example. But I think that there are many who do not, and they will not change their habitual approaches until they are hit over the head by a study in the press.

Ron wrote, "Sadly both patients and certainly hospitals prefer the quick, if expensive fix of a CABG or angioplasty."

As Arnold points out in his book Crisis of Abundance (which I just finished BTW), patients' preferences would probably change quite a bit if they weren't insulated from healthcare costs by employer insurance and government programs.

"Doctors also think that drug companies earn too much money. But it may be the doctors who earn too much money."

But statisticians do NOT make too much money.

The battle between doctors and statisticians has already begun. Check out this article: http://hypatia.ss.uci.edu/lps/psa2k/fifty-years.pdf

Here's a quote:

For a very wide range of prediction problems, statistical prediction rules (or SPRs), often rules that are very easy to implement, make more reliable predictions than human experts. This literature has been mistakenly
rejected by many otherwise reasonable folk (section 2), and it has been woefully neglected by contemporary ethicists, epistemologists and philosophers of science (sections 3−6).

The article is particularly hard on physicians and their stubborn refusal to even consider statistics.

Of course, doctors treat specific patients, not statistical patients. Even when a particular intervention might be no better on the average, over some universe of patients, than some cheaper therapy, it might be better for a specific patient (see BiDil). It may be difficult to discern, ex ante, which patients will benefit the most from the costly intervention (imagine if BiDil's differential performance by race had not been analyzed). If a patient can afford (or his third-party-payer will fund) some intervention, it may be reasonable for that patient (considering just his own interests) to request (gamble on) it.

Sure, a particular intervention might harm a patient rather than help him. But patients often face asymmetrical wagers. As scarhill pointed out, patients may rationally gamble on increased quality-of-life. There are plenty of interventions (e.g., coronary artery bypass) which seem to affect quality of life rather more than mortality rate. (Consider the extreme case of cataract surgery/lens replacement. It doesn't affect mortality much, but it affects quality-of-life a lot.)

I don't doubt that (statistically!) doctors' advice is tainted by their economic interests. It's hard to explain things like the proportion of Caesarian deliveries in America any other way, and for all I know, doctors may advise too many angioplasties. But you need stronger numbers than we see in the study you cited to make that case.

Mark Seecof wrote Even when a particular intervention might be no better on the average, over some universe of patients, than some cheaper therapy, it might be better for a specific patient.

One important point of the study was that no discernable difference existed between the two groups of patience. That would not be the case if a doctor could point to some trait in a patient which indicated that one treatment would be better for him than the other. The study didn't identify and such traits either. So without knowing such traits, the doctor should stick with the statistical odds.

However, 211 deaths versus 202 out of a population of 2,287 may not be statistically significant. (I could do the math but I'm too lazy.) Assuming the number of deaths is not statistically significant, it seems to me that the study says that the two treatments are equally effective. So the doctor should choose the cheaper one. However, what are his chances of getting sued if he does?

In response to Tyler's blog on French doctors income Marc Thomas and myself took two different approaches and found that doctor's income accounted for well under 10% of US healthcare spending. So even if you cut doctors income in half it would have a relatively small impact on total healthcare cost.

If you think that doctors earn to much what would you suggest is a reasonable level for doctors incomes. For example, a good but not a superstar
bond trader, portfolio manager or investment banker makes much more then the roughly $200,00 average income Tyler cited for doctors. Conservatives, libertarians seem to consistently take the position that CEO's high incomes are justified. But there are clearly reasons that neither CEO's and/or doctors incomes are completely determine in free markets. So why is one less then free market outcome OK, but the other less then free market outcome so bad?

Of course, if a doctor thinks drug companies make too much they can always invest in the drug company and earn some of those superior returns. But as with lawyers, it is essentially impossible for individuals to invest in doctors. Maye what we need is for some medical practice to follow wall streets example and go public.

As a provider who interacts with other doctors daily, I find this comment refreshing. I think it more likely reflects what should happen, but probably won't. Why? Because during episodes of care, patients are vulnerable, often highly devoted to their doctors, and lacking in real data. So their decisions, which influence the care delivered, are influenced by non-statistical data. For example, surgery has been described as the "ultimate placebo." [Please take a look at the long running controversy about the role of spine surgery in low back pain--i.e. Dr. Deyo's work at Univ of Washington]. Do doctors make too much money? Some do. Surgeons, oncologists, cardiologists, gastroenterologists, just to name a few. Some don't. Family practitioners, general internists, psychiatrists, pediatricians make that list. Physicians are notorious for not listening to statistical data, and instead using the argument: "In my clinical experience..." to fend off even the most convincing data. What to do? Pay more for things that have real value: Good primary care, good mental health care, and good preventive care. Will we do it? Of course not. I predict the collapse of the current system before another can take its place. By the way, University of Minnesota psychologist Paul Meehl, in a 1956 article entitled "Clinical vs actuarial prediction" made a very convincing argument for the use of statistical data in psychiatric clinical work--methods directly applicable to any area of medicine. A brilliant article, but has it influenced anyone??