"Years after the discovery that colorectal screening can decrease cancer incidence and deaths, few countries have adopted widespread colon cancer screening programs, although some are inching their way to that goal.

The reason, say many experts, is the burden that extensive colon cancer screening places on colonoscopy services. Behind every colorectal screening test, no matter what kind, is the potential need for a colonoscopy. If results from an FOBT, a barium enema, or even a flexible sigmoidoscopy to examine the lower colon are positive, patients must be referred for a colonoscopy that can view the entire colon and remove precancerous polyps, if need be. But many countries cannot yet fulfill that need, and such recommendations have huge implications for countries with national health care systems such as Canada and the United Kingdom."
-- Journal of the National Cancer Institute

Why does the United States spend so much more than other countries on health care? My conjecture is that Americans undergo more routine diagnostic procedures -- MRI's, fetal ultrasounds, colonoscopies, mammograms, etc.--than people in other countries. We have the personnel and the facilities to undertake such routine screenings. In other countries, where the availability of medical procedures depends on government policy, diagnostic tools are more carefully rationed.

I would like to see research that examines whether my impression is correct. And if the U.S. is more testy about health care, how does that translate into health outcomes?

In The Myth of Massive Health Care Waste and Do We Own Our Ailments?, I examined the conventional attacks on America's health care system and found them to be based on flimsy evidence. The data simply do not show that the difference between what we spend on health care and what other countries spend represents massive waste or inflated prices. Health care prices are high in this country, and I am sure that there is much waste, but those factors do not account for the fact that as a share of Gross Domestic Product, health care accounts for over 15 percent in the United States and less than 10 percent nearly everywhere else.

I have found only fragmentary data on the relative frequency of diagnostic testing across countries. For example, according to the joint Canada/United States survey of 2002-2003, 10.2 percent of Canadian women aged 50-69 had never had a mammogram, compared with 8.4 percent of U.S. women. Only 48.1 percent of Canadian women reported having a mammogram less than one year prior to the survey, compared with 64.4 percent of U.S. women.

Recently, at a social occasion, I met an American doctor, an internist. I asked her opinion about high medical spending in the United States. She cited MRI's, pointing out that when patients come in with lower back pain, she usually orders an MRI, even though the treatment would be the same regardless of whether or not the MRI shows a ruptured disc. Her rationale is the MRI is paid for by insurance (she would be more hesitant to order the MRI otherwise), and patients feel reassured.

Fetal ultrasounds are another diagnostic tool that is routinely ordered even though there is little or no proven value in the absence of other indications of problems. Ultrasounds are popular with prospective parents, and of course no obstetrician wants to forego any procedure and later face the wrath of an angry trial lawyer.

I should note that a fondness for ultrasounds is not unique to the United States. Some other countries' health plans call for one or two ultrasounds per pregnancy, although I have not been able to determine whether this in fact takes place. In any event, the United States has a higher rate of pregnancy than other countries, and for that reason we would tend to perform more ultrasounds overall.

A Basic Probability Calculation

To help prepare my statistics students for the Advanced Placement test, I recently gave them the following question (actually, the version below is slightly simplified):

Suppose that a medical test costs $1000, and 98 percent of the time it fails to turn up anything that would affect treatment. The other 2 percent of the time, it results in a treatment choice that extends life by 5 years. How much does a year of life have to be worth in order for the test to have an expected value that exceeds its cost?

The answer is that the "expected number of life-years saved" is .02 times 5, or one-tenth of one year. If one year is worth more than $10,000, then one-tenth of one year is worth more than $1000, so that the test is worthwhile.

Harvard economist David Cutler, author of Your Money or Your Life, argues that a year of life is worth much more than $10,000 -- perhaps ten times more. Thus, in my hypothetical example, he would contend that the test is worthwhile.

Continuing with my hypothetical example, suppose that this test were ordered for thirty percent of the American population. If it extended life by 1/10 of one year, then our average longevity would go up by just .03 years. Yet it would cost almost 1 percent of our GDP. Notwithstanding Cutler's arithmetic, such numbers would be used by critics to portray the American health care system as flawed.

A Specific Research Proposal

At this point, let me make a more specific research proposal, to try to get closer to an assessment of the cost-benefit ratio for diagnostic tests. I propose that for a sample of doctors in various countries, every time a test is ordered that costs more than $200, the following questions would be answered.

1. If you did not conduct this test, what treatment plan would you follow?

2. How might the results from this test alter your treatment plan?

3. What is the probability that the test will show the results that would alter your treatment plan: 10 percent or more? more than 5 percent, but not more than 10 percent? more than 1 percent, but not more than 5 percent? less than 1 percent?

My personal opinion is that in a non-life-threatening situation, such as back pain, ordering an expensive test that has less than a 10 percent chance of affecting treatment is a poor use of resources. On the other hand, if there is even a .1 percent chance of detecting cancer early so that it may be treated, a test may be warranted.

The purpose of the research that I propose is to see how well actual practice conforms to my views on cost-effectiveness. There are a number of ways in which it might fail to do so.

For example, doctors may mis-estimate probabilities. Suppose that out of 1000 instances in which they estimate that the probability of a treatment-altering result is more than 10 percent, only 20 tests turn up the result which alters the treatment plan. In that case, doctors are probably ordering too many tests.

On the other hand, suppose that doctors are accurate in their probability assessments. My conjecture is that patients in other countries and uninsured patients in the United States are underscreened for cancer and heart disease. However, for insured patients with non-life-threatening ailments, my guess is that American doctors order more tests than are cost-effective.

In order to make cost-effective decisions, doctors and patients will have to make careful calculations based on probabilities. Given the rapid growth of health care expenditures, I believe that we need the sort of research proposed here in order to improve the accuracy of such calculations.