Statistics lead to frustration and falsehoods: "Facts and figures"
t's 2 AM. The telephone rings. Roused out of his sleep, the man in the bed lifts the telephone receiver, listens, and then shouts into the receiver: "You got the wrong idiot, you number!"
When dealing with numbers and statistics in medicine, and in our case, in medical imaging, this slip of the tongue very often comes into my mind. I was reminded of it in particular when preparing a lecture on "Figures & Facts" for a recent congress. When I received the invitation to give this talk I thought I had finally been given an easy topic. It is not so difficult to find data on the number of MR imaging, CT, ultrasound (US), and nuclear medicine machines worldwide, how many examinations are performed per year, and how much money this costs; and you can easily compare these numbers with those of other imaging modalities.
This was the first mistake. You hardly find any figures on medical equipment or procedures in the scientific literature, although there are some in health/ general economic publications. However, even their data are scarce and mostly outdated. When you approach people selling equipment or accessories, many of them either do not have a good overview of the marketplace or they keep the numbers secret. The last resort is so-called “intelligence” companies, which gather data for the industry or other interested parties. To get some of their figures and predictions, you have to pay them between US$ 1,000 and US$ 4,500.
All numbers cited in this column are from sources commonly considered reliable. However, when I compared data compiled from different sources, I discovered that it is not so easy to find exact numbers, and those you find are wrong.
Now you ask: “How do you know that they are wrong?”
For example, one source says there are 7,000 MR imaging systems worldwide, the next says there are 9,000 in the three largest markets combined. A third source states there were exactly 6,678 MR units installed worldwide in 1995. Trusting the latter one is like believing the numbers in statements by politicians or published in clinical studies: exact numbers must be wrong, and percentages are even worse.
It seems that health politicians, trade unionists, journalists, lawyers, health system planners, people talking about cost containment, medical functionaries, company people, and medical researchers (in this order) often get their numbers wrong.
Taken together, physicians in the United States, Japan, and Europe operate approximately 150,000 ultrasound, 20,000 computed tomography, 7,350 nuclear medicine (mostly SPECT, single photon emission computed tomography), and 9,000 MR imaging systems. Anually, these machines perform some 115 million ultrasound, 63 million CT, 35 million nuclear medicine, and 17.5 million MR examinations.
For the sake of finding a common denominator, let's assume the numbers given in the last paragraph are correct within the medical range (± 25%). Then you can start looking at them in detail and assessing them.
In 1995, Germany had 24,000 US machines, France 12,750, Italy 12,400, Great Britain 9,350, and Spain 6,380. There were 1,350 CT units in Germany, 514 in France, 350 in Great Britain, 266 in Italy, and 190 in Spain. Germany had 365 MR machines, Italy 210, France 125, Great Britain 147, Spain 131.
In another source, the figures for MR equipment read: Germany 560, Italy 302, Spain 179, and Great Britain and France same as above. A third source states that there are more than 800 MR machines in Germany. This means that the figures for Germany alone differ by more than 100%.
If you turn these numbers into imaging units per head of population, Germany stays the leader with 293 US units, 16.5 CTs, and 4.5 (or 9.6) MR machines per one million inhabitants. This is followed by Italy and France in close competition with each other; and, after a while, by Spain and Great Britain.
Thus, the statistical notion (or prejudice) that the per capita density of sophisticated medical imaging equipment changes negatively in proportion to the per capita consumption of red wine is wrong; it decreases proportionally to the consumption of stale ale to the northwest of Europe and red wine to the southeast.
Even worse for some of my friends with deeply rooted preconceptions: Because more high-technology equipment is sold in Northern Italy than in the south, this region has a higher density of medical imaging equipment than most countries of the European Union.
Health-care statistics can also be used to find out the reasons for differences between countries, or to reach unsuspected conclusions: the Germans have one ultrasound machine per ten physicians; in Spain, 21 physicians share one US unit; you see more bullfights in Spain than in Germany. This must be because Spanish physicians train with bulls first to fight for the use of the few US machines.
Gathering the figures and getting the facts – or, even worse, making predictions for the development of different imaging modalities or individual markets or companies' market shares is difficult because collecting and processing the numbers is a doubtful business. The penetration and acceptance of a technique or modality do not seem to depend much on scientific results or cost considerations, but rather on local culture, politics, and lobbying within the healthcare system.
In the first half of the 1990s, and increasingly in the second half of the decade, health politics have stressed cost reduction. Although this approach showed limited success from the beginning, medical imaging has been relatively hard hit. The impact was mainly on large capital purchases (= heavy equipment such as MR imaging, CT, and angiography equipment) as well as bulk sales through healthcare purchasing groups.
X-ray accounts for more than half the worldwide sales of imaging equipment. In Europe, sales of x-ray equipment were only affected during the first half of the 1990s and have since recovered. Both the x-ray markets in the U.S.A. and Japan have increased considerably during this period. The values of the CT and MR markets have risen again after an earlier fall, and the ultrasound and nuclear medicine markets have been effected but show signs of recovery. Unit sales of ultrasound equipment have steadily grown.
Let's return to our numbers. When numbers are processed, incredible results are often obtained, but they might be true. The differences in the utilization of imaging equipment are not as pronounced as those in the regional distribution.
Whether you are in the U.S.A., Japan, or in one of the European countries mentioned, the annual number of ultrasound procedures per machine ranges between 710 and 790, with Germany being at the lower end and Japan at the higher end. Annual MR examinations in the U.S.A. range between 2,300 and 2,600, in Europe between 3,100 and 3,300. The numbers for CT examinations are higher in the U.S.A., amounting to 3,500 per year and unit; but in Europe, CT examinations balance with MR examinations, with 3,000 to 3,200 examinations per machine and year (oops, there goes the argument of higher patient throughput in CT; everything depends on how many machines there are in a region and how the machines are used).
If you calculate on the basis of 250 working days per year, you get the most interesting result of this excursion into statistics: the average number of patients examined in a CT or MR unit is about 12 per day in the U.S.A. or Europe. The average number examined with a US machine is two to three. Even considering that there might be an error rate of 50%, the number of patients examined by each ultrasound unit per year is extremely low. Ultrasound is supposedly versatile, fast, and mobile, but apparently the equipment is only marginally used.
Why do we see this discrepancy?
The answer seems to be hidden in the different user populations. CT and MR machines are only used by specialists educated and trained to perform and assess such examinations. Thus, it is possible to get accurate figures and facts on the utilization of this kind of equipment. Utilization of and indications for ultrasound are diffuse. In the best case you get some soft data of the utilization.
Ultrasound can be performed by any physician, and in some countries no training or only introductory courses are required. The price of a state-of-the-art ultrasound machine is affordable at one sixth of a CT or one tenth of an MR equipment (of course, you also can acquire ultrasound equipment for the price of a CT).
Furthermore, there are severe restrictions on buying heavy equipment. To acquire MR, CT, or angiography equipment, you need permission in many countries; and permission will not be easily granted, at least not by the politicians and the cleaning lady on the supervising board in charge of new acquisitions.
On the other hand, ultrasound machines can be freely sold and bought. In many countries the market values of ultrasound sales are at least the same or even substantially higher than CT and MR combined. In other words, the overall costs for ultrasound are as high as for all “heavy” imaging equipment. Actual ultrasound costs per examination are one fifth of an MR examination.
Is this an artificially created market which was chosen to bypass the hard-hit MR and CT markets? If so, was it chosen by the physicians themselves or by administrators and/or vendors?
At the heart of this discussion on facts and figures are the questions: who needs them, who uses them and for what purpose are they used?
Reliable figures should be compiled and used to improve health services and not to engage in games where we end up not knowing which numbers are correct. Those selling equipment and accessories should know the state of the market for their own and their clients’ interests. The same holds for medical associations, which should argue their cases with dependable and trustworthy data that have not been manipulated in favor of one or another medical discipline.
Last but not least, by definition health politicians cannot be trusted because, according to Mark Twain, they base their arguments on three pillars: lies, damned lies, and statistics.