Is questionable medical information technology putting patients at risk?

There's been a lot of good news in cancer care lately: The American Cancer society recently reported a decrease in cancer deaths, medical information delivery harnesses technologies such as interactive patient-information kiosks, and nano-technology verges on a breakthrough for more targeted treatments. All of these, combined with more vigorous prevention and detection efforts and more effective standard treatments has helped to make cancer an increasingly survivable journey.

Technology, however, is not a panacea. A striking example is the increasing usage of computer-aided detection (CAD) for both screening and detection mammography. Rather than depending on the seasoned eyes of radiologists, CAD uses a computer program to analyze radiographic images.   The current issue of the Journal of the American College of Radiology reports that 74% of screening mammograms and 50% of diagnostic mammograms performed in 2008 used CAD. That's over 5 million uses of this medical information technology.

That would be good news if CAD helped patients, but so far there's little evidence that it helps. And a recent New England Journal of Medicine study suggests that all it is doing is hurting patients and unnecessarily driving up costs. After reviewing 200,000 mammograms, the study found that CAD led to a 32% increase in follow-up mammograms, and 20% increase in biopsies, but didn't detect more cancers, or save or prolong any additional lives. These are striking figures, since mammograms themselves increase one's cancer risk, and biopsies are associated with not-insignificant stress as you wait for the pathologist's verdict. Another study recently estimated that Medicare is wasting $20 million a year on CAD. Remember, using technologies or procedures that are not proven to help, can only cause harm.

Medical information technology has many promises, and in many cases is already leading to better care and better lives. The CAD example, however, is an example of the technology cart coming before the evidence-of-effectiveness horse. How might we avoid this conundrum? And what happens if advances in CAD over the next ten years eventually lead to lives-saved, is this just the necessary process of technological evolution?