Are medical errors responsible for as many deaths as believed; how easy is it to predict the future of healthcare; and why are so few doctors embracing scientific research? This month's column examines these issues

Medicine in general, and hospital medicine in particular, is a risky business. We all have our favourite anecdotes. Mine - I came across it a few years ago in, I think, the BMJ - comprised a list of the metallic objects found among the ashes by staff working in a crematorium. An embarrassing number of these scorched remnants turned out be surgical instruments.

The image of medicine as dangerous to health is even more strongly held across the pond. As a couple of doctors working in Ann Arbor, Michigan, point out (JAMA, 286: 415-420), some of the scariest claims originate within the profession itself. 'A recent Institute of Medicine report, ' say Rodney Hayward and Timothy Hofer, 'quoted rates estimating that medical errors kill between 44,000 and 98,000 people a year in US hospitals.'

But how reliable are such unnerving figures?

Though a touch self-serving, coming as it does from doctors, there is an argument that the commonly-used methods of estimating the importance of errors tends to exaggerate their size and impact.

Hayward and Hofer have been exploring precisely this. They have done so by organising their own review of more than 100 deaths which occurred at several medical centres over the course of one year.

In addition to the conventional protocol for such studies, their reviewers were asked to make a few extra comments on specific issues, including the patients' underlying short-term prognosis, and the likelihood that they would have died anyway, irrespective of errors.

While no-one would suggest that doctors and managers should be anything less than diligent in trying to eliminate medical errors, the findings do suggest that mistakes may not be responsible for quite as many deaths as some observers have claimed. As the authors put it in the subtitle to their work, 'Preventability is in the eye of the reviewer.'