Medical practice has rapidly shifted towards an 'evidence-based' approach.
While there are acknowledged clear benefits to this, a number of pitfalls a
re frequently not appreciated. Perhaps the most important limitation is the
extent to which the current body of data is inadequate for many common cli
nical decisions. Algorithms risk being developed, frequently by third parti
es, without acknowledgement of these limitations and with substantial impli
cations for clinical independence and the quality of patient care. This pap
er discusses potential problems of the evidence-based approach and suggests
possible guidelines for the management of clinical decisions given the lim
itations of data-based guidelines.