Risk preference and laboratory use.

One hundred thirty-seven physicians were asked to choose between a certain loss of five years of life expectancy and a 50/50 gamble of losing either ten years or zero years of life expectancy. These choices were presented as hypothetical options for a patient with cancer. The 46 who chose the certain loss ordered 23% fewer laboratory tests/patient visit in our General Medicine Clinic over an eight-week period than those who chose the gamble (p less than 0.05). This risk preference was further stratified by sequentially offering five, four, three, two, and one years of certain loss against the same 50/50 gamble of ten or zero years of loss; greater preference for the gamble correlated with greater laboratory use (p less than 0.002). To exclude physician knowledge or case mix as a cause of this result, this risk preference was correlated with total score, and with the cost of tests ordered in patient management problems, of 49 physician on the American Board of Internal Medicine Certifying Examination. Cost of tests ordered was 23% less for the 17 who chose the certain loss (p less than 0.001); total score did not vary with risk preferences. There was no association between risk preference in the face of gain and laboratory use, and none between an individual's risk preferences in the face of gain and in the face of loss. The cognitive processes that determine risk preference in the face of loss, whatever they may be, appear to have a substantial influence on physician test-ordering behavior.