Variation in physicians' decision-making thresholds in management of a sexually transmitted disease.

OBJECTIVE

To gain insight into the variation in physicians' clinical decisions and further understand the factors that influence physicians' thresholds for testing and treating.

DESIGN

Written clinical scenarios were mailed to two groups of physicians who were asked to provide probability estimates of syphilis, how these estimates might change with new information, and when a diagnostic test would be ordered or treatment begun. A model was then used to calculate the probabilities at which physicians ordered tests or initiated treatment.

PARTICIPANTS

Group 1 comprised 126 board-certified internists from metropolitan Philadelphia responding from a sample of 360 such physicians randomly selected from a directory. Group 2 consisted of 31 experts in sexually transmitted disease responding from a sample of 50 experts selected by the authors.

MEASUREMENTS AND MAIN

RESULTS

Experts were willing to obtain a serologic screening test at a lower likelihood of syphilis (0.013%) than were internists (0.034%), and they were willing to obtain a lumbar puncture at a lower likelihood of neurosyphilis (0.165%) than were internists (0.393%). The difference in the groups' thresholds to begin neurosyphilis treatment was not significant. A multivariate model showed that group differences were created by individual characteristics (years in practice, subspecialty board certification, and full-time nonacademic practice) that were associated with higher thresholds for serologic screening.

CONCLUSIONS

There are differences in the diagnostic testing practices for syphilis between national experts and internists. Although status in one of these groups alone did not predict the threshold for obtaining syphilis tests, certain individual characteristics were predictive. Examination of physician characteristics helps to explain the variation observed in their practice patterns, and determination of physicians' thresholds aids in analyzing these variations.