Do physicians do as they say? The case of mammography
B. G. Saver, T. R. Taylor, J. R. Treadwell and W. G. Cole
Department of Family Medicine, University of Washington, Seattle, USA.
OBJECTIVE: To assess the utility of survey-based physician policy in
predicting actual mammography ordering behavior, as measured by medical
record abstraction. DESIGN: Cross-sectional survey of practicing community
physicians. Responses were correlated with data abstracted from the medical
records of patients in the practices of the participating physicians.
PARTICIPANTS: Family and general practitioners in Washington State. Medical
records of female patients aged 40 to 80 years provided data on actual
mammography performance. MAIN OUTCOME MEASURES: The proportions of female
patients aged 40 to 49 and 50 to 80 years who had received a screening
mammogram within the previous 2 years. RESULTS: Of the more than 100
potential predictors available, only 4 were significantly associated with
screening rates for women younger than 50 years and only 3 were associated
with screening rates for older women. Regression models explained only 21%
to 25% of the variance in screening rates. Physician estimates of screening
rates were poorly correlated with actual screening rates. CONCLUSIONS:
Practicing physicians do not know how well they screen their patients using
mammography. Extensive survey data, including direct estimates of behavior,
demographics, policy measures, and case scenario responses, were of limited
use in predicting actual screening rates. Our results underscore the
importance of using data rather than proxy measures to study physician
performance.
