132 Annual Meeting Logo - Go to APHA Meeting Page  
APHA Logo - Go to APHA Home Page

Correction for bias: Using validity measures to adjust self-reported colorectal cancer screening rates

Resa M. Jones, MPH1, Steven J. Mongin, MS2, Timothy R. Church, PhD2, and Mark W. Yeazel, MD3. (1) Division of Epidemiology, University of Minnesota, 1300 S. 2nd St., Suite 300, Minneapolis, MN 55454-1015, 612-626-8881, jones_r@epi.umn.edu, (2) Department of Environmental Health, School of Public Health, University of Minnesota, 420 Delaware St SE, MMC 807, Minneapolis, MN 55455, (3) Department of Family Practice and Community Health, University of Minnesota, Room 5-255 PWB, 516 Delaware St SE, Minneapolis, MN 55455

Studies of self-reported colorectal cancer screening indicate adults may not accurately recall whether or when they had tests. A validation study was conducted in a random sample of 200 adults age 50 and older selected from those responding to annual surveys on screening behavior as part of an intervention study. Positive (PPV) and negative predictive values (NPV) were based on adequately completed surveys, while sensitivity (SENS) and specificity (SPEC) were based on screening experience determined from medical records; therefore, subject counts vary for these estimates. To account for non-response, estimates were adjusted to the target sample of the parent intervention study by age, gender, and other covariates from the driver license database. Approximately 90% were contacted and about 60% of the eligible sample authorized medical record review. The baseline survey SENS and SPEC for having any one of the four colorectal cancer screening tests per recommendations (i.e. annual fecal occult blood test, flexible sigmoidoscopy or barium enema in five years, or colonoscopy in ten years) were 97.4% and 66.2%; PPV and NPV were 82.6% and 97.4% respectively. Follow-up survey SENS and SPEC were 95.3% and 65.6%; PPV and NPV were 77.9% and 97.1% respectively. Adjusting screening given PPV and NPV, the baseline observed rate of screening per recommendations decreased from 57.0% to 48.2%. At follow-up, correction for bias reduced the percent screening from 71.0% to 56.2%. Adjusting self-reported rates to correct for bias provides a more accurate estimate of actual screening rates and intervention effects.

Learning Objectives:

Keywords: Cancer Screening, Survey

Presenting author's disclosure statement:
I do not have any significant financial interest/arrangement or affiliation with any organization/institution whose products or services are being discussed in this session.

Epidemiologic Methods for Studies Related to Cancer

The 132nd Annual Meeting (November 6-10, 2004) of APHA