Table 3.

Example of using PPV and NPV to compute percent overreporting by applying them to the intervention study screening rates per recommendations by treatment group

Baseline
Year 1 follow-up
Percent overreporting
Percent overreporting
Rx0* (n = 436)Rx1 (n = 423)Rx2 (n = 436)Rx3§ (n = 403)Rx0* (n = 409)Rx1 (n = 404)Rx2 (n = 389)Rx3§ (n = 351)
FOBT7.97.67.18.210.810.619.524.8
Flexible sigmoidoscopy5.76.15.26.07.57.16.97.4
Colonoscopy11.011.59.711.38.88.77.210.6
Barium enema3.95.43.13.33.53.82.72.4
Overall adherence8.18.67.88.710.912.713.314.3
  • NOTE: Predictive values used in computations are averaged over all treatment groups rather than specific to treatment and are not themselves adjusted for nonresponse.

  • * Rx0 = control counties (in Wright County Colorectal Cancer Screening Project sample).

  • Rx1 = intervention county, community intervention only (in Wright County Colorectal Cancer Screening Project sample).

  • Rx2 = intervention county, direct mailing of FOBT kits with no reminders plus community intervention (in Wright County Colorectal Cancer Screening Project sample).

  • § Rx3 = intervention county, direct mailing of FOBT kits with reminders plus community intervention (in Wright County Colorectal Cancer Screening Project sample).

  • Overall screening adherence via any one of the four tests per recommendations.