Assessing impact of organised breast screening across small residential areas - development and internal validation of a prediction model

Monitoring screening mammography effects in small areas is often limited by small numbers of deaths and delayed effects. We developed a risk score for breast cancer death to circumvent these limitations. Screening, if effective, would increase post-diagnostic survivals through lead-time and related effects, as well as mortality reductions. Linked cancer and BreastScreen data at four hospitals (n = 2,039) were used to investigate whether screened cases had higher recorded survivals in 13 small areas, using breast cancer deaths as the outcome (M1), and a risk of death score derived from TNM stage, grade, histology type, hormone receptor status, and related variables (M2). M1 indicated lower risk of death in screened cases in 12 of the 13 areas, achieving statistical significance (p <.05) in 5. M2 indicated lower risk scores in screened cases in all 13 areas, achieving statistical significance in 12. For cases recently screened at diagnosis (<6 months), statistically significant reductions applied in 8 areas (M1) and all 13 areas (M2). Screening effects are more detectable in small areas using these risk scores than death itself as the outcome variable. An added advantage is the application of risk scores for providing a marker of screening effect soon after diagnosis.