Case-mix fails to explain variation in mastectomy rates: management of screen-detected breast cancer in a UK region 1997–2003

Wide variation in the surgical management of breast cancer exists at hospital, regional, national and international level. To demonstrate whether variation in surgical practice observed at aggregate level between breast units persists following adjustment for case-mix, individual patient-level data from the Trent Breast Screening Programme Quality Assurance database (1997–2003) was analysed. Expected case-mix adjusted mastectomy rates were derived by logistic regression using the variables tumour size, site and grade, patient age and year of presentation, employing the region's overall case-mix adjusted practice as the reference population. The region's 11 breast screening units detected 5109 (3989 invasive) surgically managed primary breast cancers over the 6-year period. A total of 1828 mastectomies (Mx) were performed (Mx rate 35.8%, 95% confidence interval: 34.5–37.1%). Significant variation in mastectomy rates were observed between units (range 25–45%, P<0.0001), and persists following case-mix adjustment (P<0.0001). Two-fold variation in observed to expected unit mastectomy rate coefficient is demonstrated overall (range 0.66–1.36), increasing to almost four-fold variation in cancers less than 15 mm diameter (range 0.55–1.95). Significant variation in surgery for screen-detected primary breast cancer is not explained by case-mix. Further research is required to investigate potential patient and professional causative factors.