Background: Research has not paid sufficient attention to the need for adequate case-mix adjustment in studies of the relationship between provider volume and performance. This study attempted to address this limitation by capturing and including 5-year survival outcomes and a wide range of case-mix variables in multivariate analyses of the volume-outcome relationship relating to breast cancer treatments.
Methods: All patients diagnosed with invasive primary breast cancer during 1996 (n = 809) were included. Patient, disease and treatment data were extracted from medical records; survival data were corroborated using official death registrations. A Cox proportional hazards approach was used to model relationships between patient, disease and service variables and risk of death.
Results: There were 262 deaths among 807 patients followed up; overall 5-year survival was 70%. Advancing age, higher levels of co-morbidity, late-stage disease, more positive nodes, and high-grade tumour were independently associated with lower survival (P < 0.05). Patients who received hormonal therapy (HR 0.50, 95% CI 0.28-0.89) and radiotherapy (HR 0.73, 95% CI 0.53-1.03) had a survival advantage. Using a cut-off point of > or =30 cases per annum, survival was lower for patients treated in low volume settings (HR 1.47, 95% CI 1.09-1.96) after adjustment for case mix.
Conclusions: There was some evidence to support treatment in high volume settings although patient and disease variables were the major determinants of survival for patients with breast cancer.