Methods Three cross-sectional analyses, conducted using data from the electronic medical records of all patients with diabetes registered in 23 general practices in Brent, North West London between 2004/2005 and 2006/2007. Patterns of exclusions were examined for three intermediate outcome indicators.
Results Excluded patients were less likely to achieve treatment targets for HbA1c (2004/2005, 2006/2007), blood pressure (2005/2006, 2006/2007) and cholesterol (2005/2006). Black and South Asian patients were more likely to be excluded from the HbA1c indicator than White patients [adjusted odds ratio = 1.64 (1.17-2.29) in 2005/2006]. Patients diagnosed with diabetes duration of > 10 years [adjusted odds ratio = 2.01 (1.65-2.45) for HbA1c in 2006-2007] and those with co-morbidities (adjusted odds ratio, ≥ 3 co-morbidities compared with no co-morbidity for HbA1c adjusted odds ratio = 1.90 (1.24-2.90) in 2004/2005] were more likely to be excluded. Larger practices excluded more patients from the HbA1c indicator [adjusted odds ratio, practice ≥ 7000 compared with < 3000, 3.52 (2.35-5.27) in 2005-2006]. More deprived practices consistently excluded more patients from all indicators, whilst in 2007 older patients were excluded to a larger degree [adjusted odds ratio = 2.52 (1.21-5.28) ≥ 75 compared with 18-44 for blood pressure control].
Conclusions Patients excluded from pay-for-performance programmes may be less likely to achieve treatment goals and disproportionately come from disadvantaged groups. Permitting physicians to exclude patients from pay-for-performance programmes may worsen health disparities.