Evaluating variations in medical practice between government primary care health centres

We evaluated variations in medical care for diabetes among government health centres in Trinidad and Tobago. Data were analysed for 1579 subjects with clinical diabetes attending 23 health centres concerning 12 processes of care and six case-mix or outcome variables. Random effects models were used for analysis. Health centre level variation was appreciable. Intraclass correlations ranged from 0.025 to 0.316 for process of care variables and 0.000 to 0.056 for case-mix variables. Compared with health centres with only one nurse, patients attending those with three or more nurses received 1.18 (95% confidence interval 0.27 to 2.09) more processes of care. Delivery of medical care varied among the health centres and those with higher staffing levels provided more processes of care.