Repeated measures models for prescribing change.

Linear mixed models are used to detect a change, if any, in prescribing habits at the primary care practice level due to an educational intervention given repeated measures data before and after intervention and a control group. Inferences are corrected for general practice size, fundholding status and baseline prescribing. The correlation structure is discussed and the results for multilevel modelling using MLwiN and NLME version 3.0 are compared.