Methods: The trial design is a pragmatic cluster randomised multiple interrupted time series with 760 patients from five Primary Care Trusts (PCTs) in the UK. Eligible patients were aged more than 75 years with five or more repeat prescriptions, living at home and able to give their consent to take part. The complex nature of the trial requires innovative use of econometric techniques to inform the cost-effectiveness model: the use of Generalized Linear Models in a multi-level framework; requirement to produce population-averaged results in a meaningful metric rather than subject-specific effects; formal incorporation of time in a difference-in-difference model; identification of systematic heterogeneous effects across the population; and the means to incorporate the econometrics in a probabilistic sensitivity analysis. Thus, this research also represents an important methodological bridge between two hitherto relatively distinct health-economic disciplines.
Results: The analysis indicates that PC has an incremental CE ratio of £12,187 per QALY and a probability of being CE of 74% at a £20,000 threshold and approaching 82% as the threshold approaches infinity. The analysis also identifies that the expected net benefit is not consistent across the population, with the more elderly patients on higher numbers of repeats systematically receiving less benefit and producing higher costs.
Conclusion: The analysis has successfully applied econometric techniques in a CE framework to account for the complex trial design to find evidence that PC is cost effective in an elderly population. In addition, the application paves the way for further use of econometric techniques in economic evaluation.