Background The drug supply chain is a cross-disciplinary process involving numerous actors. This can create difficulties in attempts to successfully analyse and manage its execution.

Objective To produce a tool allowing easy evaluation and optimisation of the hospital drug supply chain.

Material and methods A supervised Bayesian network was built to model a hospital drug supply chain. Two learning patterns were used: literature and experimental data. The network was tested by using data separate from the data used for its construction. Two hundred scenarios were simulated to evaluate the impact of organisational modalities.

Results The model estimates a need for 3·2 clinical pharmacists per 100 patients. The model gave an appropriate estimation of technician workforce of three hospital pharmacies and showed that one of them was understaffed. Simulations showed that a unit-dose drug distribution system and prescription analysis seem to have comparable impact on quality index, with a maximum increase of 28 for unit-dose drug distribution and 25 for prescription control (depending on control methods, the increase in quality varies between 31 and 89). In addition, changing from a global to a unit-dose distribution system results in an increase of 14 % in the hospital pharmacy's payroll whereas changing from no prescription control to the daily control of 100 % of prescriptions leads to a rise of 218 % in the pharmacy's wage bill.

Conclusions The Bayesian approach appears to be a valuable decision tool to gauge the influence of organisational changes on care quality.