Objetive: Presents a reengineering process of the distribution of drugs into the hospital, analyzing all the options available in the market, and looking for alternative solutions that may be more cost-effective. Method: The processes and subprocesses in the cycle from prescribing, distribution, and drug administration, are defined and discussed based on studies of medication errors (ME). The differential analysis is performed on the subprocesses. As technique for finding creative solutions (new cost-effective alternatives) apply the Theory of Constraints (TOC), and the TRIZ methodology. Results: Since patient safety can distinguish four processes: prescription (about 40% of ME), transcription, distribution (about 10% each), and administration (about 40% of ME). In the administration, avoided ME before they reach the patient are minimal  (only 2%). In the prescripcion/transcription there are 4 options: manual prescription, preprinted sheets, electronic prescription, and assisted prescription. In the distribution has 3 options: clasical SUD, filling carts using automated carousels, and automated dispensing systems (ADS). For administration there are other 3 options: manual record, electronic registration, and registration across the barcode. The most expensive option would be the introduction of ADS in all plants (1.4 million€ for a hospital of 280 beds). But these teams only reduces errors about 10% of all ME. Applying the TOC and TRIZ, investment in electronic prescribing, and administration with barcodes is the most cost-effective. Dose-day (sending medication for one day but not rated by patient) could be the most efficient system by simplifying processes. The error difference between Dose-day, and SDU can be annulled by the advantages of the assisted prescription, and administration with barcode Conclusions: It is surprising to invest large sums in improving distribution processes (ADS) - where the fewest mistakes occurs - instead of prescribing and administration. The dose-day with barcode administration would be the most cost-effective theoretical-model