OBJECTIVE: To assess the relevance of several pharmaceutical databases (PDBs) to be integrated in a computerized prescription system for hospitals. MATERIALS AND METHODS: Fifty medical prescriptions were designed as tests. They were supposed to answer to a security or a relevance assessment item. They were analyzed using four French PDBs (Claude-Bernard, Thériaque, Thésorimed, Vidal), available online. The outcome was the rate of conformity with the expected answers. RESULTS: The rate of conformity was: 31% for Claude-Bernard, 30% for Thériaque, 26% for Vidal, and 20% for Thésorimed (no statistical significance). DISCUSSION: The PDBs easily detect interactions between different pharmacological class drugs but redundancies and illogical situations are poorly intercepted. They seem to manage medicines only by the active ingredient, irrespective of the dose or the indication. The knowledge of the patient profile and the context are not really enriched by the content of the prescription. CONCLUSION: The efficacy of these PDB has to be improved.