4CPS-143 DrugGPT: can a large language model match clinical pharmacists in DOAC prescription validation?