In this paper we introduce the ODD protocol, proposed in ecology for easing the communication and replication of Individual Based Models (IBM). First, we use it to describe a realistic model in epidemiology, which deals with the local determinants of the propagation of the H5N1 virus. From this description we first point out the advantages of ODD. Then we focus on its weaknesses, the major ones being its ambiguity and its inability to attain replication. Finally, we propose several improvements for this protocol in terms of re-organization, better specifications (adapted from existing software engineering methodologies) and the addition of a section dedicated to experimentations.