Studies on the impact of a Health Information Technology seldom consider socio-technical characteristics of the work system in which the technology is implemented. Yet those dimensions may act as hidden variables that could explain the inconsistency of impact studies' results in terms of performance, quality and satisfaction. This paper reports on the identification of those variables in the discharge letter (DL) process. Human Factors experts performed an analysis of the work system of the DL process in 17 medical units. The DL process is composed of three sub-processes running with work system differing according to the distribution of tasks, the technology implemented and the work organization. Hidden variables identified are: verification by the physician, technology's integration, number of editing cycles, physicians' preferences etc. Those variables can be collected automatically or by questionnaire. Statistical analyses will have to be performed to know which variable explain impact indicators.