Résumé : Cost outliers account for 6 to 8% of hospital inpatient stays and concentrate 22 to 30% of inpatient costs. Explanatory factors were highlighted in various studies. They are the lenght of stay, an intensive care unit stay, the severity of illness index related to DRG and social factors. Patients are not always explained by these factors. The objective of this study is to analyse cases not explained by those factors, through a detailed analysis of medical files. In the studied hospital, there are 6,3% high cost outliers and 1,1% low cost outliers. These stays were isolated on the basis of a rule based on percentiles. Extra costs generated by high cost outliers are 6.999 euro per stay. The extra lenght of stay for these patients is 20,42 days. Among the 454 patients high cost outliers, 334 patients are explained by factors extracted from a statistical analysis based on a logistic regression (intensive care unit stay, severity of illness index, lenght of stay and social factors). The analysis of medical files of the 120 not explained inpatient stays highlights new explanatory factors (coding errors, heterogeneity of DRGs, etc.). At the end of this study, the conclusion is that a statistical analysis combined with a precise analysis of medical files allowed to explain the majority of cost outliers. An explanation is however not necessarily synonymous with medical justification.