Exacerbations of chronic obstructive pulmonary disease (COPD) are a factor associated with a poor prognosis for the course of the disease. The most severe cases require hospitalization. Identifying hospital stays due to COPD exacerbations within the data from the Hospital Information System (PMSI) is complex. On the one hand, since these hospitalizations occur in patients with frequent comorbidities, COPD is not always listed as the primary diagnosis (PD) for the stay. On the other hand, hospitalizations for evaluation are not directly identifiable in the PMSI data. The objective of this study was therefore to define the most relevant algorithms for identifying hospital stays for COPD exacerbations in the national PMSI database. The analysis focused on stays with a primary or associated diagnosis code corresponding to COPD or chronic respiratory failure, occurring in adults aged 25 years or older. Several coding combinations for the primary diagnosis and associated diagnoses were defined in collaboration with hospital pulmonologists. A descriptive analysis of hospital stays was first conducted for the year 2005. Changes in the characteristics of hospital stays between 1998 and 2007 were then described for each algorithm to test the robustness of trends over time. The number of hospital stays varied considerably depending on the algorithm used (from 66,000 to 109,000 per year for 2006–2007). The total number of hospital stays increased between 1998 and 2007, but the distribution of the different codes remained stable over these ten years. In conclusion, although the overall burden of hospitalizations for COPD exacerbations is difficult to assess, it is possible to use PMSI data to analyze variations over time. We propose using two indicators to account for potential coding shifts and to verify the robustness of trends over time. We recommend excluding hospital stays of less than 48 hours that do not result in death. (R.A.)