French hospital discharge database (PMSI) and bacterial resistance: Is coding adapted to hospital epidemiology?

Abstract

ObjectiveA preliminary analysis of data consistency on different types of bacterial resistance by infection site and causative agents was conducted using the French hospital discharge database (French acronym PMSI) to assess the use of the database in a national cartography tool. Material and methodsHospital stays in medical, surgical, and obstetrical units were extracted from the 2014 PMSI database using the ICD-10 diagnosis codes. Bacterial infections, causative agents, and resistance corresponding to these stays were also identified. ResultsData from 1258462 patients, corresponding to a total of 1617893 stays, was extracted. Among these stays, 46% were associated with a bacteria code and 7% with a resistance code. Lower respiratory tract infections were the most frequent infections (32% of stays; pneumonia in 95% of cases), followed by genitourinary infections (26%), intra-abdominal infections and diarrhoeas (24%), and skin and soft tissue infections (15%). Inconsistencies were observed between the types of infection and associated bacteria and between bacteria and associated resistance. These inconsistencies are likely due to initial coding errors. ConclusionThe cartography of bacterial infections cannot be developed using the data of the current PMSI coding. These results underline the need to improve the coding of PMSI data for its use as a complementary tool of epidemiological surveillance of bacterial infections.