Background Traumatic brain injuries (TBI) result most often from injuries which could have been avoided through preventive measures. They have very costly human and financial consequences. To contribute to the epidemiological surveillance of TBI, the objective of this project was to analyse hospitalizations for TBI in France.

Methods Each hospitalisation results in a report containing information on the diagnosis, the treatment, the health condition upon discharge, etc. This database, called the Programme for the Medicalization of Information Systems (PMSI), is comprehensive. The main hospital diagnosis (MD) and associated diagnosis (AD) are coded in the International Classification of Diseases, 10^th Revision (ICD10). The selections were made from different ICD10 code lists used in the literature: “Intracranial injury” (S06 codes) always selected for TBI analysis; “Fracture of vault of skull, etc.” (S02.0, S02.1, S02.3, S07.1) often retained (OR); “Fracture of skull and facial bones, etc.” (S02.7, S02.8, S02.9, S07.0, S07.8, S07.9, S09.7, S09.8, S09.9) sometimes retained (SR). The selection must be made on all diagnoses (MD and AD), since the TBI can be coded as an AD in case of multiple injuries.

Results In 2013, there were 134,546 hospitalizations with at least one diagnosis coded as S06, 8,292 additional hospitalizations coded as OR, and 6,821 hospitalizations coded as SR, representing a total of 149,659 hospitalizations. The fact of adding to the selection of S06 codes hospitalizations with OR codes or hospitalizations with OR and SR codes results in an increasing number of hospitalizations: 6.2% and 11.2% respectively.

Conclusions The analysis of hospitalizations for TBI should not be limited to S06 codes because they produce too many false negatives. The analysis of false positives and negatives associated with the addition of OR or OR + SR codes requires the use of a gold standard to conclude on the best selection and estimate hospital morbidity due to TBI.