Positive predictive values of selected hospital discharge diagnoses to identify infections responsible for hospitalization in the French national hospital database.

PURPOSE: The identification of infections in electronic health databases is a key issue for pharmacoepidemiology research. The aim of this study was to assess the positive predictive values (PPVs) of hospitalizations for infection in the Système National d'Information Inter-régimes de l'Assurance Maladie, that is the electronic database recording in-and-out hospital data for the entire French population (66 million inhabitants).
METHODS: The source of data was the database of hospitalizations (Programme de Médicalisation des Systèmes d'Informations) of Toulouse University hospital, South of France (2880 beds). Among all hospital stays between September and December 2014, we randomly selected 100 stays with an International Classification of Diseases, 10th revision code of infection as primary diagnosis and 100 as related diagnosis. Medical charts were reviewed to assess the PPV of infection codes, as well as the PPV of correct coding of infection type among the true positive cases.
RESULTS: The PPVs of codes of infection as reason for hospitalization were 0.97, 95% confidence interval (CI) [0.93-1.00] for primary diagnosis codes and 0.70, 95% CI [0.61-0.71] for related diagnosis codes. Among the true positive cases, the PPVs of correct coding of the type of infection were, respectively, 0.98, 95% CI [0.95-1.00] and 0.93, 95% CI [0.88-0.98].
CONCLUSIONS: Hospitalizations for infection codes have very good PPVs in the Programme de Médicalisation des Systèmes d'Informations.