Objective. To assess the sensitivity of hospital discharge diagnoses for identifying sepsis in patients with blood culture confirmation.
Methods. A cross-sectional study was conducted at the Italian 1000-bed University Hospital of Udine. The administrative databases of the Hospital were used as the source of information. Laboratory data were linked with hospital discharge data. We estimated the proportion of hospitalizations with at least 2 positive blood culture tests in which at least one discharge diagnosis indicated bloodstream infection.
Results. From 2011 to 2017, 3571 hospitalizations (1.2%) had positive blood culture tests. Of them, only 49.5% had at least one ICD-9-CM discharge diagnosis code of sepsis, with lower proportions in surgical than in medical wards.
Conclusions. The sensitivity of ICD-9-CM discharge codes for sepsis is low as compared with the blood culture gold standard. Using discharge codes for epidemiological estimates of sepsis, health planning and risk management may yield biased results. Audits and ICD coding training are needed.