Identifying possible deaths associated with nosocomial infection in a hospital by data mining

Nosocomial infection (NI) is a cause of patient morbidity and mortality. Conducting an audit of deaths due to NI is a potentially useful approach to improving professional standards. In France, these deaths are required to be reported, but the reporting is left to clinicians, who often do not comply. The aim of the present study was to assess whether linking the microbiological database with the hospital mortality database might be a suitable surveillance approach for identifying patients who died with an NI.

A total of 1,726 deaths were recorded in the mortality database of a French university hospital between September 1, 2006, and September 16, 2007. During this same period, 6,290 potential NIs (PNIs) were identified by bacteriological examination. These PNIs were generated using a computer algorithm specific to the bacteriology database. PNI information request forms were sent to the senior doctor of the unit where the samples had been obtained to determine whether the PNI was an NI, colonization, or a non-nosocomial infection. A total of 364 cases were common to both databases; from these, a sample of 135 cases was selected for further analysis. To establish the strength of evidence for NI as the cause of death, the 135 cases were analyzed using the patient record by an investigator from the hospital hygiene team.

During the study period, no deaths associated with NI were reported spontaneously. Of the 135 cases analyzed, NI was considered the main cause of death in 6 (4.4%) and a contributory factor in 51 (37.8%). Thus, NI was estimated to be the main cause of death in 0.9% of all patients who died in the hospital during the study period and a contributory cause in another 8.0% of these patients.

Linking databases from bacteriology with those containing hospital mortality records is a simple, reproducible tool for identifying the number of deaths attributable to NI. This may provide a powerful approach to help reduce the burden of disease due to NI through the auditing of such identified deaths.