Automatic detection of patients with nosocomial infection by a computer-based surveillance system: a validation study in a general hospital

Infect Control Hosp Epidemiol. 2006 May;27(5):500-3. doi: 10.1086/502685. Epub 2006 Apr 20.


Objective: To validate an automated system for the detection of patients with nosocomial infection (NI) in an intensive care unit (ICU).

Design: Retrospective analysis of data from the hospital information system. We applied 3 different NI suspicion criteria (positive microbiology reports, antibiotic administration, and diagnosis of clinical infection) and compared the results to those of a prospective NI incidence study done in the ICU during the same period (1999-2002).

Setting: A 250-bed general hospital in Barcelona, Spain.

Patients: From April 15, 1999, through June 30, 2002, 1380 patients were admitted to the ICU. Of these, 1043 had an ICU stay of more than 24 hours and were included in the study.

Results: At least one NI suspicion criterion was present for 242 patients (23.2%); 2 criteria were present for 184 patients (17.6%); and all 3 criteria were present for 112 (11.7%). Comparison of hospital information system data to the results of the prospective study indicated that the combination of 2 criteria demonstrated the most satisfactory sensitivity (94.3%; 95% confidence interval [CI], 79.5%-99.0%) and specificity (83.6%; 95% CI, 76.8%-88.9%). The positive predictive value was 55.9% (95% CI, 42.5%-68.6%); the negative predictive value was 98.5% (95% CI, 94.2%-99.7%). The system could assign a site of infection for 90.4% of the NIs detected.

Conclusion: The hospital information system was a useful tool for retrospectively detecting patients with an NI during the ICU stay. Given its high sensitivity, it may be useful as an alert for the NI team.

Publication types

  • Evaluation Study

MeSH terms

  • Aged
  • Automation
  • Computer Systems*
  • Cross Infection / diagnosis
  • Cross Infection / epidemiology*
  • Data Collection
  • Diagnosis, Computer-Assisted
  • Female
  • Hospitals, General*
  • Humans
  • Intensive Care Units
  • Male
  • Middle Aged
  • Predictive Value of Tests
  • Sensitivity and Specificity
  • Sentinel Surveillance*