Real-time identification of serious infection in geriatric patients using clinical information system surveillance

Abstract

Objectives: To develop and characterize an automated syndromic surveillance mechanism for early identification of older emergency department (ED) patients with possible life-threatening infection.

Design: Prospective, consecutive-enrollment, single-site observational study.

Setting: A large university medical center with an annual ED census of 75,273.

Participants: Patients aged 70 and older admitted to the ED and having two or more systemic inflammatory response syndrome (SIRS) criteria during their ED stay.

Measurements: A search algorithm was developed to screen the census of the ED through its clinical information system. A study coordinator confirmed all patients electronically identified as having a probable infectious explanation for their visit.

Results: Infection accounted for 28% of ED and 34% of final hospital diagnoses. Identification using the software tool alone carried a 1.63 relative risk of infection (95% confidence interval CI51.09-2.44) compared with other ED patients sufficiently ill to require admission. Follow-up confirmation by a study coordinator increased the risk to 3.06 (95% CI52.11-4.44). The sensitivity of the strategy overall wasmodest (14%), but patients identified were likely to have an infectious diagnosis (specificity 598%). The most common SIRS criterion triggering the electronic notification was the combination of tachycardia and tachypnea.

Conclusion: A simple clinical informatics algorithm can detect infection in elderly patients in real time with high specificity. The utility of this tool for research and clinical care may be substantial.