The predictive ability of a weighted systemic inflammatory response syndrome score for microbiologically confirmed infection in hospitalised patients with suspected sepsis.

BACKGROUND

The systemic inflammatory response syndrome (SIRS) concept lacks sensitivity and specificity for guiding clinical practice and sepsis research.

OBJECTIVE

To assess the performance of a weighted SIRS score, with emphasis on white cell count and temperature criteria in predicting microbiologically confirmed infection.

DESIGN AND SETTING

Prospective cohort study at Princess Alexandra Hospital, a tertiary teaching hospital in Queensland, Australia.

PARTICIPANTS

Patients aged 18 years or older who were hospitalised with suspected infection and started on antimicrobial therapy.

MAIN OUTCOME MEASURES

The utility of each SIRS criterion, the 1992 consensus conference recommendation (≤ 2 SIRS criteria) and a weighted SIRS score in predicting microbiologically confirmed infection were compared.

RESULTS

2085 patients were included in the analysis. All criteria performed poorly, with low sensitivities (27.3%-70.6%), low specificities (37.5%-77.5%), low positive predictive values (61.5%-65.3%), low negative predictive values (39.8%-45.1%), and likelihood ratios close to 1.0. Both SIRS and weighted SIRS scores did not perform better than clinicians' suspicion for infection.

CONCLUSIONS

Both SIRS and weighted SIRS score had low predictive ability for microbiologically confirmed infection. A more robust conceptual framework incorporating clinical, biochemical and immunological markers must be formulated and validated to better guide clinical practice and research. Clinicians' suspicions may be as good as any scoring system at identifying patients with infection and sepsis.