Diagnostic accuracy of criteria for urinary tract infection in a cohort of nursing home residents.

Objectives: To prospectively evaluate nursing home residents with suspected urinary tract infection (UTI) to determine whether they met the McGeer, Loeb, or revised Loeb consensus-based criteria and whether any set of criteria was associated with laboratory evidence of UTI, namely bacteriuria (>100,000 colony forming units) plus pyuria (>10 white blood cells).

Design: Prospective cohort study.

Setting: Three New Haven-area nursing homes.

Participants: Of 611 residents screened, 457 were eligible, 362 consented, and 340 enrolled.

Measurements: Participants underwent prospective surveillance from May 2005 to April 2006 for the development of suspected UTI (defined as a participant's physician or nurse clinically suspecting UTI). One hundred participants with suspected UTI and a urinalysis and urine culture performed were included in the analyses.

Results: Participants were identified who met the criteria of McGeer, Loeb, revised Loeb, and laboratory evidence of UTI. Using laboratory evidence of UTI as the outcome, the McGeer criteria demonstrated 30% sensitivity, 82% specificity, 57% positive predictive value (PPV), and 61% negative predictive value (NPV); the Loeb criteria showed 19% sensitivity, 89% specificity, 57% PPV, and 59% NPV; and the revised Loeb criteria demonstrated 30% sensitivity, 79% specificity, 52% PPV, and 60% NPV.

Conclusion: All of the consensus-based criteria have similar test characteristics. The diagnostic accuracy of UTI criteria in nursing home residents could be improved, and the data suggest that evidence-based clinical criteria associated with laboratory evidence of UTI need to be identified and validated.