Purpose: Bacteriuria is common among institutionalized elderly populations, but the contribution of urinary infection to febrile morbidity is unknown because of difficulties in clinical ascertainment. This study was undertaken to febrile morbidity using both clinical and serologic criteria.
Methods: Episodes of fever in residents of two long-term care institutions were identified prospectively for 2 years. Serum and urine specimens were obtained initially and at 4 weeks. The proportion of episodes attributable to urinary infection was determined by both standard clinical criteria proposed for use in these populations and serum antibody response to uropathogens.
Results: For 372 fewer episodes, 211 met clinical criteria for infection: 147 (40%) of the respiratory tract; 26 (7%) of the genitourinary tract; 25 (6%) of the gastrointestinal tract; and 13 (3%) of skin and soft tissue. Of the remaining 161 fever episodes, 2 (1%) were noninfectious and 159 (43%) were of unknown origin. The prevalence of bacteriuria for residents with nongenitourinary sources of fever varied from 32% to 75%. An antibody response meeting serologic criteria for urinary infection occurred in 26 (8.3%) of 314 episodes with paired sera obtained; 10 (43%) of 23 identified clinically as genitourinary infection, 14 (11%) of 132 unknown, 1 (4%) of 25 gastrointestinal, and 1 (0.8%) of 122 respiratory. The positive predictive value of bacteriuria for febrile urinary infection identified by clinical criteria was was 11% (95% confidence interval [CI] 4%, 18%) and identified by serologic criteria was 12% (95% CI 7%, 17%).
Conclusions: Urinary infection contributes to less than 10% of episodes of clinically significant fever in this high-prevalence bacteriuric population. A restrictive clinical definition for genitourinary infection has poor sensitivity and specificity compared with serologic criteria for identification of fever of urinary source, and bacteriuria has a low predictive value for identifying febrile urinary infection.