Objective: To validate a previously published clinical decision rule to predict risk of urinary tract infection in febrile young girls.
Methods: We performed a retrospective case-control study at a children's hospital emergency department in a different city than that in which the original derivation study took place. Girls younger than 2 years in whom urinalysis and urine culture were performed for evaluation of fever were eligible. Cases consisted of all patients with a positive urine culture result, defined as 50,000 or more colony-forming units per milliliter of a urinary tract pathogen (n = 98). A random sample of patients with a negative urine culture result (n = 114) was also selected as controls. The clinical prediction rule included five risk factors: age younger than 12 months, white race, temperature of 39.0 degrees C or higher, absence of any other potential source of fever, and fever for 2 days or more. The sensitivity and false-positive rate of this rule were calculated at different cutoff values.
Results: The overall discriminative ability of the rule, as indicated by the area under the receiver-operator characteristic curve (AUC), was similar in this validation sample (AUC = 0.72) to that in the original study (AUC = 0.76). However, in the validation sample, the presence of three or more risk factors (rather than two or more as in the original study) appeared to be the optimum cutoff to define a positive rule, which results in an indication for obtaining further diagnostic testing (sensitivity, 88% [95% CI, 79-94%]; false-positive rate, 70% [95% CI, 61-79%]).
Conclusion: A simple clinical decision rule previously developed to predict urinary tract infection based on five risk factors performs similarly in a different patient population.