Predicting antibiotic prescription after symptomatic treatment for urinary tract infection: development of a model using data from an RCT in general practice

Background: Uncomplicated urinary tract infection (UTI) is often treated with antibiotics, resulting in increasing resistance levels. A randomised controlled trial showed that two-thirds of females with UTI treated symptomatically recovered without subsequent antibiotic treatment.

Aim: To investigate whether there are differences between females with a UTI who were subsequently prescribed antibiotics and those who recovered with symptomatic treatment only, and to develop a model to predict those who can safely and effectively be treated symptomatically.

Design and setting: This is a subgroup analysis of females assigned to ibuprofen in a UTI trial in general practices.

Method: Multiple logistic regression analysis was used to select variables for a prediction model, The discriminative value of the model was estimated by the area under the receiver operator curve (AUC) and the effects of different thresholds were calculated within the model predicting antibiotic prescription and need for follow-up visits.

Results: Of the 235 females in the ibuprofen group, 79 were subsequently prescribed antibiotics within 28 days of follow-up. The final model included five predictors: urgency/frequency, impaired daily activities, and positive dipstick test results for erythrocytes, leucocytes, and nitrite. The AUC was 0.73 (95% CI = 0.67 to 0.80). A reasonable threshold for antibiotic initiation would result in 58% of females presenting with UTI being treated with antibiotics. Of the remaining females, only 6% would return to the practice because of symptomatic treatment failure.

Conclusion: The present model revealed moderately good accuracy and could be the basis for a decision aid for GPs and females to find the treatment option that fits best.