In The Journal of urology ; h5-index 80.0
PURPOSE : Continuous antibiotic prophylaxis (CAP) reduces the risk of recurrent urinary tract infection (rUTI) by 50% in vesicoureteral reflux (VUR) children. However, there may be subgroups in whom CAP could be used more selectively. We sought to develop a machine learning model to identify such subgroups.
MATERIALS AND METHODS : We used RIVUR data and randomly split it into train/test in 4:1 ratio. Two models were developed to predict rUTI risk in scenario with and without CAP. The test set was then used to validate rUTI events and the effectiveness of CAP. Predicted probabilities of rUTI were generated from each model. CAP was assigned at various cutoffs of rUTI risk reduction to evaluate CAP effectiveness.
RESULTS : 607 patients (558 female/49 male, median age 12 months) were included. Predictors included VUR grade, serum creatinine, race/sex, prior UTI symptoms (fever/dysuria), and weight percentiles. The AUC of the prediction model of rUTI (CAP/placebo) was 0.82 (95% CI: 0.74-0.87). Using 10% rUTI risk reduction cutoff, minimal rUTI on population-level can be achieved by giving CAP to 40% VUR patients instead of everyone. In test set (n=121), 51 patients had CAP randomization consistent with model recommendation (CAP if rUTI risk reduction >10%). rUTI incidence was significantly lower among this group compared to those whose CAP assignment differed from model suggestion (7.5% vs 19.4%, p=0.037).
CONCLUSIONS : Our predictive model identifies VUR patients who are more likely to benefit from CAP, which would allow more selective, personalized use of CAP with maximal benefit, while minimizing use in those with least need.
Wang Hsin-Hsiao Scott, Li Michael, Bertsimas Dimitri, Estrada Carlos, Caleb Nelson
Prediction model, RIVUR, antibiotic prophylaxis, machine learning, urinary tract infection, vesico-ureteral reflux