In Reply We thank Drs Bunting-Early and Figueroa for their interest in our article. Although they correctly point out that some of the data included are relatively old and that our analysis is retrospective (albeit using prospectively collected data), they do not proffer any reasons why these characteristics make the data unreliable. Individual-patient data meta-analysis is considered the best available study design to address risk prediction because it uses all available data and avoids confounding by study, which group-level meta-analyses may introduce. Their concerns regarding our noninclusion of vesicoureteral reflux in the model may result from misinterpretation of our study’s aims. Our prediction model was not designed to determine an epidemiological causal pathway for scarring, but rather, to develop a risk prediction model to assist physicians to identify children most at risk using routinely collected information. We did not include vesicoureteral reflux in model 1 (the clinical model) because vesicoureteral reflux was not known at the time of the initial diagnosis and may have never been known. However, we did present data on a model in which vesicoureteral reflux was included (model 3), which showed that adding vesicoureteral reflux to the clinical model improved accuracy only marginally. Therefore, we concluded that using a simple model with only 3 clinical variables provided a reasonable screening strategy for the identification of children with a first urinary tract infection who were at risk for renal scarring.
Shaikh N, Craig JC, Shope TR. Risk Model for Renal Scarring: Validation and Implications Still Needed for Primary Care—Reply. JAMA Pediatr. 2015;169(2):189-190. doi:10.1001/jamapediatrics.2014.2902