LAUSR

DOI: 10.1097/PCC.0000000000001820 comparison, PRISM and PICU collected data and were validated on tens of thousands of patients and significantly more deaths with the express purpose of validating predictors of PICU mortality. When taken together, the limitations in the current study by Jacobs et al (9) may account for the differences in model performance of PRISM and PIM in predicting the 90-day mortality endpoint, an endpoint for which the scores were not designed to predict in their development. Further, these limitations prohibit any judgement about the scores’ performance characteristics in predicting 90-day mortality as an outcome characteristic for use in clinical trials. Importantly, the authors attempted to overcome these limitations, to the degree that they could be using a resampling methodology, which is certainly a reasonable approach. However, this simulationbased approach depends implicitly on the data used to generate the simulations and falls short in providing confidence that the fundamental development principles listed above were adhered to. With all of this said, the authors’ work has attempted to advance our thinking about other relevant outcomes from PICU care and created a case for prospectively evaluating the longer term outcome of 90-day mortality with direction and purpose. They collected data on 1,400 children across three centers, three nations, and two continents, thus overcoming a number of operational and logistical challenges in the performance of multiinstitutional research across broad geographies. Finally, they used bootstrapping techniques with resampling to overcome the limitations posed by a relatively small sample size. All in all, their research incrementally continues to advance the work of SOI scoring and the methods that began many decades ago. For that, we need to be appreciative.