lww.com, as data

**Systematic Reviews and Meta-Analyses.** *Critical Care Medicine* requires authors to provide the following information for all submitted systematic reviews and meta-analyses as part of the submission:Scoping reviews are inherently exploratory and are ordinarily referred to *Critical Care Explorations* at the time of submission.***Statistical Considerations*:** For all studies, teams should consider engaging a statistical consultant at the design stage. Description of the project plan, power calculations, evaluation methods proposed, stopping rules, and related information is especially helpful when submitted as supplementary information.* While criteria of "significance" are the province of the study designers and analysts, the selection of tests and the thresholds for significance must be justified, including measures taken to avoid false discovery.* Avoid assertion and reporting of "trends" where data are incomplete or a predetermined threshold for significance is not achieved.* Communication of confidence intervals is preferred over listing of *P* values wherever possible.* Where *P* values are used, they should not be reported in isolation; a measure of effect size should also be included.* Sensitivity analyses should be considered when there is the potential for unmeasured confounding or numerous ways to define either exposures or outcomes.* Fragility analyses should be conducted where attribution of a different outcome to a small number of individuals might substantially change the results.* Reference (control) group selection should be appropriate for the hypothesis being tested. When selecting reference groups in retrospect, authors should carefully distinguish between covariate matching and propensity matching.***Predictive Models*:** Authors presenting predictive models must include comparison with generally accepted models that address a similar prediction problem. If the performance or usability of the new model is not substantially better than generally accepted models or models previously reported, the authors must explain in their cover letter why it is important to add their new model to the collection of previously reported models. Authors presenting predictive models must either indicate that they have implemented their model in their clinical environment and report its prospective performance or else explain in the cover letter and in the manuscript why it is important to report the performance of an unimplemented model.