What have we learned from observational studies on neonatal sepsis?

To assess how observational studies on neonatal sepsis can help define the knowledge base required for neonatal randomized, clinical trials.

Methodologic review of past observational studies and critical reviews.

Observational studies on neonatal sepsis have suffered from important limitations: failure to employ multivariate analyses, considering infection in isolation (ignoring coexisting respiratory distress), failure to provide likelihood ratios for predictors and combinations of predictors, and ignoring the phenomenologic dimension of clinicians' experience.

Future observational studies must address three key issues. They should begin with a clear analytic and sampling plan that pays careful attention to the proper use and reporting of multivariate analyses. Second, they must explicitly address two subpopulations: critically ill newborns with negative cultures and asymptomatic newborns with positive cultures. Finally, they should be theory driven and provide empirical physiologic data that permit situating their results in the context of the evolving systemic inflammatory response syndrome and PIRO (predisposition, infection, host response, organ dysfunction) models.