Neonatal sepsis causes significant mortality and morbidity worldwide. Diagnosis is usually confirmed via blood culture results. Blood culture sepsis confirmation can take days and suffer from contamination and false negatives. Empiric therapy with antibiotics is common. This study aims to retrospectively describe and compare treatments of blood culture-confirmed and unconfirmed, but suspected, sepsis within the University of Utah Hospital system. Electronic health records were obtained from 1,248 neonates from January 1, 2006, to December 31, 2017. Sepsis was categorized into early-onset (≤3 days of birth, EOS) and late-onset (>3 and ≤28 days of birth, LOS) and categorized as culture-confirmed sepsis if a pathogen was cultured from the blood and unconfirmed if all blood cultures were negative with no potentially contaminated blood cultures. Of 1,010 neonates in the EOS cohort, 23 (2.3%) were culture-confirmed, most with Escherichia coli (42%). Treatment for unconfirmed EOS lasted an average of 6.1 days with primarily gentamicin and ampicillin while confirmed patients were treated for an average of 12.3 days with increased administration of cefotaxime. Of 311 neonates in the LOS cohort, 62 (20%) were culture-confirmed, most culturing coagulase negative staphylococci (46%). Treatment courses for unconfirmed LOS lasted an average of 7.8 days while confirmed patients were treated for an average of 11.4 days, these patients were primarily treated with vancomycin and gentamicin. The use of cefotaxime for unconfirmed EOS and LOS increased throughout the study period. Cefotaxime administration was associated with an increase in neonatal mortality, even when potential confounding factors were added to the logistic regression model (adjusted odds ratio 2.8, 95%CI [1.21, 6.88], p = 0.02). These results may not be generalized to all hospitals and the use of cefotaxime may be a surrogate for other factors. Given the low rate of blood culture positive diagnosis and the high exposure rate of empiric antibiotics, this patient population might benefit from improved diagnostics with reevaluation of antibiotic use guidelines.

No abstract

No abstract

No abstract

Traditional spatial modeling approaches assume that data are second-order stationary, which is rarely true over large geographical areas. A simple way to model nonstationary data is to partition the space and build models for each region in the partition. This has the side effect of creating discontinuities in the prediction surface at region borders. The regional border smoothing approach ensures continuous predictions by using a weighted average of predictions from regional models. The R package remap is an implementation of regional border smoothing that builds a collection of spatial models. Special consideration is given to distance calculations that make remap package scalable to large problems. Using the remap package, as opposed to global spatial models, results in improved prediction accuracy on test data. These accuracy improvements, coupled with their computational feasibility, illustrate the efficacy of the remap approach to modeling nonstationary data.