Escrever uma avaliação PREreview

Background and Objectives: Most current pediatric sepsis screening tools operate in the Emergency Department to detect low incidence Phoenix criteria sepsis. We sought to create an inpatient alert with broader scope and higher predictive value to identify infectious illness cases likely to progress to major physiological decompensation. Methods: Clinician reviews of cases meeting a modified pSOFA sepsis criteria within a 10,992 inpatient cohort identified 83 cases of “gold standard” infection-related decompensation (GS-IRD). Our refinement process utilized Lasso regression to create a model identifying children likely to progress to major or extreme decompensation according to 3M’s APR-DRG severity of illness (SOI) index. Prior to live implementation, a second 4,050 patient cohort was selected and 50 GS-IRD cases identified for external validation of the model and as a baseline for assessing post-implementation patient improvements. We examined the relationship between patients’ maximum alert “score” and final SOI level. We hypothesized that post-firing decreases in score represent clinical improvement, affording a measure of the alert’s efficacy through silent versus live cohort comparisons. Results: Our final model exhibited 85.0% sensitivity and 93.6% specificity for GS-IRD cases, and positive and negative predictive values of 43.6% and 83.7% for major or extreme SOI outcomes. The live epoch showed a greater average decrease in both alert scores over the 48-hour post-firing interval (p< .0001) and maximum scores at ≥48 hours post-admission (p< .0001), compared with the silent epoch. Conclusion: Our pediatric alert shows a high predictive value for progression of infectious illness to major or extreme severity. Alert implementation was associated with significant post-firing clinical improvements.