An Evaluation of Antimicrobial Prescribing and Risk-adjusted Mortality.

Area of Science:

• Pediatric critical care medicine
• Infectious diseases
• Health informatics

Background:

• The Centers for Disease Control and Prevention (CDC) recommends tracking risk-adjusted antimicrobial prescribing.
• Previous quality improvement initiatives have used prescribing variation without adjusting for illness severity.
• Assessing antimicrobial use in pediatric intensive care units (ICUs) requires accurate risk adjustment.

Purpose of the Study:

• To determine the relationship between antimicrobial prescribing and risk-adjusted ICU mortality in the Pediatric Health Information Systems (PHIS) database.
• To evaluate the IBM-Watson risk of mortality model for this assessment.
• To explore an alternative risk model incorporating laboratory data.

Main Methods:

• A retrospective cohort study of pediatric ICU patients in the PHIS database (2010-2019).
• A nested study using PHIS+ data (2010-2012) to assess a risk model with laboratory data.
• Analysis of hospital antimicrobial prescription volumes and their association with risk-adjusted mortality.

Main Results:

• The study included 953,821 ICU encounters, with 2.7% nonsurvivors.
• A 4-fold variability in center-level antimicrobial use was observed.
• ICU antimicrobial use was not correlated with risk-adjusted mortality using the IBM-Watson model.
• A risk model incorporating laboratory data significantly outperformed the IBM-Watson model (c-statistic 0.940 vs. 0.891).

Conclusions:

• Antimicrobial prescribing in pediatric ICUs within the PHIS database shows variability but no association with risk-adjusted mortality via IBM-Watson.
• Integrating laboratory data into administrative databases offers more meaningful insights into multicenter antibiotic prescribing practices.
• Enhanced risk models are crucial for effective antibiotic stewardship programs.