Passive Clinical Decision Support Penetrance: A Comparative Analysis in Two Pediatric Intensive Care Units

Clinical decision support (CDS) tools are designed to streamline workflows and enhance care quality in pediatric intensive care units (PICUs). Prior research has largely focused on interruptive, or active, CDS, leaving little known about the usage of passive CDS (pCDS) tools such as note writing support, order sets, and laboratory result flagging across institutions.

This study aimed to quantify and compare the penetrance of pCDS across two academic PICUs, with the aim of identifying shared practices and opportunities for standardization.

We conducted a retrospective observational cohort study of all patients admitted to two large PICUs from 2016 to 2023. Using Epic electronic health record data, we defined metrics across three domains of pCDS: note writing support, order set usage, and abnormal laboratory result highlighting. We harmonized elements and compared across institutions. We report descriptive statistics and compare sites using the Wilcoxon signed rank and chi-square test of significance.

The cohort included 90,575 PICU encounters and 2.65 million notes. We measured significant intersite variability in documentation practices: Children's Healthcare of Atlanta (CHOA) notes were significantly longer (2,020 vs. 1,434 characters) related to a 257% higher use of smart documentation elements and 10% greater reliance on copy/paste features. Across 8.36 million placed orders, order set usage differed by 20% (CHOA: 41 vs. University of Rochester Golisano Children's Hospital: 21%), with consistently higher use at CHOA, although both sites showed lower usage for medication and laboratory orders compared with nursing orders. Among >20 million laboratory results, abnormal flagging differed by 7% (56 vs. 49%), although variability was observed across test categories.

Passive CDS usage in PICUs is highly variable across institutions and domains. These findings highlight the need for standardized implementation and cross-site learning to optimize efficiency, reduce cognitive burden, and improve CDS.