Disclosure(s): No relevant financial relationship(s) to disclose.
First Author: Adam C. Dziorny, MD, PhD – Golisano Children's Hospital University of Rochester Co-Author: Akira Nishisaki, MD, MSCE – Childrens Hospital of Philadelphia Co-Author: Michael Spaeder, MD, FCCM – Program Director, Fellowship in Pediatric Critical Care Medicine; Professor of Pediatrics, University of Virginia School of Medicine Co-Author: Sherry Kausch, PhD – Research Assistant Professor, University of Virginia Co-Author: Nelson Sanchez-Pinto, MD, FCCM – Associate Professor of Pediatrics (Critical Care), Ann & Robert H. Lurie Children's Hospital of Chicago Co-Author: Daniel Tawfik, MD, MS – Assistant Professor, Pediatrics, Stanford University School of Medicine Co-Author: Timothy Cornell, MD – Professor, Pediatrics, Stanford University School of Medicine Co-Author: Tellen Bennett, MD, MS – Professor, Pediatrics, University of Colorado Anschutz Medical Campus Co-Author: Mark Mai, MD, MHS – Children's Healthcare of Atlanta/Emory University Co-Author: Randall C. Wetzel, BS, MB, MBA – CEO; Professor, Childrens Hospital Los Angeles
Introduction: Scheduled laboratory tests in the pediatric intensive care unit may be overutilized and contribute to reduced care quality. The objective of this study was to quantify inter-site variation in laboratory ordering practices across PICUs. We hypothesized that significant variation exists, which may offer opportunities for testing reduction.
Methods: We developed queries to characterize laboratory ordering and shared these with 7 participating PICU Data Collaborative sites. In this federated analysis we provided identical R Markdown scripts, and each site ran the script locally without sharing row-level data. We distinguished provider-initiated (“placed”) parent orders from the child (“sent”) orders. At each site, lab names were mapped to a list of the most common names. We measured characteristics including placement frequency, order set usage, urgency level, and ordering provider. We compared proportions across sites with the Chi-squared test of significance.
Results: Among 178,520 ICU encounters (1.2 million patient-days), there were 7.0 million placed lab orders and 10.7 million child lab orders. Orders placed per patient-day differed by site (range: 2.8 – 16.4, median: 4.4). We mapped 3.3 million placed orders (47.1%) to the 22 most common lab names. Of these, the most frequent lab orders included CBC (21.5%), BMP (10.2%), Magnesium (10.2%), and ABG (9.4%). All measured lab order characteristics significantly differed across sites (p < 0.001). Few orders were placed via an order set (per-site range: 9.1-24.2%), with many of these occurring in the first 2 hours of ICU admission (16.8 – 45.6%). Placed order frequency significantly differed across sites with the two most common being “One Time” (26.0–81.2%) and “Daily” (6.5-24.2%). Use of “STAT” ordering differed by site (2.4-34.3%). Physicians were the predominant ordering provider at all sites (39.6-81.3%) followed by nurse practitioners (12.9-22.3%).
Conclusions: Laboratory ordering practices vary significantly across sites, with wide variability in the order frequency, timing, and source. Applying best practices from individual sites to others may offer opportunities for site-specific improvement. Effective decision support designed to reduce unnecessary laboratory ordering must account for site-specific variation and fit local workflows.
