545 - A Semi-Automated EPIC-Based Tool Improves Predictiveness of the Pediatric Early Warning Score for Pediatric Intensive Care Unit Transfers

Publication Number: 545.3863

Audrey Kamzan, UCLA Mattel Childrens Hospital, Los Angeles, CA, United States; Scott Hu, University of California, Los Angeles David Geffen School of Medicine, Los Angeles, CA, United States; Kevin Chhan, Information Services & Solutions UCLA, Los Angeles, CA, United States

Background: The pediatric early warning score (PEWS) is a validated scoring tool designed to predict pediatric patient decompensation in the acute care setting, with a goal of preventing delays in interventions or transfers to an intensive care unit (ICU). However, PEWS can be challenging to score accurately as it requires compiling several data points including vital sign data, physical exam findings, level of oxygen support, and other variables.

Objective: To determine whether a semi-automated electronic medical record (EMR)-based tool improves the predictiveness of PEWS in detecting need for transfer to an ICU.

Design/Methods: Prior to implementation of the semi-automated tool at our tertiary care children’s hospital, PEWS was being scored every 4 hours by the bedside nurse via manual calculation. We designed a semi-automated tool within the EMR that would automatically gather data from the patient record to help calculate PEWS including vital signs for age and level of oxygen support. The nurses would then answer six multiple-choice questions every 4 hours pertaining to physical exam findings including behavior, color, capillary refill, and respiratory effort, as well as persistent postoperative vomiting (yes/no) and need for continuous albuterol (yes/no). The tool would then combine the nurse-provided information with the information automatically pulled from the chart to produce a total score. The new tool went live at our tertiary care center on 12/8/2021. We gathered retrospective data on all admissions to our tertiary care institution pre-implementation starting in 7/1/2021, then post-implementation through 6/30/2023. We compared the PEWS scores of patients who were subsequently transferred to the ICU with the PEWS scores of patients who did not require ICU transfer. A receiver operating characteristic analysis was performed to evaluate the performance of PEWS on predicting clinical deterioration within a 12-hour period after it was scored. A 2-hour window prior to clinical deterioration was used to exclude scores that were too close to the transfer event to have impacted the clinical decision making.

Results: During the study period, there were 1263 patient admissions and 79 ICU transfers. Prior to implementation of the semi-automated tool, the area under the curve (AUC) was 0.59. After implementation of the tool, the AUC improved to 0.67 (Figs. 1 and 2) (95% confidence interval, 0.608-0.698).

Conclusion(s): A semi-automated EPIC-based tool demonstrated statistically significant improvement in the ability of the PEWS score to predict ICU transfer.

PEWS Score Receiver Operating Characteristic Post Semi-Automated Tool Implementation
Pre-implementation of the semi-automated tool, the area under the curve was 0.59 (where a tool with perfect sensitivity and specificity would have an AUC of 1.0)

PEWS Score Receiver Operating Characteristic Post Semi-Automated Tool Implementation
Pre-implementation of the semi-automated tool, the area under the curve was 0.67 (where a tool with perfect sensitivity and specificity would have an AUC of 1.0)