Predicting Subsequent High-Frequency, Low-Acuity Utilization of the Pediatric Emergency Department

Margaret E Samuels-Kalow1, Matthew W Bryan2, Kathy N Shaw3

  • 1Department of Emergency Medicine, Massachusetts General Hospital, Boston, Mass.

Academic Pediatrics
|November 24, 2016
PubMed

Insights

A predictive model can identify children likely to frequently use the pediatric emergency department for low-acuity visits. This helps target interventions to improve healthcare utilization.

Area of Science:

  • Pediatric Emergency Medicine
  • Health Services Research
  • Predictive Analytics

Background:

  • High-frequency, low-acuity visits to pediatric emergency departments pose a challenge to healthcare resource allocation.
  • Developing tools to predict such utilization is crucial for optimizing care pathways.

Purpose of the Study:

  • To derive and validate a predictive model for high-frequency (≥4 visits/year), low-acuity (Emergency Severity Index [ESI] 4 or 5) utilization of pediatric emergency departments.
  • To identify key predictors available at triage for this specific patient group.

Main Methods:

  • Utilized 3 years of data (2012-2014) from a tertiary children's hospital.
  • Developed a predictive model using data available at triage, splitting data into derivation and testing sets.
  • Included variables with significant univariate association for multivariable modeling.

Main Results:

  • A total of 590 (1%) of 61,430 index visits resulted in ≥4 low-acuity visits in the subsequent year.
  • The final model included primary care site, age, acuity, previous utilization, race, and insurance.
  • The model achieved an area under the receiver operating characteristic curve of 0.84.

Conclusions:

  • Triage data can effectively predict subsequent high-frequency, low-acuity pediatric emergency department utilization.
  • Further validation and refinement in diverse settings are recommended.
  • Electronic risk calculation can facilitate targeted interventions for appropriate healthcare utilization.
Abstract