Margaret E Samuels-Kalow1, Matthew W Bryan2, Kathy N Shaw3. 1. Department of Emergency Medicine, Massachusetts General Hospital, Boston, Mass. Electronic address: Msamuels-kalow@partners.org. 2. Department of Biostatistics and Epidemiology, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pa. 3. Department of Pediatrics, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pa; Division of Emergency Medicine, The Children's Hospital of Philadelphia, Philadelphia, Pa.
Abstract
OBJECTIVE: To derive and test a predictive model for high-frequency (4 or more visits per year), low-acuity (emergency severity index 4 or 5) utilization of the pediatric emergency department. METHODS: The study sample used 3 years of data (2012-2014) from a single tertiary-care children's hospital for patients <21 years of age. Utilization in 2013 defined the index visit; prior utilization was drawn from 2012; and 2014 was used for outcome measurement. Candidate predictor variables were those that would be available at the time of triage. Data were split into derivation and test sets randomly; variables with a significant univariate association in the derivation set were included for multivariable modeling. The final model from the derivation set was then tested in the validation set, with calculation of a receiver operating characteristic curve. RESULTS: There were 90,972 visits in 2013, of which 61,430 were first (index) visits. A total of 590 (1%) had 4 or more triage level 4 or 5 visits in the following year (2014). The final model included site of primary care, age, acuity, previous utilization, race, and insurance, and had an area under the receiver operating characteristic curve of 0.84. CONCLUSIONS: Data available to the emergency department provider at the time of initial visit triage can predict utilization for low-acuity complaints in the subsequent year. Future work should focus on validation and refinement of the model in additional settings, and electronic calculation of risk status for targeted intervention to improve appropriate utilization of health care services.
OBJECTIVE: To derive and test a predictive model for high-frequency (4 or more visits per year), low-acuity (emergency severity index 4 or 5) utilization of the pediatric emergency department. METHODS: The study sample used 3 years of data (2012-2014) from a single tertiary-care children's hospital for patients <21 years of age. Utilization in 2013 defined the index visit; prior utilization was drawn from 2012; and 2014 was used for outcome measurement. Candidate predictor variables were those that would be available at the time of triage. Data were split into derivation and test sets randomly; variables with a significant univariate association in the derivation set were included for multivariable modeling. The final model from the derivation set was then tested in the validation set, with calculation of a receiver operating characteristic curve. RESULTS: There were 90,972 visits in 2013, of which 61,430 were first (index) visits. A total of 590 (1%) had 4 or more triage level 4 or 5 visits in the following year (2014). The final model included site of primary care, age, acuity, previous utilization, race, and insurance, and had an area under the receiver operating characteristic curve of 0.84. CONCLUSIONS: Data available to the emergency department provider at the time of initial visit triage can predict utilization for low-acuity complaints in the subsequent year. Future work should focus on validation and refinement of the model in additional settings, and electronic calculation of risk status for targeted intervention to improve appropriate utilization of health care services.
Authors: Alon Peltz; Margaret E Samuels-Kalow; Jonathan Rodean; Matthew Hall; Elizabeth R Alpern; Paul L Aronson; Jay G Berry; Kathy N Shaw; Rustin B Morse; Stephen B Freedman; Eyal Cohen; Harold K Simon; Samir S Shah; Yiannis Katsogridakis; Mark I Neuman Journal: Pediatrics Date: 2017-08-01 Impact factor: 7.124
Authors: Kristin N Ray; Kristin A Yahner; Jamil Bey; Katherine C Martin; Arrianna M Planey; Alison J Culyba; Elizabeth Miller Journal: Health Aff (Millwood) Date: 2021-01 Impact factor: 6.301