OBJECTIVE: This pilot study aims to identify potential predictors of postadmission interventions of hospitalized croup patients and derive a risk model aimed at reducing hospitalizations for croup.
METHODS: Data were collected on all croup hospitalizations for patients aged 1 month to 17 years admitted through a community hospital's emergency department (ED) between 2012 and 2017. Potential predictors were obtained from the electronic medical records including demographics, vital signs, ED length of stay, preintervention and postintervention Westley Croup Score (WCS), number of racemic epinephrine nebulizations administered, time to dexamethasone administration, preexisting conditions, and additional interventions during hospitalization. Statistical analysis used the outcome "patient received racemic epinephrine after hospital admission (yes/no)" to identify characteristics of the child or ED visit associated with that outcome. Preliminary analyses using stepwise logistic regression, tree models, and random forests suggested predictors, interactions among predictors, and the form of their association with the outcome. A final analysis used logistic regression.
RESULTS: A total of 116 croup admissions were included, of which 19 (16%) received racemic epinephrine posthospitalization. These characteristics were identified as having some predictive power: sex, preexisting conditions, and preintervention and postintervention WCS, along with the interaction between sex and postintervention WCS. Logistic regression estimated an equation describing the probability of postadmission intervention, permitting the choice among admission thresholds giving different sensitivities and specificities.
CONCLUSIONS: There appear to be promising predictors in croup patients presenting to the ED, which might help stratify risk for interventions after the ED encounter and thus reduce the number of potentially avoidable admissions.
- predictive model
PubMed: MeSH publication types
- Journal Article