Model to predict risk for hospital admission and indicate the safety of reverse triage in a hospital emergency department: a prospective validation study

OBJECTIVES

To prospectively validate a model to predict hospital admission of patients given a low-priority classification on emergency department triage and to indicate the safety of reverse triage.

MATERIAL AND METHODS

Single-center observational study of a prospective cohort to validate a risk model incorporating demographic and emergency care process variables as well as vital signs. The cohort included emergency visits from patients over the age of 15 years with priority level classifications of IV and V according to the Andorran-Spanish triage system (Spanish acronym, MAT-SET) between October 2018 and June 2019. The area under the receiver operating characteristic curve (AUC) of the model was calculated to evaluate discrimination. Based on the model, we identified cut-off points to distinguish patients with low, intermediate, or high risk for hospital admission.

RESULTS

A total of 2110 emergencies were included in the validation cohort; 109 patients (5.2%) were hospitalized. The median age was 43.5 years (interquartile range, 31-60.3 years); 55.5% were female. The AUC was 0.71 (95% CI, 0.64-0.75). The model identified 357 patients (16.9%) at low risk of hospitalization and 240 (11.4%) at high risk. A total of 15.8% of the high-risk patients and 2.8% of the low-risk patients were hospitalized.

CONCLUSION

The validated model is able to identify risk for hospitalization among patients classified as low priority on triage. Patients identified as having high risk of hospitalization could be offered preferential treatment within the same level of priority at triage, while those at low risk of admission could be referred to a more appropriate care level on reverse triage.

Predicting hospital admission of patients with emergencies considered low priority according to assigned triage level

OBJECTIVES

To develop a model to predict hospital admission of patients in cases assessed as nonurgent or semiurgent on emergency department triage.

MATERIAL AND METHODS

Single-center observational study of a retrospective cohort. We included cases of patients older than 15 years whose emergency was classified as level IV-V according to the Andorran-Spanish triage model (MAT-SET, the Spanish acronym). Fourteen independent variables included demographic and care process items as well as vital signs. The dependent variable was hospital admission. The regression models were based on generalized estimating equations.

RESULTS

A total of 53 860 episodes were included; 3430 patients (6.4%) were admitted. The median (interquartile range) age was 44.5 (31.1-63.9) years, and 54.1% were female. Vital signs were recorded in 19.3% of the episodes. The model that best predicted admission included the following variables: age > 84 years (adjusted odds ratio [aOR], 6.72; 95% CI, 5.26-8.60); male sex (aOR, 1.46; 95% CI, 1.28-1.66); referral from a primary care center (aOR, 1.94; 95% CI, 1.64-2.29); referral from another acute-care hospital (aOR, 11.22; 95% CI, 4.42-28.51); arrival by ambulance (aOR, 3.72; 95% CI, 3.16-4.40); revisit 72 hours (aOR, 2.15; 95% CI, 1.60-2.87); systolic blood pressure $ 150 mmHg (aOR, 0.83; 95% CI, 0.71-0.97); diastolic blood pressure 60 mmHg (aOR, 1.57; 95% CI, 1.25-1.98); axillary temperature > 37°C (aOR, 2.29; 95% CI, 1.91-2.74); heart rate > 100 beats/min (aOR, 1.65; 95% CI, 1.40-1.96); baseline oxygen saturation in arterial blood (SaO2) 93% (aOR, 2.66; 95% CI, 1.86-3.81); and SaO2 93%-95% (aOR, 1.70; 95% CI, 1.42-2.05). The area under the receiver operating characteristic curve for the model was 0.82 (95% CI; 95% CI, 0.80-0.83).

CONCLUSION

The model predicts which patients are more likely to be admitted after their cases were initially considered nonurgent or semi-urgent on triage. Patients found to be at risk can then be given greater attention than others in the same triage level.