Prediction model for in-hospital admission in patients arriving in the Emergency Department

Background. To develop a prediction model for in-hospital admission to provide an almost "real time,. determination of hospital beds needed, so as to predict the resources required as early as possible. Material and methods. A prospective observational study in the emergency department of a university hospital. We included all consecutive patients between 8.00-22.00 hours during one month. We analyzed 7 variables taken when the patient arrived at the emergency department: age, sex, level of triage, initial disposition, first diagnosis, diagnostic test and medication, and we performed a logistic regression. Results. We included 2,476 visits of which 114 (4.6%) were admitted. A significant direct correlation was seen between: age >65 years old (odds ratio[OR]=2.1, confidence interval [Cl] 95%,1.3-3.2; p=0.001); male sex (OR=1.6, IC 95%,1.1-2.4; p=0.020); dyspnea (OR=5.2, IC 95%, 2.8-9.7; p<0.0001), abdominal pain (OR=4.7, IC 95%, 2.7-8.3; p<0.0001); acute care initial disposition (OR=8.9, IC 95%, 5.4-14.9; p<0.0001), diagnostic test (OR=1.1, IC 95%,0.9-1.3; p=0.064) and treatment prescription (OR=2.6, IC95%,1.642; p=<0.0001). The model had a sensitivity of 76% and a specificity of 82% (area under curve 0.85 [IC 95% 0.81-0.88; P<0.001]). Conclusions. The in-hospital admission prediction model is a good and useful tool for predicting the in-hospital beds needed when patients arrive at the emergency department.