Using Data Mining to Predict Hospital Admissions From the Emergency Department

Abstract

There is a risk that people may suffer serious harm as a result o f
overcrowding in emergency departments (EDs). As a result,
emergency clinics must explore employing innovative techniques
to boost patient flow while simultaneously minimising congestion
in the waiting area. One option for projecting emergency
department admissions is to use data mining and machine
learning technologies to anticipate ED admissions. This study,
which takes use of routinely obtained administrative d a ta (1 20
600 records) from two major acute hospitals in Northern Ireland,
presents a comparison of two rival machine learning algorithms
for predicting the risk of admission from the emergency
department at the hospital. Three algorithms are used in the
process of developing the prediction models: A decision tree may
be divided into three types, which are as follows: Decision trees
include: 1) decision trees, 2) gradient boosted machines, and 3 )
logistic regression, which are all types of decision trees (GBM) .
The GBM has an AUC-ROC D of 0:824 which was bet ter than
both the decision tree and the logistic regression model (accuracy
D 80:06 percent, AUC-ROC D 0:824). In this case, the accura cy
is 80:06 percent and the AUC-ROC is 0:824. In this situation, the
accuracy is 80:31 percent, and the AUC-ROC is 0:859, which
indicates a good fit. (0:849) (0:849) (AUC-ROC D 0:849)
(accuracy D 79:94 percent). We discovered a number of fa cto rs
that were connected with hospital admissions via the application
of logistic regression. These considerations included hospital
location, age, arrival mode, triage category, care group, and
previous hospitalisation during the previous month or year,
among other things. This study highlights the potential value o f
machine learning systems by using three fundamental machine
learning algorithms to predict patient admissions. Decision
support systems may be able to offer a picture of expected ED
admissions at any given moment as a consequence of this stud y,
allowing for resource planning ahead of time and avoiding
patient flow bottlenecks. This research also suggests that the
models described in this study may be utilised to perform
comparisons between projected and actual admission rates.
Generalised bivariate models (GBMs) are sufficient when
interpretability is a concern; however, if accuracy is crucial,
logistic regression models should be considered.