Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

Patient Flow Control in Emergency Departments

Using Simulation Modeling and the
Random Forest Algorithm
Pyelshak Yusuf1 Fatima Umar Zambuk2; Badamasi Imam Yau3
Department of Health Information Management Department Of Mathematical Sciences
Plateau State College of Health Tech Abubakar Tafawa Balewa University
Plateau, Nigeria Bauchi, Nigeria

Solomon Rifkatu Aaron 4 Atangs Ishaku5

School of Environmental Health School of Community Health
Plateau State College of Health Tech Plateau State College of Health Technology Zawan
Plateau, Nigeria Plateau, Nigeria

Aminu Agabus6 Solomon Panshak Dawal7

Department of Computer Sciences School of Health Information Management
Nigerian Army University Plateau State College of Health Tech
Biu Nigeria Plateau, Nigeria

Ismail Zahraddeen Yakubu8

Department of Computing Technologies
SRM Institute of Science and Technology Kattankulathur, Chennai, India, 603203

Abstract:- The proposed thesis aims to optimize patient I. INTRODUCTION

flow and reduce waiting times in emergency
departments using simulation modeling and the Emergency departments (EDs) are essential to
Random Forest algorithm. Emergency departments healthcare systems, providing medical care for patients
face significant challenges in managing patient flow and with acute illnesses and injuries. EDs operate 24 hours a
reducing waiting times, which can lead to increased day, seven days a week, and are often the first point of
patient dissatisfaction and decreased quality of care. contact for patients seeking medical attention (Zibulewsky,
The proposed solution uses simulation modeling to 2021). However, EDs are also known for long waiting
create a virtual model of the emergency department and times and overcrowding, which can lead to negative health
simulate patient flow under different scenarios. The outcomes and patient dissatisfaction. Research has shown
Random Forest algorithm is then used to analyze the that prolonged waiting times in EDs are associated with
simulation results and identify the factors impacting increased morbidity, mortality, and healthcare costs (Shen,
patient flow and waiting times. By optimizing these & Lee, 2017).
factors, the proposed solution aims to reduce waiting
times and improve the overall patient experience. The The problem of long waiting times and overcrowding
research involves the development and validation of the in EDs is a complex issue with multiple causes, including
simulation model and the implementation of the high patient demand, limited resources, and inefficiencies
Random Forest algorithm using real-world emergency in patient flow management. Despite numerous efforts to
department data. The outcomes of the implemented improve ED patient flow, existing methods have limitations
Random Forest Model in Chapter Four showcase its and may not be sufficient to address the problem
efficacy with an accuracy rate of 0.85, sensitivity rate of comprehensively. Therefore, new approaches are needed to
0.99, and other favorable metrics. The proposed optimize patient flow and reduce waiting times in EDs
solution has the potential to improve patient outcomes (Yarmohammadian, et, al., 2017).
and reduce costs associated with emergency department
overcrowding and delays. Queueing theory and Random Forest Algorithm are
two promising approaches that can be used to optimize
Keywords:- Emergency Department, Patient Flow Control, patient flow and reduce waiting times in EDs. Queueing
Machine Learning Algorithm, Simulation Model. theory provides a mathematical framework for analyzing
the behavior of systems with waiting lines. Queueing
models can be used to estimate waiting times, identify
bottlenecks in the system, and optimize resource allocation
(Green & Yih, 2011).

IJISRT24MAR1035 www.ijisrt.com 3107

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

Random Forest is a machine learning algorithm that condition severity. The proposed method enhances wait-
belongs to the category of ensemble methods. It is used for time estimates by constantly updating and altering
both classification and regression tasks. In Random Forest, predictions based on patient and ED-specific data, resulting
a large number of decision trees are created and combined in more accurate probabilistic and point forecasts.
to make a prediction. Each decision tree is trained on a
subset of the training data, and at each split, a random Various approaches, including machine learning and
subset of features is selected to determine the best split. systems thinking, have been used to anticipate ED wait
This process is repeated for each tree, and the final times, as proven by Kuo et al. (2020) and Stagge (2020).
prediction is made by aggregating the predictions of all the
individual trees. The main advantages of Random Forest Arha (2017) and Curtis et al. (2018) used machine
are its ability to handle large datasets, its robustness to learning algorithms to forecast wait times for low-acuity
noise and outliers, and its ability to capture complex patients in the emergency department, taking into account
nonlinear relationships in the data. It also provides parameters such as arrival, service completion, and
estimates of feature importance, which can be used to examination. Studies on patient waiting time before
identify the most relevant features for the task at hand. treatment use quantile regression, but this study uses multi-
Random Forest has been successfully applied in a wide DL optimization strategies and extracts new predictors
range of domains, including finance, healthcare, and from patient joining, queue waiting time, and departure
marketing. However, it may not be the best algorithm for time. Because of long wait times and congestion in many
all tasks, and other algorithms such as support vector hospitals throughout the globe, the number of emergency
machines or neural networks may be more appropriate department visits in the United States is growing year after
depending on the specifics of the problem (Breiman, 2001). year (Di et al. 2015).

Some of the obstacles experienced by existing model According to the National Center for Health, 145.6
were the imbalanced nature of the dataset in the class million individuals visit the ED each year, with rising visits
variables. A need to balance the dataset, increase the and wait times. Since 2015, the Canadian Institute for
dataset size and use an advanced model for performance Health Information has documented considerable growth.
enhancement is necessary. The proposed model tries to These problems might be addressed by evaluating ER
take that measures for the betterment of the system. The efficiency. (Rasouli et al. 2019). By tracking patient arrival
remaining sections of the article are related work, method times, some hospitals utilize queuing models to enhance
findings and conclusion. staff allocation. (Kaushal et al. 2015; Sasanfar et al. 2020).
Predictive models are critical in the medical business for
Alenany & Cadi. (2020) combines machine learning anticipating patient wait times utilizing past data and
(ML) and simulation models to model patient flow in an efficiently handling seasonal arrival and wait times. (Ruben
emergency department (ED). The ML model predicts et al. 2010; Cai et al. 2016). Electronic Health Record EHR
whether a patient will be admitted to the inpatient unit after data is critical for uncovering hidden healthcare concerns
receiving treatment at the ED, based on patient data. The and improving queuing systems, especially in predictive
simulation model uses this output to assess the expected models for future behavior analysis. (Eiset et al. 2019).
reduction in patient length of stay (LOS) and door-to-
doctor time (DTDT) if patients are admitted directly to Machine learning approaches were used in the
inpatient units at an early stage of their ED journey. The research on queuing behavior projection, however their
study shows that using ML and simulation can help manage time series analysis on queue data prediction study is
ED patient flow and reduce congestion, with the potential faulty. (Srivastava 2016; Stagge 2020). According to Dong
for further improvements through increased data size and et al.'s 2019 research, ED waiting time is an important
the use of other ML models. However, the effect of other aspect people evaluate when selecting their medical care
related measures on patient quality should also be provider. The previously released data assists in operational
considered. Hence, the proposed model attempt to improve choices targeted at minimizing wait times and congestion in
this model using the Random Forest Algorithm instead of the Emergency Room. (Abir et al. 2019).
the Decision tree used in the existing model. The data size
will also be increased to enable an improvement in the Kroer et al. (2018) and Meersman and Maenhout
performance of the proposed model. (2022) investigated capacity allocation for elective and
emergency patients to decrease wait times and OR and
II. RELATED WORK overtime expenditures.

E James W. T. (2023) investigates the estimation of For allocating COVID-19 patients and speciality
patient waiting times in emergency rooms, emphasizing the teams, Arab Momeni et al. (2022) offered a mixed-integer
need for more precise and nuanced projections to increase mathematical programming technique, while Wang et al.
patient satisfaction and decrease abandonment. To produce (2016) employed a discrete simulation model.
probabilistic predictions from huge patient-level data sets, a
quantile regression forest machine learning technique was Tuwatananurak et al. (2019) used a 15,000 surgical
utilized, extracting predictor parameters such as calendar case data set to predict patient surgery duration using leap
influences, demographics, staff count, ED load, and patient Rail, a customized machine learning algorithm.

IJISRT24MAR1035 www.ijisrt.com 3108

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

Fairley et al. (2019) forecasted PACU time using surgery scheduling. Machine learning models were helpful
machine learning, resulting in lower holdings and cost in OT scheduling.
reductions. Schiele et al. (2021) combined ORs and units to
schedule master operations. Shuvo et al. (2020) created a III. METHODOLOGY
deep reinforcement learning strategy.
The proposed model aims to optimize patient flow and
Luo and Wang (2019) used machine learning reduce waiting times in emergency departments using a
algorithms to identify canceled procedures, with the simulation model and random forest algorithm. The
random forest model proving to be the most successful, simulation model is used to simulate patient flow and
allowing preventative actions to be taken to lower predict patient volumes, acuity levels, and waiting times.
cancellation rates. To forecast surgical cancellations at The random forest algorithm is then used to analyze the
West China Hospital, Luo et al. (2016) used machine simulation results and make predictions based on patient
learning approaches such as boosting, Bayesian additive data, such as age, gender, and presenting complaint. The
regression trees, and random forest. performance of the model is evaluated using metrics such
as accuracy, specificity, and sensitivity. The goal is to
Erekat et al. (2020) and Zhao et al. (2019) employed improve patient flow and reduce waiting times in
data mining approaches to estimate surgery cancellations, emergency departments, ultimately leading to better patient
resulting in cost reductions and more efficient robotic outcomes and higher patient satisfaction.

Fig 1: Framework of the Proposed System

A simulation model of the emergency department's emergency department's patient flow process to reduce
patient flow process will be developed using the Rockwell waiting times, improve patient flow, and predict if the
Arena simulator V15. The model includes; relevant patient patient can move to the Inpatient Unit or should stay in the
flow components, such as arrival patterns, triage, ED. The effectiveness of the optimization strategies is
registration, examination, diagnosis, treatment, and evaluated by comparing the simulation results before and
discharge. The developed model will be verified to ensure after applying the strategies.
that it accurately represents the actual emergency
department's patient flow process. Different scenarios are A. Proposed Model Dataset
created to simulate various patient flow process changes, The summary statistics provided in recent research
such as changes in staffing levels, triage processes, (for instance, Graham et al., 2018) served as an inspiration
examination processes, and treatment processes. The for the data gathered for the various aspects. There are
simulation model is run using the created scenarios to created 500 patient records with six characteristics. In order
generate data on patient flow and waiting times for each to build the prediction model, patient records that were
scenario. created and sent to the ED are summarized. The produced
data's rate of admitted and non-admitted patients is
The simulation output data is analyzed using the consistent with previous research (such as Graham et al.,
random forest algorithm to identify key factors that affect 2018), which is the same dataset used in the existing system
patient flow and waiting times in the emergency and shall be use by the proposed system to justify the
department. Based on the results of the data analysis, comparison.
optimization strategies can be developed and applied to the

IJISRT24MAR1035 www.ijisrt.com 3109

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

B. Proposed System Algorithm

The proposed model algorithm consists of the
following steps;

 Step 1: Start
 Step 2: Problem Identification and Data Collection
 Using the same dataset as in the existing Model
 Step 3: Model Development: Using the Rockwell Arena
simulator V15
 Includes the relevant patient flow components, such as
arrival patterns, triage, registration, examination,
diagnosis, treatment, and discharge.
 Step 4: Model Verification and Validation
 To represent the actual emergency department patient
flow
 Compare the simulated result with the real result,
 If accurate goto to step 5 else step 3
 Step 5: Scenario Creation, Such as
 Staffing levels
 Triage processes
 Examination processes, and
 Treatment processes.
 Step 6: Simulation Runs: With the created Scenario
 To generate data on patient flow and waiting times for
each scenario.
 Step 7: Data analysis: With Random Forest Model
 The Simulation data is then analyzed to identify the key
factors in the effects patient flow process to reduce
waiting time
 Step 8: Optimization: Best on the result of the analysed
in step 7,
 To reduce waiting time
 Improve patient flow
 Step 9: Evaluation; Measure the model performance
 With Accuracy, Sensitivity and Specificity
 Step 10: Stop
 The flowchart of the proposed system is depicted in
Figure 3.3.

Fig 2: The Flowchart of the Proposed System

C. Implementation and Evaluation Metric age as input factors that affect patient waiting times. By
The proposed model is an integration of the including the statistical distributions of these processes in
Simulation model and Machine learning technique (known the DES, accurate predictions of waiting times can be
as the Random Forest Algorithm). The System requirement obtained. The RF model, in particular, has shown good
of the model consists of the requirments of the simulation performance with low Root mean square error (RMSE),
Model (i.e Rockwell Arena simulator V15) and that of the Mean Square Error (MSE), Mean Absolute Error (MAE),
Machine learning model. The model is evaluated against and high R2 values. This integration of DES and ML
existing method using Accuracy, Sensitivity and models can help overcome various factors, such as
Specificity. satisfaction, cost, and quality, in service sectors with
dynamic structures.
IV. EXPERIMENTAL SETUP AND RESULTS
RMSE, MSE, and MAE serve as customary metrics or
The study attempts to Integrate a discrete event measures employed in the evaluation of the efficacy of
simulation (DES) model with machine learning (ML) predictive model precision, particularly within the
algorithms, such as random forest (RF), which had a framework of regression analysis. These metrics or
positive effect on patient flow and waiting times in the measures effectively gauge the disparity between
emergency departments (EDs). The DES model considers anticipated values and factual values, thereby furnishing a
factors such as length of stay, resource efficiency rates, metric indicative of the model's precision and its ability to
patient genders, walking distance, time of processes, and capture the inherent patterns within the data set.

IJISRT24MAR1035 www.ijisrt.com 3110

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

Table 1: Statistical Summary of the Patient Records

SN Features Categories % per categories
1 Gender Female 67.3
Male 32.7
2 Arrival Day Monday 26.1
Tuesday 15.3
Wednesday 15.6
Thursday 14.2
Friday 15.8
Saturday 10.7
Sunday 2.3
3 Triage High 42.85
Medium 28.57
Low 28.57
4 X_RAY 0 57.14
1 42.85
5 Lab 0 57.14
1 42.85
6 Status (Admission) Normal 57.14
Critical 42.85

The dataset comprised two distinct types: specifically, A. The Result of the Proposed System
the categorical attributes (comprising gender, arrival day, After the triumphant execution of the Random Forest
triage x-ray, and lab) were presented in Table 1, while the Classifier to diminish the duration of waiting, and refine the
numerical attributes consisted of patient registration movement of patients in a critical care center, Figure 2
number, age, and arrival time. portrays the random forest tree of the Innovative model.
The depiction bestows a profound understanding of how
the model arrives at conclusions grounded on various
attributes, which are recorded in Table 2.

Table 2: The Performance Result of the Proposed Random Forest Model

The algorithm Result
1 Accuracy 0.85
2 Specificity 0.58
3 Sensitivity 0.99
4 Precision 0.86
5 F1-Score 0.92

The model's exceptional sensitivity, with a value of B. Graph of the Random Forest
0.99, showcases its remarkable ability to accurately predict Figure 2 embodies the ethereal Forest Tree Model
a whopping 99% of the actual positive instances. With a within a mesmerizing graph, showcasing the graceful dance
precision of 0.86, the model impressively identifies 86% of of patients as they traverse the labyrinthine corridors of the
the instances predicted as positive, which indeed proves its ATBU Teaching Hospital. A captivating masterpiece, it
proficiency. A remarkable F1-Score of 0.92 indicates a beckons the inquisitive mind to delve deeper into its
harmonious equilibrium between precision and recall, enigmatic depths, where limitless interpretations lie in wait,
solidifying the model's prowess. Furthermore, an accuracy ready to be unraveled in the forthcoming section.
of 0.85 signifies that the model astutely predicted 85% of
the instances, making it commendable in its capabilities.

The model's superior performance in identifying

positive instances is reflected in its elevated sensitivity and
remarkable F1-Score. However, the precision reveals that
among the predicted positive instances, there are some
inaccuracies. To holistically assess the model's
performance, the overall accuracy is provided, offering a
comprehensive perspective encompassing both positive and
negative instances.

IJISRT24MAR1035 www.ijisrt.com 3111

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

Fig 3: The Forest Plot of the Proposed Random Forest Model

Fig 3: Showcases the Manifesto Obtained from the Graphical Arrangement of the Proposed RF Model

IJISRT24MAR1035 www.ijisrt.com 3112

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

Figure 3 showcases the Manifesto obtained from the  Left-Left Branch:

graphical arrangement of the proposed RF model.
 Feature: arrival_time_08:58:00
The provided Manifesto in Figure 4.2 is an artistic  Decision: If the value is less than or equal to 0.50,
portrayal of judgmental foliage. Each strand signifies a proceed to the left branch; otherwise, predict the class
critical criterion for a distinct attribute, while the as 0.0.
subsequent strands, gracefully indented below, epitomize
sub-decisions based on the preceding criterion. The  Left-Left-Left Branch:
structure of this verdant tree aids in comprehending the
modus operandi of the model, as it expertly predicts  Feature: arrival_time_11:39:00
outcomes based on the input attributes.  Decision: If the value is less than or equal to 0.50,
proceed to the left branch; otherwise, predict the class
 To Understand it Better, Let's Interpret from the Roots; as 0.0.

 Root Node:  ... and so on.

The decision tree continues to split based on different
 Feature: arrival_time_01:09:00 features until it reaches a point where it assigns a class label
 Decision: If the value is less than or equal to 0.50, (0.0 or 1.0). Each path from the root to a leaf represents a
proceed to the left branch; otherwise, go to the right set of conditions that lead to a specific prediction.
branch.
C. Comparing the Model's Performance
 Left Branch: The performance analysis of the proposed model
described in Table 4.2 will be compared with the results of
 Feature: arrival_time_10:53:00 the existing approaches discussed in (Alenany & Cadi.,
 Decision: If the value is less than or equal to 0.50, 2020) which can be demonstrated in Table 3.
proceed to the left branch; otherwise, predict the class
as 0.0.

Table 3: Comparing the Predicted Model Performance

The algorithm Accuracy Sensitivity Specificity
1 DT1 0.81 0.89 0.39
2 DT1 0.72 0.88 0.26
3 kNN (k=1) 0.74 0.79 0.40
4 Graham, et al., (2018) DT 0.80 0.90 0.53
5 Proposed RF Model 0.85 0.99 0.58

D. Result Discussions a specificity of 0.58, suggesting its moderate effectiveness

The Proposed RF Model presents itself as the epitome in correctly identifying negative instances. The Proposed
of accuracy (0.85) when compared to the other models RF Model truly stands out due to its extraordinary accuracy
listed, implying its exceptional performance overall in and sensitivity. However, it is of utmost importance to take
contrast to the existing methods showcased in Table 4.3. into account the specific requirements of the application
This Model also demonstrates a remarkable level of when interpreting and selecting models. The comparative
sensitivity (0.99), signifying its innate ability to accurately performance of the Models is depicted in Figure 4.
identify positive instances. Moreover, the Model possesses

Fig 4: The Comparative Analysis of the Proposed System with the Existing Systems

IJISRT24MAR1035 www.ijisrt.com 3113

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

E. Model Development foundation for accurately modeling the movement of

A modeling software was utilized to craft an patients across hospital.
ingenious model of the hospital emergency department.
Meticulous attention was paid to the precise layout of the Figure 5, is a visual representation showcasing the
hospital, which proved instrumental in the development of hospital's layout and the seamless transition of patients
the said model. A thorough analysis of the patients' flow in between different departments. The blue line elegantly
the section as mentioned earlier served as a valuable illustrates the enchanting split flow pattern that guides
patients throughout the entirety of the hospital.

Fig 5: Visual Representation of the Hospital's Layout

V. CONCLUSION Based on the discoveries and accomplishments

delineated in the investigation, it is highly recommended
In conclusion, this research endeavors to tackle the that healthcare establishments, particularly emergency
obstacles faced by emergency departments, particularly the departments, contemplate the adoption and implementation
issue of lengthy waiting times and overcrowding, by of the proposed system that integrates the principles of
employing innovative methodologies such as queueing queueing theory with the Random Forest Algorithm. The
theory and the Random Forest Algorithm. The significance triumphant application of this model in the optimization of
of this problem is well-established, highlighting its global patient flow and reduction of waiting times, as evidenced
impact on healthcare systems. The extensive review of by an accuracy of 0.85, sensitivity of 0.99, and other
existing literature identifies gaps and informs the favorable metrics, indicates its potential to augment
development of the proposed system, which incorporates a operational efficiency. Institutions should prioritize the
simulation model with the Random Forest Algorithm to fusion of both simulation modeling and machine learning
enhance the optimization of patient flow. Chapter Three algorithms, considering the specific attributes outlined in
outlines the methodology, showcasing the step-by-step the study, such as patient registration, age, gender, arrival
process and emphasizing improvements over the current particulars, triage, x-ray, lab, and admission status.
model. The outcomes of the implemented Random Forest However, it is imperative for decision-makers to
Model in Chapter Four showcase its efficacy with an judiciously assess and customize the model to the
accuracy rate of 0.85, sensitivity rate of 0.99, and other distinctive characteristics and requisites of their respective
favorable metrics. While the proposed model surpasses healthcare environments. Regular evaluations and
existing approaches, it is crucial to emphasize the comparisons with existing models should be conducted to
importance of considering specific application ensure continued improvement and pertinence in
requirements. This research not only contributes to the addressing the ever-evolving challenges in the management
advancement of patient flow optimization but also provides of emergency healthcare.
a framework for future evaluations and comparisons
between systems, ensuring ongoing refinement and
adaptability in addressing emerging healthcare challenges.

IJISRT24MAR1035 www.ijisrt.com 3114

Volume 9, Issue 3, March – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24MAR1035

ACKNOWLEDGMENT [12]. Oh, S. H., Park, J., Lee, S. J., Kang, S., & Mo, J.
(2022). Reinforcement learning-based expanded
We wish to appreciate the expert support of our personalized diabetes treatment recommendation
supervisors Dr. F. U. Zambuk and Dr. B. I Ya’u towards using South Korean electronic health records.
the success of this research. Expert Systems with Applications, 206, 117932.
[13]. Razavi, S., de la Hoz, E., & Akhavan-Tabatabaei,
REFERENCES R. (2021). Optimizing Patient Flow in Emergency
Departments Using Discrete Event Simulation and
[1]. Alenany, E., & Cadi, A. A. E. (2020). Modeling Deep Reinforcement Learning. International
patient flow in the emergency department using Journal of Environmental Research and Public
machine learning and simulation. arXiv preprint Health, 18(9), 4722.
arXiv:2012.01192. https://doi.org/10.3390/ijerph18094722
[2]. Elalouf, A., & Wachtel, G. (2021, December).
Queueing problems in emergency departments: a
review of practical approaches and research
methodologies. In Operations Research Forum
(Vol. 3, No. 1, p. 2). Cham: Springer International
Publishing.
[3]. Finkelstein, S. M., Wong, E., Lin, H., & Li, Y.
(2021). Reducing Patient Waiting Time in
Emergency Departments Using Reinforcement
Learning and Queueing Theory. INFORMS
Journal on Applied Analytics, 51(3), 162–179.
https://doi.org/10.1287/inte.2020.1064
[4]. Haripriya, G., Abinaya, K., Aarthi, N., & Kumar,
P. P. (2021). Random Forest Algorithms in Health
Care Sectors: A Review of Applications.
[5]. Jarvis, P. R. E. (2016). Improving emergency
department patient flow. Clinical and experimental
emergency medicine, 3(2), 63.
[6]. Khalilnejad Tabari, M., Aghajari, S., &
Salahshour, S. (2022). Reinforcement Learning for
Scheduling in Emergency Departments. In 2022
IEEE International Conference on Healthcare
Informatics (ICHI) (pp. 10-16). IEEE.
[7]. Lee, H., Lee, J., Lee, H., Lee, J., Lee, Y., & Kim,
Y. (2021). Reducing Patient Waiting Time in
Emergency Departments Using Reinforcement
Learning and Multi-Agent Systems. Applied
Sciences, 11(8), 3429.
https://doi.org/10.3390/app11083429
[8]. Li, J., Cao, W., Yang, Y., Hu, S., & Wei, S.
(2021). Data-driven optimization of emergency
department operations using machine learning
algorithms. Health Care Management Science,
24(1), 32-46.
[9]. Liu, J., Xie, J., Cheng, X., Zhang, Z., Li, X., & Li,
Z. (2022). A Novel Reinforcement Learning
Algorithm for Emergency Department Patient
Flow Optimization. Journal of Healthcare
Engineering, 2022, 1-11.
[10]. Liu, Y., Liu, X., Li, Q., Zhu, Y., Ma, J., & Wang,
Y. (2021). An intelligent triage system for an
emergency department based on machine learning
and queuing theory. Journal of Medical Systems,
45(9), 1-10.
[11]. Mahesh, B. (2020). Machine learning algorithms-a
review. International Journal of Science and
Research (IJSR). [Internet], 9, 381-386.

IJISRT24MAR1035 www.ijisrt.com 3115